283 files changed, 87895 insertions, 12285 deletions

diff --git a/drivers/gpu/drm/Kconfig b/drivers/gpu/drm/Kconfig
index f831ea159291..305c59003963 100644
--- a/drivers/gpu/drm/Kconfig
+++ b/drivers/gpu/drm/Kconfig
@@ -66,6 +66,8 @@ config DRM_RADEON
 
 	  If M is selected, the module will be called radeon.
 
+source "drivers/gpu/drm/radeon/Kconfig"
+
 config DRM_I810
 	tristate "Intel I810"
 	depends on DRM && AGP && AGP_INTEL
@@ -92,6 +94,7 @@ config DRM_I830
 config DRM_I915
 	tristate "i915 driver"
 	depends on AGP_INTEL
+	select SHMEM
 	select DRM_KMS_HELPER
 	select FB_CFB_FILLRECT
 	select FB_CFB_COPYAREA
diff --git a/drivers/gpu/drm/Makefile b/drivers/gpu/drm/Makefile
index 3c8827a7aabd..abe3f446ca48 100644
--- a/drivers/gpu/drm/Makefile
+++ b/drivers/gpu/drm/Makefile
@@ -4,7 +4,7 @@
 
 ccflags-y := -Iinclude/drm
 
-drm-y       :=	drm_auth.o drm_bufs.o drm_cache.o \
+drm-y       :=	drm_auth.o drm_buffer.o drm_bufs.o drm_cache.o \
 		drm_context.o drm_dma.o drm_drawable.o \
 		drm_drv.o drm_fops.o drm_gem.o drm_ioctl.o drm_irq.o \
 		drm_lock.o drm_memory.o drm_proc.o drm_stub.o drm_vm.o \
@@ -15,7 +15,7 @@ drm-y       :=	drm_auth.o drm_bufs.o drm_cache.o \
 
 drm-$(CONFIG_COMPAT) += drm_ioc32.o
 
-drm_kms_helper-y := drm_fb_helper.o drm_crtc_helper.o
+drm_kms_helper-y := drm_fb_helper.o drm_crtc_helper.o drm_dp_i2c_helper.o
 
 obj-$(CONFIG_DRM_KMS_HELPER) += drm_kms_helper.o
 
@@ -30,4 +30,7 @@ obj-$(CONFIG_DRM_I830)	+= i830/
 obj-$(CONFIG_DRM_I915)  += i915/
 obj-$(CONFIG_DRM_SIS)   += sis/
 obj-$(CONFIG_DRM_SAVAGE)+= savage/
+obj-$(CONFIG_DRM_VMWGFX)+= vmwgfx/
 obj-$(CONFIG_DRM_VIA)	+=via/
+obj-$(CONFIG_DRM_NOUVEAU) +=nouveau/
+obj-y			+= i2c/
diff --git a/drivers/gpu/drm/ati_pcigart.c b/drivers/gpu/drm/ati_pcigart.c
index 628eae3e9b83..17be051b7aa3 100644
--- a/drivers/gpu/drm/ati_pcigart.c
+++ b/drivers/gpu/drm/ati_pcigart.c
@@ -39,8 +39,7 @@ static int drm_ati_alloc_pcigart_table(struct drm_device *dev,
 				       struct drm_ati_pcigart_info *gart_info)
 {
 	gart_info->table_handle = drm_pci_alloc(dev, gart_info->table_size,
-						PAGE_SIZE,
-						gart_info->table_mask);
+						PAGE_SIZE);
 	if (gart_info->table_handle == NULL)
 		return -ENOMEM;
 
@@ -112,6 +111,13 @@ int drm_ati_pcigart_init(struct drm_device *dev, struct drm_ati_pcigart_info *ga
 	if (gart_info->gart_table_location == DRM_ATI_GART_MAIN) {
 		DRM_DEBUG("PCI: no table in VRAM: using normal RAM\n");
 
+		if (pci_set_dma_mask(dev->pdev, gart_info->table_mask)) {
+			DRM_ERROR("fail to set dma mask to 0x%Lx\n",
+				  (unsigned long long)gart_info->table_mask);
+			ret = 1;
+			goto done;
+		}
+
 		ret = drm_ati_alloc_pcigart_table(dev, gart_info);
 		if (ret) {
 			DRM_ERROR("cannot allocate PCI GART page!\n");
diff --git a/drivers/gpu/drm/drm_buffer.c b/drivers/gpu/drm/drm_buffer.c
new file mode 100644
index 000000000000..55d03ed05000
--- /dev/null
+++ b/drivers/gpu/drm/drm_buffer.c
@@ -0,0 +1,184 @@
+/**************************************************************************
+ *
+ * Copyright 2010 Pauli Nieminen.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ *
+ **************************************************************************/
+/*
+ * Multipart buffer for coping data which is larger than the page size.
+ *
+ * Authors:
+ * Pauli Nieminen <suokkos-at-gmail-dot-com>
+ */
+
+#include "drm_buffer.h"
+
+/**
+ * Allocate the drm buffer object.
+ *
+ *   buf: Pointer to a pointer where the object is stored.
+ *   size: The number of bytes to allocate.
+ */
+int drm_buffer_alloc(struct drm_buffer **buf, int size)
+{
+	int nr_pages = size / PAGE_SIZE + 1;
+	int idx;
+
+	/* Allocating pointer table to end of structure makes drm_buffer
+	 * variable sized */
+	*buf = kzalloc(sizeof(struct drm_buffer) + nr_pages*sizeof(char *),
+			GFP_KERNEL);
+
+	if (*buf == NULL) {
+		DRM_ERROR("Failed to allocate drm buffer object to hold"
+				" %d bytes in %d pages.\n",
+				size, nr_pages);
+		return -ENOMEM;
+	}
+
+	(*buf)->size = size;
+
+	for (idx = 0; idx < nr_pages; ++idx) {
+
+		(*buf)->data[idx] =
+			kmalloc(min(PAGE_SIZE, size - idx * PAGE_SIZE),
+				GFP_KERNEL);
+
+
+		if ((*buf)->data[idx] == NULL) {
+			DRM_ERROR("Failed to allocate %dth page for drm"
+					" buffer with %d bytes and %d pages.\n",
+					idx + 1, size, nr_pages);
+			goto error_out;
+		}
+
+	}
+
+	return 0;
+
+error_out:
+
+	/* Only last element can be null pointer so check for it first. */
+	if ((*buf)->data[idx])
+		kfree((*buf)->data[idx]);
+
+	for (--idx; idx >= 0; --idx)
+		kfree((*buf)->data[idx]);
+
+	kfree(*buf);
+	return -ENOMEM;
+}
+EXPORT_SYMBOL(drm_buffer_alloc);
+
+/**
+ * Copy the user data to the begin of the buffer and reset the processing
+ * iterator.
+ *
+ *   user_data: A pointer the data that is copied to the buffer.
+ *   size: The Number of bytes to copy.
+ */
+extern int drm_buffer_copy_from_user(struct drm_buffer *buf,
+		void __user *user_data, int size)
+{
+	int nr_pages = size / PAGE_SIZE + 1;
+	int idx;
+
+	if (size > buf->size) {
+		DRM_ERROR("Requesting to copy %d bytes to a drm buffer with"
+				" %d bytes space\n",
+				size, buf->size);
+		return -EFAULT;
+	}
+
+	for (idx = 0; idx < nr_pages; ++idx) {
+
+		if (DRM_COPY_FROM_USER(buf->data[idx],
+			user_data + idx * PAGE_SIZE,
+			min(PAGE_SIZE, size - idx * PAGE_SIZE))) {
+			DRM_ERROR("Failed to copy user data (%p) to drm buffer"
+					" (%p) %dth page.\n",
+					user_data, buf, idx);
+			return -EFAULT;
+
+		}
+	}
+	buf->iterator = 0;
+	return 0;
+}
+EXPORT_SYMBOL(drm_buffer_copy_from_user);
+
+/**
+ * Free the drm buffer object
+ */
+void drm_buffer_free(struct drm_buffer *buf)
+{
+
+	if (buf != NULL) {
+
+		int nr_pages = buf->size / PAGE_SIZE + 1;
+		int idx;
+		for (idx = 0; idx < nr_pages; ++idx)
+			kfree(buf->data[idx]);
+
+		kfree(buf);
+	}
+}
+EXPORT_SYMBOL(drm_buffer_free);
+
+/**
+ * Read an object from buffer that may be split to multiple parts. If object
+ * is not split function just returns the pointer to object in buffer. But in
+ * case of split object data is copied to given stack object that is suplied
+ * by caller.
+ *
+ * The processing location of the buffer is also advanced to the next byte
+ * after the object.
+ *
+ *   objsize: The size of the objet in bytes.
+ *   stack_obj: A pointer to a memory location where object can be copied.
+ */
+void *drm_buffer_read_object(struct drm_buffer *buf,
+		int objsize, void *stack_obj)
+{
+	int idx = drm_buffer_index(buf);
+	int page = drm_buffer_page(buf);
+	void *obj = 0;
+
+	if (idx + objsize <= PAGE_SIZE) {
+		obj = &buf->data[page][idx];
+	} else {
+		/* The object is split which forces copy to temporary object.*/
+		int beginsz = PAGE_SIZE - idx;
+		memcpy(stack_obj, &buf->data[page][idx], beginsz);
+
+		memcpy(stack_obj + beginsz, &buf->data[page + 1][0],
+				objsize - beginsz);
+
+		obj = stack_obj;
+	}
+
+	drm_buffer_advance(buf, objsize);
+	return obj;
+}
+EXPORT_SYMBOL(drm_buffer_read_object);
diff --git a/drivers/gpu/drm/drm_bufs.c b/drivers/gpu/drm/drm_bufs.c
index 3d09e304f6f4..8417cc4c43f1 100644
--- a/drivers/gpu/drm/drm_bufs.c
+++ b/drivers/gpu/drm/drm_bufs.c
@@ -326,7 +326,7 @@ static int drm_addmap_core(struct drm_device * dev, resource_size_t offset,
 		 * As we're limiting the address to 2^32-1 (or less),
 		 * casting it down to 32 bits is no problem, but we
 		 * need to point to a 64bit variable first. */
-		dmah = drm_pci_alloc(dev, map->size, map->size, 0xffffffffUL);
+		dmah = drm_pci_alloc(dev, map->size, map->size);
 		if (!dmah) {
 			kfree(map);
 			return -ENOMEM;
@@ -885,7 +885,7 @@ int drm_addbufs_pci(struct drm_device * dev, struct drm_buf_desc * request)
 
 	while (entry->buf_count < count) {
 
-		dmah = drm_pci_alloc(dev, PAGE_SIZE << page_order, 0x1000, 0xfffffffful);
+		dmah = drm_pci_alloc(dev, PAGE_SIZE << page_order, 0x1000);
 
 		if (!dmah) {
 			/* Set count correctly so we free the proper amount. */
diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c
index ba728ad77f2a..d91fb8c0b7b3 100644
--- a/drivers/gpu/drm/drm_crtc.c
+++ b/drivers/gpu/drm/drm_crtc.c
@@ -125,6 +125,15 @@ static struct drm_prop_enum_list drm_tv_subconnector_enum_list[] =
 DRM_ENUM_NAME_FN(drm_get_tv_subconnector_name,
 		 drm_tv_subconnector_enum_list)
 
+static struct drm_prop_enum_list drm_dirty_info_enum_list[] = {
+	{ DRM_MODE_DIRTY_OFF,      "Off"      },
+	{ DRM_MODE_DIRTY_ON,       "On"       },
+	{ DRM_MODE_DIRTY_ANNOTATE, "Annotate" },
+};
+
+DRM_ENUM_NAME_FN(drm_get_dirty_info_name,
+		 drm_dirty_info_enum_list)
+
 struct drm_conn_prop_enum_list {
 	int type;
 	char *name;
@@ -149,6 +158,7 @@ static struct drm_conn_prop_enum_list drm_connector_enum_list[] =
 	{ DRM_MODE_CONNECTOR_HDMIA, "HDMI Type A", 0 },
 	{ DRM_MODE_CONNECTOR_HDMIB, "HDMI Type B", 0 },
 	{ DRM_MODE_CONNECTOR_TV, "TV", 0 },
+	{ DRM_MODE_CONNECTOR_eDP, "Embedded DisplayPort", 0 },
 };
 
 static struct drm_prop_enum_list drm_encoder_enum_list[] =
@@ -247,7 +257,8 @@ static void drm_mode_object_put(struct drm_device *dev,
 	mutex_unlock(&dev->mode_config.idr_mutex);
 }
 
-void *drm_mode_object_find(struct drm_device *dev, uint32_t id, uint32_t type)
+struct drm_mode_object *drm_mode_object_find(struct drm_device *dev,
+		uint32_t id, uint32_t type)
 {
 	struct drm_mode_object *obj = NULL;
 
@@ -272,7 +283,7 @@ EXPORT_SYMBOL(drm_mode_object_find);
  * functions & device file and adds it to the master fd list.
  *
  * RETURNS:
- * Zero on success, error code on falure.
+ * Zero on success, error code on failure.
  */
 int drm_framebuffer_init(struct drm_device *dev, struct drm_framebuffer *fb,
 			 const struct drm_framebuffer_funcs *funcs)
@@ -482,6 +493,7 @@ void drm_connector_cleanup(struct drm_connector *connector)
 	list_for_each_entry_safe(mode, t, &connector->user_modes, head)
 		drm_mode_remove(connector, mode);
 
+	kfree(connector->fb_helper_private);
 	mutex_lock(&dev->mode_config.mutex);
 	drm_mode_object_put(dev, &connector->base);
 	list_del(&connector->head);
@@ -801,6 +813,36 @@ int drm_mode_create_dithering_property(struct drm_device *dev)
 EXPORT_SYMBOL(drm_mode_create_dithering_property);
 
 /**
+ * drm_mode_create_dirty_property - create dirty property
+ * @dev: DRM device
+ *
+ * Called by a driver the first time it's needed, must be attached to desired
+ * connectors.
+ */
+int drm_mode_create_dirty_info_property(struct drm_device *dev)
+{
+	struct drm_property *dirty_info;
+	int i;
+
+	if (dev->mode_config.dirty_info_property)
+		return 0;
+
+	dirty_info =
+		drm_property_create(dev, DRM_MODE_PROP_ENUM |
+				    DRM_MODE_PROP_IMMUTABLE,
+				    "dirty",
+				    ARRAY_SIZE(drm_dirty_info_enum_list));
+	for (i = 0; i < ARRAY_SIZE(drm_dirty_info_enum_list); i++)
+		drm_property_add_enum(dirty_info, i,
+				      drm_dirty_info_enum_list[i].type,
+				      drm_dirty_info_enum_list[i].name);
+	dev->mode_config.dirty_info_property = dirty_info;
+
+	return 0;
+}
+EXPORT_SYMBOL(drm_mode_create_dirty_info_property);
+
+/**
  * drm_mode_config_init - initialize DRM mode_configuration structure
  * @dev: DRM device
  *
@@ -1555,8 +1597,6 @@ int drm_mode_cursor_ioctl(struct drm_device *dev,
 	struct drm_crtc *crtc;
 	int ret = 0;
 
-	DRM_DEBUG_KMS("\n");
-
 	if (!req->flags) {
 		DRM_ERROR("no operation set\n");
 		return -EINVAL;
@@ -1754,6 +1794,71 @@ out:
 	return ret;
 }
 
+int drm_mode_dirtyfb_ioctl(struct drm_device *dev,
+			   void *data, struct drm_file *file_priv)
+{
+	struct drm_clip_rect __user *clips_ptr;
+	struct drm_clip_rect *clips = NULL;
+	struct drm_mode_fb_dirty_cmd *r = data;
+	struct drm_mode_object *obj;
+	struct drm_framebuffer *fb;
+	unsigned flags;
+	int num_clips;
+	int ret = 0;
+
+	mutex_lock(&dev->mode_config.mutex);
+	obj = drm_mode_object_find(dev, r->fb_id, DRM_MODE_OBJECT_FB);
+	if (!obj) {
+		DRM_ERROR("invalid framebuffer id\n");
+		ret = -EINVAL;
+		goto out_err1;
+	}
+	fb = obj_to_fb(obj);
+
+	num_clips = r->num_clips;
+	clips_ptr = (struct drm_clip_rect *)(unsigned long)r->clips_ptr;
+
+	if (!num_clips != !clips_ptr) {
+		ret = -EINVAL;
+		goto out_err1;
+	}
+
+	flags = DRM_MODE_FB_DIRTY_FLAGS & r->flags;
+
+	/* If userspace annotates copy, clips must come in pairs */
+	if (flags & DRM_MODE_FB_DIRTY_ANNOTATE_COPY && (num_clips % 2)) {
+		ret = -EINVAL;
+		goto out_err1;
+	}
+
+	if (num_clips && clips_ptr) {
+		clips = kzalloc(num_clips * sizeof(*clips), GFP_KERNEL);
+		if (!clips) {
+			ret = -ENOMEM;
+			goto out_err1;
+		}
+
+		ret = copy_from_user(clips, clips_ptr,
+				     num_clips * sizeof(*clips));
+		if (ret)
+			goto out_err2;
+	}
+
+	if (fb->funcs->dirty) {
+		ret = fb->funcs->dirty(fb, flags, r->color, clips, num_clips);
+	} else {
+		ret = -ENOSYS;
+		goto out_err2;
+	}
+
+out_err2:
+	kfree(clips);
+out_err1:
+	mutex_unlock(&dev->mode_config.mutex);
+	return ret;
+}
+
+
 /**
  * drm_fb_release - remove and free the FBs on this file
  * @filp: file * from the ioctl
@@ -2329,7 +2434,7 @@ int drm_mode_connector_property_set_ioctl(struct drm_device *dev,
 	} else if (connector->funcs->set_property)
 		ret = connector->funcs->set_property(connector, property, out_resp->value);
 
-	/* store the property value if succesful */
+	/* store the property value if successful */
 	if (!ret)
 		drm_connector_property_set_value(connector, property, out_resp->value);
 out:
@@ -2479,3 +2584,72 @@ out:
 	mutex_unlock(&dev->mode_config.mutex);
 	return ret;
 }
+
+int drm_mode_page_flip_ioctl(struct drm_device *dev,
+			     void *data, struct drm_file *file_priv)
+{
+	struct drm_mode_crtc_page_flip *page_flip = data;
+	struct drm_mode_object *obj;
+	struct drm_crtc *crtc;
+	struct drm_framebuffer *fb;
+	struct drm_pending_vblank_event *e = NULL;
+	unsigned long flags;
+	int ret = -EINVAL;
+
+	if (page_flip->flags & ~DRM_MODE_PAGE_FLIP_FLAGS ||
+	    page_flip->reserved != 0)
+		return -EINVAL;
+
+	mutex_lock(&dev->mode_config.mutex);
+	obj = drm_mode_object_find(dev, page_flip->crtc_id, DRM_MODE_OBJECT_CRTC);
+	if (!obj)
+		goto out;
+	crtc = obj_to_crtc(obj);
+
+	if (crtc->funcs->page_flip == NULL)
+		goto out;
+
+	obj = drm_mode_object_find(dev, page_flip->fb_id, DRM_MODE_OBJECT_FB);
+	if (!obj)
+		goto out;
+	fb = obj_to_fb(obj);
+
+	if (page_flip->flags & DRM_MODE_PAGE_FLIP_EVENT) {
+		ret = -ENOMEM;
+		spin_lock_irqsave(&dev->event_lock, flags);
+		if (file_priv->event_space < sizeof e->event) {
+			spin_unlock_irqrestore(&dev->event_lock, flags);
+			goto out;
+		}
+		file_priv->event_space -= sizeof e->event;
+		spin_unlock_irqrestore(&dev->event_lock, flags);
+
+		e = kzalloc(sizeof *e, GFP_KERNEL);
+		if (e == NULL) {
+			spin_lock_irqsave(&dev->event_lock, flags);
+			file_priv->event_space += sizeof e->event;
+			spin_unlock_irqrestore(&dev->event_lock, flags);
+			goto out;
+		}
+
+		e->event.base.type = DRM_EVENT_FLIP_COMPLETE;
+		e->event.base.length = sizeof e->event;
+		e->event.user_data = page_flip->user_data;
+		e->base.event = &e->event.base;
+		e->base.file_priv = file_priv;
+		e->base.destroy =
+			(void (*) (struct drm_pending_event *)) kfree;
+	}
+
+	ret = crtc->funcs->page_flip(crtc, fb, e);
+	if (ret) {
+		spin_lock_irqsave(&dev->event_lock, flags);
+		file_priv->event_space += sizeof e->event;
+		spin_unlock_irqrestore(&dev->event_lock, flags);
+		kfree(e);
+	}
+
+out:
+	mutex_unlock(&dev->mode_config.mutex);
+	return ret;
+}
diff --git a/drivers/gpu/drm/drm_crtc_helper.c b/drivers/gpu/drm/drm_crtc_helper.c
index fe8697447f32..f2aaf39be398 100644
--- a/drivers/gpu/drm/drm_crtc_helper.c
+++ b/drivers/gpu/drm/drm_crtc_helper.c
@@ -32,6 +32,7 @@
 #include "drmP.h"
 #include "drm_crtc.h"
 #include "drm_crtc_helper.h"
+#include "drm_fb_helper.h"
 
 static void drm_mode_validate_flag(struct drm_connector *connector,
 				   int flags)
@@ -90,7 +91,15 @@ int drm_helper_probe_single_connector_modes(struct drm_connector *connector,
 	list_for_each_entry_safe(mode, t, &connector->modes, head)
 		mode->status = MODE_UNVERIFIED;
 
-	connector->status = connector->funcs->detect(connector);
+	if (connector->force) {
+		if (connector->force == DRM_FORCE_ON)
+			connector->status = connector_status_connected;
+		else
+			connector->status = connector_status_disconnected;
+		if (connector->funcs->force)
+			connector->funcs->force(connector);
+	} else
+		connector->status = connector->funcs->detect(connector);
 
 	if (connector->status == connector_status_disconnected) {
 		DRM_DEBUG_KMS("%s is disconnected\n",
@@ -100,7 +109,7 @@ int drm_helper_probe_single_connector_modes(struct drm_connector *connector,
 
 	count = (*connector_funcs->get_modes)(connector);
 	if (!count) {
-		count = drm_add_modes_noedid(connector, 800, 600);
+		count = drm_add_modes_noedid(connector, 1024, 768);
 		if (!count)
 			return 0;
 	}
@@ -207,7 +216,7 @@ bool drm_helper_crtc_in_use(struct drm_crtc *crtc)
 EXPORT_SYMBOL(drm_helper_crtc_in_use);
 
 /**
- * drm_disable_unused_functions - disable unused objects
+ * drm_helper_disable_unused_functions - disable unused objects
  * @dev: DRM device
  *
  * LOCKING:
@@ -267,6 +276,66 @@ static struct drm_display_mode *drm_has_preferred_mode(struct drm_connector *con
 	return NULL;
 }
 
+static bool drm_has_cmdline_mode(struct drm_connector *connector)
+{
+	struct drm_fb_helper_connector *fb_help_conn = connector->fb_helper_private;
+	struct drm_fb_helper_cmdline_mode *cmdline_mode;
+
+	if (!fb_help_conn)
+		return false;
+
+	cmdline_mode = &fb_help_conn->cmdline_mode;
+	return cmdline_mode->specified;
+}
+
+static struct drm_display_mode *drm_pick_cmdline_mode(struct drm_connector *connector, int width, int height)
+{
+	struct drm_fb_helper_connector *fb_help_conn = connector->fb_helper_private;
+	struct drm_fb_helper_cmdline_mode *cmdline_mode;
+	struct drm_display_mode *mode = NULL;
+
+	if (!fb_help_conn)
+		return mode;
+
+	cmdline_mode = &fb_help_conn->cmdline_mode;
+	if (cmdline_mode->specified == false)
+		return mode;
+
+	/* attempt to find a matching mode in the list of modes
+	 *  we have gotten so far, if not add a CVT mode that conforms
+	 */
+	if (cmdline_mode->rb || cmdline_mode->margins)
+		goto create_mode;
+
+	list_for_each_entry(mode, &connector->modes, head) {
+		/* check width/height */
+		if (mode->hdisplay != cmdline_mode->xres ||
+		    mode->vdisplay != cmdline_mode->yres)
+			continue;
+
+		if (cmdline_mode->refresh_specified) {
+			if (mode->vrefresh != cmdline_mode->refresh)
+				continue;
+		}
+
+		if (cmdline_mode->interlace) {
+			if (!(mode->flags & DRM_MODE_FLAG_INTERLACE))
+				continue;
+		}
+		return mode;
+	}
+
+create_mode:
+	mode = drm_cvt_mode(connector->dev, cmdline_mode->xres,
+			    cmdline_mode->yres,
+			    cmdline_mode->refresh_specified ? cmdline_mode->refresh : 60,
+			    cmdline_mode->rb, cmdline_mode->interlace,
+			    cmdline_mode->margins);
+	drm_mode_set_crtcinfo(mode, CRTC_INTERLACE_HALVE_V);
+	list_add(&mode->head, &connector->modes);
+	return mode;
+}
+
 static bool drm_connector_enabled(struct drm_connector *connector, bool strict)
 {
 	bool enable;
@@ -317,10 +386,16 @@ static bool drm_target_preferred(struct drm_device *dev,
 			continue;
 		}
 
-		DRM_DEBUG_KMS("looking for preferred mode on connector %d\n",
-			  connector->base.id);
+		DRM_DEBUG_KMS("looking for cmdline mode on connector %d\n",
+			      connector->base.id);
 
-		modes[i] = drm_has_preferred_mode(connector, width, height);
+		/* got for command line mode first */
+		modes[i] = drm_pick_cmdline_mode(connector, width, height);
+		if (!modes[i]) {
+			DRM_DEBUG_KMS("looking for preferred mode on connector %d\n",
+				      connector->base.id);
+			modes[i] = drm_has_preferred_mode(connector, width, height);
+		}
 		/* No preferred modes, pick one off the list */
 		if (!modes[i] && !list_empty(&connector->modes)) {
 			list_for_each_entry(modes[i], &connector->modes, head)
@@ -369,6 +444,8 @@ static int drm_pick_crtcs(struct drm_device *dev,
 	my_score = 1;
 	if (connector->status == connector_status_connected)
 		my_score++;
+	if (drm_has_cmdline_mode(connector))
+		my_score++;
 	if (drm_has_preferred_mode(connector, width, height))
 		my_score++;
 
@@ -625,7 +702,7 @@ bool drm_crtc_helper_set_mode(struct drm_crtc *crtc,
 		if (encoder->crtc != crtc)
 			continue;
 
-		DRM_INFO("%s: set mode %s %x\n", drm_get_encoder_name(encoder),
+		DRM_DEBUG("%s: set mode %s %x\n", drm_get_encoder_name(encoder),
 			 mode->name, mode->base.id);
 		encoder_funcs = encoder->helper_private;
 		encoder_funcs->mode_set(encoder, mode, adjusted_mode);
@@ -759,11 +836,7 @@ int drm_crtc_helper_set_config(struct drm_mode_set *set)
 			mode_changed = true;
 		} else if (set->fb == NULL) {
 			mode_changed = true;
-		} else if ((set->fb->bits_per_pixel !=
-			 set->crtc->fb->bits_per_pixel) ||
-			 set->fb->depth != set->crtc->fb->depth)
-			fb_changed = true;
-		else
+		} else
 			fb_changed = true;
 	}
 
@@ -943,6 +1016,11 @@ bool drm_helper_initial_config(struct drm_device *dev)
 {
 	int count = 0;
 
+	/* disable all the possible outputs/crtcs before entering KMS mode */
+	drm_helper_disable_unused_functions(dev);
+
+	drm_fb_helper_parse_command_line(dev);
+
 	count = drm_helper_probe_connector_modes(dev,
 						 dev->mode_config.max_width,
 						 dev->mode_config.max_height);
@@ -950,7 +1028,8 @@ bool drm_helper_initial_config(struct drm_device *dev)
 	/*
 	 * we shouldn't end up with no modes here.
 	 */
-	WARN(!count, "Connected connector with 0 modes\n");
+	if (count == 0)
+		printk(KERN_INFO "No connectors reported connected with modes\n");
 
 	drm_setup_crtcs(dev);
 
@@ -1080,6 +1159,9 @@ EXPORT_SYMBOL(drm_helper_mode_fill_fb_struct);
 int drm_helper_resume_force_mode(struct drm_device *dev)
 {
 	struct drm_crtc *crtc;
+	struct drm_encoder *encoder;
+	struct drm_encoder_helper_funcs *encoder_funcs;
+	struct drm_crtc_helper_funcs *crtc_funcs;
 	int ret;
 
 	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
@@ -1092,6 +1174,25 @@ int drm_helper_resume_force_mode(struct drm_device *dev)
 
 		if (ret == false)
 			DRM_ERROR("failed to set mode on crtc %p\n", crtc);
+
+		/* Turn off outputs that were already powered off */
+		if (drm_helper_choose_crtc_dpms(crtc)) {
+			list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) {
+
+				if(encoder->crtc != crtc)
+					continue;
+
+				encoder_funcs = encoder->helper_private;
+				if (encoder_funcs->dpms)
+					(*encoder_funcs->dpms) (encoder,
+								drm_helper_choose_encoder_dpms(encoder));
+
+				crtc_funcs = crtc->helper_private;
+				if (crtc_funcs->dpms)
+					(*crtc_funcs->dpms) (crtc,
+							     drm_helper_choose_crtc_dpms(crtc));
+			}
+		}
 	}
 	/* disable the unused connectors while restoring the modesetting */
 	drm_helper_disable_unused_functions(dev);
diff --git a/drivers/gpu/drm/i915/intel_dp_i2c.c b/drivers/gpu/drm/drm_dp_i2c_helper.c
index a63b6f57d2d4..548887c8506f 100644
--- a/drivers/gpu/drm/i915/intel_dp_i2c.c
+++ b/drivers/gpu/drm/drm_dp_i2c_helper.c
@@ -28,84 +28,20 @@
 #include <linux/errno.h>
 #include <linux/sched.h>
 #include <linux/i2c.h>
-#include "intel_dp.h"
+#include "drm_dp_helper.h"
 #include "drmP.h"
 
 /* Run a single AUX_CH I2C transaction, writing/reading data as necessary */
-
-#define MODE_I2C_START	1
-#define MODE_I2C_WRITE	2
-#define MODE_I2C_READ	4
-#define MODE_I2C_STOP	8
-
 static int
 i2c_algo_dp_aux_transaction(struct i2c_adapter *adapter, int mode,
 			    uint8_t write_byte, uint8_t *read_byte)
 {
 	struct i2c_algo_dp_aux_data *algo_data = adapter->algo_data;
-	uint16_t address = algo_data->address;
-	uint8_t msg[5];
-	uint8_t reply[2];
-	int msg_bytes;
-	int reply_bytes;
 	int ret;
-
-	/* Set up the command byte */
-	if (mode & MODE_I2C_READ)
-		msg[0] = AUX_I2C_READ << 4;
-	else
-		msg[0] = AUX_I2C_WRITE << 4;
-
-	if (!(mode & MODE_I2C_STOP))
-		msg[0] |= AUX_I2C_MOT << 4;
-
-	msg[1] = address >> 8;
-	msg[2] = address;
-
-	switch (mode) {
-	case MODE_I2C_WRITE:
-		msg[3] = 0;
-		msg[4] = write_byte;
-		msg_bytes = 5;
-		reply_bytes = 1;
-		break;
-	case MODE_I2C_READ:
-		msg[3] = 0;
-		msg_bytes = 4;
-		reply_bytes = 2;
-		break;
-	default:
-		msg_bytes = 3;
-		reply_bytes = 1;
-		break;
-	}
-
-	for (;;) {
-		ret = (*algo_data->aux_ch)(adapter,
-					   msg, msg_bytes,
-					   reply, reply_bytes);
-		if (ret < 0) {
-			DRM_DEBUG("aux_ch failed %d\n", ret);
-			return ret;
-		}
-		switch (reply[0] & AUX_I2C_REPLY_MASK) {
-		case AUX_I2C_REPLY_ACK:
-			if (mode == MODE_I2C_READ) {
-				*read_byte = reply[1];
-			}
-			return reply_bytes - 1;
-		case AUX_I2C_REPLY_NACK:
-			DRM_DEBUG("aux_ch nack\n");
-			return -EREMOTEIO;
-		case AUX_I2C_REPLY_DEFER:
-			DRM_DEBUG("aux_ch defer\n");
-			udelay(100);
-			break;
-		default:
-			DRM_ERROR("aux_ch invalid reply 0x%02x\n", reply[0]);
-			return -EREMOTEIO;
-		}
-	}
+	
+	ret = (*algo_data->aux_ch)(adapter, mode,
+				   write_byte, read_byte);
+	return ret;
 }
 
 /*
@@ -224,7 +160,7 @@ i2c_algo_dp_aux_xfer(struct i2c_adapter *adapter,
 	if (ret >= 0)
 		ret = num;
 	i2c_algo_dp_aux_stop(adapter, reading);
-	DRM_DEBUG("dp_aux_xfer return %d\n", ret);
+	DRM_DEBUG_KMS("dp_aux_xfer return %d\n", ret);
 	return ret;
 }
 
diff --git a/drivers/gpu/drm/drm_drv.c b/drivers/gpu/drm/drm_drv.c
index a75ca63deea6..f3c58e2bd75c 100644
--- a/drivers/gpu/drm/drm_drv.c
+++ b/drivers/gpu/drm/drm_drv.c
@@ -125,26 +125,28 @@ static struct drm_ioctl_desc drm_ioctls[] = {
 
 	DRM_IOCTL_DEF(DRM_IOCTL_UPDATE_DRAW, drm_update_drawable_info, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY),
 
-	DRM_IOCTL_DEF(DRM_IOCTL_GEM_CLOSE, drm_gem_close_ioctl, 0),
-	DRM_IOCTL_DEF(DRM_IOCTL_GEM_FLINK, drm_gem_flink_ioctl, DRM_AUTH),
-	DRM_IOCTL_DEF(DRM_IOCTL_GEM_OPEN, drm_gem_open_ioctl, DRM_AUTH),
-
-	DRM_IOCTL_DEF(DRM_IOCTL_MODE_GETRESOURCES, drm_mode_getresources, DRM_MASTER|DRM_CONTROL_ALLOW),
-	DRM_IOCTL_DEF(DRM_IOCTL_MODE_GETCRTC, drm_mode_getcrtc, DRM_MASTER|DRM_CONTROL_ALLOW),
-	DRM_IOCTL_DEF(DRM_IOCTL_MODE_SETCRTC, drm_mode_setcrtc, DRM_MASTER|DRM_CONTROL_ALLOW),
-	DRM_IOCTL_DEF(DRM_IOCTL_MODE_CURSOR, drm_mode_cursor_ioctl, DRM_MASTER|DRM_CONTROL_ALLOW),
-	DRM_IOCTL_DEF(DRM_IOCTL_MODE_GETGAMMA, drm_mode_gamma_get_ioctl, DRM_MASTER),
-	DRM_IOCTL_DEF(DRM_IOCTL_MODE_SETGAMMA, drm_mode_gamma_set_ioctl, DRM_MASTER),
-	DRM_IOCTL_DEF(DRM_IOCTL_MODE_GETENCODER, drm_mode_getencoder, DRM_MASTER|DRM_CONTROL_ALLOW),
-	DRM_IOCTL_DEF(DRM_IOCTL_MODE_GETCONNECTOR, drm_mode_getconnector, DRM_MASTER|DRM_CONTROL_ALLOW),
-	DRM_IOCTL_DEF(DRM_IOCTL_MODE_ATTACHMODE, drm_mode_attachmode_ioctl, DRM_MASTER|DRM_CONTROL_ALLOW),
-	DRM_IOCTL_DEF(DRM_IOCTL_MODE_DETACHMODE, drm_mode_detachmode_ioctl, DRM_MASTER|DRM_CONTROL_ALLOW),
-	DRM_IOCTL_DEF(DRM_IOCTL_MODE_GETPROPERTY, drm_mode_getproperty_ioctl, DRM_MASTER | DRM_CONTROL_ALLOW),
-	DRM_IOCTL_DEF(DRM_IOCTL_MODE_SETPROPERTY, drm_mode_connector_property_set_ioctl, DRM_MASTER|DRM_CONTROL_ALLOW),
-	DRM_IOCTL_DEF(DRM_IOCTL_MODE_GETPROPBLOB, drm_mode_getblob_ioctl, DRM_MASTER|DRM_CONTROL_ALLOW),
-	DRM_IOCTL_DEF(DRM_IOCTL_MODE_GETFB, drm_mode_getfb, DRM_MASTER|DRM_CONTROL_ALLOW),
-	DRM_IOCTL_DEF(DRM_IOCTL_MODE_ADDFB, drm_mode_addfb, DRM_MASTER|DRM_CONTROL_ALLOW),
-	DRM_IOCTL_DEF(DRM_IOCTL_MODE_RMFB, drm_mode_rmfb, DRM_MASTER|DRM_CONTROL_ALLOW),
+	DRM_IOCTL_DEF(DRM_IOCTL_GEM_CLOSE, drm_gem_close_ioctl, DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_IOCTL_GEM_FLINK, drm_gem_flink_ioctl, DRM_AUTH|DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_IOCTL_GEM_OPEN, drm_gem_open_ioctl, DRM_AUTH|DRM_UNLOCKED),
+
+	DRM_IOCTL_DEF(DRM_IOCTL_MODE_GETRESOURCES, drm_mode_getresources, DRM_MASTER|DRM_CONTROL_ALLOW|DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_IOCTL_MODE_GETCRTC, drm_mode_getcrtc, DRM_MASTER|DRM_CONTROL_ALLOW|DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_IOCTL_MODE_SETCRTC, drm_mode_setcrtc, DRM_MASTER|DRM_CONTROL_ALLOW|DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_IOCTL_MODE_CURSOR, drm_mode_cursor_ioctl, DRM_MASTER|DRM_CONTROL_ALLOW|DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_IOCTL_MODE_GETGAMMA, drm_mode_gamma_get_ioctl, DRM_MASTER|DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_IOCTL_MODE_SETGAMMA, drm_mode_gamma_set_ioctl, DRM_MASTER|DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_IOCTL_MODE_GETENCODER, drm_mode_getencoder, DRM_MASTER|DRM_CONTROL_ALLOW|DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_IOCTL_MODE_GETCONNECTOR, drm_mode_getconnector, DRM_MASTER|DRM_CONTROL_ALLOW|DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_IOCTL_MODE_ATTACHMODE, drm_mode_attachmode_ioctl, DRM_MASTER|DRM_CONTROL_ALLOW|DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_IOCTL_MODE_DETACHMODE, drm_mode_detachmode_ioctl, DRM_MASTER|DRM_CONTROL_ALLOW|DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_IOCTL_MODE_GETPROPERTY, drm_mode_getproperty_ioctl, DRM_MASTER | DRM_CONTROL_ALLOW|DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_IOCTL_MODE_SETPROPERTY, drm_mode_connector_property_set_ioctl, DRM_MASTER|DRM_CONTROL_ALLOW|DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_IOCTL_MODE_GETPROPBLOB, drm_mode_getblob_ioctl, DRM_MASTER|DRM_CONTROL_ALLOW|DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_IOCTL_MODE_GETFB, drm_mode_getfb, DRM_MASTER|DRM_CONTROL_ALLOW|DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_IOCTL_MODE_ADDFB, drm_mode_addfb, DRM_MASTER|DRM_CONTROL_ALLOW|DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_IOCTL_MODE_RMFB, drm_mode_rmfb, DRM_MASTER|DRM_CONTROL_ALLOW|DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_IOCTL_MODE_PAGE_FLIP, drm_mode_page_flip_ioctl, DRM_MASTER|DRM_CONTROL_ALLOW|DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_IOCTL_MODE_DIRTYFB, drm_mode_dirtyfb_ioctl, DRM_MASTER|DRM_CONTROL_ALLOW|DRM_UNLOCKED)
 };
 
 #define DRM_CORE_IOCTL_COUNT	ARRAY_SIZE( drm_ioctls )
@@ -366,6 +368,29 @@ module_init(drm_core_init);
 module_exit(drm_core_exit);
 
 /**
+ * Copy and IOCTL return string to user space
+ */
+static int drm_copy_field(char *buf, size_t *buf_len, const char *value)
+{
+	int len;
+
+	/* don't overflow userbuf */
+	len = strlen(value);
+	if (len > *buf_len)
+		len = *buf_len;
+
+	/* let userspace know exact length of driver value (which could be
+	 * larger than the userspace-supplied buffer) */
+	*buf_len = strlen(value);
+
+	/* finally, try filling in the userbuf */
+	if (len && buf)
+		if (copy_to_user(buf, value, len))
+			return -EFAULT;
+	return 0;
+}
+
+/**
  * Get version information
  *
  * \param inode device inode.
@@ -380,16 +405,21 @@ static int drm_version(struct drm_device *dev, void *data,
 		       struct drm_file *file_priv)
 {
 	struct drm_version *version = data;
-	int len;
+	int err;
 
 	version->version_major = dev->driver->major;
 	version->version_minor = dev->driver->minor;
 	version->version_patchlevel = dev->driver->patchlevel;
-	DRM_COPY(version->name, dev->driver->name);
-	DRM_COPY(version->date, dev->driver->date);
-	DRM_COPY(version->desc, dev->driver->desc);
-
-	return 0;
+	err = drm_copy_field(version->name, &version->name_len,
+			dev->driver->name);
+	if (!err)
+		err = drm_copy_field(version->date, &version->date_len,
+				dev->driver->date);
+	if (!err)
+		err = drm_copy_field(version->desc, &version->desc_len,
+				dev->driver->desc);
+
+	return err;
 }
 
 /**
@@ -404,11 +434,11 @@ static int drm_version(struct drm_device *dev, void *data,
  * Looks up the ioctl function in the ::ioctls table, checking for root
  * previleges if so required, and dispatches to the respective function.
  */
-int drm_ioctl(struct inode *inode, struct file *filp,
+long drm_ioctl(struct file *filp,
 	      unsigned int cmd, unsigned long arg)
 {
 	struct drm_file *file_priv = filp->private_data;
-	struct drm_device *dev = file_priv->minor->dev;
+	struct drm_device *dev;
 	struct drm_ioctl_desc *ioctl;
 	drm_ioctl_t *func;
 	unsigned int nr = DRM_IOCTL_NR(cmd);
@@ -416,6 +446,7 @@ int drm_ioctl(struct inode *inode, struct file *filp,
 	char stack_kdata[128];
 	char *kdata = NULL;
 
+	dev = file_priv->minor->dev;
 	atomic_inc(&dev->ioctl_count);
 	atomic_inc(&dev->counts[_DRM_STAT_IOCTLS]);
 	++file_priv->ioctl_count;
@@ -471,7 +502,13 @@ int drm_ioctl(struct inode *inode, struct file *filp,
 				goto err_i1;
 			}
 		}
-		retcode = func(dev, kdata, file_priv);
+		if (ioctl->flags & DRM_UNLOCKED)
+			retcode = func(dev, kdata, file_priv);
+		else {
+			lock_kernel();
+			retcode = func(dev, kdata, file_priv);
+			unlock_kernel();
+		}
 
 		if (cmd & IOC_OUT) {
 			if (copy_to_user((void __user *)arg, kdata,
diff --git a/drivers/gpu/drm/drm_edid.c b/drivers/gpu/drm/drm_edid.c
index 90d76bacff17..f97e7c42ac8e 100644
--- a/drivers/gpu/drm/drm_edid.c
+++ b/drivers/gpu/drm/drm_edid.c
@@ -60,8 +60,7 @@
 #define EDID_QUIRK_FIRST_DETAILED_PREFERRED	(1 << 5)
 /* use +hsync +vsync for detailed mode */
 #define EDID_QUIRK_DETAILED_SYNC_PP		(1 << 6)
-/* define the number of Extension EDID block */
-#define MAX_EDID_EXT_NUM 4
+
 
 #define LEVEL_DMT	0
 #define LEVEL_GTF	1
@@ -109,30 +108,34 @@ static struct edid_quirk {
 
 
 /* Valid EDID header has these bytes */
-static u8 edid_header[] = { 0x00, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0x00 };
+static const u8 edid_header[] = {
+	0x00, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0x00
+};
 
 /**
- * edid_is_valid - sanity check EDID data
+ * drm_edid_is_valid - sanity check EDID data
  * @edid: EDID data
  *
  * Sanity check the EDID block by looking at the header, the version number
  * and the checksum.  Return 0 if the EDID doesn't check out, or 1 if it's
  * valid.
  */
-static bool edid_is_valid(struct edid *edid)
+bool drm_edid_is_valid(struct edid *edid)
 {
-	int i;
+	int i, score = 0;
 	u8 csum = 0;
 	u8 *raw_edid = (u8 *)edid;
 
-	if (memcmp(edid->header, edid_header, sizeof(edid_header)))
-		goto bad;
-	if (edid->version != 1) {
-		DRM_ERROR("EDID has major version %d, instead of 1\n", edid->version);
+	for (i = 0; i < sizeof(edid_header); i++)
+		if (raw_edid[i] == edid_header[i])
+			score++;
+
+	if (score == 8) ;
+	else if (score >= 6) {
+		DRM_DEBUG("Fixing EDID header, your hardware may be failing\n");
+		memcpy(raw_edid, edid_header, sizeof(edid_header));
+	} else
 		goto bad;
-	}
-	if (edid->revision > 4)
-		DRM_DEBUG("EDID minor > 4, assuming backward compatibility\n");
 
 	for (i = 0; i < EDID_LENGTH; i++)
 		csum += raw_edid[i];
@@ -141,6 +144,14 @@ static bool edid_is_valid(struct edid *edid)
 		goto bad;
 	}
 
+	if (edid->version != 1) {
+		DRM_ERROR("EDID has major version %d, instead of 1\n", edid->version);
+		goto bad;
+	}
+
+	if (edid->revision > 4)
+		DRM_DEBUG("EDID minor > 4, assuming backward compatibility\n");
+
 	return 1;
 
 bad:
@@ -151,6 +162,7 @@ bad:
 	}
 	return 0;
 }
+EXPORT_SYMBOL(drm_edid_is_valid);
 
 /**
  * edid_vendor - match a string against EDID's obfuscated vendor field
@@ -479,16 +491,17 @@ static struct drm_display_mode drm_dmt_modes[] = {
 		   3048, 3536, 0, 1600, 1603, 1609, 1682, 0,
 		   DRM_MODE_FLAG_NHSYNC | DRM_MODE_FLAG_PVSYNC) },
 };
+static const int drm_num_dmt_modes =
+	sizeof(drm_dmt_modes) / sizeof(struct drm_display_mode);
 
 static struct drm_display_mode *drm_find_dmt(struct drm_device *dev,
 			int hsize, int vsize, int fresh)
 {
-	int i, count;
+	int i;
 	struct drm_display_mode *ptr, *mode;
 
-	count = sizeof(drm_dmt_modes) / sizeof(struct drm_display_mode);
 	mode = NULL;
-	for (i = 0; i < count; i++) {
+	for (i = 0; i < drm_num_dmt_modes; i++) {
 		ptr = &drm_dmt_modes[i];
 		if (hsize == ptr->hdisplay &&
 			vsize == ptr->vdisplay &&
@@ -500,6 +513,19 @@ static struct drm_display_mode *drm_find_dmt(struct drm_device *dev,
 	}
 	return mode;
 }
+
+/*
+ * 0 is reserved.  The spec says 0x01 fill for unused timings.  Some old
+ * monitors fill with ascii space (0x20) instead.
+ */
+static int
+bad_std_timing(u8 a, u8 b)
+{
+	return (a == 0x00 && b == 0x00) ||
+	       (a == 0x01 && b == 0x01) ||
+	       (a == 0x20 && b == 0x20);
+}
+
 /**
  * drm_mode_std - convert standard mode info (width, height, refresh) into mode
  * @t: standard timing params
@@ -513,6 +539,7 @@ static struct drm_display_mode *drm_find_dmt(struct drm_device *dev,
  */
 struct drm_display_mode *drm_mode_std(struct drm_device *dev,
 				      struct std_timing *t,
+				      int revision,
 				      int timing_level)
 {
 	struct drm_display_mode *mode;
@@ -523,14 +550,20 @@ struct drm_display_mode *drm_mode_std(struct drm_device *dev,
 	unsigned vfreq = (t->vfreq_aspect & EDID_TIMING_VFREQ_MASK)
 		>> EDID_TIMING_VFREQ_SHIFT;
 
+	if (bad_std_timing(t->hsize, t->vfreq_aspect))
+		return NULL;
+
 	/* According to the EDID spec, the hdisplay = hsize * 8 + 248 */
 	hsize = t->hsize * 8 + 248;
 	/* vrefresh_rate = vfreq + 60 */
 	vrefresh_rate = vfreq + 60;
 	/* the vdisplay is calculated based on the aspect ratio */
-	if (aspect_ratio == 0)
-		vsize = (hsize * 10) / 16;
-	else if (aspect_ratio == 1)
+	if (aspect_ratio == 0) {
+		if (revision < 3)
+			vsize = hsize;
+		else
+			vsize = (hsize * 10) / 16;
+	} else if (aspect_ratio == 1)
 		vsize = (hsize * 3) / 4;
 	else if (aspect_ratio == 2)
 		vsize = (hsize * 4) / 5;
@@ -538,7 +571,8 @@ struct drm_display_mode *drm_mode_std(struct drm_device *dev,
 		vsize = (hsize * 9) / 16;
 	/* HDTV hack */
 	if (hsize == 1360 && vsize == 765 && vrefresh_rate == 60) {
-		mode = drm_cvt_mode(dev, hsize, vsize, vrefresh_rate, 0, 0);
+		mode = drm_cvt_mode(dev, hsize, vsize, vrefresh_rate, 0, 0,
+				    false);
 		mode->hdisplay = 1366;
 		mode->vsync_start = mode->vsync_start - 1;
 		mode->vsync_end = mode->vsync_end - 1;
@@ -557,12 +591,57 @@ struct drm_display_mode *drm_mode_std(struct drm_device *dev,
 		mode = drm_gtf_mode(dev, hsize, vsize, vrefresh_rate, 0, 0);
 		break;
 	case LEVEL_CVT:
-		mode = drm_cvt_mode(dev, hsize, vsize, vrefresh_rate, 0, 0);
+		mode = drm_cvt_mode(dev, hsize, vsize, vrefresh_rate, 0, 0,
+				    false);
 		break;
 	}
 	return mode;
 }
 
+/*
+ * EDID is delightfully ambiguous about how interlaced modes are to be
+ * encoded.  Our internal representation is of frame height, but some
+ * HDTV detailed timings are encoded as field height.
+ *
+ * The format list here is from CEA, in frame size.  Technically we
+ * should be checking refresh rate too.  Whatever.
+ */
+static void
+drm_mode_do_interlace_quirk(struct drm_display_mode *mode,
+			    struct detailed_pixel_timing *pt)
+{
+	int i;
+	static const struct {
+		int w, h;
+	} cea_interlaced[] = {
+		{ 1920, 1080 },
+		{  720,  480 },
+		{ 1440,  480 },
+		{ 2880,  480 },
+		{  720,  576 },
+		{ 1440,  576 },
+		{ 2880,  576 },
+	};
+	static const int n_sizes =
+		sizeof(cea_interlaced)/sizeof(cea_interlaced[0]);
+
+	if (!(pt->misc & DRM_EDID_PT_INTERLACED))
+		return;
+
+	for (i = 0; i < n_sizes; i++) {
+		if ((mode->hdisplay == cea_interlaced[i].w) &&
+		    (mode->vdisplay == cea_interlaced[i].h / 2)) {
+			mode->vdisplay *= 2;
+			mode->vsync_start *= 2;
+			mode->vsync_end *= 2;
+			mode->vtotal *= 2;
+			mode->vtotal |= 1;
+		}
+	}
+
+	mode->flags |= DRM_MODE_FLAG_INTERLACE;
+}
+
 /**
  * drm_mode_detailed - create a new mode from an EDID detailed timing section
  * @dev: DRM device (needed to create new mode)
@@ -598,10 +677,15 @@ static struct drm_display_mode *drm_mode_detailed(struct drm_device *dev,
 		return NULL;
 	}
 	if (!(pt->misc & DRM_EDID_PT_SEPARATE_SYNC)) {
-		printk(KERN_WARNING "integrated sync not supported\n");
-		return NULL;
+		printk(KERN_WARNING "composite sync not supported\n");
 	}
 
+	/* it is incorrect if hsync/vsync width is zero */
+	if (!hsync_pulse_width || !vsync_pulse_width) {
+		DRM_DEBUG_KMS("Incorrect Detailed timing. "
+				"Wrong Hsync/Vsync pulse width\n");
+		return NULL;
+	}
 	mode = drm_mode_create(dev);
 	if (!mode)
 		return NULL;
@@ -623,10 +707,24 @@ static struct drm_display_mode *drm_mode_detailed(struct drm_device *dev,
 	mode->vsync_end = mode->vsync_start + vsync_pulse_width;
 	mode->vtotal = mode->vdisplay + vblank;
 
+	/* perform the basic check for the detailed timing */
+	if (mode->hsync_end > mode->htotal ||
+		mode->vsync_end > mode->vtotal) {
+		drm_mode_destroy(dev, mode);
+		DRM_DEBUG_KMS("Incorrect detailed timing. "
+				"Sync is beyond the blank.\n");
+		return NULL;
+	}
+
+	/* Some EDIDs have bogus h/vtotal values */
+	if (mode->hsync_end > mode->htotal)
+		mode->htotal = mode->hsync_end + 1;
+	if (mode->vsync_end > mode->vtotal)
+		mode->vtotal = mode->vsync_end + 1;
+
 	drm_mode_set_name(mode);
 
-	if (pt->misc & DRM_EDID_PT_INTERLACED)
-		mode->flags |= DRM_MODE_FLAG_INTERLACE;
+	drm_mode_do_interlace_quirk(mode, pt);
 
 	if (quirks & EDID_QUIRK_DETAILED_SYNC_PP) {
 		pt->misc |= DRM_EDID_PT_HSYNC_POSITIVE | DRM_EDID_PT_VSYNC_POSITIVE;
@@ -779,7 +877,7 @@ static int add_standard_modes(struct drm_connector *connector, struct edid *edid
 			continue;
 
 		newmode = drm_mode_std(dev, &edid->standard_timings[i],
-					timing_level);
+				       edid->revision, timing_level);
 		if (newmode) {
 			drm_mode_probed_add(connector, newmode);
 			modes++;
@@ -789,8 +887,169 @@ static int add_standard_modes(struct drm_connector *connector, struct edid *edid
 	return modes;
 }
 
+/*
+ * XXX fix this for:
+ * - GTF secondary curve formula
+ * - EDID 1.4 range offsets
+ * - CVT extended bits
+ */
+static bool
+mode_in_range(struct drm_display_mode *mode, struct detailed_timing *timing)
+{
+	struct detailed_data_monitor_range *range;
+	int hsync, vrefresh;
+
+	range = &timing->data.other_data.data.range;
+
+	hsync = drm_mode_hsync(mode);
+	vrefresh = drm_mode_vrefresh(mode);
+
+	if (hsync < range->min_hfreq_khz || hsync > range->max_hfreq_khz)
+		return false;
+
+	if (vrefresh < range->min_vfreq || vrefresh > range->max_vfreq)
+		return false;
+
+	if (range->pixel_clock_mhz && range->pixel_clock_mhz != 0xff) {
+		/* be forgiving since it's in units of 10MHz */
+		int max_clock = range->pixel_clock_mhz * 10 + 9;
+		max_clock *= 1000;
+		if (mode->clock > max_clock)
+			return false;
+	}
+
+	return true;
+}
+
+/*
+ * XXX If drm_dmt_modes ever regrows the CVT-R modes (and it will) this will
+ * need to account for them.
+ */
+static int drm_gtf_modes_for_range(struct drm_connector *connector,
+				   struct detailed_timing *timing)
+{
+	int i, modes = 0;
+	struct drm_display_mode *newmode;
+	struct drm_device *dev = connector->dev;
+
+	for (i = 0; i < drm_num_dmt_modes; i++) {
+		if (mode_in_range(drm_dmt_modes + i, timing)) {
+			newmode = drm_mode_duplicate(dev, &drm_dmt_modes[i]);
+			if (newmode) {
+				drm_mode_probed_add(connector, newmode);
+				modes++;
+			}
+		}
+	}
+
+	return modes;
+}
+
+static int drm_cvt_modes(struct drm_connector *connector,
+			 struct detailed_timing *timing)
+{
+	int i, j, modes = 0;
+	struct drm_display_mode *newmode;
+	struct drm_device *dev = connector->dev;
+	struct cvt_timing *cvt;
+	const int rates[] = { 60, 85, 75, 60, 50 };
+	const u8 empty[3] = { 0, 0, 0 };
+
+	for (i = 0; i < 4; i++) {
+		int uninitialized_var(width), height;
+		cvt = &(timing->data.other_data.data.cvt[i]);
+
+		if (!memcmp(cvt->code, empty, 3))
+			continue;
+
+		height = (cvt->code[0] + ((cvt->code[1] & 0xf0) << 4) + 1) * 2;
+		switch (cvt->code[1] & 0x0c) {
+		case 0x00:
+			width = height * 4 / 3;
+			break;
+		case 0x04:
+			width = height * 16 / 9;
+			break;
+		case 0x08:
+			width = height * 16 / 10;
+			break;
+		case 0x0c:
+			width = height * 15 / 9;
+			break;
+		}
+
+		for (j = 1; j < 5; j++) {
+			if (cvt->code[2] & (1 << j)) {
+				newmode = drm_cvt_mode(dev, width, height,
+						       rates[j], j == 0,
+						       false, false);
+				if (newmode) {
+					drm_mode_probed_add(connector, newmode);
+					modes++;
+				}
+			}
+		}
+	}
+
+	return modes;
+}
+
+static int add_detailed_modes(struct drm_connector *connector,
+			      struct detailed_timing *timing,
+			      struct edid *edid, u32 quirks, int preferred)
+{
+	int i, modes = 0;
+	struct detailed_non_pixel *data = &timing->data.other_data;
+	int timing_level = standard_timing_level(edid);
+	int gtf = (edid->features & DRM_EDID_FEATURE_DEFAULT_GTF);
+	struct drm_display_mode *newmode;
+	struct drm_device *dev = connector->dev;
+
+	if (timing->pixel_clock) {
+		newmode = drm_mode_detailed(dev, edid, timing, quirks);
+		if (!newmode)
+			return 0;
+
+		if (preferred)
+			newmode->type |= DRM_MODE_TYPE_PREFERRED;
+
+		drm_mode_probed_add(connector, newmode);
+		return 1;
+	}
+
+	/* other timing types */
+	switch (data->type) {
+	case EDID_DETAIL_MONITOR_RANGE:
+		if (gtf)
+			modes += drm_gtf_modes_for_range(connector, timing);
+		break;
+	case EDID_DETAIL_STD_MODES:
+		/* Six modes per detailed section */
+		for (i = 0; i < 6; i++) {
+			struct std_timing *std;
+			struct drm_display_mode *newmode;
+
+			std = &data->data.timings[i];
+			newmode = drm_mode_std(dev, std, edid->revision,
+					       timing_level);
+			if (newmode) {
+				drm_mode_probed_add(connector, newmode);
+				modes++;
+			}
+		}
+		break;
+	case EDID_DETAIL_CVT_3BYTE:
+		modes += drm_cvt_modes(connector, timing);
+		break;
+	default:
+		break;
+	}
+
+	return modes;
+}
+
 /**
- * add_detailed_modes - get detailed mode info from EDID data
+ * add_detailed_info - get detailed mode info from EDID data
  * @connector: attached connector
  * @edid: EDID block to scan
  * @quirks: quirks to apply
@@ -801,67 +1060,24 @@ static int add_standard_modes(struct drm_connector *connector, struct edid *edid
 static int add_detailed_info(struct drm_connector *connector,
 			     struct edid *edid, u32 quirks)
 {
-	struct drm_device *dev = connector->dev;
-	int i, j, modes = 0;
-	int timing_level;
-
-	timing_level = standard_timing_level(edid);
+	int i, modes = 0;
 
 	for (i = 0; i < EDID_DETAILED_TIMINGS; i++) {
 		struct detailed_timing *timing = &edid->detailed_timings[i];
-		struct detailed_non_pixel *data = &timing->data.other_data;
-		struct drm_display_mode *newmode;
-
-		/* X server check is version 1.1 or higher */
-		if (edid->version == 1 && edid->revision >= 1 &&
-		    !timing->pixel_clock) {
-			/* Other timing or info */
-			switch (data->type) {
-			case EDID_DETAIL_MONITOR_SERIAL:
-				break;
-			case EDID_DETAIL_MONITOR_STRING:
-				break;
-			case EDID_DETAIL_MONITOR_RANGE:
-				/* Get monitor range data */
-				break;
-			case EDID_DETAIL_MONITOR_NAME:
-				break;
-			case EDID_DETAIL_MONITOR_CPDATA:
-				break;
-			case EDID_DETAIL_STD_MODES:
-				/* Five modes per detailed section */
-				for (j = 0; j < 5; i++) {
-					struct std_timing *std;
-					struct drm_display_mode *newmode;
-
-					std = &data->data.timings[j];
-					newmode = drm_mode_std(dev, std,
-							       timing_level);
-					if (newmode) {
-						drm_mode_probed_add(connector, newmode);
-						modes++;
-					}
-				}
-				break;
-			default:
-				break;
-			}
-		} else {
-			newmode = drm_mode_detailed(dev, edid, timing, quirks);
-			if (!newmode)
-				continue;
+		int preferred = (i == 0) && (edid->features & DRM_EDID_FEATURE_PREFERRED_TIMING);
 
-			/* First detailed mode is preferred */
-			if (i == 0 && (edid->features & DRM_EDID_FEATURE_PREFERRED_TIMING))
-				newmode->type |= DRM_MODE_TYPE_PREFERRED;
-			drm_mode_probed_add(connector, newmode);
+		/* In 1.0, only timings are allowed */
+		if (!timing->pixel_clock && edid->version == 1 &&
+			edid->revision == 0)
+			continue;
 
-			modes++;
-		}
+		modes += add_detailed_modes(connector, timing, edid, quirks,
+					    preferred);
 	}
 
 	return modes;
 }
+
 /**
  * add_detailed_mode_eedid - get detailed mode info from addtional timing
  * 			EDID block
@@ -875,12 +1091,9 @@ static int add_detailed_info(struct drm_connector *connector,
 static int add_detailed_info_eedid(struct drm_connector *connector,
 			     struct edid *edid, u32 quirks)
 {
-	struct drm_device *dev = connector->dev;
-	int i, j, modes = 0;
+	int i, modes = 0;
 	char *edid_ext = NULL;
 	struct detailed_timing *timing;
-	struct detailed_non_pixel *data;
-	struct drm_display_mode *newmode;
 	int edid_ext_num;
 	int start_offset, end_offset;
 	int timing_level;
@@ -899,8 +1112,8 @@ static int add_detailed_info_eedid(struct drm_connector *connector,
 	}
 
 	/* Chose real EDID extension number */
-	edid_ext_num = edid->extensions > MAX_EDID_EXT_NUM ?
-		       MAX_EDID_EXT_NUM : edid->extensions;
+	edid_ext_num = edid->extensions > DRM_MAX_EDID_EXT_NUM ?
+		DRM_MAX_EDID_EXT_NUM : edid->extensions;
 
 	/* Find CEA extension */
 	for (i = 0; i < edid_ext_num; i++) {
@@ -931,49 +1144,7 @@ static int add_detailed_info_eedid(struct drm_connector *connector,
 	for (i = start_offset; i < end_offset;
 			i += sizeof(struct detailed_timing)) {
 		timing = (struct detailed_timing *)(edid_ext + i);
-		data = &timing->data.other_data;
-		/* Detailed mode timing */
-		if (timing->pixel_clock) {
-			newmode = drm_mode_detailed(dev, edid, timing, quirks);
-			if (!newmode)
-				continue;
-
-			drm_mode_probed_add(connector, newmode);
-
-			modes++;
-			continue;
-		}
-
-		/* Other timing or info */
-		switch (data->type) {
-		case EDID_DETAIL_MONITOR_SERIAL:
-			break;
-		case EDID_DETAIL_MONITOR_STRING:
-			break;
-		case EDID_DETAIL_MONITOR_RANGE:
-			/* Get monitor range data */
-			break;
-		case EDID_DETAIL_MONITOR_NAME:
-			break;
-		case EDID_DETAIL_MONITOR_CPDATA:
-			break;
-		case EDID_DETAIL_STD_MODES:
-			/* Five modes per detailed section */
-			for (j = 0; j < 5; i++) {
-				struct std_timing *std;
-				struct drm_display_mode *newmode;
-
-				std = &data->data.timings[j];
-				newmode = drm_mode_std(dev, std, timing_level);
-				if (newmode) {
-					drm_mode_probed_add(connector, newmode);
-					modes++;
-				}
-			}
-			break;
-		default:
-			break;
-		}
+		modes += add_detailed_modes(connector, timing, edid, quirks, 0);
 	}
 
 	return modes;
@@ -1019,19 +1190,19 @@ static int drm_ddc_read_edid(struct drm_connector *connector,
 			     struct i2c_adapter *adapter,
 			     char *buf, int len)
 {
-	int ret;
+	int i;
 
-	ret = drm_do_probe_ddc_edid(adapter, buf, len);
-	if (ret != 0) {
-		goto end;
+	for (i = 0; i < 4; i++) {
+		if (drm_do_probe_ddc_edid(adapter, buf, len))
+			return -1;
+		if (drm_edid_is_valid((struct edid *)buf))
+			return 0;
 	}
-	if (!edid_is_valid((struct edid *)buf)) {
-		dev_warn(&connector->dev->pdev->dev, "%s: EDID invalid.\n",
-			 drm_get_connector_name(connector));
-		ret = -1;
-	}
-end:
-	return ret;
+
+	/* repeated checksum failures; warn, but carry on */
+	dev_warn(&connector->dev->pdev->dev, "%s: EDID invalid.\n",
+		 drm_get_connector_name(connector));
+	return -1;
 }
 
 /**
@@ -1049,7 +1220,7 @@ struct edid *drm_get_edid(struct drm_connector *connector,
 	int ret;
 	struct edid *edid;
 
-	edid = kmalloc(EDID_LENGTH * (MAX_EDID_EXT_NUM + 1),
+	edid = kmalloc(EDID_LENGTH * (DRM_MAX_EDID_EXT_NUM + 1),
 		       GFP_KERNEL);
 	if (edid == NULL) {
 		dev_warn(&connector->dev->pdev->dev,
@@ -1067,14 +1238,14 @@ struct edid *drm_get_edid(struct drm_connector *connector,
 	if (edid->extensions != 0) {
 		int edid_ext_num = edid->extensions;
 
-		if (edid_ext_num > MAX_EDID_EXT_NUM) {
+		if (edid_ext_num > DRM_MAX_EDID_EXT_NUM) {
 			dev_warn(&connector->dev->pdev->dev,
 				 "The number of extension(%d) is "
 				 "over max (%d), actually read number (%d)\n",
-				 edid_ext_num, MAX_EDID_EXT_NUM,
-				 MAX_EDID_EXT_NUM);
+				 edid_ext_num, DRM_MAX_EDID_EXT_NUM,
+				 DRM_MAX_EDID_EXT_NUM);
 			/* Reset EDID extension number to be read */
-			edid_ext_num = MAX_EDID_EXT_NUM;
+			edid_ext_num = DRM_MAX_EDID_EXT_NUM;
 		}
 		/* Read EDID including extensions too */
 		ret = drm_ddc_read_edid(connector, adapter, (char *)edid,
@@ -1117,8 +1288,8 @@ bool drm_detect_hdmi_monitor(struct edid *edid)
 		goto end;
 
 	/* Chose real EDID extension number */
-	edid_ext_num = edid->extensions > MAX_EDID_EXT_NUM ?
-		       MAX_EDID_EXT_NUM : edid->extensions;
+	edid_ext_num = edid->extensions > DRM_MAX_EDID_EXT_NUM ?
+		       DRM_MAX_EDID_EXT_NUM : edid->extensions;
 
 	/* Find CEA extension */
 	for (i = 0; i < edid_ext_num; i++) {
@@ -1175,7 +1346,7 @@ int drm_add_edid_modes(struct drm_connector *connector, struct edid *edid)
 	if (edid == NULL) {
 		return 0;
 	}
-	if (!edid_is_valid(edid)) {
+	if (!drm_edid_is_valid(edid)) {
 		dev_warn(&connector->dev->pdev->dev, "%s: EDID invalid.\n",
 			 drm_get_connector_name(connector));
 		return 0;
@@ -1249,6 +1420,8 @@ int drm_add_modes_noedid(struct drm_connector *connector,
 					ptr->vdisplay > vdisplay)
 				continue;
 		}
+		if (drm_mode_vrefresh(ptr) > 61)
+			continue;
 		mode = drm_mode_duplicate(dev, ptr);
 		if (mode) {
 			drm_mode_probed_add(connector, mode);
diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c
index 2c4671314884..50549703584f 100644
--- a/drivers/gpu/drm/drm_fb_helper.c
+++ b/drivers/gpu/drm/drm_fb_helper.c
@@ -27,6 +27,7 @@
  *      Dave Airlie <airlied@linux.ie>
  *      Jesse Barnes <jesse.barnes@intel.com>
  */
+#include <linux/kernel.h>
 #include <linux/sysrq.h>
 #include <linux/fb.h>
 #include "drmP.h"
@@ -40,6 +41,184 @@ MODULE_LICENSE("GPL and additional rights");
 
 static LIST_HEAD(kernel_fb_helper_list);
 
+int drm_fb_helper_add_connector(struct drm_connector *connector)
+{
+	connector->fb_helper_private = kzalloc(sizeof(struct drm_fb_helper_connector), GFP_KERNEL);
+	if (!connector->fb_helper_private)
+		return -ENOMEM;
+
+	return 0;
+}
+EXPORT_SYMBOL(drm_fb_helper_add_connector);
+
+/**
+ * drm_fb_helper_connector_parse_command_line - parse command line for connector
+ * @connector - connector to parse line for
+ * @mode_option - per connector mode option
+ *
+ * This parses the connector specific then generic command lines for
+ * modes and options to configure the connector.
+ *
+ * This uses the same parameters as the fb modedb.c, except for extra
+ *	<xres>x<yres>[M][R][-<bpp>][@<refresh>][i][m][eDd]
+ *
+ * enable/enable Digital/disable bit at the end
+ */
+static bool drm_fb_helper_connector_parse_command_line(struct drm_connector *connector,
+						       const char *mode_option)
+{
+	const char *name;
+	unsigned int namelen;
+	int res_specified = 0, bpp_specified = 0, refresh_specified = 0;
+	unsigned int xres = 0, yres = 0, bpp = 32, refresh = 0;
+	int yres_specified = 0, cvt = 0, rb = 0, interlace = 0, margins = 0;
+	int i;
+	enum drm_connector_force force = DRM_FORCE_UNSPECIFIED;
+	struct drm_fb_helper_connector *fb_help_conn = connector->fb_helper_private;
+	struct drm_fb_helper_cmdline_mode *cmdline_mode;
+
+	if (!fb_help_conn)
+		return false;
+
+	cmdline_mode = &fb_help_conn->cmdline_mode;
+	if (!mode_option)
+		mode_option = fb_mode_option;
+
+	if (!mode_option) {
+		cmdline_mode->specified = false;
+		return false;
+	}
+
+	name = mode_option;
+	namelen = strlen(name);
+	for (i = namelen-1; i >= 0; i--) {
+		switch (name[i]) {
+		case '@':
+			namelen = i;
+			if (!refresh_specified && !bpp_specified &&
+			    !yres_specified) {
+				refresh = simple_strtol(&name[i+1], NULL, 10);
+				refresh_specified = 1;
+				if (cvt || rb)
+					cvt = 0;
+			} else
+				goto done;
+			break;
+		case '-':
+			namelen = i;
+			if (!bpp_specified && !yres_specified) {
+				bpp = simple_strtol(&name[i+1], NULL, 10);
+				bpp_specified = 1;
+				if (cvt || rb)
+					cvt = 0;
+			} else
+				goto done;
+			break;
+		case 'x':
+			if (!yres_specified) {
+				yres = simple_strtol(&name[i+1], NULL, 10);
+				yres_specified = 1;
+			} else
+				goto done;
+		case '0' ... '9':
+			break;
+		case 'M':
+			if (!yres_specified)
+				cvt = 1;
+			break;
+		case 'R':
+			if (!cvt)
+				rb = 1;
+			break;
+		case 'm':
+			if (!cvt)
+				margins = 1;
+			break;
+		case 'i':
+			if (!cvt)
+				interlace = 1;
+			break;
+		case 'e':
+			force = DRM_FORCE_ON;
+			break;
+		case 'D':
+			if ((connector->connector_type != DRM_MODE_CONNECTOR_DVII) &&
+			    (connector->connector_type != DRM_MODE_CONNECTOR_HDMIB))
+				force = DRM_FORCE_ON;
+			else
+				force = DRM_FORCE_ON_DIGITAL;
+			break;
+		case 'd':
+			force = DRM_FORCE_OFF;
+			break;
+		default:
+			goto done;
+		}
+	}
+	if (i < 0 && yres_specified) {
+		xres = simple_strtol(name, NULL, 10);
+		res_specified = 1;
+	}
+done:
+
+	DRM_DEBUG_KMS("cmdline mode for connector %s %dx%d@%dHz%s%s%s\n",
+		drm_get_connector_name(connector), xres, yres,
+		(refresh) ? refresh : 60, (rb) ? " reduced blanking" :
+		"", (margins) ? " with margins" : "", (interlace) ?
+		" interlaced" : "");
+
+	if (force) {
+		const char *s;
+		switch (force) {
+		case DRM_FORCE_OFF: s = "OFF"; break;
+		case DRM_FORCE_ON_DIGITAL: s = "ON - dig"; break;
+		default:
+		case DRM_FORCE_ON: s = "ON"; break;
+		}
+
+		DRM_INFO("forcing %s connector %s\n",
+			 drm_get_connector_name(connector), s);
+		connector->force = force;
+	}
+
+	if (res_specified) {
+		cmdline_mode->specified = true;
+		cmdline_mode->xres = xres;
+		cmdline_mode->yres = yres;
+	}
+
+	if (refresh_specified) {
+		cmdline_mode->refresh_specified = true;
+		cmdline_mode->refresh = refresh;
+	}
+
+	if (bpp_specified) {
+		cmdline_mode->bpp_specified = true;
+		cmdline_mode->bpp = bpp;
+	}
+	cmdline_mode->rb = rb ? true : false;
+	cmdline_mode->cvt = cvt  ? true : false;
+	cmdline_mode->interlace = interlace ? true : false;
+
+	return true;
+}
+
+int drm_fb_helper_parse_command_line(struct drm_device *dev)
+{
+	struct drm_connector *connector;
+
+	list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
+		char *option = NULL;
+
+		/* do something on return - turn off connector maybe */
+		if (fb_get_options(drm_get_connector_name(connector), &option))
+			continue;
+
+		drm_fb_helper_connector_parse_command_line(connector, option);
+	}
+	return 0;
+}
+
 bool drm_fb_helper_force_kernel_mode(void)
 {
 	int i = 0;
@@ -87,6 +266,7 @@ void drm_fb_helper_restore(void)
 }
 EXPORT_SYMBOL(drm_fb_helper_restore);
 
+#ifdef CONFIG_MAGIC_SYSRQ
 static void drm_fb_helper_restore_work_fn(struct work_struct *ignored)
 {
 	drm_fb_helper_restore();
@@ -103,6 +283,7 @@ static struct sysrq_key_op sysrq_drm_fb_helper_restore_op = {
 	.help_msg = "force-fb(V)",
 	.action_msg = "Restore framebuffer console",
 };
+#endif
 
 static void drm_fb_helper_on(struct fb_info *info)
 {
@@ -178,11 +359,9 @@ static void drm_fb_helper_off(struct fb_info *info, int dpms_mode)
 					mutex_unlock(&dev->mode_config.mutex);
 				}
 			}
-			if (dpms_mode == DRM_MODE_DPMS_OFF) {
-				mutex_lock(&dev->mode_config.mutex);
-				crtc_funcs->dpms(crtc, dpms_mode);
-				mutex_unlock(&dev->mode_config.mutex);
-			}
+			mutex_lock(&dev->mode_config.mutex);
+			crtc_funcs->dpms(crtc, DRM_MODE_DPMS_OFF);
+			mutex_unlock(&dev->mode_config.mutex);
 		}
 	}
 }
@@ -190,18 +369,23 @@ static void drm_fb_helper_off(struct fb_info *info, int dpms_mode)
 int drm_fb_helper_blank(int blank, struct fb_info *info)
 {
 	switch (blank) {
+	/* Display: On; HSync: On, VSync: On */
 	case FB_BLANK_UNBLANK:
 		drm_fb_helper_on(info);
 		break;
+	/* Display: Off; HSync: On, VSync: On */
 	case FB_BLANK_NORMAL:
 		drm_fb_helper_off(info, DRM_MODE_DPMS_STANDBY);
 		break;
+	/* Display: Off; HSync: Off, VSync: On */
 	case FB_BLANK_HSYNC_SUSPEND:
 		drm_fb_helper_off(info, DRM_MODE_DPMS_STANDBY);
 		break;
+	/* Display: Off; HSync: On, VSync: Off */
 	case FB_BLANK_VSYNC_SUSPEND:
 		drm_fb_helper_off(info, DRM_MODE_DPMS_SUSPEND);
 		break;
+	/* Display: Off; HSync: Off, VSync: Off */
 	case FB_BLANK_POWERDOWN:
 		drm_fb_helper_off(info, DRM_MODE_DPMS_OFF);
 		break;
@@ -259,6 +443,109 @@ out_free:
 }
 EXPORT_SYMBOL(drm_fb_helper_init_crtc_count);
 
+static int setcolreg(struct drm_crtc *crtc, u16 red, u16 green,
+		     u16 blue, u16 regno, struct fb_info *info)
+{
+	struct drm_fb_helper *fb_helper = info->par;
+	struct drm_framebuffer *fb = fb_helper->fb;
+	int pindex;
+
+	if (info->fix.visual == FB_VISUAL_TRUECOLOR) {
+		u32 *palette;
+		u32 value;
+		/* place color in psuedopalette */
+		if (regno > 16)
+			return -EINVAL;
+		palette = (u32 *)info->pseudo_palette;
+		red >>= (16 - info->var.red.length);
+		green >>= (16 - info->var.green.length);
+		blue >>= (16 - info->var.blue.length);
+		value = (red << info->var.red.offset) |
+			(green << info->var.green.offset) |
+			(blue << info->var.blue.offset);
+		palette[regno] = value;
+		return 0;
+	}
+
+	pindex = regno;
+
+	if (fb->bits_per_pixel == 16) {
+		pindex = regno << 3;
+
+		if (fb->depth == 16 && regno > 63)
+			return -EINVAL;
+		if (fb->depth == 15 && regno > 31)
+			return -EINVAL;
+
+		if (fb->depth == 16) {
+			u16 r, g, b;
+			int i;
+			if (regno < 32) {
+				for (i = 0; i < 8; i++)
+					fb_helper->funcs->gamma_set(crtc, red,
+						green, blue, pindex + i);
+			}
+
+			fb_helper->funcs->gamma_get(crtc, &r,
+						    &g, &b,
+						    pindex >> 1);
+
+			for (i = 0; i < 4; i++)
+				fb_helper->funcs->gamma_set(crtc, r,
+							    green, b,
+							    (pindex >> 1) + i);
+		}
+	}
+
+	if (fb->depth != 16)
+		fb_helper->funcs->gamma_set(crtc, red, green, blue, pindex);
+	return 0;
+}
+
+int drm_fb_helper_setcmap(struct fb_cmap *cmap, struct fb_info *info)
+{
+	struct drm_fb_helper *fb_helper = info->par;
+	struct drm_device *dev = fb_helper->dev;
+	u16 *red, *green, *blue, *transp;
+	struct drm_crtc *crtc;
+	int i, rc = 0;
+	int start;
+
+	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
+		struct drm_crtc_helper_funcs *crtc_funcs = crtc->helper_private;
+		for (i = 0; i < fb_helper->crtc_count; i++) {
+			if (crtc->base.id == fb_helper->crtc_info[i].crtc_id)
+				break;
+		}
+		if (i == fb_helper->crtc_count)
+			continue;
+
+		red = cmap->red;
+		green = cmap->green;
+		blue = cmap->blue;
+		transp = cmap->transp;
+		start = cmap->start;
+
+		for (i = 0; i < cmap->len; i++) {
+			u16 hred, hgreen, hblue, htransp = 0xffff;
+
+			hred = *red++;
+			hgreen = *green++;
+			hblue = *blue++;
+
+			if (transp)
+				htransp = *transp++;
+
+			rc = setcolreg(crtc, hred, hgreen, hblue, start++, info);
+			if (rc)
+				return rc;
+		}
+		crtc_funcs->load_lut(crtc);
+	}
+	return rc;
+}
+EXPORT_SYMBOL(drm_fb_helper_setcmap);
+
 int drm_fb_helper_setcolreg(unsigned regno,
 			    unsigned red,
 			    unsigned green,
@@ -270,10 +557,13 @@ int drm_fb_helper_setcolreg(unsigned regno,
 	struct drm_device *dev = fb_helper->dev;
 	struct drm_crtc *crtc;
 	int i;
+	int ret;
 
-	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
-		struct drm_framebuffer *fb = fb_helper->fb;
+	if (regno > 255)
+		return 1;
 
+	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
+		struct drm_crtc_helper_funcs *crtc_funcs = crtc->helper_private;
 		for (i = 0; i < fb_helper->crtc_count; i++) {
 			if (crtc->base.id == fb_helper->crtc_info[i].crtc_id)
 				break;
@@ -281,35 +571,11 @@ int drm_fb_helper_setcolreg(unsigned regno,
 		if (i == fb_helper->crtc_count)
 			continue;
 
-		if (regno > 255)
-			return 1;
-
-		if (fb->depth == 8) {
-			fb_helper->funcs->gamma_set(crtc, red, green, blue, regno);
-			return 0;
-		}
+		ret = setcolreg(crtc, red, green, blue, regno, info);
+		if (ret)
+			return ret;
 
-		if (regno < 16) {
-			switch (fb->depth) {
-			case 15:
-				fb->pseudo_palette[regno] = ((red & 0xf800) >> 1) |
-					((green & 0xf800) >>  6) |
-					((blue & 0xf800) >> 11);
-				break;
-			case 16:
-				fb->pseudo_palette[regno] = (red & 0xf800) |
-					((green & 0xfc00) >>  5) |
-					((blue  & 0xf800) >> 11);
-				break;
-			case 24:
-			case 32:
-				fb->pseudo_palette[regno] =
-					(((red >> 8) & 0xff) << info->var.red.offset) |
-					(((green >> 8) & 0xff) << info->var.green.offset) |
-					(((blue >> 8) & 0xff) << info->var.blue.offset);
-				break;
-			}
-		}
+		crtc_funcs->load_lut(crtc);
 	}
 	return 0;
 }
@@ -322,15 +588,14 @@ int drm_fb_helper_check_var(struct fb_var_screeninfo *var,
 	struct drm_framebuffer *fb = fb_helper->fb;
 	int depth;
 
-	if (var->pixclock == -1 || !var->pixclock)
+	if (var->pixclock != 0)
 		return -EINVAL;
 
 	/* Need to resize the fb object !!! */
-	if (var->xres > fb->width || var->yres > fb->height) {
-		DRM_ERROR("Requested width/height is greater than current fb "
-			   "object %dx%d > %dx%d\n", var->xres, var->yres,
-			   fb->width, fb->height);
-		DRM_ERROR("Need resizing code.\n");
+	if (var->bits_per_pixel > fb->bits_per_pixel || var->xres > fb->width || var->yres > fb->height) {
+		DRM_DEBUG("fb userspace requested width/height/bpp is greater than current fb "
+			  "object %dx%d-%d > %dx%d-%d\n", var->xres, var->yres, var->bits_per_pixel,
+			  fb->width, fb->height, fb->bits_per_pixel);
 		return -EINVAL;
 	}
 
@@ -414,8 +679,8 @@ int drm_fb_helper_set_par(struct fb_info *info)
 	int ret;
 	int i;
 
-	if (var->pixclock != -1) {
-		DRM_ERROR("PIXEL CLCOK SET\n");
+	if (var->pixclock != 0) {
+		DRM_ERROR("PIXEL CLOCK SET\n");
 		return -EINVAL;
 	}
 
@@ -430,7 +695,7 @@ int drm_fb_helper_set_par(struct fb_info *info)
 
 		if (crtc->fb == fb_helper->crtc_info[i].mode_set.fb) {
 			mutex_lock(&dev->mode_config.mutex);
-			ret = crtc->funcs->set_config(&fb_helper->crtc_info->mode_set);
+			ret = crtc->funcs->set_config(&fb_helper->crtc_info[i].mode_set);
 			mutex_unlock(&dev->mode_config.mutex);
 			if (ret)
 				return ret;
@@ -479,11 +744,14 @@ int drm_fb_helper_pan_display(struct fb_var_screeninfo *var,
 EXPORT_SYMBOL(drm_fb_helper_pan_display);
 
 int drm_fb_helper_single_fb_probe(struct drm_device *dev,
+				  int preferred_bpp,
 				  int (*fb_create)(struct drm_device *dev,
 						   uint32_t fb_width,
 						   uint32_t fb_height,
 						   uint32_t surface_width,
 						   uint32_t surface_height,
+						   uint32_t surface_depth,
+						   uint32_t surface_bpp,
 						   struct drm_framebuffer **fb_ptr))
 {
 	struct drm_crtc *crtc;
@@ -497,8 +765,48 @@ int drm_fb_helper_single_fb_probe(struct drm_device *dev,
 	struct drm_framebuffer *fb;
 	struct drm_mode_set *modeset = NULL;
 	struct drm_fb_helper *fb_helper;
+	uint32_t surface_depth = 24, surface_bpp = 32;
 
+	/* if driver picks 8 or 16 by default use that
+	   for both depth/bpp */
+	if (preferred_bpp != surface_bpp) {
+		surface_depth = surface_bpp = preferred_bpp;
+	}
 	/* first up get a count of crtcs now in use and new min/maxes width/heights */
+	list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
+		struct drm_fb_helper_connector *fb_help_conn = connector->fb_helper_private;
+
+		struct drm_fb_helper_cmdline_mode *cmdline_mode;
+
+		if (!fb_help_conn)
+			continue;
+		
+		cmdline_mode = &fb_help_conn->cmdline_mode;
+
+		if (cmdline_mode->bpp_specified) {
+			switch (cmdline_mode->bpp) {
+			case 8:
+				surface_depth = surface_bpp = 8;
+				break;
+			case 15:
+				surface_depth = 15;
+				surface_bpp = 16;
+				break;
+			case 16:
+				surface_depth = surface_bpp = 16;
+				break;
+			case 24:
+				surface_depth = surface_bpp = 24;
+				break;
+			case 32:
+				surface_depth = 24;
+				surface_bpp = 32;
+				break;
+			}
+			break;
+		}
+	}
+
 	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
 		if (drm_helper_crtc_in_use(crtc)) {
 			if (crtc->desired_mode) {
@@ -527,7 +835,8 @@ int drm_fb_helper_single_fb_probe(struct drm_device *dev,
 	/* do we have an fb already? */
 	if (list_empty(&dev->mode_config.fb_kernel_list)) {
 		ret = (*fb_create)(dev, fb_width, fb_height, surface_width,
-				   surface_height, &fb);
+				   surface_height, surface_depth, surface_bpp,
+				   &fb);
 		if (ret)
 			return -EINVAL;
 		new_fb = 1;
@@ -583,9 +892,14 @@ int drm_fb_helper_single_fb_probe(struct drm_device *dev,
 	fb_helper->fb = fb;
 
 	if (new_fb) {
-		info->var.pixclock = -1;
-		if (register_framebuffer(info) < 0)
+		info->var.pixclock = 0;
+		ret = fb_alloc_cmap(&info->cmap, modeset->crtc->gamma_size, 0);
+		if (ret)
+			return ret;
+		if (register_framebuffer(info) < 0) {
+			fb_dealloc_cmap(&info->cmap);
 			return -EINVAL;
+		}
 	} else {
 		drm_fb_helper_set_par(info);
 	}
@@ -615,13 +929,16 @@ void drm_fb_helper_free(struct drm_fb_helper *helper)
 		unregister_sysrq_key('v', &sysrq_drm_fb_helper_restore_op);
 	}
 	drm_fb_helper_crtc_free(helper);
+	fb_dealloc_cmap(&helper->fb->fbdev->cmap);
 }
 EXPORT_SYMBOL(drm_fb_helper_free);
 
-void drm_fb_helper_fill_fix(struct fb_info *info, uint32_t pitch)
+void drm_fb_helper_fill_fix(struct fb_info *info, uint32_t pitch,
+			    uint32_t depth)
 {
 	info->fix.type = FB_TYPE_PACKED_PIXELS;
-	info->fix.visual = FB_VISUAL_TRUECOLOR;
+	info->fix.visual = depth == 8 ? FB_VISUAL_PSEUDOCOLOR :
+		FB_VISUAL_TRUECOLOR;
 	info->fix.type_aux = 0;
 	info->fix.xpanstep = 1; /* doing it in hw */
 	info->fix.ypanstep = 1; /* doing it in hw */
diff --git a/drivers/gpu/drm/drm_fops.c b/drivers/gpu/drm/drm_fops.c
index 251bc0e3b5ec..08d14df3bb42 100644
--- a/drivers/gpu/drm/drm_fops.c
+++ b/drivers/gpu/drm/drm_fops.c
@@ -257,6 +257,9 @@ static int drm_open_helper(struct inode *inode, struct file *filp,
 
 	INIT_LIST_HEAD(&priv->lhead);
 	INIT_LIST_HEAD(&priv->fbs);
+	INIT_LIST_HEAD(&priv->event_list);
+	init_waitqueue_head(&priv->event_wait);
+	priv->event_space = 4096; /* set aside 4k for event buffer */
 
 	if (dev->driver->driver_features & DRIVER_GEM)
 		drm_gem_open(dev, priv);
@@ -297,6 +300,18 @@ static int drm_open_helper(struct inode *inode, struct file *filp,
 				goto out_free;
 			}
 		}
+		mutex_lock(&dev->struct_mutex);
+		if (dev->driver->master_set) {
+			ret = dev->driver->master_set(dev, priv, true);
+			if (ret) {
+				/* drop both references if this fails */
+				drm_master_put(&priv->minor->master);
+				drm_master_put(&priv->master);
+				mutex_unlock(&dev->struct_mutex);
+				goto out_free;
+			}
+		}
+		mutex_unlock(&dev->struct_mutex);
 	} else {
 		/* get a reference to the master */
 		priv->master = drm_master_get(priv->minor->master);
@@ -413,6 +428,30 @@ static void drm_master_release(struct drm_device *dev, struct file *filp)
 	}
 }
 
+static void drm_events_release(struct drm_file *file_priv)
+{
+	struct drm_device *dev = file_priv->minor->dev;
+	struct drm_pending_event *e, *et;
+	struct drm_pending_vblank_event *v, *vt;
+	unsigned long flags;
+
+	spin_lock_irqsave(&dev->event_lock, flags);
+
+	/* Remove pending flips */
+	list_for_each_entry_safe(v, vt, &dev->vblank_event_list, base.link)
+		if (v->base.file_priv == file_priv) {
+			list_del(&v->base.link);
+			drm_vblank_put(dev, v->pipe);
+			v->base.destroy(&v->base);
+		}
+
+	/* Remove unconsumed events */
+	list_for_each_entry_safe(e, et, &file_priv->event_list, link)
+		e->destroy(e);
+
+	spin_unlock_irqrestore(&dev->event_lock, flags);
+}
+
 /**
  * Release file.
  *
@@ -451,6 +490,8 @@ int drm_release(struct inode *inode, struct file *filp)
 	if (file_priv->minor->master)
 		drm_master_release(dev, filp);
 
+	drm_events_release(file_priv);
+
 	if (dev->driver->driver_features & DRIVER_GEM)
 		drm_gem_release(dev, file_priv);
 
@@ -504,6 +545,8 @@ int drm_release(struct inode *inode, struct file *filp)
 
 		if (file_priv->minor->master == file_priv->master) {
 			/* drop the reference held my the minor */
+			if (dev->driver->master_drop)
+				dev->driver->master_drop(dev, file_priv, true);
 			drm_master_put(&file_priv->minor->master);
 		}
 	}
@@ -544,9 +587,74 @@ int drm_release(struct inode *inode, struct file *filp)
 }
 EXPORT_SYMBOL(drm_release);
 
-/** No-op. */
+static bool
+drm_dequeue_event(struct drm_file *file_priv,
+		  size_t total, size_t max, struct drm_pending_event **out)
+{
+	struct drm_device *dev = file_priv->minor->dev;
+	struct drm_pending_event *e;
+	unsigned long flags;
+	bool ret = false;
+
+	spin_lock_irqsave(&dev->event_lock, flags);
+
+	*out = NULL;
+	if (list_empty(&file_priv->event_list))
+		goto out;
+	e = list_first_entry(&file_priv->event_list,
+			     struct drm_pending_event, link);
+	if (e->event->length + total > max)
+		goto out;
+
+	file_priv->event_space += e->event->length;
+	list_del(&e->link);
+	*out = e;
+	ret = true;
+
+out:
+	spin_unlock_irqrestore(&dev->event_lock, flags);
+	return ret;
+}
+
+ssize_t drm_read(struct file *filp, char __user *buffer,
+		 size_t count, loff_t *offset)
+{
+	struct drm_file *file_priv = filp->private_data;
+	struct drm_pending_event *e;
+	size_t total;
+	ssize_t ret;
+
+	ret = wait_event_interruptible(file_priv->event_wait,
+				       !list_empty(&file_priv->event_list));
+	if (ret < 0)
+		return ret;
+
+	total = 0;
+	while (drm_dequeue_event(file_priv, total, count, &e)) {
+		if (copy_to_user(buffer + total,
+				 e->event, e->event->length)) {
+			total = -EFAULT;
+			break;
+		}
+
+		total += e->event->length;
+		e->destroy(e);
+	}
+
+	return total;
+}
+EXPORT_SYMBOL(drm_read);
+
 unsigned int drm_poll(struct file *filp, struct poll_table_struct *wait)
 {
-	return 0;
+	struct drm_file *file_priv = filp->private_data;
+	unsigned int mask = 0;
+
+	poll_wait(filp, &file_priv->event_wait, wait);
+
+	if (!list_empty(&file_priv->event_list))
+		mask |= POLLIN | POLLRDNORM;
+
+	return mask;
 }
 EXPORT_SYMBOL(drm_poll);
diff --git a/drivers/gpu/drm/drm_gem.c b/drivers/gpu/drm/drm_gem.c
index 80391995bdec..aa89d4b0b4c4 100644
--- a/drivers/gpu/drm/drm_gem.c
+++ b/drivers/gpu/drm/drm_gem.c
@@ -142,19 +142,6 @@ drm_gem_object_alloc(struct drm_device *dev, size_t size)
 	if (IS_ERR(obj->filp))
 		goto free;
 
-	/* Basically we want to disable the OOM killer and handle ENOMEM
-	 * ourselves by sacrificing pages from cached buffers.
-	 * XXX shmem_file_[gs]et_gfp_mask()
-	 */
-	mapping_set_gfp_mask(obj->filp->f_path.dentry->d_inode->i_mapping,
-			     GFP_HIGHUSER |
-			     __GFP_COLD |
-			     __GFP_FS |
-			     __GFP_RECLAIMABLE |
-			     __GFP_NORETRY |
-			     __GFP_NOWARN |
-			     __GFP_NOMEMALLOC);
-
 	kref_init(&obj->refcount);
 	kref_init(&obj->handlecount);
 	obj->size = size;
@@ -205,9 +192,7 @@ drm_gem_handle_delete(struct drm_file *filp, u32 handle)
 	idr_remove(&filp->object_idr, handle);
 	spin_unlock(&filp->table_lock);
 
-	mutex_lock(&dev->struct_mutex);
-	drm_gem_object_handle_unreference(obj);
-	mutex_unlock(&dev->struct_mutex);
+	drm_gem_object_handle_unreference_unlocked(obj);
 
 	return 0;
 }
@@ -338,9 +323,7 @@ again:
 	}
 
 err:
-	mutex_lock(&dev->struct_mutex);
-	drm_gem_object_unreference(obj);
-	mutex_unlock(&dev->struct_mutex);
+	drm_gem_object_unreference_unlocked(obj);
 	return ret;
 }
 
@@ -371,9 +354,7 @@ drm_gem_open_ioctl(struct drm_device *dev, void *data,
 		return -ENOENT;
 
 	ret = drm_gem_handle_create(file_priv, obj, &handle);
-	mutex_lock(&dev->struct_mutex);
-	drm_gem_object_unreference(obj);
-	mutex_unlock(&dev->struct_mutex);
+	drm_gem_object_unreference_unlocked(obj);
 	if (ret)
 		return ret;
 
@@ -403,7 +384,7 @@ drm_gem_object_release_handle(int id, void *ptr, void *data)
 {
 	struct drm_gem_object *obj = ptr;
 
-	drm_gem_object_handle_unreference(obj);
+	drm_gem_object_handle_unreference_unlocked(obj);
 
 	return 0;
 }
@@ -416,16 +397,25 @@ drm_gem_object_release_handle(int id, void *ptr, void *data)
 void
 drm_gem_release(struct drm_device *dev, struct drm_file *file_private)
 {
-	mutex_lock(&dev->struct_mutex);
 	idr_for_each(&file_private->object_idr,
 		     &drm_gem_object_release_handle, NULL);
 
 	idr_destroy(&file_private->object_idr);
-	mutex_unlock(&dev->struct_mutex);
+}
+
+static void
+drm_gem_object_free_common(struct drm_gem_object *obj)
+{
+	struct drm_device *dev = obj->dev;
+	fput(obj->filp);
+	atomic_dec(&dev->object_count);
+	atomic_sub(obj->size, &dev->object_memory);
+	kfree(obj);
 }
 
 /**
  * Called after the last reference to the object has been lost.
+ * Must be called holding struct_ mutex
  *
  * Frees the object
  */
@@ -440,14 +430,40 @@ drm_gem_object_free(struct kref *kref)
 	if (dev->driver->gem_free_object != NULL)
 		dev->driver->gem_free_object(obj);
 
-	fput(obj->filp);
-	atomic_dec(&dev->object_count);
-	atomic_sub(obj->size, &dev->object_memory);
-	kfree(obj);
+	drm_gem_object_free_common(obj);
 }
 EXPORT_SYMBOL(drm_gem_object_free);
 
 /**
+ * Called after the last reference to the object has been lost.
+ * Must be called without holding struct_mutex
+ *
+ * Frees the object
+ */
+void
+drm_gem_object_free_unlocked(struct kref *kref)
+{
+	struct drm_gem_object *obj = (struct drm_gem_object *) kref;
+	struct drm_device *dev = obj->dev;
+
+	if (dev->driver->gem_free_object_unlocked != NULL)
+		dev->driver->gem_free_object_unlocked(obj);
+	else if (dev->driver->gem_free_object != NULL) {
+		mutex_lock(&dev->struct_mutex);
+		dev->driver->gem_free_object(obj);
+		mutex_unlock(&dev->struct_mutex);
+	}
+
+	drm_gem_object_free_common(obj);
+}
+EXPORT_SYMBOL(drm_gem_object_free_unlocked);
+
+static void drm_gem_object_ref_bug(struct kref *list_kref)
+{
+	BUG();
+}
+
+/**
  * Called after the last handle to the object has been closed
  *
  * Removes any name for the object. Note that this must be
@@ -471,8 +487,10 @@ drm_gem_object_handle_free(struct kref *kref)
 		/*
 		 * The object name held a reference to this object, drop
 		 * that now.
+		*
+		* This cannot be the last reference, since the handle holds one too.
 		 */
-		drm_gem_object_unreference(obj);
+		kref_put(&obj->refcount, drm_gem_object_ref_bug);
 	} else
 		spin_unlock(&dev->object_name_lock);
 
@@ -490,11 +508,8 @@ EXPORT_SYMBOL(drm_gem_vm_open);
 void drm_gem_vm_close(struct vm_area_struct *vma)
 {
 	struct drm_gem_object *obj = vma->vm_private_data;
-	struct drm_device *dev = obj->dev;
 
-	mutex_lock(&dev->struct_mutex);
-	drm_gem_object_unreference(obj);
-	mutex_unlock(&dev->struct_mutex);
+	drm_gem_object_unreference_unlocked(obj);
 }
 EXPORT_SYMBOL(drm_gem_vm_close);
 
@@ -552,7 +567,7 @@ int drm_gem_mmap(struct file *filp, struct vm_area_struct *vma)
 	vma->vm_flags |= VM_RESERVED | VM_IO | VM_PFNMAP | VM_DONTEXPAND;
 	vma->vm_ops = obj->dev->driver->gem_vm_ops;
 	vma->vm_private_data = map->handle;
-	vma->vm_page_prot = pgprot_writecombine(vma->vm_page_prot);
+	vma->vm_page_prot =  pgprot_writecombine(vm_get_page_prot(vma->vm_flags));
 
 	/* Take a ref for this mapping of the object, so that the fault
 	 * handler can dereference the mmap offset's pointer to the object.
diff --git a/drivers/gpu/drm/drm_ioc32.c b/drivers/gpu/drm/drm_ioc32.c
index 282d9fdf9f4e..d61d185cf040 100644
--- a/drivers/gpu/drm/drm_ioc32.c
+++ b/drivers/gpu/drm/drm_ioc32.c
@@ -104,7 +104,7 @@ static int compat_drm_version(struct file *file, unsigned int cmd,
 			  &version->desc))
 		return -EFAULT;
 
-	err = drm_ioctl(file->f_path.dentry->d_inode, file,
+	err = drm_ioctl(file,
 			DRM_IOCTL_VERSION, (unsigned long)version);
 	if (err)
 		return err;
@@ -145,8 +145,7 @@ static int compat_drm_getunique(struct file *file, unsigned int cmd,
 			  &u->unique))
 		return -EFAULT;
 
-	err = drm_ioctl(file->f_path.dentry->d_inode, file,
-			DRM_IOCTL_GET_UNIQUE, (unsigned long)u);
+	err = drm_ioctl(file, DRM_IOCTL_GET_UNIQUE, (unsigned long)u);
 	if (err)
 		return err;
 
@@ -174,8 +173,7 @@ static int compat_drm_setunique(struct file *file, unsigned int cmd,
 			  &u->unique))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_SET_UNIQUE, (unsigned long)u);
+	return drm_ioctl(file, DRM_IOCTL_SET_UNIQUE, (unsigned long)u);
 }
 
 typedef struct drm_map32 {
@@ -205,8 +203,7 @@ static int compat_drm_getmap(struct file *file, unsigned int cmd,
 	if (__put_user(idx, &map->offset))
 		return -EFAULT;
 
-	err = drm_ioctl(file->f_path.dentry->d_inode, file,
-			DRM_IOCTL_GET_MAP, (unsigned long)map);
+	err = drm_ioctl(file, DRM_IOCTL_GET_MAP, (unsigned long)map);
 	if (err)
 		return err;
 
@@ -246,8 +243,7 @@ static int compat_drm_addmap(struct file *file, unsigned int cmd,
 	    || __put_user(m32.flags, &map->flags))
 		return -EFAULT;
 
-	err = drm_ioctl(file->f_path.dentry->d_inode, file,
-			DRM_IOCTL_ADD_MAP, (unsigned long)map);
+	err = drm_ioctl(file, DRM_IOCTL_ADD_MAP, (unsigned long)map);
 	if (err)
 		return err;
 
@@ -284,8 +280,7 @@ static int compat_drm_rmmap(struct file *file, unsigned int cmd,
 	if (__put_user((void *)(unsigned long)handle, &map->handle))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_RM_MAP, (unsigned long)map);
+	return drm_ioctl(file, DRM_IOCTL_RM_MAP, (unsigned long)map);
 }
 
 typedef struct drm_client32 {
@@ -314,8 +309,7 @@ static int compat_drm_getclient(struct file *file, unsigned int cmd,
 	if (__put_user(idx, &client->idx))
 		return -EFAULT;
 
-	err = drm_ioctl(file->f_path.dentry->d_inode, file,
-			DRM_IOCTL_GET_CLIENT, (unsigned long)client);
+	err = drm_ioctl(file, DRM_IOCTL_GET_CLIENT, (unsigned long)client);
 	if (err)
 		return err;
 
@@ -351,8 +345,7 @@ static int compat_drm_getstats(struct file *file, unsigned int cmd,
 	if (!access_ok(VERIFY_WRITE, stats, sizeof(*stats)))
 		return -EFAULT;
 
-	err = drm_ioctl(file->f_path.dentry->d_inode, file,
-			DRM_IOCTL_GET_STATS, (unsigned long)stats);
+	err = drm_ioctl(file, DRM_IOCTL_GET_STATS, (unsigned long)stats);
 	if (err)
 		return err;
 
@@ -395,8 +388,7 @@ static int compat_drm_addbufs(struct file *file, unsigned int cmd,
 	    || __put_user(agp_start, &buf->agp_start))
 		return -EFAULT;
 
-	err = drm_ioctl(file->f_path.dentry->d_inode, file,
-			DRM_IOCTL_ADD_BUFS, (unsigned long)buf);
+	err = drm_ioctl(file, DRM_IOCTL_ADD_BUFS, (unsigned long)buf);
 	if (err)
 		return err;
 
@@ -427,8 +419,7 @@ static int compat_drm_markbufs(struct file *file, unsigned int cmd,
 	    || __put_user(b32.high_mark, &buf->high_mark))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_MARK_BUFS, (unsigned long)buf);
+	return drm_ioctl(file, DRM_IOCTL_MARK_BUFS, (unsigned long)buf);
 }
 
 typedef struct drm_buf_info32 {
@@ -469,8 +460,7 @@ static int compat_drm_infobufs(struct file *file, unsigned int cmd,
 	    || __put_user(list, &request->list))
 		return -EFAULT;
 
-	err = drm_ioctl(file->f_path.dentry->d_inode, file,
-			DRM_IOCTL_INFO_BUFS, (unsigned long)request);
+	err = drm_ioctl(file, DRM_IOCTL_INFO_BUFS, (unsigned long)request);
 	if (err)
 		return err;
 
@@ -531,8 +521,7 @@ static int compat_drm_mapbufs(struct file *file, unsigned int cmd,
 	    || __put_user(list, &request->list))
 		return -EFAULT;
 
-	err = drm_ioctl(file->f_path.dentry->d_inode, file,
-			DRM_IOCTL_MAP_BUFS, (unsigned long)request);
+	err = drm_ioctl(file, DRM_IOCTL_MAP_BUFS, (unsigned long)request);
 	if (err)
 		return err;
 
@@ -578,8 +567,7 @@ static int compat_drm_freebufs(struct file *file, unsigned int cmd,
 			  &request->list))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_FREE_BUFS, (unsigned long)request);
+	return drm_ioctl(file, DRM_IOCTL_FREE_BUFS, (unsigned long)request);
 }
 
 typedef struct drm_ctx_priv_map32 {
@@ -605,8 +593,7 @@ static int compat_drm_setsareactx(struct file *file, unsigned int cmd,
 			  &request->handle))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_SET_SAREA_CTX, (unsigned long)request);
+	return drm_ioctl(file, DRM_IOCTL_SET_SAREA_CTX, (unsigned long)request);
 }
 
 static int compat_drm_getsareactx(struct file *file, unsigned int cmd,
@@ -628,8 +615,7 @@ static int compat_drm_getsareactx(struct file *file, unsigned int cmd,
 	if (__put_user(ctx_id, &request->ctx_id))
 		return -EFAULT;
 
-	err = drm_ioctl(file->f_path.dentry->d_inode, file,
-			DRM_IOCTL_GET_SAREA_CTX, (unsigned long)request);
+	err = drm_ioctl(file, DRM_IOCTL_GET_SAREA_CTX, (unsigned long)request);
 	if (err)
 		return err;
 
@@ -664,8 +650,7 @@ static int compat_drm_resctx(struct file *file, unsigned int cmd,
 			  &res->contexts))
 		return -EFAULT;
 
-	err = drm_ioctl(file->f_path.dentry->d_inode, file,
-			DRM_IOCTL_RES_CTX, (unsigned long)res);
+	err = drm_ioctl(file, DRM_IOCTL_RES_CTX, (unsigned long)res);
 	if (err)
 		return err;
 
@@ -718,8 +703,7 @@ static int compat_drm_dma(struct file *file, unsigned int cmd,
 			  &d->request_sizes))
 		return -EFAULT;
 
-	err = drm_ioctl(file->f_path.dentry->d_inode, file,
-			DRM_IOCTL_DMA, (unsigned long)d);
+	err = drm_ioctl(file, DRM_IOCTL_DMA, (unsigned long)d);
 	if (err)
 		return err;
 
@@ -751,8 +735,7 @@ static int compat_drm_agp_enable(struct file *file, unsigned int cmd,
 	if (put_user(m32.mode, &mode->mode))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_AGP_ENABLE, (unsigned long)mode);
+	return drm_ioctl(file, DRM_IOCTL_AGP_ENABLE, (unsigned long)mode);
 }
 
 typedef struct drm_agp_info32 {
@@ -781,8 +764,7 @@ static int compat_drm_agp_info(struct file *file, unsigned int cmd,
 	if (!access_ok(VERIFY_WRITE, info, sizeof(*info)))
 		return -EFAULT;
 
-	err = drm_ioctl(file->f_path.dentry->d_inode, file,
-			DRM_IOCTL_AGP_INFO, (unsigned long)info);
+	err = drm_ioctl(file, DRM_IOCTL_AGP_INFO, (unsigned long)info);
 	if (err)
 		return err;
 
@@ -827,16 +809,14 @@ static int compat_drm_agp_alloc(struct file *file, unsigned int cmd,
 	    || __put_user(req32.type, &request->type))
 		return -EFAULT;
 
-	err = drm_ioctl(file->f_path.dentry->d_inode, file,
-			DRM_IOCTL_AGP_ALLOC, (unsigned long)request);
+	err = drm_ioctl(file, DRM_IOCTL_AGP_ALLOC, (unsigned long)request);
 	if (err)
 		return err;
 
 	if (__get_user(req32.handle, &request->handle)
 	    || __get_user(req32.physical, &request->physical)
 	    || copy_to_user(argp, &req32, sizeof(req32))) {
-		drm_ioctl(file->f_path.dentry->d_inode, file,
-			  DRM_IOCTL_AGP_FREE, (unsigned long)request);
+		drm_ioctl(file, DRM_IOCTL_AGP_FREE, (unsigned long)request);
 		return -EFAULT;
 	}
 
@@ -856,8 +836,7 @@ static int compat_drm_agp_free(struct file *file, unsigned int cmd,
 	    || __put_user(handle, &request->handle))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_AGP_FREE, (unsigned long)request);
+	return drm_ioctl(file, DRM_IOCTL_AGP_FREE, (unsigned long)request);
 }
 
 typedef struct drm_agp_binding32 {
@@ -881,8 +860,7 @@ static int compat_drm_agp_bind(struct file *file, unsigned int cmd,
 	    || __put_user(req32.offset, &request->offset))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_AGP_BIND, (unsigned long)request);
+	return drm_ioctl(file, DRM_IOCTL_AGP_BIND, (unsigned long)request);
 }
 
 static int compat_drm_agp_unbind(struct file *file, unsigned int cmd,
@@ -898,8 +876,7 @@ static int compat_drm_agp_unbind(struct file *file, unsigned int cmd,
 	    || __put_user(handle, &request->handle))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_AGP_UNBIND, (unsigned long)request);
+	return drm_ioctl(file, DRM_IOCTL_AGP_UNBIND, (unsigned long)request);
 }
 #endif				/* __OS_HAS_AGP */
 
@@ -923,8 +900,7 @@ static int compat_drm_sg_alloc(struct file *file, unsigned int cmd,
 	    || __put_user(x, &request->size))
 		return -EFAULT;
 
-	err = drm_ioctl(file->f_path.dentry->d_inode, file,
-			DRM_IOCTL_SG_ALLOC, (unsigned long)request);
+	err = drm_ioctl(file, DRM_IOCTL_SG_ALLOC, (unsigned long)request);
 	if (err)
 		return err;
 
@@ -950,8 +926,7 @@ static int compat_drm_sg_free(struct file *file, unsigned int cmd,
 	    || __put_user(x << PAGE_SHIFT, &request->handle))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_SG_FREE, (unsigned long)request);
+	return drm_ioctl(file, DRM_IOCTL_SG_FREE, (unsigned long)request);
 }
 
 #if defined(CONFIG_X86) || defined(CONFIG_IA64)
@@ -981,8 +956,7 @@ static int compat_drm_update_draw(struct file *file, unsigned int cmd,
 	    __put_user(update32.data, &request->data))
 		return -EFAULT;
 
-	err = drm_ioctl(file->f_path.dentry->d_inode, file,
-			DRM_IOCTL_UPDATE_DRAW, (unsigned long)request);
+	err = drm_ioctl(file, DRM_IOCTL_UPDATE_DRAW, (unsigned long)request);
 	return err;
 }
 #endif
@@ -1023,8 +997,7 @@ static int compat_drm_wait_vblank(struct file *file, unsigned int cmd,
 	    || __put_user(req32.request.signal, &request->request.signal))
 		return -EFAULT;
 
-	err = drm_ioctl(file->f_path.dentry->d_inode, file,
-			DRM_IOCTL_WAIT_VBLANK, (unsigned long)request);
+	err = drm_ioctl(file, DRM_IOCTL_WAIT_VBLANK, (unsigned long)request);
 	if (err)
 		return err;
 
@@ -1094,16 +1067,14 @@ long drm_compat_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
 	 * than always failing.
 	 */
 	if (nr >= ARRAY_SIZE(drm_compat_ioctls))
-		return drm_ioctl(filp->f_dentry->d_inode, filp, cmd, arg);
+		return drm_ioctl(filp, cmd, arg);
 
 	fn = drm_compat_ioctls[nr];
 
-	lock_kernel();		/* XXX for now */
 	if (fn != NULL)
 		ret = (*fn) (filp, cmd, arg);
 	else
-		ret = drm_ioctl(filp->f_path.dentry->d_inode, filp, cmd, arg);
-	unlock_kernel();
+		ret = drm_ioctl(filp, cmd, arg);
 
 	return ret;
 }
diff --git a/drivers/gpu/drm/drm_irq.c b/drivers/gpu/drm/drm_irq.c
index 0a6f0b3bdc78..b98384dbd9a7 100644
--- a/drivers/gpu/drm/drm_irq.c
+++ b/drivers/gpu/drm/drm_irq.c
@@ -115,6 +115,7 @@ void drm_vblank_cleanup(struct drm_device *dev)
 
 	dev->num_crtcs = 0;
 }
+EXPORT_SYMBOL(drm_vblank_cleanup);
 
 int drm_vblank_init(struct drm_device *dev, int num_crtcs)
 {
@@ -163,7 +164,6 @@ int drm_vblank_init(struct drm_device *dev, int num_crtcs)
 	}
 
 	dev->vblank_disable_allowed = 0;
-
 	return 0;
 
 err:
@@ -429,15 +429,21 @@ int drm_vblank_get(struct drm_device *dev, int crtc)
 
 	spin_lock_irqsave(&dev->vbl_lock, irqflags);
 	/* Going from 0->1 means we have to enable interrupts again */
-	if (atomic_add_return(1, &dev->vblank_refcount[crtc]) == 1 &&
-	    !dev->vblank_enabled[crtc]) {
-		ret = dev->driver->enable_vblank(dev, crtc);
-		DRM_DEBUG("enabling vblank on crtc %d, ret: %d\n", crtc, ret);
-		if (ret)
+	if (atomic_add_return(1, &dev->vblank_refcount[crtc]) == 1) {
+		if (!dev->vblank_enabled[crtc]) {
+			ret = dev->driver->enable_vblank(dev, crtc);
+			DRM_DEBUG("enabling vblank on crtc %d, ret: %d\n", crtc, ret);
+			if (ret)
+				atomic_dec(&dev->vblank_refcount[crtc]);
+			else {
+				dev->vblank_enabled[crtc] = 1;
+				drm_update_vblank_count(dev, crtc);
+			}
+		}
+	} else {
+		if (!dev->vblank_enabled[crtc]) {
 			atomic_dec(&dev->vblank_refcount[crtc]);
-		else {
-			dev->vblank_enabled[crtc] = 1;
-			drm_update_vblank_count(dev, crtc);
+			ret = -EINVAL;
 		}
 	}
 	spin_unlock_irqrestore(&dev->vbl_lock, irqflags);
@@ -464,6 +470,18 @@ void drm_vblank_put(struct drm_device *dev, int crtc)
 }
 EXPORT_SYMBOL(drm_vblank_put);
 
+void drm_vblank_off(struct drm_device *dev, int crtc)
+{
+	unsigned long irqflags;
+
+	spin_lock_irqsave(&dev->vbl_lock, irqflags);
+	DRM_WAKEUP(&dev->vbl_queue[crtc]);
+	dev->vblank_enabled[crtc] = 0;
+	dev->last_vblank[crtc] = dev->driver->get_vblank_counter(dev, crtc);
+	spin_unlock_irqrestore(&dev->vbl_lock, irqflags);
+}
+EXPORT_SYMBOL(drm_vblank_off);
+
 /**
  * drm_vblank_pre_modeset - account for vblanks across mode sets
  * @dev: DRM device
@@ -475,6 +493,9 @@ EXPORT_SYMBOL(drm_vblank_put);
  */
 void drm_vblank_pre_modeset(struct drm_device *dev, int crtc)
 {
+	/* vblank is not initialized (IRQ not installed ?) */
+	if (!dev->num_crtcs)
+		return;
 	/*
 	 * To avoid all the problems that might happen if interrupts
 	 * were enabled/disabled around or between these calls, we just
@@ -550,6 +571,63 @@ out:
 	return ret;
 }
 
+static int drm_queue_vblank_event(struct drm_device *dev, int pipe,
+				  union drm_wait_vblank *vblwait,
+				  struct drm_file *file_priv)
+{
+	struct drm_pending_vblank_event *e;
+	struct timeval now;
+	unsigned long flags;
+	unsigned int seq;
+
+	e = kzalloc(sizeof *e, GFP_KERNEL);
+	if (e == NULL)
+		return -ENOMEM;
+
+	e->pipe = pipe;
+	e->event.base.type = DRM_EVENT_VBLANK;
+	e->event.base.length = sizeof e->event;
+	e->event.user_data = vblwait->request.signal;
+	e->base.event = &e->event.base;
+	e->base.file_priv = file_priv;
+	e->base.destroy = (void (*) (struct drm_pending_event *)) kfree;
+
+	do_gettimeofday(&now);
+	spin_lock_irqsave(&dev->event_lock, flags);
+
+	if (file_priv->event_space < sizeof e->event) {
+		spin_unlock_irqrestore(&dev->event_lock, flags);
+		kfree(e);
+		return -ENOMEM;
+	}
+
+	file_priv->event_space -= sizeof e->event;
+	seq = drm_vblank_count(dev, pipe);
+	if ((vblwait->request.type & _DRM_VBLANK_NEXTONMISS) &&
+	    (seq - vblwait->request.sequence) <= (1 << 23)) {
+		vblwait->request.sequence = seq + 1;
+		vblwait->reply.sequence = vblwait->request.sequence;
+	}
+
+	DRM_DEBUG("event on vblank count %d, current %d, crtc %d\n",
+		  vblwait->request.sequence, seq, pipe);
+
+	e->event.sequence = vblwait->request.sequence;
+	if ((seq - vblwait->request.sequence) <= (1 << 23)) {
+		e->event.tv_sec = now.tv_sec;
+		e->event.tv_usec = now.tv_usec;
+		drm_vblank_put(dev, e->pipe);
+		list_add_tail(&e->base.link, &e->base.file_priv->event_list);
+		wake_up_interruptible(&e->base.file_priv->event_wait);
+	} else {
+		list_add_tail(&e->base.link, &dev->vblank_event_list);
+	}
+
+	spin_unlock_irqrestore(&dev->event_lock, flags);
+
+	return 0;
+}
+
 /**
  * Wait for VBLANK.
  *
@@ -609,6 +687,9 @@ int drm_wait_vblank(struct drm_device *dev, void *data,
 		goto done;
 	}
 
+	if (flags & _DRM_VBLANK_EVENT)
+		return drm_queue_vblank_event(dev, crtc, vblwait, file_priv);
+
 	if ((flags & _DRM_VBLANK_NEXTONMISS) &&
 	    (seq - vblwait->request.sequence) <= (1<<23)) {
 		vblwait->request.sequence = seq + 1;
@@ -641,6 +722,38 @@ done:
 	return ret;
 }
 
+void drm_handle_vblank_events(struct drm_device *dev, int crtc)
+{
+	struct drm_pending_vblank_event *e, *t;
+	struct timeval now;
+	unsigned long flags;
+	unsigned int seq;
+
+	do_gettimeofday(&now);
+	seq = drm_vblank_count(dev, crtc);
+
+	spin_lock_irqsave(&dev->event_lock, flags);
+
+	list_for_each_entry_safe(e, t, &dev->vblank_event_list, base.link) {
+		if (e->pipe != crtc)
+			continue;
+		if ((seq - e->event.sequence) > (1<<23))
+			continue;
+
+		DRM_DEBUG("vblank event on %d, current %d\n",
+			  e->event.sequence, seq);
+
+		e->event.sequence = seq;
+		e->event.tv_sec = now.tv_sec;
+		e->event.tv_usec = now.tv_usec;
+		drm_vblank_put(dev, e->pipe);
+		list_move_tail(&e->base.link, &e->base.file_priv->event_list);
+		wake_up_interruptible(&e->base.file_priv->event_wait);
+	}
+
+	spin_unlock_irqrestore(&dev->event_lock, flags);
+}
+
 /**
  * drm_handle_vblank - handle a vblank event
  * @dev: DRM device
@@ -651,7 +764,11 @@ done:
  */
 void drm_handle_vblank(struct drm_device *dev, int crtc)
 {
+	if (!dev->num_crtcs)
+		return;
+
 	atomic_inc(&dev->_vblank_count[crtc]);
 	DRM_WAKEUP(&dev->vbl_queue[crtc]);
+	drm_handle_vblank_events(dev, crtc);
 }
 EXPORT_SYMBOL(drm_handle_vblank);
diff --git a/drivers/gpu/drm/drm_mm.c b/drivers/gpu/drm/drm_mm.c
index c861d80fd779..2ac074c8f5d2 100644
--- a/drivers/gpu/drm/drm_mm.c
+++ b/drivers/gpu/drm/drm_mm.c
@@ -103,6 +103,11 @@ static struct drm_mm_node *drm_mm_kmalloc(struct drm_mm *mm, int atomic)
 	return child;
 }
 
+/* drm_mm_pre_get() - pre allocate drm_mm_node structure
+ * drm_mm:	memory manager struct we are pre-allocating for
+ *
+ * Returns 0 on success or -ENOMEM if allocation fails.
+ */
 int drm_mm_pre_get(struct drm_mm *mm)
 {
 	struct drm_mm_node *node;
@@ -221,6 +226,44 @@ struct drm_mm_node *drm_mm_get_block_generic(struct drm_mm_node *node,
 }
 EXPORT_SYMBOL(drm_mm_get_block_generic);
 
+struct drm_mm_node *drm_mm_get_block_range_generic(struct drm_mm_node *node,
+						unsigned long size,
+						unsigned alignment,
+						unsigned long start,
+						unsigned long end,
+						int atomic)
+{
+	struct drm_mm_node *align_splitoff = NULL;
+	unsigned tmp = 0;
+	unsigned wasted = 0;
+
+	if (node->start < start)
+		wasted += start - node->start;
+	if (alignment)
+		tmp = ((node->start + wasted) % alignment);
+
+	if (tmp)
+		wasted += alignment - tmp;
+	if (wasted) {
+		align_splitoff = drm_mm_split_at_start(node, wasted, atomic);
+		if (unlikely(align_splitoff == NULL))
+			return NULL;
+	}
+
+	if (node->size == size) {
+		list_del_init(&node->fl_entry);
+		node->free = 0;
+	} else {
+		node = drm_mm_split_at_start(node, size, atomic);
+	}
+
+	if (align_splitoff)
+		drm_mm_put_block(align_splitoff);
+
+	return node;
+}
+EXPORT_SYMBOL(drm_mm_get_block_range_generic);
+
 /*
  * Put a block. Merge with the previous and / or next block if they are free.
  * Otherwise add to the free stack.
@@ -253,12 +296,14 @@ void drm_mm_put_block(struct drm_mm_node *cur)
 				prev_node->size += next_node->size;
 				list_del(&next_node->ml_entry);
 				list_del(&next_node->fl_entry);
+				spin_lock(&mm->unused_lock);
 				if (mm->num_unused < MM_UNUSED_TARGET) {
 					list_add(&next_node->fl_entry,
 						 &mm->unused_nodes);
 					++mm->num_unused;
 				} else
 					kfree(next_node);
+				spin_unlock(&mm->unused_lock);
 			} else {
 				next_node->size += cur->size;
 				next_node->start = cur->start;
@@ -271,11 +316,13 @@ void drm_mm_put_block(struct drm_mm_node *cur)
 		list_add(&cur->fl_entry, &mm->fl_entry);
 	} else {
 		list_del(&cur->ml_entry);
+		spin_lock(&mm->unused_lock);
 		if (mm->num_unused < MM_UNUSED_TARGET) {
 			list_add(&cur->fl_entry, &mm->unused_nodes);
 			++mm->num_unused;
 		} else
 			kfree(cur);
+		spin_unlock(&mm->unused_lock);
 	}
 }
 
@@ -311,7 +358,7 @@ struct drm_mm_node *drm_mm_search_free(const struct drm_mm *mm,
 		if (entry->size >= size + wasted) {
 			if (!best_match)
 				return entry;
-			if (size < best_size) {
+			if (entry->size < best_size) {
 				best = entry;
 				best_size = entry->size;
 			}
@@ -322,6 +369,57 @@ struct drm_mm_node *drm_mm_search_free(const struct drm_mm *mm,
 }
 EXPORT_SYMBOL(drm_mm_search_free);
 
+struct drm_mm_node *drm_mm_search_free_in_range(const struct drm_mm *mm,
+						unsigned long size,
+						unsigned alignment,
+						unsigned long start,
+						unsigned long end,
+						int best_match)
+{
+	struct list_head *list;
+	const struct list_head *free_stack = &mm->fl_entry;
+	struct drm_mm_node *entry;
+	struct drm_mm_node *best;
+	unsigned long best_size;
+	unsigned wasted;
+
+	best = NULL;
+	best_size = ~0UL;
+
+	list_for_each(list, free_stack) {
+		entry = list_entry(list, struct drm_mm_node, fl_entry);
+		wasted = 0;
+
+		if (entry->size < size)
+			continue;
+
+		if (entry->start > end || (entry->start+entry->size) < start)
+			continue;
+
+		if (entry->start < start)
+			wasted += start - entry->start;
+
+		if (alignment) {
+			register unsigned tmp = (entry->start + wasted) % alignment;
+			if (tmp)
+				wasted += alignment - tmp;
+		}
+
+		if (entry->size >= size + wasted &&
+		    (entry->start + wasted + size) <= end) {
+			if (!best_match)
+				return entry;
+			if (entry->size < best_size) {
+				best = entry;
+				best_size = entry->size;
+			}
+		}
+	}
+
+	return best;
+}
+EXPORT_SYMBOL(drm_mm_search_free_in_range);
+
 int drm_mm_clean(struct drm_mm * mm)
 {
 	struct list_head *head = &mm->ml_entry;
@@ -372,6 +470,26 @@ void drm_mm_takedown(struct drm_mm * mm)
 }
 EXPORT_SYMBOL(drm_mm_takedown);
 
+void drm_mm_debug_table(struct drm_mm *mm, const char *prefix)
+{
+	struct drm_mm_node *entry;
+	int total_used = 0, total_free = 0, total = 0;
+
+	list_for_each_entry(entry, &mm->ml_entry, ml_entry) {
+		printk(KERN_DEBUG "%s 0x%08lx-0x%08lx: %8ld: %s\n",
+			prefix, entry->start, entry->start + entry->size,
+			entry->size, entry->free ? "free" : "used");
+		total += entry->size;
+		if (entry->free)
+			total_free += entry->size;
+		else
+			total_used += entry->size;
+	}
+	printk(KERN_DEBUG "%s total: %d, used %d free %d\n", prefix, total,
+		total_used, total_free);
+}
+EXPORT_SYMBOL(drm_mm_debug_table);
+
 #if defined(CONFIG_DEBUG_FS)
 int drm_mm_dump_table(struct seq_file *m, struct drm_mm *mm)
 {
@@ -386,7 +504,7 @@ int drm_mm_dump_table(struct seq_file *m, struct drm_mm *mm)
 		else
 			total_used += entry->size;
 	}
-	seq_printf(m, "total: %d, used %d free %d\n", total, total_free, total_used);
+	seq_printf(m, "total: %d, used %d free %d\n", total, total_used, total_free);
 	return 0;
 }
 EXPORT_SYMBOL(drm_mm_dump_table);
diff --git a/drivers/gpu/drm/drm_modes.c b/drivers/gpu/drm/drm_modes.c
index 49404ce1666e..76d63394c776 100644
--- a/drivers/gpu/drm/drm_modes.c
+++ b/drivers/gpu/drm/drm_modes.c
@@ -1,9 +1,4 @@
 /*
- * The list_sort function is (presumably) licensed under the GPL (see the
- * top level "COPYING" file for details).
- *
- * The remainder of this file is:
- *
  * Copyright © 1997-2003 by The XFree86 Project, Inc.
  * Copyright © 2007 Dave Airlie
  * Copyright © 2007-2008 Intel Corporation
@@ -36,6 +31,7 @@
  */
 
 #include <linux/list.h>
+#include <linux/list_sort.h>
 #include "drmP.h"
 #include "drm.h"
 #include "drm_crtc.h"
@@ -88,7 +84,7 @@ EXPORT_SYMBOL(drm_mode_debug_printmodeline);
 #define HV_FACTOR			1000
 struct drm_display_mode *drm_cvt_mode(struct drm_device *dev, int hdisplay,
 				      int vdisplay, int vrefresh,
-				      bool reduced, bool interlaced)
+				      bool reduced, bool interlaced, bool margins)
 {
 	/* 1) top/bottom margin size (% of height) - default: 1.8, */
 #define	CVT_MARGIN_PERCENTAGE		18
@@ -101,7 +97,6 @@ struct drm_display_mode *drm_cvt_mode(struct drm_device *dev, int hdisplay,
 	/* Pixel Clock step (kHz) */
 #define CVT_CLOCK_STEP			250
 	struct drm_display_mode *drm_mode;
-	bool margins = false;
 	unsigned int vfieldrate, hperiod;
 	int hdisplay_rnd, hmargin, vdisplay_rnd, vmargin, vsync;
 	int interlace;
@@ -554,6 +549,32 @@ int drm_mode_height(struct drm_display_mode *mode)
 }
 EXPORT_SYMBOL(drm_mode_height);
 
+/** drm_mode_hsync - get the hsync of a mode
+ * @mode: mode
+ *
+ * LOCKING:
+ * None.
+ *
+ * Return @modes's hsync rate in kHz, rounded to the nearest int.
+ */
+int drm_mode_hsync(struct drm_display_mode *mode)
+{
+	unsigned int calc_val;
+
+	if (mode->hsync)
+		return mode->hsync;
+
+	if (mode->htotal < 0)
+		return 0;
+
+	calc_val = (mode->clock * 1000) / mode->htotal; /* hsync in Hz */
+	calc_val += 500;				/* round to 1000Hz */
+	calc_val /= 1000;				/* truncate to kHz */
+
+	return calc_val;
+}
+EXPORT_SYMBOL(drm_mode_hsync);
+
 /**
  * drm_mode_vrefresh - get the vrefresh of a mode
  * @mode: mode
@@ -561,7 +582,7 @@ EXPORT_SYMBOL(drm_mode_height);
  * LOCKING:
  * None.
  *
- * Return @mode's vrefresh rate or calculate it if necessary.
+ * Return @mode's vrefresh rate in Hz or calculate it if necessary.
  *
  * FIXME: why is this needed?  shouldn't vrefresh be set already?
  *
@@ -830,6 +851,7 @@ EXPORT_SYMBOL(drm_mode_prune_invalid);
 
 /**
  * drm_mode_compare - compare modes for favorability
+ * @priv: unused
  * @lh_a: list_head for first mode
  * @lh_b: list_head for second mode
  *
@@ -843,7 +865,7 @@ EXPORT_SYMBOL(drm_mode_prune_invalid);
  * Negative if @lh_a is better than @lh_b, zero if they're equivalent, or
  * positive if @lh_b is better than @lh_a.
  */
-static int drm_mode_compare(struct list_head *lh_a, struct list_head *lh_b)
+static int drm_mode_compare(void *priv, struct list_head *lh_a, struct list_head *lh_b)
 {
 	struct drm_display_mode *a = list_entry(lh_a, struct drm_display_mode, head);
 	struct drm_display_mode *b = list_entry(lh_b, struct drm_display_mode, head);
@@ -860,85 +882,6 @@ static int drm_mode_compare(struct list_head *lh_a, struct list_head *lh_b)
 	return diff;
 }
 
-/* FIXME: what we don't have a list sort function? */
-/* list sort from Mark J Roberts (mjr@znex.org) */
-void list_sort(struct list_head *head,
-	       int (*cmp)(struct list_head *a, struct list_head *b))
-{
-	struct list_head *p, *q, *e, *list, *tail, *oldhead;
-	int insize, nmerges, psize, qsize, i;
-
-	list = head->next;
-	list_del(head);
-	insize = 1;
-	for (;;) {
-		p = oldhead = list;
-		list = tail = NULL;
-		nmerges = 0;
-
-		while (p) {
-			nmerges++;
-			q = p;
-			psize = 0;
-			for (i = 0; i < insize; i++) {
-				psize++;
-				q = q->next == oldhead ? NULL : q->next;
-				if (!q)
-					break;
-			}
-
-			qsize = insize;
-			while (psize > 0 || (qsize > 0 && q)) {
-				if (!psize) {
-					e = q;
-					q = q->next;
-					qsize--;
-					if (q == oldhead)
-						q = NULL;
-				} else if (!qsize || !q) {
-					e = p;
-					p = p->next;
-					psize--;
-					if (p == oldhead)
-						p = NULL;
-				} else if (cmp(p, q) <= 0) {
-					e = p;
-					p = p->next;
-					psize--;
-					if (p == oldhead)
-						p = NULL;
-				} else {
-					e = q;
-					q = q->next;
-					qsize--;
-					if (q == oldhead)
-						q = NULL;
-				}
-				if (tail)
-					tail->next = e;
-				else
-					list = e;
-				e->prev = tail;
-				tail = e;
-			}
-			p = q;
-		}
-
-		tail->next = list;
-		list->prev = tail;
-
-		if (nmerges <= 1)
-			break;
-
-		insize *= 2;
-	}
-
-	head->next = list;
-	head->prev = list->prev;
-	list->prev->next = head;
-	list->prev = head;
-}
-
 /**
  * drm_mode_sort - sort mode list
  * @mode_list: list to sort
@@ -950,7 +893,7 @@ void list_sort(struct list_head *head,
  */
 void drm_mode_sort(struct list_head *mode_list)
 {
-	list_sort(mode_list, drm_mode_compare);
+	list_sort(NULL, mode_list, drm_mode_compare);
 }
 EXPORT_SYMBOL(drm_mode_sort);
 
diff --git a/drivers/gpu/drm/drm_pci.c b/drivers/gpu/drm/drm_pci.c
index 577094fb1995..e68ebf92fa2a 100644
--- a/drivers/gpu/drm/drm_pci.c
+++ b/drivers/gpu/drm/drm_pci.c
@@ -47,8 +47,7 @@
 /**
  * \brief Allocate a PCI consistent memory block, for DMA.
  */
-drm_dma_handle_t *drm_pci_alloc(struct drm_device * dev, size_t size, size_t align,
-				dma_addr_t maxaddr)
+drm_dma_handle_t *drm_pci_alloc(struct drm_device * dev, size_t size, size_t align)
 {
 	drm_dma_handle_t *dmah;
 #if 1
@@ -63,11 +62,6 @@ drm_dma_handle_t *drm_pci_alloc(struct drm_device * dev, size_t size, size_t ali
 	if (align > size)
 		return NULL;
 
-	if (pci_set_dma_mask(dev->pdev, maxaddr) != 0) {
-		DRM_ERROR("Setting pci dma mask failed\n");
-		return NULL;
-	}
-
 	dmah = kmalloc(sizeof(drm_dma_handle_t), GFP_KERNEL);
 	if (!dmah)
 		return NULL;
diff --git a/drivers/gpu/drm/drm_stub.c b/drivers/gpu/drm/drm_stub.c
index 55bb8a82d612..ad73e141afdb 100644
--- a/drivers/gpu/drm/drm_stub.c
+++ b/drivers/gpu/drm/drm_stub.c
@@ -128,6 +128,7 @@ struct drm_master *drm_master_get(struct drm_master *master)
 	kref_get(&master->refcount);
 	return master;
 }
+EXPORT_SYMBOL(drm_master_get);
 
 static void drm_master_destroy(struct kref *kref)
 {
@@ -170,10 +171,13 @@ void drm_master_put(struct drm_master **master)
 	kref_put(&(*master)->refcount, drm_master_destroy);
 	*master = NULL;
 }
+EXPORT_SYMBOL(drm_master_put);
 
 int drm_setmaster_ioctl(struct drm_device *dev, void *data,
 			struct drm_file *file_priv)
 {
+	int ret = 0;
+
 	if (file_priv->is_master)
 		return 0;
 
@@ -188,6 +192,13 @@ int drm_setmaster_ioctl(struct drm_device *dev, void *data,
 		mutex_lock(&dev->struct_mutex);
 		file_priv->minor->master = drm_master_get(file_priv->master);
 		file_priv->is_master = 1;
+		if (dev->driver->master_set) {
+			ret = dev->driver->master_set(dev, file_priv, false);
+			if (unlikely(ret != 0)) {
+				file_priv->is_master = 0;
+				drm_master_put(&file_priv->minor->master);
+			}
+		}
 		mutex_unlock(&dev->struct_mutex);
 	}
 
@@ -204,6 +215,8 @@ int drm_dropmaster_ioctl(struct drm_device *dev, void *data,
 		return -EINVAL;
 
 	mutex_lock(&dev->struct_mutex);
+	if (dev->driver->master_drop)
+		dev->driver->master_drop(dev, file_priv, false);
 	drm_master_put(&file_priv->minor->master);
 	file_priv->is_master = 0;
 	mutex_unlock(&dev->struct_mutex);
@@ -220,9 +233,11 @@ static int drm_fill_in_dev(struct drm_device * dev, struct pci_dev *pdev,
 	INIT_LIST_HEAD(&dev->ctxlist);
 	INIT_LIST_HEAD(&dev->vmalist);
 	INIT_LIST_HEAD(&dev->maplist);
+	INIT_LIST_HEAD(&dev->vblank_event_list);
 
 	spin_lock_init(&dev->count_lock);
 	spin_lock_init(&dev->drw_lock);
+	spin_lock_init(&dev->event_lock);
 	init_timer(&dev->timer);
 	mutex_init(&dev->struct_mutex);
 	mutex_init(&dev->ctxlist_mutex);
diff --git a/drivers/gpu/drm/drm_sysfs.c b/drivers/gpu/drm/drm_sysfs.c
index 7e42b7e9d43a..014ce24761b9 100644
--- a/drivers/gpu/drm/drm_sysfs.c
+++ b/drivers/gpu/drm/drm_sysfs.c
@@ -70,19 +70,17 @@ static int drm_class_resume(struct device *dev)
 	return 0;
 }
 
-/* Display the version of drm_core. This doesn't work right in current design */
-static ssize_t version_show(struct class *dev, char *buf)
-{
-	return sprintf(buf, "%s %d.%d.%d %s\n", CORE_NAME, CORE_MAJOR,
-		       CORE_MINOR, CORE_PATCHLEVEL, CORE_DATE);
-}
-
 static char *drm_devnode(struct device *dev, mode_t *mode)
 {
 	return kasprintf(GFP_KERNEL, "dri/%s", dev_name(dev));
 }
 
-static CLASS_ATTR(version, S_IRUGO, version_show, NULL);
+static CLASS_ATTR_STRING(version, S_IRUGO,
+		CORE_NAME " "
+		__stringify(CORE_MAJOR) "."
+		__stringify(CORE_MINOR) "."
+		__stringify(CORE_PATCHLEVEL) " "
+		CORE_DATE);
 
 /**
  * drm_sysfs_create - create a struct drm_sysfs_class structure
@@ -109,7 +107,7 @@ struct class *drm_sysfs_create(struct module *owner, char *name)
 	class->suspend = drm_class_suspend;
 	class->resume = drm_class_resume;
 
-	err = class_create_file(class, &class_attr_version);
+	err = class_create_file(class, &class_attr_version.attr);
 	if (err)
 		goto err_out_class;
 
@@ -132,7 +130,7 @@ void drm_sysfs_destroy(void)
 {
 	if ((drm_class == NULL) || (IS_ERR(drm_class)))
 		return;
-	class_remove_file(drm_class, &class_attr_version);
+	class_remove_file(drm_class, &class_attr_version.attr);
 	class_destroy(drm_class);
 }
 
diff --git a/drivers/gpu/drm/i2c/Makefile b/drivers/gpu/drm/i2c/Makefile
new file mode 100644
index 000000000000..6d2abaf35ba2
--- /dev/null
+++ b/drivers/gpu/drm/i2c/Makefile
@@ -0,0 +1,4 @@
+ccflags-y := -Iinclude/drm
+
+ch7006-y := ch7006_drv.o ch7006_mode.o
+obj-$(CONFIG_DRM_I2C_CH7006) += ch7006.o
diff --git a/drivers/gpu/drm/i2c/ch7006_drv.c b/drivers/gpu/drm/i2c/ch7006_drv.c
new file mode 100644
index 000000000000..81681a07a806
--- /dev/null
+++ b/drivers/gpu/drm/i2c/ch7006_drv.c
@@ -0,0 +1,536 @@
+/*
+ * Copyright (C) 2009 Francisco Jerez.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "ch7006_priv.h"
+
+/* DRM encoder functions */
+
+static void ch7006_encoder_set_config(struct drm_encoder *encoder,
+				      void *params)
+{
+	struct ch7006_priv *priv = to_ch7006_priv(encoder);
+
+	priv->params = params;
+}
+
+static void ch7006_encoder_destroy(struct drm_encoder *encoder)
+{
+	struct ch7006_priv *priv = to_ch7006_priv(encoder);
+
+	drm_property_destroy(encoder->dev, priv->scale_property);
+
+	kfree(priv);
+	to_encoder_slave(encoder)->slave_priv = NULL;
+
+	drm_i2c_encoder_destroy(encoder);
+}
+
+static void  ch7006_encoder_dpms(struct drm_encoder *encoder, int mode)
+{
+	struct i2c_client *client = drm_i2c_encoder_get_client(encoder);
+	struct ch7006_priv *priv = to_ch7006_priv(encoder);
+	struct ch7006_state *state = &priv->state;
+
+	ch7006_dbg(client, "\n");
+
+	if (mode == priv->last_dpms)
+		return;
+	priv->last_dpms = mode;
+
+	ch7006_setup_power_state(encoder);
+
+	ch7006_load_reg(client, state, CH7006_POWER);
+}
+
+static void ch7006_encoder_save(struct drm_encoder *encoder)
+{
+	struct i2c_client *client = drm_i2c_encoder_get_client(encoder);
+	struct ch7006_priv *priv = to_ch7006_priv(encoder);
+
+	ch7006_dbg(client, "\n");
+
+	ch7006_state_save(client, &priv->saved_state);
+}
+
+static void ch7006_encoder_restore(struct drm_encoder *encoder)
+{
+	struct i2c_client *client = drm_i2c_encoder_get_client(encoder);
+	struct ch7006_priv *priv = to_ch7006_priv(encoder);
+
+	ch7006_dbg(client, "\n");
+
+	ch7006_state_load(client, &priv->saved_state);
+}
+
+static bool ch7006_encoder_mode_fixup(struct drm_encoder *encoder,
+				      struct drm_display_mode *mode,
+				      struct drm_display_mode *adjusted_mode)
+{
+	struct ch7006_priv *priv = to_ch7006_priv(encoder);
+
+	/* The ch7006 is painfully picky with the input timings so no
+	 * custom modes for now... */
+
+	priv->mode = ch7006_lookup_mode(encoder, mode);
+
+	return !!priv->mode;
+}
+
+static int ch7006_encoder_mode_valid(struct drm_encoder *encoder,
+				     struct drm_display_mode *mode)
+{
+	if (ch7006_lookup_mode(encoder, mode))
+		return MODE_OK;
+	else
+		return MODE_BAD;
+}
+
+static void ch7006_encoder_mode_set(struct drm_encoder *encoder,
+				     struct drm_display_mode *drm_mode,
+				     struct drm_display_mode *adjusted_mode)
+{
+	struct i2c_client *client = drm_i2c_encoder_get_client(encoder);
+	struct ch7006_priv *priv = to_ch7006_priv(encoder);
+	struct ch7006_encoder_params *params = priv->params;
+	struct ch7006_state *state = &priv->state;
+	uint8_t *regs = state->regs;
+	struct ch7006_mode *mode = priv->mode;
+	struct ch7006_tv_norm_info *norm = &ch7006_tv_norms[priv->norm];
+	int start_active;
+
+	ch7006_dbg(client, "\n");
+
+	regs[CH7006_DISPMODE] = norm->dispmode | mode->dispmode;
+	regs[CH7006_BWIDTH] = 0;
+	regs[CH7006_INPUT_FORMAT] = bitf(CH7006_INPUT_FORMAT_FORMAT,
+					 params->input_format);
+
+	regs[CH7006_CLKMODE] = CH7006_CLKMODE_SUBC_LOCK
+		| bitf(CH7006_CLKMODE_XCM, params->xcm)
+		| bitf(CH7006_CLKMODE_PCM, params->pcm);
+	if (params->clock_mode)
+		regs[CH7006_CLKMODE] |= CH7006_CLKMODE_MASTER;
+	if (params->clock_edge)
+		regs[CH7006_CLKMODE] |= CH7006_CLKMODE_POS_EDGE;
+
+	start_active = (drm_mode->htotal & ~0x7) - (drm_mode->hsync_start & ~0x7);
+	regs[CH7006_POV] = bitf(CH7006_POV_START_ACTIVE_8, start_active);
+	regs[CH7006_START_ACTIVE] = bitf(CH7006_START_ACTIVE_0, start_active);
+
+	regs[CH7006_INPUT_SYNC] = 0;
+	if (params->sync_direction)
+		regs[CH7006_INPUT_SYNC] |= CH7006_INPUT_SYNC_OUTPUT;
+	if (params->sync_encoding)
+		regs[CH7006_INPUT_SYNC] |= CH7006_INPUT_SYNC_EMBEDDED;
+	if (drm_mode->flags & DRM_MODE_FLAG_PVSYNC)
+		regs[CH7006_INPUT_SYNC] |= CH7006_INPUT_SYNC_PVSYNC;
+	if (drm_mode->flags & DRM_MODE_FLAG_PHSYNC)
+		regs[CH7006_INPUT_SYNC] |= CH7006_INPUT_SYNC_PHSYNC;
+
+	regs[CH7006_DETECT] = 0;
+	regs[CH7006_BCLKOUT] = 0;
+
+	regs[CH7006_SUBC_INC3] = 0;
+	if (params->pout_level)
+		regs[CH7006_SUBC_INC3] |= CH7006_SUBC_INC3_POUT_3_3V;
+
+	regs[CH7006_SUBC_INC4] = 0;
+	if (params->active_detect)
+		regs[CH7006_SUBC_INC4] |= CH7006_SUBC_INC4_DS_INPUT;
+
+	regs[CH7006_PLL_CONTROL] = priv->saved_state.regs[CH7006_PLL_CONTROL];
+
+	ch7006_setup_levels(encoder);
+	ch7006_setup_subcarrier(encoder);
+	ch7006_setup_pll(encoder);
+	ch7006_setup_power_state(encoder);
+	ch7006_setup_properties(encoder);
+
+	ch7006_state_load(client, state);
+}
+
+static enum drm_connector_status ch7006_encoder_detect(struct drm_encoder *encoder,
+						       struct drm_connector *connector)
+{
+	struct i2c_client *client = drm_i2c_encoder_get_client(encoder);
+	struct ch7006_priv *priv = to_ch7006_priv(encoder);
+	struct ch7006_state *state = &priv->state;
+	int det;
+
+	ch7006_dbg(client, "\n");
+
+	ch7006_save_reg(client, state, CH7006_DETECT);
+	ch7006_save_reg(client, state, CH7006_POWER);
+	ch7006_save_reg(client, state, CH7006_CLKMODE);
+
+	ch7006_write(client, CH7006_POWER, CH7006_POWER_RESET |
+					   bitfs(CH7006_POWER_LEVEL, NORMAL));
+	ch7006_write(client, CH7006_CLKMODE, CH7006_CLKMODE_MASTER);
+
+	ch7006_write(client, CH7006_DETECT, CH7006_DETECT_SENSE);
+
+	ch7006_write(client, CH7006_DETECT, 0);
+
+	det = ch7006_read(client, CH7006_DETECT);
+
+	ch7006_load_reg(client, state, CH7006_CLKMODE);
+	ch7006_load_reg(client, state, CH7006_POWER);
+	ch7006_load_reg(client, state, CH7006_DETECT);
+
+	if ((det & (CH7006_DETECT_SVIDEO_Y_TEST|
+		    CH7006_DETECT_SVIDEO_C_TEST|
+		    CH7006_DETECT_CVBS_TEST)) == 0)
+		priv->subconnector = DRM_MODE_SUBCONNECTOR_SCART;
+	else if ((det & (CH7006_DETECT_SVIDEO_Y_TEST|
+			 CH7006_DETECT_SVIDEO_C_TEST)) == 0)
+		priv->subconnector = DRM_MODE_SUBCONNECTOR_SVIDEO;
+	else if ((det & CH7006_DETECT_CVBS_TEST) == 0)
+		priv->subconnector = DRM_MODE_SUBCONNECTOR_Composite;
+	else
+		priv->subconnector = DRM_MODE_SUBCONNECTOR_Unknown;
+
+	drm_connector_property_set_value(connector,
+			encoder->dev->mode_config.tv_subconnector_property,
+							priv->subconnector);
+
+	return priv->subconnector ? connector_status_connected :
+					connector_status_disconnected;
+}
+
+static int ch7006_encoder_get_modes(struct drm_encoder *encoder,
+				    struct drm_connector *connector)
+{
+	struct ch7006_priv *priv = to_ch7006_priv(encoder);
+	struct ch7006_mode *mode;
+	int n = 0;
+
+	for (mode = ch7006_modes; mode->mode.clock; mode++) {
+		if (~mode->valid_scales & 1<<priv->scale ||
+		    ~mode->valid_norms & 1<<priv->norm)
+			continue;
+
+		drm_mode_probed_add(connector,
+				drm_mode_duplicate(encoder->dev, &mode->mode));
+
+		n++;
+	}
+
+	return n;
+}
+
+static int ch7006_encoder_create_resources(struct drm_encoder *encoder,
+					   struct drm_connector *connector)
+{
+	struct ch7006_priv *priv = to_ch7006_priv(encoder);
+	struct drm_device *dev = encoder->dev;
+	struct drm_mode_config *conf = &dev->mode_config;
+
+	drm_mode_create_tv_properties(dev, NUM_TV_NORMS, ch7006_tv_norm_names);
+
+	priv->scale_property = drm_property_create(dev, DRM_MODE_PROP_RANGE,
+						   "scale", 2);
+	priv->scale_property->values[0] = 0;
+	priv->scale_property->values[1] = 2;
+
+	drm_connector_attach_property(connector, conf->tv_select_subconnector_property,
+				      priv->select_subconnector);
+	drm_connector_attach_property(connector, conf->tv_subconnector_property,
+				      priv->subconnector);
+	drm_connector_attach_property(connector, conf->tv_left_margin_property,
+				      priv->hmargin);
+	drm_connector_attach_property(connector, conf->tv_bottom_margin_property,
+				      priv->vmargin);
+	drm_connector_attach_property(connector, conf->tv_mode_property,
+				      priv->norm);
+	drm_connector_attach_property(connector, conf->tv_brightness_property,
+				      priv->brightness);
+	drm_connector_attach_property(connector, conf->tv_contrast_property,
+				      priv->contrast);
+	drm_connector_attach_property(connector, conf->tv_flicker_reduction_property,
+				      priv->flicker);
+	drm_connector_attach_property(connector, priv->scale_property,
+				      priv->scale);
+
+	return 0;
+}
+
+static int ch7006_encoder_set_property(struct drm_encoder *encoder,
+				       struct drm_connector *connector,
+				       struct drm_property *property,
+				       uint64_t val)
+{
+	struct i2c_client *client = drm_i2c_encoder_get_client(encoder);
+	struct ch7006_priv *priv = to_ch7006_priv(encoder);
+	struct ch7006_state *state = &priv->state;
+	struct drm_mode_config *conf = &encoder->dev->mode_config;
+	struct drm_crtc *crtc = encoder->crtc;
+	bool modes_changed = false;
+
+	ch7006_dbg(client, "\n");
+
+	if (property == conf->tv_select_subconnector_property) {
+		priv->select_subconnector = val;
+
+		ch7006_setup_power_state(encoder);
+
+		ch7006_load_reg(client, state, CH7006_POWER);
+
+	} else if (property == conf->tv_left_margin_property) {
+		priv->hmargin = val;
+
+		ch7006_setup_properties(encoder);
+
+		ch7006_load_reg(client, state, CH7006_POV);
+		ch7006_load_reg(client, state, CH7006_HPOS);
+
+	} else if (property == conf->tv_bottom_margin_property) {
+		priv->vmargin = val;
+
+		ch7006_setup_properties(encoder);
+
+		ch7006_load_reg(client, state, CH7006_POV);
+		ch7006_load_reg(client, state, CH7006_VPOS);
+
+	} else if (property == conf->tv_mode_property) {
+		if (connector->dpms != DRM_MODE_DPMS_OFF)
+			return -EINVAL;
+
+		priv->norm = val;
+
+		modes_changed = true;
+
+	} else if (property == conf->tv_brightness_property) {
+		priv->brightness = val;
+
+		ch7006_setup_levels(encoder);
+
+		ch7006_load_reg(client, state, CH7006_BLACK_LEVEL);
+
+	} else if (property == conf->tv_contrast_property) {
+		priv->contrast = val;
+
+		ch7006_setup_properties(encoder);
+
+		ch7006_load_reg(client, state, CH7006_CONTRAST);
+
+	} else if (property == conf->tv_flicker_reduction_property) {
+		priv->flicker = val;
+
+		ch7006_setup_properties(encoder);
+
+		ch7006_load_reg(client, state, CH7006_FFILTER);
+
+	} else if (property == priv->scale_property) {
+		if (connector->dpms != DRM_MODE_DPMS_OFF)
+			return -EINVAL;
+
+		priv->scale = val;
+
+		modes_changed = true;
+
+	} else {
+		return -EINVAL;
+	}
+
+	if (modes_changed) {
+		drm_helper_probe_single_connector_modes(connector, 0, 0);
+
+		/* Disable the crtc to ensure a full modeset is
+		 * performed whenever it's turned on again. */
+		if (crtc) {
+			struct drm_mode_set modeset = {
+				.crtc = crtc,
+			};
+
+			crtc->funcs->set_config(&modeset);
+		}
+	}
+
+	return 0;
+}
+
+static struct drm_encoder_slave_funcs ch7006_encoder_funcs = {
+	.set_config = ch7006_encoder_set_config,
+	.destroy = ch7006_encoder_destroy,
+	.dpms = ch7006_encoder_dpms,
+	.save = ch7006_encoder_save,
+	.restore = ch7006_encoder_restore,
+	.mode_fixup = ch7006_encoder_mode_fixup,
+	.mode_valid = ch7006_encoder_mode_valid,
+	.mode_set = ch7006_encoder_mode_set,
+	.detect = ch7006_encoder_detect,
+	.get_modes = ch7006_encoder_get_modes,
+	.create_resources = ch7006_encoder_create_resources,
+	.set_property = ch7006_encoder_set_property,
+};
+
+
+/* I2C driver functions */
+
+static int ch7006_probe(struct i2c_client *client, const struct i2c_device_id *id)
+{
+	uint8_t addr = CH7006_VERSION_ID;
+	uint8_t val;
+	int ret;
+
+	ch7006_dbg(client, "\n");
+
+	ret = i2c_master_send(client, &addr, sizeof(addr));
+	if (ret < 0)
+		goto fail;
+
+	ret = i2c_master_recv(client, &val, sizeof(val));
+	if (ret < 0)
+		goto fail;
+
+	ch7006_info(client, "Detected version ID: %x\n", val);
+
+	/* I don't know what this is for, but otherwise I get no
+	 * signal.
+	 */
+	ch7006_write(client, 0x3d, 0x0);
+
+	return 0;
+
+fail:
+	ch7006_err(client, "Error %d reading version ID\n", ret);
+
+	return -ENODEV;
+}
+
+static int ch7006_remove(struct i2c_client *client)
+{
+	ch7006_dbg(client, "\n");
+
+	return 0;
+}
+
+static int ch7006_encoder_init(struct i2c_client *client,
+			       struct drm_device *dev,
+			       struct drm_encoder_slave *encoder)
+{
+	struct ch7006_priv *priv;
+	int i;
+
+	ch7006_dbg(client, "\n");
+
+	priv = kzalloc(sizeof(*priv), GFP_KERNEL);
+	if (!priv)
+		return -ENOMEM;
+
+	encoder->slave_priv = priv;
+	encoder->slave_funcs = &ch7006_encoder_funcs;
+
+	priv->norm = TV_NORM_PAL;
+	priv->select_subconnector = DRM_MODE_SUBCONNECTOR_Automatic;
+	priv->subconnector = DRM_MODE_SUBCONNECTOR_Unknown;
+	priv->scale = 1;
+	priv->contrast = 50;
+	priv->brightness = 50;
+	priv->flicker = 50;
+	priv->hmargin = 50;
+	priv->vmargin = 50;
+	priv->last_dpms = -1;
+
+	if (ch7006_tv_norm) {
+		for (i = 0; i < NUM_TV_NORMS; i++) {
+			if (!strcmp(ch7006_tv_norm_names[i], ch7006_tv_norm)) {
+				priv->norm = i;
+				break;
+			}
+		}
+
+		if (i == NUM_TV_NORMS)
+			ch7006_err(client, "Invalid TV norm setting \"%s\".\n",
+				   ch7006_tv_norm);
+	}
+
+	if (ch7006_scale >= 0 && ch7006_scale <= 2)
+		priv->scale = ch7006_scale;
+	else
+		ch7006_err(client, "Invalid scale setting \"%d\".\n",
+			   ch7006_scale);
+
+	return 0;
+}
+
+static struct i2c_device_id ch7006_ids[] = {
+	{ "ch7006", 0 },
+	{ }
+};
+MODULE_DEVICE_TABLE(i2c, ch7006_ids);
+
+static struct drm_i2c_encoder_driver ch7006_driver = {
+	.i2c_driver = {
+		.probe = ch7006_probe,
+		.remove = ch7006_remove,
+
+		.driver = {
+			.name = "ch7006",
+		},
+
+		.id_table = ch7006_ids,
+	},
+
+	.encoder_init = ch7006_encoder_init,
+};
+
+
+/* Module initialization */
+
+static int __init ch7006_init(void)
+{
+	return drm_i2c_encoder_register(THIS_MODULE, &ch7006_driver);
+}
+
+static void __exit ch7006_exit(void)
+{
+	drm_i2c_encoder_unregister(&ch7006_driver);
+}
+
+int ch7006_debug;
+module_param_named(debug, ch7006_debug, int, 0600);
+MODULE_PARM_DESC(debug, "Enable debug output.");
+
+char *ch7006_tv_norm;
+module_param_named(tv_norm, ch7006_tv_norm, charp, 0600);
+MODULE_PARM_DESC(tv_norm, "Default TV norm.\n"
+		 "\t\tSupported: PAL, PAL-M, PAL-N, PAL-Nc, PAL-60, NTSC-M, NTSC-J.\n"
+		 "\t\tDefault: PAL");
+
+int ch7006_scale = 1;
+module_param_named(scale, ch7006_scale, int, 0600);
+MODULE_PARM_DESC(scale, "Default scale.\n"
+		 "\t\tSupported: 0 -> Select video modes with a higher blanking ratio.\n"
+		 "\t\t\t1 -> Select default video modes.\n"
+		 "\t\t\t2 -> Select video modes with a lower blanking ratio.");
+
+MODULE_AUTHOR("Francisco Jerez <currojerez@riseup.net>");
+MODULE_DESCRIPTION("Chrontel ch7006 TV encoder driver");
+MODULE_LICENSE("GPL and additional rights");
+
+module_init(ch7006_init);
+module_exit(ch7006_exit);
diff --git a/drivers/gpu/drm/i2c/ch7006_mode.c b/drivers/gpu/drm/i2c/ch7006_mode.c
new file mode 100644
index 000000000000..e447dfb63890
--- /dev/null
+++ b/drivers/gpu/drm/i2c/ch7006_mode.c
@@ -0,0 +1,468 @@
+/*
+ * Copyright (C) 2009 Francisco Jerez.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "ch7006_priv.h"
+
+char *ch7006_tv_norm_names[] = {
+	[TV_NORM_PAL] = "PAL",
+	[TV_NORM_PAL_M] = "PAL-M",
+	[TV_NORM_PAL_N] = "PAL-N",
+	[TV_NORM_PAL_NC] = "PAL-Nc",
+	[TV_NORM_PAL_60] = "PAL-60",
+	[TV_NORM_NTSC_M] = "NTSC-M",
+	[TV_NORM_NTSC_J] = "NTSC-J",
+};
+
+#define NTSC_LIKE_TIMINGS .vrefresh = 60 * fixed1/1.001,		\
+		.vdisplay = 480,					\
+		.vtotal = 525,						\
+		.hvirtual = 660
+
+#define PAL_LIKE_TIMINGS .vrefresh = 50 * fixed1,		\
+		.vdisplay = 576,				\
+		.vtotal = 625,					\
+		.hvirtual = 810
+
+struct ch7006_tv_norm_info ch7006_tv_norms[] = {
+	[TV_NORM_NTSC_M] = {
+		NTSC_LIKE_TIMINGS,
+		.black_level = 0.339 * fixed1,
+		.subc_freq = 3579545 * fixed1,
+		.dispmode = bitfs(CH7006_DISPMODE_OUTPUT_STD, NTSC),
+		.voffset = 0,
+	},
+	[TV_NORM_NTSC_J] = {
+		NTSC_LIKE_TIMINGS,
+		.black_level = 0.286 * fixed1,
+		.subc_freq = 3579545 * fixed1,
+		.dispmode = bitfs(CH7006_DISPMODE_OUTPUT_STD, NTSC_J),
+		.voffset = 0,
+	},
+	[TV_NORM_PAL] = {
+		PAL_LIKE_TIMINGS,
+		.black_level = 0.3 * fixed1,
+		.subc_freq = 4433618.75 * fixed1,
+		.dispmode = bitfs(CH7006_DISPMODE_OUTPUT_STD, PAL),
+		.voffset = 0,
+	},
+	[TV_NORM_PAL_M] = {
+		NTSC_LIKE_TIMINGS,
+		.black_level = 0.339 * fixed1,
+		.subc_freq = 3575611.433 * fixed1,
+		.dispmode = bitfs(CH7006_DISPMODE_OUTPUT_STD, PAL_M),
+		.voffset = 16,
+	},
+
+	/* The following modes seem to work right but they're
+	 * undocumented */
+
+	[TV_NORM_PAL_N] = {
+		PAL_LIKE_TIMINGS,
+		.black_level = 0.339 * fixed1,
+		.subc_freq = 4433618.75 * fixed1,
+		.dispmode = bitfs(CH7006_DISPMODE_OUTPUT_STD, PAL),
+		.voffset = 0,
+	},
+	[TV_NORM_PAL_NC] = {
+		PAL_LIKE_TIMINGS,
+		.black_level = 0.3 * fixed1,
+		.subc_freq = 3582056.25 * fixed1,
+		.dispmode = bitfs(CH7006_DISPMODE_OUTPUT_STD, PAL),
+		.voffset = 0,
+	},
+	[TV_NORM_PAL_60] = {
+		NTSC_LIKE_TIMINGS,
+		.black_level = 0.3 * fixed1,
+		.subc_freq = 4433618.75 * fixed1,
+		.dispmode = bitfs(CH7006_DISPMODE_OUTPUT_STD, PAL_M),
+		.voffset = 16,
+	},
+};
+
+#define __MODE(f, hd, vd, ht, vt, hsynp, vsynp,				\
+	       subc, scale, scale_mask, norm_mask, e_hd, e_vd) {	\
+		.mode = {						\
+			.name = #hd "x" #vd,				\
+			.status = 0,					\
+			.type = DRM_MODE_TYPE_DRIVER,			\
+			.clock = f,					\
+			.hdisplay = hd,					\
+			.hsync_start = e_hd + 16,			\
+			.hsync_end = e_hd + 80,				\
+			.htotal = ht,					\
+			.hskew = 0,					\
+			.vdisplay = vd,					\
+			.vsync_start = vd + 10,				\
+			.vsync_end = vd + 26,				\
+			.vtotal = vt,					\
+			.vscan = 0,					\
+			.flags = DRM_MODE_FLAG_##hsynp##HSYNC |		\
+				DRM_MODE_FLAG_##vsynp##VSYNC,		\
+			.vrefresh = 0,					\
+		},							\
+		.enc_hdisp = e_hd,					\
+		.enc_vdisp = e_vd,					\
+		.subc_coeff = subc * fixed1,				\
+		.dispmode = bitfs(CH7006_DISPMODE_SCALING_RATIO, scale) | \
+			    bitfs(CH7006_DISPMODE_INPUT_RES, e_hd##x##e_vd), \
+		.valid_scales = scale_mask,				\
+		.valid_norms = norm_mask				\
+	 }
+
+#define MODE(f, hd, vd, ht, vt, hsynp, vsynp,				\
+	     subc, scale, scale_mask, norm_mask)			\
+	__MODE(f, hd, vd, ht, vt, hsynp, vsynp, subc, scale,		\
+	       scale_mask, norm_mask, hd, vd)
+
+#define NTSC_LIKE (1 << TV_NORM_NTSC_M | 1 << TV_NORM_NTSC_J |		\
+		   1 << TV_NORM_PAL_M | 1 << TV_NORM_PAL_60)
+
+#define PAL_LIKE (1 << TV_NORM_PAL | 1 << TV_NORM_PAL_N | 1 << TV_NORM_PAL_NC)
+
+struct ch7006_mode ch7006_modes[] = {
+	MODE(21000, 512, 384, 840, 500, N, N, 181.797557582, 5_4, 0x6, PAL_LIKE),
+	MODE(26250, 512, 384, 840, 625, N, N, 145.438046066, 1_1, 0x1, PAL_LIKE),
+	MODE(20140, 512, 384, 800, 420, N, N, 213.257083791, 5_4, 0x4, NTSC_LIKE),
+	MODE(24671, 512, 384, 784, 525, N, N, 174.0874153, 1_1, 0x3, NTSC_LIKE),
+	MODE(28125, 720, 400, 1125, 500, N, N, 135.742176298, 5_4, 0x6, PAL_LIKE),
+	MODE(34875, 720, 400, 1116, 625, N, N, 109.469496898, 1_1, 0x1, PAL_LIKE),
+	MODE(23790, 720, 400, 945, 420, N, N, 160.475642016, 5_4, 0x4, NTSC_LIKE),
+	MODE(29455, 720, 400, 936, 525, N, N, 129.614941843, 1_1, 0x3, NTSC_LIKE),
+	MODE(25000, 640, 400, 1000, 500, N, N, 152.709948279, 5_4, 0x6, PAL_LIKE),
+	MODE(31500, 640, 400, 1008, 625, N, N, 121.198371646, 1_1, 0x1, PAL_LIKE),
+	MODE(21147, 640, 400, 840, 420, N, N, 180.535097338, 5_4, 0x4, NTSC_LIKE),
+	MODE(26434, 640, 400, 840, 525, N, N, 144.42807787, 1_1, 0x2, NTSC_LIKE),
+	MODE(30210, 640, 400, 840, 600, N, N, 126.374568276, 7_8, 0x1, NTSC_LIKE),
+	MODE(21000, 640, 480, 840, 500, N, N, 181.797557582, 5_4, 0x4, PAL_LIKE),
+	MODE(26250, 640, 480, 840, 625, N, N, 145.438046066, 1_1, 0x2, PAL_LIKE),
+	MODE(31500, 640, 480, 840, 750, N, N, 121.198371646, 5_6, 0x1, PAL_LIKE),
+	MODE(24671, 640, 480, 784, 525, N, N, 174.0874153, 1_1, 0x4, NTSC_LIKE),
+	MODE(28196, 640, 480, 784, 600, N, N, 152.326488422, 7_8, 0x2, NTSC_LIKE),
+	MODE(30210, 640, 480, 800, 630, N, N, 142.171389101, 5_6, 0x1, NTSC_LIKE),
+	__MODE(29500, 720, 576, 944, 625, P, P, 145.592111636, 1_1, 0x7, PAL_LIKE, 800, 600),
+	MODE(36000, 800, 600, 960, 750, P, P, 119.304647022, 5_6, 0x6, PAL_LIKE),
+	MODE(39000, 800, 600, 936, 836, P, P, 110.127366499, 3_4, 0x1, PAL_LIKE),
+	MODE(39273, 800, 600, 1040, 630, P, P, 145.816809399, 5_6, 0x4, NTSC_LIKE),
+	MODE(43636, 800, 600, 1040, 700, P, P, 131.235128487, 3_4, 0x2, NTSC_LIKE),
+	MODE(47832, 800, 600, 1064, 750, P, P, 119.723275165, 7_10, 0x1, NTSC_LIKE),
+	{}
+};
+
+struct ch7006_mode *ch7006_lookup_mode(struct drm_encoder *encoder,
+				       struct drm_display_mode *drm_mode)
+{
+	struct ch7006_priv *priv = to_ch7006_priv(encoder);
+	struct ch7006_mode *mode;
+
+	for (mode = ch7006_modes; mode->mode.clock; mode++) {
+
+		if (~mode->valid_norms & 1<<priv->norm)
+			continue;
+
+		if (mode->mode.hdisplay != drm_mode->hdisplay ||
+		    mode->mode.vdisplay != drm_mode->vdisplay ||
+		    mode->mode.vtotal != drm_mode->vtotal ||
+		    mode->mode.htotal != drm_mode->htotal ||
+		    mode->mode.clock != drm_mode->clock)
+			continue;
+
+		return mode;
+	}
+
+	return NULL;
+}
+
+/* Some common HW state calculation code */
+
+void ch7006_setup_levels(struct drm_encoder *encoder)
+{
+	struct i2c_client *client = drm_i2c_encoder_get_client(encoder);
+	struct ch7006_priv *priv = to_ch7006_priv(encoder);
+	uint8_t *regs = priv->state.regs;
+	struct ch7006_tv_norm_info *norm = &ch7006_tv_norms[priv->norm];
+	int gain;
+	int black_level;
+
+	/* Set DAC_GAIN if the voltage drop between white and black is
+	 * high enough. */
+	if (norm->black_level < 339*fixed1/1000) {
+		gain = 76;
+
+		regs[CH7006_INPUT_FORMAT] |= CH7006_INPUT_FORMAT_DAC_GAIN;
+	} else {
+		gain = 71;
+
+		regs[CH7006_INPUT_FORMAT] &= ~CH7006_INPUT_FORMAT_DAC_GAIN;
+	}
+
+	black_level = round_fixed(norm->black_level*26625)/gain;
+
+	/* Correct it with the specified brightness. */
+	black_level = interpolate(90, black_level, 208, priv->brightness);
+
+	regs[CH7006_BLACK_LEVEL] = bitf(CH7006_BLACK_LEVEL_0, black_level);
+
+	ch7006_dbg(client, "black level: %d\n", black_level);
+}
+
+void ch7006_setup_subcarrier(struct drm_encoder *encoder)
+{
+	struct i2c_client *client = drm_i2c_encoder_get_client(encoder);
+	struct ch7006_priv *priv = to_ch7006_priv(encoder);
+	struct ch7006_state *state = &priv->state;
+	struct ch7006_tv_norm_info *norm = &ch7006_tv_norms[priv->norm];
+	struct ch7006_mode *mode = priv->mode;
+	uint32_t subc_inc;
+
+	subc_inc = round_fixed((mode->subc_coeff >> 8)
+			       * (norm->subc_freq >> 24));
+
+	setbitf(state, CH7006_SUBC_INC0, 28, subc_inc);
+	setbitf(state, CH7006_SUBC_INC1, 24, subc_inc);
+	setbitf(state, CH7006_SUBC_INC2, 20, subc_inc);
+	setbitf(state, CH7006_SUBC_INC3, 16, subc_inc);
+	setbitf(state, CH7006_SUBC_INC4, 12, subc_inc);
+	setbitf(state, CH7006_SUBC_INC5, 8, subc_inc);
+	setbitf(state, CH7006_SUBC_INC6, 4, subc_inc);
+	setbitf(state, CH7006_SUBC_INC7, 0, subc_inc);
+
+	ch7006_dbg(client, "subcarrier inc: %u\n", subc_inc);
+}
+
+void ch7006_setup_pll(struct drm_encoder *encoder)
+{
+	struct i2c_client *client = drm_i2c_encoder_get_client(encoder);
+	struct ch7006_priv *priv = to_ch7006_priv(encoder);
+	uint8_t *regs = priv->state.regs;
+	struct ch7006_mode *mode = priv->mode;
+	int n, best_n = 0;
+	int m, best_m = 0;
+	int freq, best_freq = 0;
+
+	for (n = 0; n < CH7006_MAXN; n++) {
+		for (m = 0; m < CH7006_MAXM; m++) {
+			freq = CH7006_FREQ0*(n+2)/(m+2);
+
+			if (abs(freq - mode->mode.clock) <
+			    abs(best_freq - mode->mode.clock)) {
+				best_freq = freq;
+				best_n = n;
+				best_m = m;
+			}
+		}
+	}
+
+	regs[CH7006_PLLOV] = bitf(CH7006_PLLOV_N_8, best_n) |
+		bitf(CH7006_PLLOV_M_8, best_m);
+
+	regs[CH7006_PLLM] = bitf(CH7006_PLLM_0, best_m);
+	regs[CH7006_PLLN] = bitf(CH7006_PLLN_0, best_n);
+
+	if (best_n < 108)
+		regs[CH7006_PLL_CONTROL] |= CH7006_PLL_CONTROL_CAPACITOR;
+	else
+		regs[CH7006_PLL_CONTROL] &= ~CH7006_PLL_CONTROL_CAPACITOR;
+
+	ch7006_dbg(client, "n=%d m=%d f=%d c=%d\n",
+		   best_n, best_m, best_freq, best_n < 108);
+}
+
+void ch7006_setup_power_state(struct drm_encoder *encoder)
+{
+	struct ch7006_priv *priv = to_ch7006_priv(encoder);
+	uint8_t *power = &priv->state.regs[CH7006_POWER];
+	int subconnector;
+
+	subconnector = priv->select_subconnector ? priv->select_subconnector :
+							priv->subconnector;
+
+	*power = CH7006_POWER_RESET;
+
+	if (priv->last_dpms == DRM_MODE_DPMS_ON) {
+		switch (subconnector) {
+		case DRM_MODE_SUBCONNECTOR_SVIDEO:
+			*power |= bitfs(CH7006_POWER_LEVEL, CVBS_OFF);
+			break;
+		case DRM_MODE_SUBCONNECTOR_Composite:
+			*power |= bitfs(CH7006_POWER_LEVEL, SVIDEO_OFF);
+			break;
+		case DRM_MODE_SUBCONNECTOR_SCART:
+			*power |= bitfs(CH7006_POWER_LEVEL, NORMAL) |
+				CH7006_POWER_SCART;
+			break;
+		}
+
+	} else {
+		*power |= bitfs(CH7006_POWER_LEVEL, FULL_POWER_OFF);
+	}
+}
+
+void ch7006_setup_properties(struct drm_encoder *encoder)
+{
+	struct i2c_client *client = drm_i2c_encoder_get_client(encoder);
+	struct ch7006_priv *priv = to_ch7006_priv(encoder);
+	struct ch7006_state *state = &priv->state;
+	struct ch7006_tv_norm_info *norm = &ch7006_tv_norms[priv->norm];
+	struct ch7006_mode *ch_mode = priv->mode;
+	struct drm_display_mode *mode = &ch_mode->mode;
+	uint8_t *regs = state->regs;
+	int flicker, contrast, hpos, vpos;
+	uint64_t scale, aspect;
+
+	flicker = interpolate(0, 2, 3, priv->flicker);
+	regs[CH7006_FFILTER] = bitf(CH7006_FFILTER_TEXT, flicker) |
+		bitf(CH7006_FFILTER_LUMA, flicker) |
+		bitf(CH7006_FFILTER_CHROMA, 1);
+
+	contrast = interpolate(0, 5, 7, priv->contrast);
+	regs[CH7006_CONTRAST] = bitf(CH7006_CONTRAST_0, contrast);
+
+	scale = norm->vtotal*fixed1;
+	do_div(scale, mode->vtotal);
+
+	aspect = ch_mode->enc_hdisp*fixed1;
+	do_div(aspect, ch_mode->enc_vdisp);
+
+	hpos = round_fixed((norm->hvirtual * aspect - mode->hdisplay * scale)
+			   * priv->hmargin * mode->vtotal) / norm->vtotal / 100 / 4;
+
+	setbitf(state, CH7006_POV, HPOS_8, hpos);
+	setbitf(state, CH7006_HPOS, 0, hpos);
+
+	vpos = max(0, norm->vdisplay - round_fixed(mode->vdisplay*scale)
+		   + norm->voffset) * priv->vmargin / 100 / 2;
+
+	setbitf(state, CH7006_POV, VPOS_8, vpos);
+	setbitf(state, CH7006_VPOS, 0, vpos);
+
+	ch7006_dbg(client, "hpos: %d, vpos: %d\n", hpos, vpos);
+}
+
+/* HW access functions */
+
+void ch7006_write(struct i2c_client *client, uint8_t addr, uint8_t val)
+{
+	uint8_t buf[] = {addr, val};
+	int ret;
+
+	ret = i2c_master_send(client, buf, ARRAY_SIZE(buf));
+	if (ret < 0)
+		ch7006_err(client, "Error %d writing to subaddress 0x%x\n",
+			   ret, addr);
+}
+
+uint8_t ch7006_read(struct i2c_client *client, uint8_t addr)
+{
+	uint8_t val;
+	int ret;
+
+	ret = i2c_master_send(client, &addr, sizeof(addr));
+	if (ret < 0)
+		goto fail;
+
+	ret = i2c_master_recv(client, &val, sizeof(val));
+	if (ret < 0)
+		goto fail;
+
+	return val;
+
+fail:
+	ch7006_err(client, "Error %d reading from subaddress 0x%x\n",
+		   ret, addr);
+	return 0;
+}
+
+void ch7006_state_load(struct i2c_client *client,
+		       struct ch7006_state *state)
+{
+	ch7006_load_reg(client, state, CH7006_POWER);
+
+	ch7006_load_reg(client, state, CH7006_DISPMODE);
+	ch7006_load_reg(client, state, CH7006_FFILTER);
+	ch7006_load_reg(client, state, CH7006_BWIDTH);
+	ch7006_load_reg(client, state, CH7006_INPUT_FORMAT);
+	ch7006_load_reg(client, state, CH7006_CLKMODE);
+	ch7006_load_reg(client, state, CH7006_START_ACTIVE);
+	ch7006_load_reg(client, state, CH7006_POV);
+	ch7006_load_reg(client, state, CH7006_BLACK_LEVEL);
+	ch7006_load_reg(client, state, CH7006_HPOS);
+	ch7006_load_reg(client, state, CH7006_VPOS);
+	ch7006_load_reg(client, state, CH7006_INPUT_SYNC);
+	ch7006_load_reg(client, state, CH7006_DETECT);
+	ch7006_load_reg(client, state, CH7006_CONTRAST);
+	ch7006_load_reg(client, state, CH7006_PLLOV);
+	ch7006_load_reg(client, state, CH7006_PLLM);
+	ch7006_load_reg(client, state, CH7006_PLLN);
+	ch7006_load_reg(client, state, CH7006_BCLKOUT);
+	ch7006_load_reg(client, state, CH7006_SUBC_INC0);
+	ch7006_load_reg(client, state, CH7006_SUBC_INC1);
+	ch7006_load_reg(client, state, CH7006_SUBC_INC2);
+	ch7006_load_reg(client, state, CH7006_SUBC_INC3);
+	ch7006_load_reg(client, state, CH7006_SUBC_INC4);
+	ch7006_load_reg(client, state, CH7006_SUBC_INC5);
+	ch7006_load_reg(client, state, CH7006_SUBC_INC6);
+	ch7006_load_reg(client, state, CH7006_SUBC_INC7);
+	ch7006_load_reg(client, state, CH7006_PLL_CONTROL);
+	ch7006_load_reg(client, state, CH7006_CALC_SUBC_INC0);
+}
+
+void ch7006_state_save(struct i2c_client *client,
+		       struct ch7006_state *state)
+{
+	ch7006_save_reg(client, state, CH7006_POWER);
+
+	ch7006_save_reg(client, state, CH7006_DISPMODE);
+	ch7006_save_reg(client, state, CH7006_FFILTER);
+	ch7006_save_reg(client, state, CH7006_BWIDTH);
+	ch7006_save_reg(client, state, CH7006_INPUT_FORMAT);
+	ch7006_save_reg(client, state, CH7006_CLKMODE);
+	ch7006_save_reg(client, state, CH7006_START_ACTIVE);
+	ch7006_save_reg(client, state, CH7006_POV);
+	ch7006_save_reg(client, state, CH7006_BLACK_LEVEL);
+	ch7006_save_reg(client, state, CH7006_HPOS);
+	ch7006_save_reg(client, state, CH7006_VPOS);
+	ch7006_save_reg(client, state, CH7006_INPUT_SYNC);
+	ch7006_save_reg(client, state, CH7006_DETECT);
+	ch7006_save_reg(client, state, CH7006_CONTRAST);
+	ch7006_save_reg(client, state, CH7006_PLLOV);
+	ch7006_save_reg(client, state, CH7006_PLLM);
+	ch7006_save_reg(client, state, CH7006_PLLN);
+	ch7006_save_reg(client, state, CH7006_BCLKOUT);
+	ch7006_save_reg(client, state, CH7006_SUBC_INC0);
+	ch7006_save_reg(client, state, CH7006_SUBC_INC1);
+	ch7006_save_reg(client, state, CH7006_SUBC_INC2);
+	ch7006_save_reg(client, state, CH7006_SUBC_INC3);
+	ch7006_save_reg(client, state, CH7006_SUBC_INC4);
+	ch7006_save_reg(client, state, CH7006_SUBC_INC5);
+	ch7006_save_reg(client, state, CH7006_SUBC_INC6);
+	ch7006_save_reg(client, state, CH7006_SUBC_INC7);
+	ch7006_save_reg(client, state, CH7006_PLL_CONTROL);
+	ch7006_save_reg(client, state, CH7006_CALC_SUBC_INC0);
+
+	state->regs[CH7006_FFILTER] = (state->regs[CH7006_FFILTER] & 0xf0) |
+		(state->regs[CH7006_FFILTER] & 0x0c) >> 2 |
+		(state->regs[CH7006_FFILTER] & 0x03) << 2;
+}
diff --git a/drivers/gpu/drm/i2c/ch7006_priv.h b/drivers/gpu/drm/i2c/ch7006_priv.h
new file mode 100644
index 000000000000..b06d3d93d8ac
--- /dev/null
+++ b/drivers/gpu/drm/i2c/ch7006_priv.h
@@ -0,0 +1,344 @@
+/*
+ * Copyright (C) 2009 Francisco Jerez.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __DRM_I2C_CH7006_PRIV_H__
+#define __DRM_I2C_CH7006_PRIV_H__
+
+#include "drmP.h"
+#include "drm_crtc_helper.h"
+#include "drm_encoder_slave.h"
+#include "i2c/ch7006.h"
+
+typedef int64_t fixed;
+#define fixed1 (1LL << 32)
+
+enum ch7006_tv_norm {
+	TV_NORM_PAL,
+	TV_NORM_PAL_M,
+	TV_NORM_PAL_N,
+	TV_NORM_PAL_NC,
+	TV_NORM_PAL_60,
+	TV_NORM_NTSC_M,
+	TV_NORM_NTSC_J,
+	NUM_TV_NORMS
+};
+
+struct ch7006_tv_norm_info {
+	fixed vrefresh;
+	int vdisplay;
+	int vtotal;
+	int hvirtual;
+
+	fixed subc_freq;
+	fixed black_level;
+
+	uint32_t dispmode;
+	int voffset;
+};
+
+struct ch7006_mode {
+	struct drm_display_mode mode;
+
+	int enc_hdisp;
+	int enc_vdisp;
+
+	fixed subc_coeff;
+	uint32_t dispmode;
+
+	uint32_t valid_scales;
+	uint32_t valid_norms;
+};
+
+struct ch7006_state {
+	uint8_t regs[0x26];
+};
+
+struct ch7006_priv {
+	struct ch7006_encoder_params *params;
+	struct ch7006_mode *mode;
+
+	struct ch7006_state state;
+	struct ch7006_state saved_state;
+
+	struct drm_property *scale_property;
+
+	int select_subconnector;
+	int subconnector;
+	int hmargin;
+	int vmargin;
+	enum ch7006_tv_norm norm;
+	int brightness;
+	int contrast;
+	int flicker;
+	int scale;
+
+	int last_dpms;
+};
+
+#define to_ch7006_priv(x) \
+	((struct ch7006_priv *)to_encoder_slave(x)->slave_priv)
+
+extern int ch7006_debug;
+extern char *ch7006_tv_norm;
+extern int ch7006_scale;
+
+extern char *ch7006_tv_norm_names[];
+extern struct ch7006_tv_norm_info ch7006_tv_norms[];
+extern struct ch7006_mode ch7006_modes[];
+
+struct ch7006_mode *ch7006_lookup_mode(struct drm_encoder *encoder,
+				       struct drm_display_mode *drm_mode);
+
+void ch7006_setup_levels(struct drm_encoder *encoder);
+void ch7006_setup_subcarrier(struct drm_encoder *encoder);
+void ch7006_setup_pll(struct drm_encoder *encoder);
+void ch7006_setup_power_state(struct drm_encoder *encoder);
+void ch7006_setup_properties(struct drm_encoder *encoder);
+
+void ch7006_write(struct i2c_client *client, uint8_t addr, uint8_t val);
+uint8_t ch7006_read(struct i2c_client *client, uint8_t addr);
+
+void ch7006_state_load(struct i2c_client *client,
+		       struct ch7006_state *state);
+void ch7006_state_save(struct i2c_client *client,
+		       struct ch7006_state *state);
+
+/* Some helper macros */
+
+#define ch7006_dbg(client, format, ...) do {				\
+		if (ch7006_debug)					\
+			dev_printk(KERN_DEBUG, &client->dev,		\
+				   "%s: " format, __func__, ## __VA_ARGS__); \
+	} while (0)
+#define ch7006_info(client, format, ...) \
+				dev_info(&client->dev, format, __VA_ARGS__)
+#define ch7006_err(client, format, ...) \
+				dev_err(&client->dev, format, __VA_ARGS__)
+
+#define __mask(src, bitfield) \
+		(((2 << (1 ? bitfield)) - 1) & ~((1 << (0 ? bitfield)) - 1))
+#define mask(bitfield) __mask(bitfield)
+
+#define __bitf(src, bitfield, x) \
+		(((x) >> (src) << (0 ? bitfield)) &  __mask(src, bitfield))
+#define bitf(bitfield, x) __bitf(bitfield, x)
+#define bitfs(bitfield, s) __bitf(bitfield, bitfield##_##s)
+#define setbitf(state, reg, bitfield, x)				\
+	state->regs[reg] = (state->regs[reg] & ~mask(reg##_##bitfield))	\
+		| bitf(reg##_##bitfield, x)
+
+#define __unbitf(src, bitfield, x) \
+		((x & __mask(src, bitfield)) >> (0 ? bitfield) << (src))
+#define unbitf(bitfield, x) __unbitf(bitfield, x)
+
+static inline int interpolate(int y0, int y1, int y2, int x)
+{
+	return y1 + (x < 50 ? y1 - y0 : y2 - y1) * (x - 50) / 50;
+}
+
+static inline int32_t round_fixed(fixed x)
+{
+	return (x + fixed1/2) >> 32;
+}
+
+#define ch7006_load_reg(client, state, reg) ch7006_write(client, reg, state->regs[reg])
+#define ch7006_save_reg(client, state, reg) state->regs[reg] = ch7006_read(client, reg)
+
+/* Fixed hardware specs */
+
+#define CH7006_FREQ0				14318
+#define CH7006_MAXN				650
+#define CH7006_MAXM				315
+
+/* Register definitions */
+
+#define CH7006_DISPMODE				0x00
+#define CH7006_DISPMODE_INPUT_RES		0, 7:5
+#define CH7006_DISPMODE_INPUT_RES_512x384	0x0
+#define CH7006_DISPMODE_INPUT_RES_720x400	0x1
+#define CH7006_DISPMODE_INPUT_RES_640x400	0x2
+#define CH7006_DISPMODE_INPUT_RES_640x480	0x3
+#define CH7006_DISPMODE_INPUT_RES_800x600	0x4
+#define CH7006_DISPMODE_INPUT_RES_NATIVE	0x5
+#define CH7006_DISPMODE_OUTPUT_STD		0, 4:3
+#define CH7006_DISPMODE_OUTPUT_STD_PAL		0x0
+#define CH7006_DISPMODE_OUTPUT_STD_NTSC		0x1
+#define CH7006_DISPMODE_OUTPUT_STD_PAL_M	0x2
+#define CH7006_DISPMODE_OUTPUT_STD_NTSC_J	0x3
+#define CH7006_DISPMODE_SCALING_RATIO		0, 2:0
+#define CH7006_DISPMODE_SCALING_RATIO_5_4	0x0
+#define CH7006_DISPMODE_SCALING_RATIO_1_1	0x1
+#define CH7006_DISPMODE_SCALING_RATIO_7_8	0x2
+#define CH7006_DISPMODE_SCALING_RATIO_5_6	0x3
+#define CH7006_DISPMODE_SCALING_RATIO_3_4	0x4
+#define CH7006_DISPMODE_SCALING_RATIO_7_10	0x5
+
+#define CH7006_FFILTER				0x01
+#define CH7006_FFILTER_TEXT			0, 5:4
+#define CH7006_FFILTER_LUMA			0, 3:2
+#define CH7006_FFILTER_CHROMA			0, 1:0
+#define CH7006_FFILTER_CHROMA_NO_DCRAWL		0x3
+
+#define CH7006_BWIDTH				0x03
+#define CH7006_BWIDTH_5L_FFILER			(1 << 7)
+#define CH7006_BWIDTH_CVBS_NO_CHROMA		(1 << 6)
+#define CH7006_BWIDTH_CHROMA			0, 5:4
+#define CH7006_BWIDTH_SVIDEO_YPEAK		(1 << 3)
+#define CH7006_BWIDTH_SVIDEO_LUMA		0, 2:1
+#define CH7006_BWIDTH_CVBS_LUMA			0, 0:0
+
+#define CH7006_INPUT_FORMAT			0x04
+#define CH7006_INPUT_FORMAT_DAC_GAIN		(1 << 6)
+#define CH7006_INPUT_FORMAT_RGB_PASS_THROUGH	(1 << 5)
+#define CH7006_INPUT_FORMAT_FORMAT		0, 3:0
+#define CH7006_INPUT_FORMAT_FORMAT_RGB16	0x0
+#define CH7006_INPUT_FORMAT_FORMAT_YCrCb24m16	0x1
+#define CH7006_INPUT_FORMAT_FORMAT_RGB24m16	0x2
+#define CH7006_INPUT_FORMAT_FORMAT_RGB15	0x3
+#define CH7006_INPUT_FORMAT_FORMAT_RGB24m12C	0x4
+#define CH7006_INPUT_FORMAT_FORMAT_RGB24m12I	0x5
+#define CH7006_INPUT_FORMAT_FORMAT_RGB24m8	0x6
+#define CH7006_INPUT_FORMAT_FORMAT_RGB16m8	0x7
+#define CH7006_INPUT_FORMAT_FORMAT_RGB15m8	0x8
+#define CH7006_INPUT_FORMAT_FORMAT_YCrCb24m8	0x9
+
+#define CH7006_CLKMODE				0x06
+#define CH7006_CLKMODE_SUBC_LOCK		(1 << 7)
+#define CH7006_CLKMODE_MASTER			(1 << 6)
+#define CH7006_CLKMODE_POS_EDGE			(1 << 4)
+#define CH7006_CLKMODE_XCM			0, 3:2
+#define CH7006_CLKMODE_PCM			0, 1:0
+
+#define CH7006_START_ACTIVE			0x07
+#define CH7006_START_ACTIVE_0			0, 7:0
+
+#define CH7006_POV				0x08
+#define CH7006_POV_START_ACTIVE_8		8, 2:2
+#define CH7006_POV_HPOS_8			8, 1:1
+#define CH7006_POV_VPOS_8			8, 0:0
+
+#define CH7006_BLACK_LEVEL			0x09
+#define CH7006_BLACK_LEVEL_0			0, 7:0
+
+#define CH7006_HPOS				0x0a
+#define CH7006_HPOS_0				0, 7:0
+
+#define CH7006_VPOS				0x0b
+#define CH7006_VPOS_0				0, 7:0
+
+#define CH7006_INPUT_SYNC			0x0d
+#define CH7006_INPUT_SYNC_EMBEDDED		(1 << 3)
+#define CH7006_INPUT_SYNC_OUTPUT		(1 << 2)
+#define CH7006_INPUT_SYNC_PVSYNC		(1 << 1)
+#define CH7006_INPUT_SYNC_PHSYNC		(1 << 0)
+
+#define CH7006_POWER				0x0e
+#define CH7006_POWER_SCART			(1 << 4)
+#define CH7006_POWER_RESET			(1 << 3)
+#define CH7006_POWER_LEVEL			0, 2:0
+#define CH7006_POWER_LEVEL_CVBS_OFF		0x0
+#define CH7006_POWER_LEVEL_POWER_OFF		0x1
+#define CH7006_POWER_LEVEL_SVIDEO_OFF		0x2
+#define CH7006_POWER_LEVEL_NORMAL		0x3
+#define CH7006_POWER_LEVEL_FULL_POWER_OFF	0x4
+
+#define CH7006_DETECT				0x10
+#define CH7006_DETECT_SVIDEO_Y_TEST		(1 << 3)
+#define CH7006_DETECT_SVIDEO_C_TEST		(1 << 2)
+#define CH7006_DETECT_CVBS_TEST			(1 << 1)
+#define CH7006_DETECT_SENSE			(1 << 0)
+
+#define CH7006_CONTRAST				0x11
+#define CH7006_CONTRAST_0			0, 2:0
+
+#define CH7006_PLLOV	 			0x13
+#define CH7006_PLLOV_N_8	 		8, 2:1
+#define CH7006_PLLOV_M_8	 		8, 0:0
+
+#define CH7006_PLLM	 			0x14
+#define CH7006_PLLM_0	 			0, 7:0
+
+#define CH7006_PLLN	 			0x15
+#define CH7006_PLLN_0	 			0, 7:0
+
+#define CH7006_BCLKOUT	 			0x17
+
+#define CH7006_SUBC_INC0			0x18
+#define CH7006_SUBC_INC0_28			28, 3:0
+
+#define CH7006_SUBC_INC1			0x19
+#define CH7006_SUBC_INC1_24			24, 3:0
+
+#define CH7006_SUBC_INC2			0x1a
+#define CH7006_SUBC_INC2_20			20, 3:0
+
+#define CH7006_SUBC_INC3			0x1b
+#define CH7006_SUBC_INC3_GPIO1_VAL		(1 << 7)
+#define CH7006_SUBC_INC3_GPIO0_VAL		(1 << 6)
+#define CH7006_SUBC_INC3_POUT_3_3V		(1 << 5)
+#define CH7006_SUBC_INC3_POUT_INV		(1 << 4)
+#define CH7006_SUBC_INC3_16			16, 3:0
+
+#define CH7006_SUBC_INC4			0x1c
+#define CH7006_SUBC_INC4_GPIO1_IN		(1 << 7)
+#define CH7006_SUBC_INC4_GPIO0_IN		(1 << 6)
+#define CH7006_SUBC_INC4_DS_INPUT		(1 << 4)
+#define CH7006_SUBC_INC4_12			12, 3:0
+
+#define CH7006_SUBC_INC5			0x1d
+#define CH7006_SUBC_INC5_8			8, 3:0
+
+#define CH7006_SUBC_INC6			0x1e
+#define CH7006_SUBC_INC6_4			4, 3:0
+
+#define CH7006_SUBC_INC7			0x1f
+#define CH7006_SUBC_INC7_0			0, 3:0
+
+#define CH7006_PLL_CONTROL			0x20
+#define CH7006_PLL_CONTROL_CPI			(1 << 5)
+#define CH7006_PLL_CONTROL_CAPACITOR		(1 << 4)
+#define CH7006_PLL_CONTROL_7STAGES		(1 << 3)
+#define CH7006_PLL_CONTROL_DIGITAL_5V		(1 << 2)
+#define CH7006_PLL_CONTROL_ANALOG_5V		(1 << 1)
+#define CH7006_PLL_CONTROL_MEMORY_5V		(1 << 0)
+
+#define CH7006_CALC_SUBC_INC0			0x21
+#define CH7006_CALC_SUBC_INC0_24		24, 4:3
+#define CH7006_CALC_SUBC_INC0_HYST		0, 2:1
+#define CH7006_CALC_SUBC_INC0_AUTO		(1 << 0)
+
+#define CH7006_CALC_SUBC_INC1			0x22
+#define CH7006_CALC_SUBC_INC1_16		16, 7:0
+
+#define CH7006_CALC_SUBC_INC2			0x23
+#define CH7006_CALC_SUBC_INC2_8			8, 7:0
+
+#define CH7006_CALC_SUBC_INC3			0x24
+#define CH7006_CALC_SUBC_INC3_0			0, 7:0
+
+#define CH7006_VERSION_ID			0x25
+
+#endif
diff --git a/drivers/gpu/drm/i810/i810_dma.c b/drivers/gpu/drm/i810/i810_dma.c
index 7d1d88cdf2dc..de32d22a8c39 100644
--- a/drivers/gpu/drm/i810/i810_dma.c
+++ b/drivers/gpu/drm/i810/i810_dma.c
@@ -115,7 +115,7 @@ static int i810_mmap_buffers(struct file *filp, struct vm_area_struct *vma)
 static const struct file_operations i810_buffer_fops = {
 	.open = drm_open,
 	.release = drm_release,
-	.ioctl = drm_ioctl,
+	.unlocked_ioctl = drm_ioctl,
 	.mmap = i810_mmap_buffers,
 	.fasync = drm_fasync,
 };
diff --git a/drivers/gpu/drm/i810/i810_drv.c b/drivers/gpu/drm/i810/i810_drv.c
index fabb9a817966..c1e02752e023 100644
--- a/drivers/gpu/drm/i810/i810_drv.c
+++ b/drivers/gpu/drm/i810/i810_drv.c
@@ -59,7 +59,7 @@ static struct drm_driver driver = {
 		 .owner = THIS_MODULE,
 		 .open = drm_open,
 		 .release = drm_release,
-		 .ioctl = drm_ioctl,
+		 .unlocked_ioctl = drm_ioctl,
 		 .mmap = drm_mmap,
 		 .poll = drm_poll,
 		 .fasync = drm_fasync,
diff --git a/drivers/gpu/drm/i830/i830_dma.c b/drivers/gpu/drm/i830/i830_dma.c
index 877bf6cb14a4..06bd732e6463 100644
--- a/drivers/gpu/drm/i830/i830_dma.c
+++ b/drivers/gpu/drm/i830/i830_dma.c
@@ -117,7 +117,7 @@ static int i830_mmap_buffers(struct file *filp, struct vm_area_struct *vma)
 static const struct file_operations i830_buffer_fops = {
 	.open = drm_open,
 	.release = drm_release,
-	.ioctl = drm_ioctl,
+	.unlocked_ioctl = drm_ioctl,
 	.mmap = i830_mmap_buffers,
 	.fasync = drm_fasync,
 };
diff --git a/drivers/gpu/drm/i830/i830_drv.c b/drivers/gpu/drm/i830/i830_drv.c
index 389597e4a623..44f990bed8f4 100644
--- a/drivers/gpu/drm/i830/i830_drv.c
+++ b/drivers/gpu/drm/i830/i830_drv.c
@@ -70,7 +70,7 @@ static struct drm_driver driver = {
 		 .owner = THIS_MODULE,
 		 .open = drm_open,
 		 .release = drm_release,
-		 .ioctl = drm_ioctl,
+		 .unlocked_ioctl = drm_ioctl,
 		 .mmap = drm_mmap,
 		 .poll = drm_poll,
 		 .fasync = drm_fasync,
diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
index fa7b9be096bc..9929f84ec3e1 100644
--- a/drivers/gpu/drm/i915/Makefile
+++ b/drivers/gpu/drm/i915/Makefile
@@ -15,7 +15,6 @@ i915-y := i915_drv.o i915_dma.o i915_irq.o i915_mem.o \
 	  intel_lvds.o \
 	  intel_bios.o \
 	  intel_dp.o \
-	  intel_dp_i2c.o \
 	  intel_hdmi.o \
 	  intel_sdvo.o \
 	  intel_modes.o \
@@ -23,6 +22,7 @@ i915-y := i915_drv.o i915_dma.o i915_irq.o i915_mem.o \
 	  intel_fb.o \
 	  intel_tv.o \
 	  intel_dvo.o \
+	  intel_overlay.o \
 	  dvo_ch7xxx.o \
 	  dvo_ch7017.o \
 	  dvo_ivch.o \
diff --git a/drivers/gpu/drm/i915/dvo_ch7017.c b/drivers/gpu/drm/i915/dvo_ch7017.c
index 621815b531db..1184c14ba87d 100644
--- a/drivers/gpu/drm/i915/dvo_ch7017.c
+++ b/drivers/gpu/drm/i915/dvo_ch7017.c
@@ -249,7 +249,8 @@ static bool ch7017_init(struct intel_dvo_device *dvo,
 	if (val != CH7017_DEVICE_ID_VALUE &&
 	    val != CH7018_DEVICE_ID_VALUE &&
 	    val != CH7019_DEVICE_ID_VALUE) {
-		DRM_DEBUG("ch701x not detected, got %d: from %s Slave %d.\n",
+		DRM_DEBUG_KMS("ch701x not detected, got %d: from %s "
+				"Slave %d.\n",
 			  val, i2cbus->adapter.name,dvo->slave_addr);
 		goto fail;
 	}
@@ -284,7 +285,7 @@ static void ch7017_mode_set(struct intel_dvo_device *dvo,
 	uint8_t horizontal_active_pixel_output, vertical_active_line_output;
 	uint8_t active_input_line_output;
 
-	DRM_DEBUG("Registers before mode setting\n");
+	DRM_DEBUG_KMS("Registers before mode setting\n");
 	ch7017_dump_regs(dvo);
 
 	/* LVDS PLL settings from page 75 of 7017-7017ds.pdf*/
@@ -346,7 +347,7 @@ static void ch7017_mode_set(struct intel_dvo_device *dvo,
 	/* Turn the LVDS back on with new settings. */
 	ch7017_write(dvo, CH7017_LVDS_POWER_DOWN, lvds_power_down);
 
-	DRM_DEBUG("Registers after mode setting\n");
+	DRM_DEBUG_KMS("Registers after mode setting\n");
 	ch7017_dump_regs(dvo);
 }
 
@@ -386,7 +387,7 @@ static void ch7017_dump_regs(struct intel_dvo_device *dvo)
 #define DUMP(reg)					\
 do {							\
 	ch7017_read(dvo, reg, &val);			\
-	DRM_DEBUG(#reg ": %02x\n", val);		\
+	DRM_DEBUG_KMS(#reg ": %02x\n", val);		\
 } while (0)
 
 	DUMP(CH7017_HORIZONTAL_ACTIVE_PIXEL_INPUT);
diff --git a/drivers/gpu/drm/i915/dvo_ch7xxx.c b/drivers/gpu/drm/i915/dvo_ch7xxx.c
index a9b896289680..d56ff5cc22b2 100644
--- a/drivers/gpu/drm/i915/dvo_ch7xxx.c
+++ b/drivers/gpu/drm/i915/dvo_ch7xxx.c
@@ -152,7 +152,7 @@ static bool ch7xxx_readb(struct intel_dvo_device *dvo, int addr, uint8_t *ch)
 	};
 
 	if (!ch7xxx->quiet) {
-		DRM_DEBUG("Unable to read register 0x%02x from %s:%02x.\n",
+		DRM_DEBUG_KMS("Unable to read register 0x%02x from %s:%02x.\n",
 			  addr, i2cbus->adapter.name, dvo->slave_addr);
 	}
 	return false;
@@ -179,7 +179,7 @@ static bool ch7xxx_writeb(struct intel_dvo_device *dvo, int addr, uint8_t ch)
 		return true;
 
 	if (!ch7xxx->quiet) {
-		DRM_DEBUG("Unable to write register 0x%02x to %s:%d.\n",
+		DRM_DEBUG_KMS("Unable to write register 0x%02x to %s:%d.\n",
 			  addr, i2cbus->adapter.name, dvo->slave_addr);
 	}
 
@@ -207,7 +207,8 @@ static bool ch7xxx_init(struct intel_dvo_device *dvo,
 
 	name = ch7xxx_get_id(vendor);
 	if (!name) {
-		DRM_DEBUG("ch7xxx not detected; got 0x%02x from %s slave %d.\n",
+		DRM_DEBUG_KMS("ch7xxx not detected; got 0x%02x from %s "
+				"slave %d.\n",
 			  vendor, adapter->name, dvo->slave_addr);
 		goto out;
 	}
@@ -217,13 +218,14 @@ static bool ch7xxx_init(struct intel_dvo_device *dvo,
 		goto out;
 
 	if (device != CH7xxx_DID) {
-		DRM_DEBUG("ch7xxx not detected; got 0x%02x from %s slave %d.\n",
+		DRM_DEBUG_KMS("ch7xxx not detected; got 0x%02x from %s "
+				"slave %d.\n",
 			  vendor, adapter->name, dvo->slave_addr);
 		goto out;
 	}
 
 	ch7xxx->quiet = false;
-	DRM_DEBUG("Detected %s chipset, vendor/device ID 0x%02x/0x%02x\n",
+	DRM_DEBUG_KMS("Detected %s chipset, vendor/device ID 0x%02x/0x%02x\n",
 		  name, vendor, device);
 	return true;
 out:
@@ -315,8 +317,8 @@ static void ch7xxx_dump_regs(struct intel_dvo_device *dvo)
 
 	for (i = 0; i < CH7xxx_NUM_REGS; i++) {
 		if ((i % 8) == 0 )
-			DRM_DEBUG("\n %02X: ", i);
-		DRM_DEBUG("%02X ", ch7xxx->mode_reg.regs[i]);
+			DRM_LOG_KMS("\n %02X: ", i);
+		DRM_LOG_KMS("%02X ", ch7xxx->mode_reg.regs[i]);
 	}
 }
 
diff --git a/drivers/gpu/drm/i915/dvo_ivch.c b/drivers/gpu/drm/i915/dvo_ivch.c
index aa176f9921fe..24169e528f0f 100644
--- a/drivers/gpu/drm/i915/dvo_ivch.c
+++ b/drivers/gpu/drm/i915/dvo_ivch.c
@@ -202,7 +202,8 @@ static bool ivch_read(struct intel_dvo_device *dvo, int addr, uint16_t *data)
 	};
 
 	if (!priv->quiet) {
-		DRM_DEBUG("Unable to read register 0x%02x from %s:%02x.\n",
+		DRM_DEBUG_KMS("Unable to read register 0x%02x from "
+				"%s:%02x.\n",
 			  addr, i2cbus->adapter.name, dvo->slave_addr);
 	}
 	return false;
@@ -230,7 +231,7 @@ static bool ivch_write(struct intel_dvo_device *dvo, int addr, uint16_t data)
 		return true;
 
 	if (!priv->quiet) {
-		DRM_DEBUG("Unable to write register 0x%02x to %s:%d.\n",
+		DRM_DEBUG_KMS("Unable to write register 0x%02x to %s:%d.\n",
 			  addr, i2cbus->adapter.name, dvo->slave_addr);
 	}
 
@@ -261,7 +262,7 @@ static bool ivch_init(struct intel_dvo_device *dvo,
 	 * the address it's responding on.
 	 */
 	if ((temp & VR00_BASE_ADDRESS_MASK) != dvo->slave_addr) {
-		DRM_DEBUG("ivch detect failed due to address mismatch "
+		DRM_DEBUG_KMS("ivch detect failed due to address mismatch "
 			  "(%d vs %d)\n",
 			  (temp & VR00_BASE_ADDRESS_MASK), dvo->slave_addr);
 		goto out;
@@ -367,41 +368,41 @@ static void ivch_dump_regs(struct intel_dvo_device *dvo)
 	uint16_t val;
 
 	ivch_read(dvo, VR00, &val);
-	DRM_DEBUG("VR00: 0x%04x\n", val);
+	DRM_LOG_KMS("VR00: 0x%04x\n", val);
 	ivch_read(dvo, VR01, &val);
-	DRM_DEBUG("VR01: 0x%04x\n", val);
+	DRM_LOG_KMS("VR01: 0x%04x\n", val);
 	ivch_read(dvo, VR30, &val);
-	DRM_DEBUG("VR30: 0x%04x\n", val);
+	DRM_LOG_KMS("VR30: 0x%04x\n", val);
 	ivch_read(dvo, VR40, &val);
-	DRM_DEBUG("VR40: 0x%04x\n", val);
+	DRM_LOG_KMS("VR40: 0x%04x\n", val);
 
 	/* GPIO registers */
 	ivch_read(dvo, VR80, &val);
-	DRM_DEBUG("VR80: 0x%04x\n", val);
+	DRM_LOG_KMS("VR80: 0x%04x\n", val);
 	ivch_read(dvo, VR81, &val);
-	DRM_DEBUG("VR81: 0x%04x\n", val);
+	DRM_LOG_KMS("VR81: 0x%04x\n", val);
 	ivch_read(dvo, VR82, &val);
-	DRM_DEBUG("VR82: 0x%04x\n", val);
+	DRM_LOG_KMS("VR82: 0x%04x\n", val);
 	ivch_read(dvo, VR83, &val);
-	DRM_DEBUG("VR83: 0x%04x\n", val);
+	DRM_LOG_KMS("VR83: 0x%04x\n", val);
 	ivch_read(dvo, VR84, &val);
-	DRM_DEBUG("VR84: 0x%04x\n", val);
+	DRM_LOG_KMS("VR84: 0x%04x\n", val);
 	ivch_read(dvo, VR85, &val);
-	DRM_DEBUG("VR85: 0x%04x\n", val);
+	DRM_LOG_KMS("VR85: 0x%04x\n", val);
 	ivch_read(dvo, VR86, &val);
-	DRM_DEBUG("VR86: 0x%04x\n", val);
+	DRM_LOG_KMS("VR86: 0x%04x\n", val);
 	ivch_read(dvo, VR87, &val);
-	DRM_DEBUG("VR87: 0x%04x\n", val);
+	DRM_LOG_KMS("VR87: 0x%04x\n", val);
 	ivch_read(dvo, VR88, &val);
-	DRM_DEBUG("VR88: 0x%04x\n", val);
+	DRM_LOG_KMS("VR88: 0x%04x\n", val);
 
 	/* Scratch register 0 - AIM Panel type */
 	ivch_read(dvo, VR8E, &val);
-	DRM_DEBUG("VR8E: 0x%04x\n", val);
+	DRM_LOG_KMS("VR8E: 0x%04x\n", val);
 
 	/* Scratch register 1 - Status register */
 	ivch_read(dvo, VR8F, &val);
-	DRM_DEBUG("VR8F: 0x%04x\n", val);
+	DRM_LOG_KMS("VR8F: 0x%04x\n", val);
 }
 
 static void ivch_save(struct intel_dvo_device *dvo)
diff --git a/drivers/gpu/drm/i915/dvo_sil164.c b/drivers/gpu/drm/i915/dvo_sil164.c
index e1c1f7341e5c..0001c13f0a80 100644
--- a/drivers/gpu/drm/i915/dvo_sil164.c
+++ b/drivers/gpu/drm/i915/dvo_sil164.c
@@ -105,7 +105,7 @@ static bool sil164_readb(struct intel_dvo_device *dvo, int addr, uint8_t *ch)
 	};
 
 	if (!sil->quiet) {
-		DRM_DEBUG("Unable to read register 0x%02x from %s:%02x.\n",
+		DRM_DEBUG_KMS("Unable to read register 0x%02x from %s:%02x.\n",
 			  addr, i2cbus->adapter.name, dvo->slave_addr);
 	}
 	return false;
@@ -131,7 +131,7 @@ static bool sil164_writeb(struct intel_dvo_device *dvo, int addr, uint8_t ch)
 		return true;
 
 	if (!sil->quiet) {
-		DRM_DEBUG("Unable to write register 0x%02x to %s:%d.\n",
+		DRM_DEBUG_KMS("Unable to write register 0x%02x to %s:%d.\n",
 			  addr, i2cbus->adapter.name, dvo->slave_addr);
 	}
 
@@ -158,7 +158,7 @@ static bool sil164_init(struct intel_dvo_device *dvo,
 		goto out;
 
 	if (ch != (SIL164_VID & 0xff)) {
-		DRM_DEBUG("sil164 not detected got %d: from %s Slave %d.\n",
+		DRM_DEBUG_KMS("sil164 not detected got %d: from %s Slave %d.\n",
 			  ch, adapter->name, dvo->slave_addr);
 		goto out;
 	}
@@ -167,13 +167,13 @@ static bool sil164_init(struct intel_dvo_device *dvo,
 		goto out;
 
 	if (ch != (SIL164_DID & 0xff)) {
-		DRM_DEBUG("sil164 not detected got %d: from %s Slave %d.\n",
+		DRM_DEBUG_KMS("sil164 not detected got %d: from %s Slave %d.\n",
 			  ch, adapter->name, dvo->slave_addr);
 		goto out;
 	}
 	sil->quiet = false;
 
-	DRM_DEBUG("init sil164 dvo controller successfully!\n");
+	DRM_DEBUG_KMS("init sil164 dvo controller successfully!\n");
 	return true;
 
 out:
@@ -241,15 +241,15 @@ static void sil164_dump_regs(struct intel_dvo_device *dvo)
 	uint8_t val;
 
 	sil164_readb(dvo, SIL164_FREQ_LO, &val);
-	DRM_DEBUG("SIL164_FREQ_LO: 0x%02x\n", val);
+	DRM_LOG_KMS("SIL164_FREQ_LO: 0x%02x\n", val);
 	sil164_readb(dvo, SIL164_FREQ_HI, &val);
-	DRM_DEBUG("SIL164_FREQ_HI: 0x%02x\n", val);
+	DRM_LOG_KMS("SIL164_FREQ_HI: 0x%02x\n", val);
 	sil164_readb(dvo, SIL164_REG8, &val);
-	DRM_DEBUG("SIL164_REG8: 0x%02x\n", val);
+	DRM_LOG_KMS("SIL164_REG8: 0x%02x\n", val);
 	sil164_readb(dvo, SIL164_REG9, &val);
-	DRM_DEBUG("SIL164_REG9: 0x%02x\n", val);
+	DRM_LOG_KMS("SIL164_REG9: 0x%02x\n", val);
 	sil164_readb(dvo, SIL164_REGC, &val);
-	DRM_DEBUG("SIL164_REGC: 0x%02x\n", val);
+	DRM_LOG_KMS("SIL164_REGC: 0x%02x\n", val);
 }
 
 static void sil164_save(struct intel_dvo_device *dvo)
diff --git a/drivers/gpu/drm/i915/dvo_tfp410.c b/drivers/gpu/drm/i915/dvo_tfp410.c
index 9ecc907384ec..c7c391bc116a 100644
--- a/drivers/gpu/drm/i915/dvo_tfp410.c
+++ b/drivers/gpu/drm/i915/dvo_tfp410.c
@@ -130,7 +130,7 @@ static bool tfp410_readb(struct intel_dvo_device *dvo, int addr, uint8_t *ch)
 	};
 
 	if (!tfp->quiet) {
-		DRM_DEBUG("Unable to read register 0x%02x from %s:%02x.\n",
+		DRM_DEBUG_KMS("Unable to read register 0x%02x from %s:%02x.\n",
 			  addr, i2cbus->adapter.name, dvo->slave_addr);
 	}
 	return false;
@@ -156,7 +156,7 @@ static bool tfp410_writeb(struct intel_dvo_device *dvo, int addr, uint8_t ch)
 		return true;
 
 	if (!tfp->quiet) {
-		DRM_DEBUG("Unable to write register 0x%02x to %s:%d.\n",
+		DRM_DEBUG_KMS("Unable to write register 0x%02x to %s:%d.\n",
 			  addr, i2cbus->adapter.name, dvo->slave_addr);
 	}
 
@@ -191,13 +191,15 @@ static bool tfp410_init(struct intel_dvo_device *dvo,
 	tfp->quiet = true;
 
 	if ((id = tfp410_getid(dvo, TFP410_VID_LO)) != TFP410_VID) {
-		DRM_DEBUG("tfp410 not detected got VID %X: from %s Slave %d.\n",
+		DRM_DEBUG_KMS("tfp410 not detected got VID %X: from %s "
+				"Slave %d.\n",
 			  id, adapter->name, dvo->slave_addr);
 		goto out;
 	}
 
 	if ((id = tfp410_getid(dvo, TFP410_DID_LO)) != TFP410_DID) {
-		DRM_DEBUG("tfp410 not detected got DID %X: from %s Slave %d.\n",
+		DRM_DEBUG_KMS("tfp410 not detected got DID %X: from %s "
+				"Slave %d.\n",
 			  id, adapter->name, dvo->slave_addr);
 		goto out;
 	}
@@ -262,33 +264,33 @@ static void tfp410_dump_regs(struct intel_dvo_device *dvo)
 	uint8_t val, val2;
 
 	tfp410_readb(dvo, TFP410_REV, &val);
-	DRM_DEBUG("TFP410_REV: 0x%02X\n", val);
+	DRM_LOG_KMS("TFP410_REV: 0x%02X\n", val);
 	tfp410_readb(dvo, TFP410_CTL_1, &val);
-	DRM_DEBUG("TFP410_CTL1: 0x%02X\n", val);
+	DRM_LOG_KMS("TFP410_CTL1: 0x%02X\n", val);
 	tfp410_readb(dvo, TFP410_CTL_2, &val);
-	DRM_DEBUG("TFP410_CTL2: 0x%02X\n", val);
+	DRM_LOG_KMS("TFP410_CTL2: 0x%02X\n", val);
 	tfp410_readb(dvo, TFP410_CTL_3, &val);
-	DRM_DEBUG("TFP410_CTL3: 0x%02X\n", val);
+	DRM_LOG_KMS("TFP410_CTL3: 0x%02X\n", val);
 	tfp410_readb(dvo, TFP410_USERCFG, &val);
-	DRM_DEBUG("TFP410_USERCFG: 0x%02X\n", val);
+	DRM_LOG_KMS("TFP410_USERCFG: 0x%02X\n", val);
 	tfp410_readb(dvo, TFP410_DE_DLY, &val);
-	DRM_DEBUG("TFP410_DE_DLY: 0x%02X\n", val);
+	DRM_LOG_KMS("TFP410_DE_DLY: 0x%02X\n", val);
 	tfp410_readb(dvo, TFP410_DE_CTL, &val);
-	DRM_DEBUG("TFP410_DE_CTL: 0x%02X\n", val);
+	DRM_LOG_KMS("TFP410_DE_CTL: 0x%02X\n", val);
 	tfp410_readb(dvo, TFP410_DE_TOP, &val);
-	DRM_DEBUG("TFP410_DE_TOP: 0x%02X\n", val);
+	DRM_LOG_KMS("TFP410_DE_TOP: 0x%02X\n", val);
 	tfp410_readb(dvo, TFP410_DE_CNT_LO, &val);
 	tfp410_readb(dvo, TFP410_DE_CNT_HI, &val2);
-	DRM_DEBUG("TFP410_DE_CNT: 0x%02X%02X\n", val2, val);
+	DRM_LOG_KMS("TFP410_DE_CNT: 0x%02X%02X\n", val2, val);
 	tfp410_readb(dvo, TFP410_DE_LIN_LO, &val);
 	tfp410_readb(dvo, TFP410_DE_LIN_HI, &val2);
-	DRM_DEBUG("TFP410_DE_LIN: 0x%02X%02X\n", val2, val);
+	DRM_LOG_KMS("TFP410_DE_LIN: 0x%02X%02X\n", val2, val);
 	tfp410_readb(dvo, TFP410_H_RES_LO, &val);
 	tfp410_readb(dvo, TFP410_H_RES_HI, &val2);
-	DRM_DEBUG("TFP410_H_RES: 0x%02X%02X\n", val2, val);
+	DRM_LOG_KMS("TFP410_H_RES: 0x%02X%02X\n", val2, val);
 	tfp410_readb(dvo, TFP410_V_RES_LO, &val);
 	tfp410_readb(dvo, TFP410_V_RES_HI, &val2);
-	DRM_DEBUG("TFP410_V_RES: 0x%02X%02X\n", val2, val);
+	DRM_LOG_KMS("TFP410_V_RES: 0x%02X%02X\n", val2, val);
 }
 
 static void tfp410_save(struct intel_dvo_device *dvo)
diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index f8ce9a3a420d..1376dfe44c95 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -27,6 +27,7 @@
  */
 
 #include <linux/seq_file.h>
+#include <linux/debugfs.h>
 #include "drmP.h"
 #include "drm.h"
 #include "i915_drm.h"
@@ -96,13 +97,14 @@ static int i915_gem_object_list_info(struct seq_file *m, void *data)
 	{
 		struct drm_gem_object *obj = obj_priv->obj;
 
-		seq_printf(m, "    %p: %s %8zd %08x %08x %d %s",
+		seq_printf(m, "    %p: %s %8zd %08x %08x %d%s%s",
 			   obj,
 			   get_pin_flag(obj_priv),
 			   obj->size,
 			   obj->read_domains, obj->write_domain,
 			   obj_priv->last_rendering_seqno,
-			   obj_priv->dirty ? "dirty" : "");
+			   obj_priv->dirty ? " dirty" : "",
+			   obj_priv->madv == I915_MADV_DONTNEED ? " purgeable" : "");
 
 		if (obj->name)
 			seq_printf(m, " (name: %d)", obj->name);
@@ -160,7 +162,7 @@ static int i915_interrupt_info(struct seq_file *m, void *data)
 	struct drm_device *dev = node->minor->dev;
 	drm_i915_private_t *dev_priv = dev->dev_private;
 
-	if (!IS_IGDNG(dev)) {
+	if (!HAS_PCH_SPLIT(dev)) {
 		seq_printf(m, "Interrupt enable:    %08x\n",
 			   I915_READ(IER));
 		seq_printf(m, "Interrupt identity:  %08x\n",
@@ -267,10 +269,10 @@ static void i915_dump_pages(struct seq_file *m, struct page **pages, int page_co
 	uint32_t *mem;
 
 	for (page = 0; page < page_count; page++) {
-		mem = kmap(pages[page]);
+		mem = kmap_atomic(pages[page], KM_USER0);
 		for (i = 0; i < PAGE_SIZE; i += 4)
 			seq_printf(m, "%08x :  %08x\n", i, mem[i / 4]);
-		kunmap(pages[page]);
+		kunmap_atomic(mem, KM_USER0);
 	}
 }
 
@@ -288,7 +290,7 @@ static int i915_batchbuffer_info(struct seq_file *m, void *data)
 	list_for_each_entry(obj_priv, &dev_priv->mm.active_list, list) {
 		obj = obj_priv->obj;
 		if (obj->read_domains & I915_GEM_DOMAIN_COMMAND) {
-		    ret = i915_gem_object_get_pages(obj);
+		    ret = i915_gem_object_get_pages(obj, 0);
 		    if (ret) {
 			    DRM_ERROR("Failed to get pages: %d\n", ret);
 			    spin_unlock(&dev_priv->mm.active_list_lock);
@@ -348,6 +350,36 @@ static int i915_ringbuffer_info(struct seq_file *m, void *data)
 	return 0;
 }
 
+static const char *pin_flag(int pinned)
+{
+	if (pinned > 0)
+		return " P";
+	else if (pinned < 0)
+		return " p";
+	else
+		return "";
+}
+
+static const char *tiling_flag(int tiling)
+{
+	switch (tiling) {
+	default:
+	case I915_TILING_NONE: return "";
+	case I915_TILING_X: return " X";
+	case I915_TILING_Y: return " Y";
+	}
+}
+
+static const char *dirty_flag(int dirty)
+{
+	return dirty ? " dirty" : "";
+}
+
+static const char *purgeable_flag(int purgeable)
+{
+	return purgeable ? " purgeable" : "";
+}
+
 static int i915_error_state(struct seq_file *m, void *unused)
 {
 	struct drm_info_node *node = (struct drm_info_node *) m->private;
@@ -355,6 +387,7 @@ static int i915_error_state(struct seq_file *m, void *unused)
 	drm_i915_private_t *dev_priv = dev->dev_private;
 	struct drm_i915_error_state *error;
 	unsigned long flags;
+	int i, page, offset, elt;
 
 	spin_lock_irqsave(&dev_priv->error_lock, flags);
 	if (!dev_priv->first_error) {
@@ -366,6 +399,7 @@ static int i915_error_state(struct seq_file *m, void *unused)
 
 	seq_printf(m, "Time: %ld s %ld us\n", error->time.tv_sec,
 		   error->time.tv_usec);
+	seq_printf(m, "PCI ID: 0x%04x\n", dev->pci_device);
 	seq_printf(m, "EIR: 0x%08x\n", error->eir);
 	seq_printf(m, "  PGTBL_ER: 0x%08x\n", error->pgtbl_er);
 	seq_printf(m, "  INSTPM: 0x%08x\n", error->instpm);
@@ -377,6 +411,59 @@ static int i915_error_state(struct seq_file *m, void *unused)
 		seq_printf(m, "  INSTPS: 0x%08x\n", error->instps);
 		seq_printf(m, "  INSTDONE1: 0x%08x\n", error->instdone1);
 	}
+	seq_printf(m, "seqno: 0x%08x\n", error->seqno);
+
+	if (error->active_bo_count) {
+		seq_printf(m, "Buffers [%d]:\n", error->active_bo_count);
+
+		for (i = 0; i < error->active_bo_count; i++) {
+			seq_printf(m, "  %08x %8zd %08x %08x %08x%s%s%s%s",
+				   error->active_bo[i].gtt_offset,
+				   error->active_bo[i].size,
+				   error->active_bo[i].read_domains,
+				   error->active_bo[i].write_domain,
+				   error->active_bo[i].seqno,
+				   pin_flag(error->active_bo[i].pinned),
+				   tiling_flag(error->active_bo[i].tiling),
+				   dirty_flag(error->active_bo[i].dirty),
+				   purgeable_flag(error->active_bo[i].purgeable));
+
+			if (error->active_bo[i].name)
+				seq_printf(m, " (name: %d)", error->active_bo[i].name);
+			if (error->active_bo[i].fence_reg != I915_FENCE_REG_NONE)
+				seq_printf(m, " (fence: %d)", error->active_bo[i].fence_reg);
+
+			seq_printf(m, "\n");
+		}
+	}
+
+	for (i = 0; i < ARRAY_SIZE(error->batchbuffer); i++) {
+		if (error->batchbuffer[i]) {
+			struct drm_i915_error_object *obj = error->batchbuffer[i];
+
+			seq_printf(m, "--- gtt_offset = 0x%08x\n", obj->gtt_offset);
+			offset = 0;
+			for (page = 0; page < obj->page_count; page++) {
+				for (elt = 0; elt < PAGE_SIZE/4; elt++) {
+					seq_printf(m, "%08x :  %08x\n", offset, obj->pages[page][elt]);
+					offset += 4;
+				}
+			}
+		}
+	}
+
+	if (error->ringbuffer) {
+		struct drm_i915_error_object *obj = error->ringbuffer;
+
+		seq_printf(m, "--- ringbuffer = 0x%08x\n", obj->gtt_offset);
+		offset = 0;
+		for (page = 0; page < obj->page_count; page++) {
+			for (elt = 0; elt < PAGE_SIZE/4; elt++) {
+				seq_printf(m, "%08x :  %08x\n", offset, obj->pages[page][elt]);
+				offset += 4;
+			}
+		}
+	}
 
 out:
 	spin_unlock_irqrestore(&dev_priv->error_lock, flags);
@@ -384,37 +471,270 @@ out:
 	return 0;
 }
 
-static int i915_registers_info(struct seq_file *m, void *data) {
+static int i915_rstdby_delays(struct seq_file *m, void *unused)
+{
 	struct drm_info_node *node = (struct drm_info_node *) m->private;
 	struct drm_device *dev = node->minor->dev;
 	drm_i915_private_t *dev_priv = dev->dev_private;
-	uint32_t reg;
-
-#define DUMP_RANGE(start, end) \
-	for (reg=start; reg < end; reg += 4) \
-	seq_printf(m, "%08x\t%08x\n", reg, I915_READ(reg));
-
-	DUMP_RANGE(0x00000, 0x00fff);   /* VGA registers */
-	DUMP_RANGE(0x02000, 0x02fff);   /* instruction, memory, interrupt control registers */
-	DUMP_RANGE(0x03000, 0x031ff);   /* FENCE and PPGTT control registers */
-	DUMP_RANGE(0x03200, 0x03fff);   /* frame buffer compression registers */
-	DUMP_RANGE(0x05000, 0x05fff);   /* I/O control registers */
-	DUMP_RANGE(0x06000, 0x06fff);   /* clock control registers */
-	DUMP_RANGE(0x07000, 0x07fff);   /* 3D internal debug registers */
-	DUMP_RANGE(0x07400, 0x088ff);   /* GPE debug registers */
-	DUMP_RANGE(0x0a000, 0x0afff);   /* display palette registers */
-	DUMP_RANGE(0x10000, 0x13fff);   /* MMIO MCHBAR */
-	DUMP_RANGE(0x30000, 0x3ffff);   /* overlay registers */
-	DUMP_RANGE(0x60000, 0x6ffff);   /* display engine pipeline registers */
-	DUMP_RANGE(0x70000, 0x72fff);   /* display and cursor registers */
-	DUMP_RANGE(0x73000, 0x73fff);   /* performance counters */
+	u16 crstanddelay = I915_READ16(CRSTANDVID);
+
+	seq_printf(m, "w/ctx: %d, w/o ctx: %d\n", (crstanddelay >> 8) & 0x3f, (crstanddelay & 0x3f));
 
 	return 0;
 }
 
+static int i915_cur_delayinfo(struct seq_file *m, void *unused)
+{
+	struct drm_info_node *node = (struct drm_info_node *) m->private;
+	struct drm_device *dev = node->minor->dev;
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	u16 rgvswctl = I915_READ16(MEMSWCTL);
+
+	seq_printf(m, "Last command: 0x%01x\n", (rgvswctl >> 13) & 0x3);
+	seq_printf(m, "Command status: %d\n", (rgvswctl >> 12) & 1);
+	seq_printf(m, "P%d DELAY 0x%02x\n", (rgvswctl >> 8) & 0xf,
+		   rgvswctl & 0x3f);
+
+	return 0;
+}
+
+static int i915_delayfreq_table(struct seq_file *m, void *unused)
+{
+	struct drm_info_node *node = (struct drm_info_node *) m->private;
+	struct drm_device *dev = node->minor->dev;
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	u32 delayfreq;
+	int i;
+
+	for (i = 0; i < 16; i++) {
+		delayfreq = I915_READ(PXVFREQ_BASE + i * 4);
+		seq_printf(m, "P%02dVIDFREQ: 0x%08x\n", i, delayfreq);
+	}
+
+	return 0;
+}
+
+static inline int MAP_TO_MV(int map)
+{
+	return 1250 - (map * 25);
+}
+
+static int i915_inttoext_table(struct seq_file *m, void *unused)
+{
+	struct drm_info_node *node = (struct drm_info_node *) m->private;
+	struct drm_device *dev = node->minor->dev;
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	u32 inttoext;
+	int i;
+
+	for (i = 1; i <= 32; i++) {
+		inttoext = I915_READ(INTTOEXT_BASE_ILK + i * 4);
+		seq_printf(m, "INTTOEXT%02d: 0x%08x\n", i, inttoext);
+	}
+
+	return 0;
+}
+
+static int i915_drpc_info(struct seq_file *m, void *unused)
+{
+	struct drm_info_node *node = (struct drm_info_node *) m->private;
+	struct drm_device *dev = node->minor->dev;
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	u32 rgvmodectl = I915_READ(MEMMODECTL);
+
+	seq_printf(m, "HD boost: %s\n", (rgvmodectl & MEMMODE_BOOST_EN) ?
+		   "yes" : "no");
+	seq_printf(m, "Boost freq: %d\n",
+		   (rgvmodectl & MEMMODE_BOOST_FREQ_MASK) >>
+		   MEMMODE_BOOST_FREQ_SHIFT);
+	seq_printf(m, "HW control enabled: %s\n",
+		   rgvmodectl & MEMMODE_HWIDLE_EN ? "yes" : "no");
+	seq_printf(m, "SW control enabled: %s\n",
+		   rgvmodectl & MEMMODE_SWMODE_EN ? "yes" : "no");
+	seq_printf(m, "Gated voltage change: %s\n",
+		   rgvmodectl & MEMMODE_RCLK_GATE ? "yes" : "no");
+	seq_printf(m, "Starting frequency: P%d\n",
+		   (rgvmodectl & MEMMODE_FSTART_MASK) >> MEMMODE_FSTART_SHIFT);
+	seq_printf(m, "Max frequency: P%d\n",
+		   (rgvmodectl & MEMMODE_FMAX_MASK) >> MEMMODE_FMAX_SHIFT);
+	seq_printf(m, "Min frequency: P%d\n", (rgvmodectl & MEMMODE_FMIN_MASK));
+
+	return 0;
+}
+
+static int i915_fbc_status(struct seq_file *m, void *unused)
+{
+	struct drm_info_node *node = (struct drm_info_node *) m->private;
+	struct drm_device *dev = node->minor->dev;
+	struct drm_crtc *crtc;
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	bool fbc_enabled = false;
+
+	if (!dev_priv->display.fbc_enabled) {
+		seq_printf(m, "FBC unsupported on this chipset\n");
+		return 0;
+	}
+
+	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
+		if (!crtc->enabled)
+			continue;
+		if (dev_priv->display.fbc_enabled(crtc))
+			fbc_enabled = true;
+	}
+
+	if (fbc_enabled) {
+		seq_printf(m, "FBC enabled\n");
+	} else {
+		seq_printf(m, "FBC disabled: ");
+		switch (dev_priv->no_fbc_reason) {
+		case FBC_STOLEN_TOO_SMALL:
+			seq_printf(m, "not enough stolen memory");
+			break;
+		case FBC_UNSUPPORTED_MODE:
+			seq_printf(m, "mode not supported");
+			break;
+		case FBC_MODE_TOO_LARGE:
+			seq_printf(m, "mode too large");
+			break;
+		case FBC_BAD_PLANE:
+			seq_printf(m, "FBC unsupported on plane");
+			break;
+		case FBC_NOT_TILED:
+			seq_printf(m, "scanout buffer not tiled");
+			break;
+		default:
+			seq_printf(m, "unknown reason");
+		}
+		seq_printf(m, "\n");
+	}
+	return 0;
+}
+
+static int i915_sr_status(struct seq_file *m, void *unused)
+{
+	struct drm_info_node *node = (struct drm_info_node *) m->private;
+	struct drm_device *dev = node->minor->dev;
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	bool sr_enabled = false;
+
+	if (IS_I965G(dev) || IS_I945G(dev) || IS_I945GM(dev))
+		sr_enabled = I915_READ(FW_BLC_SELF) & FW_BLC_SELF_EN;
+	else if (IS_I915GM(dev))
+		sr_enabled = I915_READ(INSTPM) & INSTPM_SELF_EN;
+	else if (IS_PINEVIEW(dev))
+		sr_enabled = I915_READ(DSPFW3) & PINEVIEW_SELF_REFRESH_EN;
+
+	seq_printf(m, "self-refresh: %s\n", sr_enabled ? "enabled" :
+		   "disabled");
+
+	return 0;
+}
+
+static int
+i915_wedged_open(struct inode *inode,
+		 struct file *filp)
+{
+	filp->private_data = inode->i_private;
+	return 0;
+}
+
+static ssize_t
+i915_wedged_read(struct file *filp,
+		 char __user *ubuf,
+		 size_t max,
+		 loff_t *ppos)
+{
+	struct drm_device *dev = filp->private_data;
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	char buf[80];
+	int len;
+
+	len = snprintf(buf, sizeof (buf),
+		       "wedged :  %d\n",
+		       atomic_read(&dev_priv->mm.wedged));
+
+	return simple_read_from_buffer(ubuf, max, ppos, buf, len);
+}
+
+static ssize_t
+i915_wedged_write(struct file *filp,
+		  const char __user *ubuf,
+		  size_t cnt,
+		  loff_t *ppos)
+{
+	struct drm_device *dev = filp->private_data;
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	char buf[20];
+	int val = 1;
+
+	if (cnt > 0) {
+		if (cnt > sizeof (buf) - 1)
+			return -EINVAL;
+
+		if (copy_from_user(buf, ubuf, cnt))
+			return -EFAULT;
+		buf[cnt] = 0;
+
+		val = simple_strtoul(buf, NULL, 0);
+	}
+
+	DRM_INFO("Manually setting wedged to %d\n", val);
+
+	atomic_set(&dev_priv->mm.wedged, val);
+	if (val) {
+		DRM_WAKEUP(&dev_priv->irq_queue);
+		queue_work(dev_priv->wq, &dev_priv->error_work);
+	}
+
+	return cnt;
+}
+
+static const struct file_operations i915_wedged_fops = {
+	.owner = THIS_MODULE,
+	.open = i915_wedged_open,
+	.read = i915_wedged_read,
+	.write = i915_wedged_write,
+};
+
+/* As the drm_debugfs_init() routines are called before dev->dev_private is
+ * allocated we need to hook into the minor for release. */
+static int
+drm_add_fake_info_node(struct drm_minor *minor,
+		       struct dentry *ent,
+		       const void *key)
+{
+	struct drm_info_node *node;
+
+	node = kmalloc(sizeof(struct drm_info_node), GFP_KERNEL);
+	if (node == NULL) {
+		debugfs_remove(ent);
+		return -ENOMEM;
+	}
+
+	node->minor = minor;
+	node->dent = ent;
+	node->info_ent = (void *) key;
+	list_add(&node->list, &minor->debugfs_nodes.list);
+
+	return 0;
+}
+
+static int i915_wedged_create(struct dentry *root, struct drm_minor *minor)
+{
+	struct drm_device *dev = minor->dev;
+	struct dentry *ent;
+
+	ent = debugfs_create_file("i915_wedged",
+				  S_IRUGO | S_IWUSR,
+				  root, dev,
+				  &i915_wedged_fops);
+	if (IS_ERR(ent))
+		return PTR_ERR(ent);
+
+	return drm_add_fake_info_node(minor, ent, &i915_wedged_fops);
+}
 
 static struct drm_info_list i915_debugfs_list[] = {
-	{"i915_regs", i915_registers_info, 0},
 	{"i915_gem_active", i915_gem_object_list_info, 0, (void *) ACTIVE_LIST},
 	{"i915_gem_flushing", i915_gem_object_list_info, 0, (void *) FLUSHING_LIST},
 	{"i915_gem_inactive", i915_gem_object_list_info, 0, (void *) INACTIVE_LIST},
@@ -427,11 +747,24 @@ static struct drm_info_list i915_debugfs_list[] = {
 	{"i915_ringbuffer_info", i915_ringbuffer_info, 0},
 	{"i915_batchbuffers", i915_batchbuffer_info, 0},
 	{"i915_error_state", i915_error_state, 0},
+	{"i915_rstdby_delays", i915_rstdby_delays, 0},
+	{"i915_cur_delayinfo", i915_cur_delayinfo, 0},
+	{"i915_delayfreq_table", i915_delayfreq_table, 0},
+	{"i915_inttoext_table", i915_inttoext_table, 0},
+	{"i915_drpc_info", i915_drpc_info, 0},
+	{"i915_fbc_status", i915_fbc_status, 0},
+	{"i915_sr_status", i915_sr_status, 0},
 };
 #define I915_DEBUGFS_ENTRIES ARRAY_SIZE(i915_debugfs_list)
 
 int i915_debugfs_init(struct drm_minor *minor)
 {
+	int ret;
+
+	ret = i915_wedged_create(minor->debugfs_root, minor);
+	if (ret)
+		return ret;
+
 	return drm_debugfs_create_files(i915_debugfs_list,
 					I915_DEBUGFS_ENTRIES,
 					minor->debugfs_root, minor);
@@ -441,7 +774,8 @@ void i915_debugfs_cleanup(struct drm_minor *minor)
 {
 	drm_debugfs_remove_files(i915_debugfs_list,
 				 I915_DEBUGFS_ENTRIES, minor);
+	drm_debugfs_remove_files((struct drm_info_list *) &i915_wedged_fops,
+				 1, minor);
 }
 
 #endif /* CONFIG_DEBUG_FS */
-
diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c
index 45d507ebd3ff..8bfc0bbf13e6 100644
--- a/drivers/gpu/drm/i915/i915_dma.c
+++ b/drivers/gpu/drm/i915/i915_dma.c
@@ -35,6 +35,9 @@
 #include "i915_drv.h"
 #include "i915_trace.h"
 #include <linux/vgaarb.h>
+#include <linux/acpi.h>
+#include <linux/pnp.h>
+#include <linux/vga_switcheroo.h>
 
 /* Really want an OS-independent resettable timer.  Would like to have
  * this loop run for (eg) 3 sec, but have the timer reset every time
@@ -123,7 +126,7 @@ static int i915_init_phys_hws(struct drm_device *dev)
 	drm_i915_private_t *dev_priv = dev->dev_private;
 	/* Program Hardware Status Page */
 	dev_priv->status_page_dmah =
-		drm_pci_alloc(dev, PAGE_SIZE, PAGE_SIZE, 0xffffffff);
+		drm_pci_alloc(dev, PAGE_SIZE, PAGE_SIZE);
 
 	if (!dev_priv->status_page_dmah) {
 		DRM_ERROR("Can not allocate hardware status page\n");
@@ -134,6 +137,10 @@ static int i915_init_phys_hws(struct drm_device *dev)
 
 	memset(dev_priv->hw_status_page, 0, PAGE_SIZE);
 
+	if (IS_I965G(dev))
+		dev_priv->dma_status_page |= (dev_priv->dma_status_page >> 28) &
+					     0xf0;
+
 	I915_WRITE(HWS_PGA, dev_priv->dma_status_page);
 	DRM_DEBUG_DRIVER("Enabled hardware status page\n");
 	return 0;
@@ -731,8 +738,10 @@ static int i915_cmdbuffer(struct drm_device *dev, void *data,
 	if (cmdbuf->num_cliprects) {
 		cliprects = kcalloc(cmdbuf->num_cliprects,
 				    sizeof(struct drm_clip_rect), GFP_KERNEL);
-		if (cliprects == NULL)
+		if (cliprects == NULL) {
+			ret = -ENOMEM;
 			goto fail_batch_free;
+		}
 
 		ret = copy_from_user(cliprects, cmdbuf->cliprects,
 				     cmdbuf->num_cliprects *
@@ -807,9 +816,19 @@ static int i915_getparam(struct drm_device *dev, void *data,
 	case I915_PARAM_NUM_FENCES_AVAIL:
 		value = dev_priv->num_fence_regs - dev_priv->fence_reg_start;
 		break;
+	case I915_PARAM_HAS_OVERLAY:
+		value = dev_priv->overlay ? 1 : 0;
+		break;
+	case I915_PARAM_HAS_PAGEFLIPPING:
+		value = 1;
+		break;
+	case I915_PARAM_HAS_EXECBUF2:
+		/* depends on GEM */
+		value = dev_priv->has_gem;
+		break;
 	default:
 		DRM_DEBUG_DRIVER("Unknown parameter %d\n",
-					param->param);
+				 param->param);
 		return -EINVAL;
 	}
 
@@ -917,6 +936,120 @@ static int i915_get_bridge_dev(struct drm_device *dev)
 	return 0;
 }
 
+#define MCHBAR_I915 0x44
+#define MCHBAR_I965 0x48
+#define MCHBAR_SIZE (4*4096)
+
+#define DEVEN_REG 0x54
+#define   DEVEN_MCHBAR_EN (1 << 28)
+
+/* Allocate space for the MCH regs if needed, return nonzero on error */
+static int
+intel_alloc_mchbar_resource(struct drm_device *dev)
+{
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	int reg = IS_I965G(dev) ? MCHBAR_I965 : MCHBAR_I915;
+	u32 temp_lo, temp_hi = 0;
+	u64 mchbar_addr;
+	int ret = 0;
+
+	if (IS_I965G(dev))
+		pci_read_config_dword(dev_priv->bridge_dev, reg + 4, &temp_hi);
+	pci_read_config_dword(dev_priv->bridge_dev, reg, &temp_lo);
+	mchbar_addr = ((u64)temp_hi << 32) | temp_lo;
+
+	/* If ACPI doesn't have it, assume we need to allocate it ourselves */
+#ifdef CONFIG_PNP
+	if (mchbar_addr &&
+	    pnp_range_reserved(mchbar_addr, mchbar_addr + MCHBAR_SIZE)) {
+		ret = 0;
+		goto out;
+	}
+#endif
+
+	/* Get some space for it */
+	ret = pci_bus_alloc_resource(dev_priv->bridge_dev->bus, &dev_priv->mch_res,
+				     MCHBAR_SIZE, MCHBAR_SIZE,
+				     PCIBIOS_MIN_MEM,
+				     0,   pcibios_align_resource,
+				     dev_priv->bridge_dev);
+	if (ret) {
+		DRM_DEBUG_DRIVER("failed bus alloc: %d\n", ret);
+		dev_priv->mch_res.start = 0;
+		goto out;
+	}
+
+	if (IS_I965G(dev))
+		pci_write_config_dword(dev_priv->bridge_dev, reg + 4,
+				       upper_32_bits(dev_priv->mch_res.start));
+
+	pci_write_config_dword(dev_priv->bridge_dev, reg,
+			       lower_32_bits(dev_priv->mch_res.start));
+out:
+	return ret;
+}
+
+/* Setup MCHBAR if possible, return true if we should disable it again */
+static void
+intel_setup_mchbar(struct drm_device *dev)
+{
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	int mchbar_reg = IS_I965G(dev) ? MCHBAR_I965 : MCHBAR_I915;
+	u32 temp;
+	bool enabled;
+
+	dev_priv->mchbar_need_disable = false;
+
+	if (IS_I915G(dev) || IS_I915GM(dev)) {
+		pci_read_config_dword(dev_priv->bridge_dev, DEVEN_REG, &temp);
+		enabled = !!(temp & DEVEN_MCHBAR_EN);
+	} else {
+		pci_read_config_dword(dev_priv->bridge_dev, mchbar_reg, &temp);
+		enabled = temp & 1;
+	}
+
+	/* If it's already enabled, don't have to do anything */
+	if (enabled)
+		return;
+
+	if (intel_alloc_mchbar_resource(dev))
+		return;
+
+	dev_priv->mchbar_need_disable = true;
+
+	/* Space is allocated or reserved, so enable it. */
+	if (IS_I915G(dev) || IS_I915GM(dev)) {
+		pci_write_config_dword(dev_priv->bridge_dev, DEVEN_REG,
+				       temp | DEVEN_MCHBAR_EN);
+	} else {
+		pci_read_config_dword(dev_priv->bridge_dev, mchbar_reg, &temp);
+		pci_write_config_dword(dev_priv->bridge_dev, mchbar_reg, temp | 1);
+	}
+}
+
+static void
+intel_teardown_mchbar(struct drm_device *dev)
+{
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	int mchbar_reg = IS_I965G(dev) ? MCHBAR_I965 : MCHBAR_I915;
+	u32 temp;
+
+	if (dev_priv->mchbar_need_disable) {
+		if (IS_I915G(dev) || IS_I915GM(dev)) {
+			pci_read_config_dword(dev_priv->bridge_dev, DEVEN_REG, &temp);
+			temp &= ~DEVEN_MCHBAR_EN;
+			pci_write_config_dword(dev_priv->bridge_dev, DEVEN_REG, temp);
+		} else {
+			pci_read_config_dword(dev_priv->bridge_dev, mchbar_reg, &temp);
+			temp &= ~1;
+			pci_write_config_dword(dev_priv->bridge_dev, mchbar_reg, temp);
+		}
+	}
+
+	if (dev_priv->mch_res.start)
+		release_resource(&dev_priv->mch_res);
+}
+
 /**
  * i915_probe_agp - get AGP bootup configuration
  * @pdev: PCI device
@@ -962,59 +1095,123 @@ static int i915_probe_agp(struct drm_device *dev, uint32_t *aperture_size,
 	 * Some of the preallocated space is taken by the GTT
 	 * and popup.  GTT is 1K per MB of aperture size, and popup is 4K.
 	 */
-	if (IS_G4X(dev) || IS_IGD(dev) || IS_IGDNG(dev))
+	if (IS_G4X(dev) || IS_PINEVIEW(dev) || IS_IRONLAKE(dev) || IS_GEN6(dev))
 		overhead = 4096;
 	else
 		overhead = (*aperture_size / 1024) + 4096;
 
-	switch (tmp & INTEL_GMCH_GMS_MASK) {
-	case INTEL_855_GMCH_GMS_DISABLED:
-		DRM_ERROR("video memory is disabled\n");
-		return -1;
-	case INTEL_855_GMCH_GMS_STOLEN_1M:
-		stolen = 1 * 1024 * 1024;
-		break;
-	case INTEL_855_GMCH_GMS_STOLEN_4M:
-		stolen = 4 * 1024 * 1024;
-		break;
-	case INTEL_855_GMCH_GMS_STOLEN_8M:
-		stolen = 8 * 1024 * 1024;
-		break;
-	case INTEL_855_GMCH_GMS_STOLEN_16M:
-		stolen = 16 * 1024 * 1024;
-		break;
-	case INTEL_855_GMCH_GMS_STOLEN_32M:
-		stolen = 32 * 1024 * 1024;
-		break;
-	case INTEL_915G_GMCH_GMS_STOLEN_48M:
-		stolen = 48 * 1024 * 1024;
-		break;
-	case INTEL_915G_GMCH_GMS_STOLEN_64M:
-		stolen = 64 * 1024 * 1024;
-		break;
-	case INTEL_GMCH_GMS_STOLEN_128M:
-		stolen = 128 * 1024 * 1024;
-		break;
-	case INTEL_GMCH_GMS_STOLEN_256M:
-		stolen = 256 * 1024 * 1024;
-		break;
-	case INTEL_GMCH_GMS_STOLEN_96M:
-		stolen = 96 * 1024 * 1024;
-		break;
-	case INTEL_GMCH_GMS_STOLEN_160M:
-		stolen = 160 * 1024 * 1024;
-		break;
-	case INTEL_GMCH_GMS_STOLEN_224M:
-		stolen = 224 * 1024 * 1024;
-		break;
-	case INTEL_GMCH_GMS_STOLEN_352M:
-		stolen = 352 * 1024 * 1024;
-		break;
-	default:
-		DRM_ERROR("unexpected GMCH_GMS value: 0x%02x\n",
-			tmp & INTEL_GMCH_GMS_MASK);
-		return -1;
+	if (IS_GEN6(dev)) {
+		/* SNB has memory control reg at 0x50.w */
+		pci_read_config_word(dev->pdev, SNB_GMCH_CTRL, &tmp);
+
+		switch (tmp & SNB_GMCH_GMS_STOLEN_MASK) {
+		case INTEL_855_GMCH_GMS_DISABLED:
+			DRM_ERROR("video memory is disabled\n");
+			return -1;
+		case SNB_GMCH_GMS_STOLEN_32M:
+			stolen = 32 * 1024 * 1024;
+			break;
+		case SNB_GMCH_GMS_STOLEN_64M:
+			stolen = 64 * 1024 * 1024;
+			break;
+		case SNB_GMCH_GMS_STOLEN_96M:
+			stolen = 96 * 1024 * 1024;
+			break;
+		case SNB_GMCH_GMS_STOLEN_128M:
+			stolen = 128 * 1024 * 1024;
+			break;
+		case SNB_GMCH_GMS_STOLEN_160M:
+			stolen = 160 * 1024 * 1024;
+			break;
+		case SNB_GMCH_GMS_STOLEN_192M:
+			stolen = 192 * 1024 * 1024;
+			break;
+		case SNB_GMCH_GMS_STOLEN_224M:
+			stolen = 224 * 1024 * 1024;
+			break;
+		case SNB_GMCH_GMS_STOLEN_256M:
+			stolen = 256 * 1024 * 1024;
+			break;
+		case SNB_GMCH_GMS_STOLEN_288M:
+			stolen = 288 * 1024 * 1024;
+			break;
+		case SNB_GMCH_GMS_STOLEN_320M:
+			stolen = 320 * 1024 * 1024;
+			break;
+		case SNB_GMCH_GMS_STOLEN_352M:
+			stolen = 352 * 1024 * 1024;
+			break;
+		case SNB_GMCH_GMS_STOLEN_384M:
+			stolen = 384 * 1024 * 1024;
+			break;
+		case SNB_GMCH_GMS_STOLEN_416M:
+			stolen = 416 * 1024 * 1024;
+			break;
+		case SNB_GMCH_GMS_STOLEN_448M:
+			stolen = 448 * 1024 * 1024;
+			break;
+		case SNB_GMCH_GMS_STOLEN_480M:
+			stolen = 480 * 1024 * 1024;
+			break;
+		case SNB_GMCH_GMS_STOLEN_512M:
+			stolen = 512 * 1024 * 1024;
+			break;
+		default:
+			DRM_ERROR("unexpected GMCH_GMS value: 0x%02x\n",
+				  tmp & SNB_GMCH_GMS_STOLEN_MASK);
+			return -1;
+		}
+	} else {
+		switch (tmp & INTEL_GMCH_GMS_MASK) {
+		case INTEL_855_GMCH_GMS_DISABLED:
+			DRM_ERROR("video memory is disabled\n");
+			return -1;
+		case INTEL_855_GMCH_GMS_STOLEN_1M:
+			stolen = 1 * 1024 * 1024;
+			break;
+		case INTEL_855_GMCH_GMS_STOLEN_4M:
+			stolen = 4 * 1024 * 1024;
+			break;
+		case INTEL_855_GMCH_GMS_STOLEN_8M:
+			stolen = 8 * 1024 * 1024;
+			break;
+		case INTEL_855_GMCH_GMS_STOLEN_16M:
+			stolen = 16 * 1024 * 1024;
+			break;
+		case INTEL_855_GMCH_GMS_STOLEN_32M:
+			stolen = 32 * 1024 * 1024;
+			break;
+		case INTEL_915G_GMCH_GMS_STOLEN_48M:
+			stolen = 48 * 1024 * 1024;
+			break;
+		case INTEL_915G_GMCH_GMS_STOLEN_64M:
+			stolen = 64 * 1024 * 1024;
+			break;
+		case INTEL_GMCH_GMS_STOLEN_128M:
+			stolen = 128 * 1024 * 1024;
+			break;
+		case INTEL_GMCH_GMS_STOLEN_256M:
+			stolen = 256 * 1024 * 1024;
+			break;
+		case INTEL_GMCH_GMS_STOLEN_96M:
+			stolen = 96 * 1024 * 1024;
+			break;
+		case INTEL_GMCH_GMS_STOLEN_160M:
+			stolen = 160 * 1024 * 1024;
+			break;
+		case INTEL_GMCH_GMS_STOLEN_224M:
+			stolen = 224 * 1024 * 1024;
+			break;
+		case INTEL_GMCH_GMS_STOLEN_352M:
+			stolen = 352 * 1024 * 1024;
+			break;
+		default:
+			DRM_ERROR("unexpected GMCH_GMS value: 0x%02x\n",
+				  tmp & INTEL_GMCH_GMS_MASK);
+			return -1;
+		}
 	}
+
 	*preallocated_size = stolen - overhead;
 	*start = overhead;
 
@@ -1048,7 +1245,7 @@ static unsigned long i915_gtt_to_phys(struct drm_device *dev,
 	int gtt_offset, gtt_size;
 
 	if (IS_I965G(dev)) {
-		if (IS_G4X(dev) || IS_IGDNG(dev)) {
+		if (IS_G4X(dev) || IS_IRONLAKE(dev) || IS_GEN6(dev)) {
 			gtt_offset = 2*1024*1024;
 			gtt_size = 2*1024*1024;
 		} else {
@@ -1070,7 +1267,7 @@ static unsigned long i915_gtt_to_phys(struct drm_device *dev,
 
 	entry = *(volatile u32 *)(gtt + (gtt_addr / 1024));
 
-	DRM_DEBUG("GTT addr: 0x%08lx, PTE: 0x%08lx\n", gtt_addr, entry);
+	DRM_DEBUG_DRIVER("GTT addr: 0x%08lx, PTE: 0x%08lx\n", gtt_addr, entry);
 
 	/* Mask out these reserved bits on this hardware. */
 	if (!IS_I9XX(dev) || IS_I915G(dev) || IS_I915GM(dev) ||
@@ -1096,7 +1293,7 @@ static unsigned long i915_gtt_to_phys(struct drm_device *dev,
 	phys =(entry & PTE_ADDRESS_MASK) |
 		((uint64_t)(entry & PTE_ADDRESS_MASK_HIGH) << (32 - 4));
 
-	DRM_DEBUG("GTT addr: 0x%08lx, phys addr: 0x%08lx\n", gtt_addr, phys);
+	DRM_DEBUG_DRIVER("GTT addr: 0x%08lx, phys addr: 0x%08lx\n", gtt_addr, phys);
 
 	return phys;
 }
@@ -1111,11 +1308,13 @@ static void i915_setup_compression(struct drm_device *dev, int size)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct drm_mm_node *compressed_fb, *compressed_llb;
-	unsigned long cfb_base, ll_base;
+	unsigned long cfb_base;
+	unsigned long ll_base = 0;
 
 	/* Leave 1M for line length buffer & misc. */
 	compressed_fb = drm_mm_search_free(&dev_priv->vram, size, 4096, 0);
 	if (!compressed_fb) {
+		dev_priv->no_fbc_reason = FBC_STOLEN_TOO_SMALL;
 		i915_warn_stolen(dev);
 		return;
 	}
@@ -1123,6 +1322,7 @@ static void i915_setup_compression(struct drm_device *dev, int size)
 	compressed_fb = drm_mm_get_block(compressed_fb, size, 4096);
 	if (!compressed_fb) {
 		i915_warn_stolen(dev);
+		dev_priv->no_fbc_reason = FBC_STOLEN_TOO_SMALL;
 		return;
 	}
 
@@ -1182,6 +1382,32 @@ static unsigned int i915_vga_set_decode(void *cookie, bool state)
 		return VGA_RSRC_NORMAL_IO | VGA_RSRC_NORMAL_MEM;
 }
 
+static void i915_switcheroo_set_state(struct pci_dev *pdev, enum vga_switcheroo_state state)
+{
+	struct drm_device *dev = pci_get_drvdata(pdev);
+	pm_message_t pmm = { .event = PM_EVENT_SUSPEND };
+	if (state == VGA_SWITCHEROO_ON) {
+		printk(KERN_INFO "i915: switched off\n");
+		/* i915 resume handler doesn't set to D0 */
+		pci_set_power_state(dev->pdev, PCI_D0);
+		i915_resume(dev);
+	} else {
+		printk(KERN_ERR "i915: switched off\n");
+		i915_suspend(dev, pmm);
+	}
+}
+
+static bool i915_switcheroo_can_switch(struct pci_dev *pdev)
+{
+	struct drm_device *dev = pci_get_drvdata(pdev);
+	bool can_switch;
+
+	spin_lock(&dev->count_lock);
+	can_switch = (dev->open_count == 0);
+	spin_unlock(&dev->count_lock);
+	return can_switch;
+}
+
 static int i915_load_modeset_init(struct drm_device *dev,
 				  unsigned long prealloc_start,
 				  unsigned long prealloc_size,
@@ -1194,14 +1420,6 @@ static int i915_load_modeset_init(struct drm_device *dev,
 	dev->mode_config.fb_base = drm_get_resource_start(dev, fb_bar) &
 		0xff000000;
 
-	if (IS_MOBILE(dev) || IS_I9XX(dev))
-		dev_priv->cursor_needs_physical = true;
-	else
-		dev_priv->cursor_needs_physical = false;
-
-	if (IS_I965G(dev) || IS_G33(dev))
-		dev_priv->cursor_needs_physical = false;
-
 	/* Basic memrange allocator for stolen space (aka vram) */
 	drm_mm_init(&dev_priv->vram, 0, prealloc_size);
 	DRM_INFO("set up %ldM of stolen space\n", prealloc_size / (1024*1024));
@@ -1227,8 +1445,7 @@ static int i915_load_modeset_init(struct drm_device *dev,
 		goto out;
 
 	/* Try to set up FBC with a reasonable compressed buffer size */
-	if (IS_MOBILE(dev) && (IS_I9XX(dev) || IS_I965G(dev) || IS_GM45(dev)) &&
-	    i915_powersave) {
+	if (I915_HAS_FBC(dev) && i915_powersave) {
 		int cfb_size;
 
 		/* Try to get an 8M buffer... */
@@ -1252,6 +1469,14 @@ static int i915_load_modeset_init(struct drm_device *dev,
 	if (ret)
 		goto destroy_ringbuffer;
 
+	ret = vga_switcheroo_register_client(dev->pdev,
+					     i915_switcheroo_set_state,
+					     i915_switcheroo_can_switch);
+	if (ret)
+		goto destroy_ringbuffer;
+
+	intel_modeset_init(dev);
+
 	ret = drm_irq_install(dev);
 	if (ret)
 		goto destroy_ringbuffer;
@@ -1266,14 +1491,14 @@ static int i915_load_modeset_init(struct drm_device *dev,
 
 	I915_WRITE(INSTPM, (1 << 5) | (1 << 21));
 
-	intel_modeset_init(dev);
-
 	drm_helper_initial_config(dev);
 
 	return 0;
 
 destroy_ringbuffer:
+	mutex_lock(&dev->struct_mutex);
 	i915_gem_cleanup_ringbuffer(dev);
+	mutex_unlock(&dev->struct_mutex);
 out:
 	return ret;
 }
@@ -1307,7 +1532,7 @@ static void i915_get_mem_freq(struct drm_device *dev)
 	drm_i915_private_t *dev_priv = dev->dev_private;
 	u32 tmp;
 
-	if (!IS_IGD(dev))
+	if (!IS_PINEVIEW(dev))
 		return;
 
 	tmp = I915_READ(CLKCFG);
@@ -1355,7 +1580,7 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	resource_size_t base, size;
-	int ret = 0, mmio_bar = IS_I9XX(dev) ? 0 : 1;
+	int ret = 0, mmio_bar;
 	uint32_t agp_size, prealloc_size, prealloc_start;
 
 	/* i915 has 4 more counters */
@@ -1371,8 +1596,10 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags)
 
 	dev->dev_private = (void *)dev_priv;
 	dev_priv->dev = dev;
+	dev_priv->info = (struct intel_device_info *) flags;
 
 	/* Add register map (needed for suspend/resume) */
+	mmio_bar = IS_I9XX(dev) ? 0 : 1;
 	base = drm_get_resource_start(dev, mmio_bar);
 	size = drm_get_resource_len(dev, mmio_bar);
 
@@ -1414,7 +1641,7 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags)
 	if (ret)
 		goto out_iomapfree;
 
-	dev_priv->wq = create_workqueue("i915");
+	dev_priv->wq = create_singlethread_workqueue("i915");
 	if (dev_priv->wq == NULL) {
 		DRM_ERROR("Failed to create our workqueue.\n");
 		ret = -ENOMEM;
@@ -1435,11 +1662,14 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags)
 
 	dev->driver->get_vblank_counter = i915_get_vblank_counter;
 	dev->max_vblank_count = 0xffffff; /* only 24 bits of frame count */
-	if (IS_G4X(dev) || IS_IGDNG(dev)) {
+	if (IS_G4X(dev) || IS_IRONLAKE(dev) || IS_GEN6(dev)) {
 		dev->max_vblank_count = 0xffffffff; /* full 32 bit counter */
 		dev->driver->get_vblank_counter = gm45_get_vblank_counter;
 	}
 
+	/* Try to make sure MCHBAR is enabled before poking at it */
+	intel_setup_mchbar(dev);
+
 	i915_gem_load(dev);
 
 	/* Init HWS */
@@ -1468,6 +1698,7 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags)
 	spin_lock_init(&dev_priv->user_irq_lock);
 	spin_lock_init(&dev_priv->error_lock);
 	dev_priv->user_irq_refcount = 0;
+	dev_priv->trace_irq_seqno = 0;
 
 	ret = drm_vblank_init(dev, I915_NUM_PIPE);
 
@@ -1489,9 +1720,7 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags)
 	}
 
 	/* Must be done after probing outputs */
-	/* FIXME: verify on IGDNG */
-	if (!IS_IGDNG(dev))
-		intel_opregion_init(dev, 0);
+	intel_opregion_init(dev, 0);
 
 	setup_timer(&dev_priv->hangcheck_timer, i915_hangcheck_elapsed,
 		    (unsigned long) dev);
@@ -1514,6 +1743,8 @@ int i915_driver_unload(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
 
+	i915_destroy_error_state(dev);
+
 	destroy_workqueue(dev_priv->wq);
 	del_timer_sync(&dev_priv->hangcheck_timer);
 
@@ -1525,7 +1756,17 @@ int i915_driver_unload(struct drm_device *dev)
 	}
 
 	if (drm_core_check_feature(dev, DRIVER_MODESET)) {
+		/*
+		 * free the memory space allocated for the child device
+		 * config parsed from VBT
+		 */
+		if (dev_priv->child_dev && dev_priv->child_dev_num) {
+			kfree(dev_priv->child_dev);
+			dev_priv->child_dev = NULL;
+			dev_priv->child_dev_num = 0;
+		}
 		drm_irq_uninstall(dev);
+		vga_switcheroo_unregister_client(dev->pdev);
 		vga_client_register(dev->pdev, NULL, NULL, NULL);
 	}
 
@@ -1535,8 +1776,7 @@ int i915_driver_unload(struct drm_device *dev)
 	if (dev_priv->regs != NULL)
 		iounmap(dev_priv->regs);
 
-	if (!IS_IGDNG(dev))
-		intel_opregion_free(dev, 0);
+	intel_opregion_free(dev, 0);
 
 	if (drm_core_check_feature(dev, DRIVER_MODESET)) {
 		intel_modeset_cleanup(dev);
@@ -1548,8 +1788,12 @@ int i915_driver_unload(struct drm_device *dev)
 		mutex_unlock(&dev->struct_mutex);
 		drm_mm_takedown(&dev_priv->vram);
 		i915_gem_lastclose(dev);
+
+		intel_cleanup_overlay(dev);
 	}
 
+	intel_teardown_mchbar(dev);
+
 	pci_dev_put(dev_priv->bridge_dev);
 	kfree(dev->dev_private);
 
@@ -1592,6 +1836,7 @@ void i915_driver_lastclose(struct drm_device * dev)
 
 	if (!dev_priv || drm_core_check_feature(dev, DRIVER_MODESET)) {
 		drm_fb_helper_restore();
+		vga_switcheroo_process_delayed_switch();
 		return;
 	}
 
@@ -1638,6 +1883,7 @@ struct drm_ioctl_desc i915_ioctls[] = {
 	DRM_IOCTL_DEF(DRM_I915_HWS_ADDR, i915_set_status_page, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY),
 	DRM_IOCTL_DEF(DRM_I915_GEM_INIT, i915_gem_init_ioctl, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY),
 	DRM_IOCTL_DEF(DRM_I915_GEM_EXECBUFFER, i915_gem_execbuffer, DRM_AUTH),
+	DRM_IOCTL_DEF(DRM_I915_GEM_EXECBUFFER2, i915_gem_execbuffer2, DRM_AUTH),
 	DRM_IOCTL_DEF(DRM_I915_GEM_PIN, i915_gem_pin_ioctl, DRM_AUTH|DRM_ROOT_ONLY),
 	DRM_IOCTL_DEF(DRM_I915_GEM_UNPIN, i915_gem_unpin_ioctl, DRM_AUTH|DRM_ROOT_ONLY),
 	DRM_IOCTL_DEF(DRM_I915_GEM_BUSY, i915_gem_busy_ioctl, DRM_AUTH),
@@ -1656,6 +1902,8 @@ struct drm_ioctl_desc i915_ioctls[] = {
 	DRM_IOCTL_DEF(DRM_I915_GEM_GET_APERTURE, i915_gem_get_aperture_ioctl, 0),
 	DRM_IOCTL_DEF(DRM_I915_GET_PIPE_FROM_CRTC_ID, intel_get_pipe_from_crtc_id, 0),
 	DRM_IOCTL_DEF(DRM_I915_GEM_MADVISE, i915_gem_madvise_ioctl, 0),
+	DRM_IOCTL_DEF(DRM_I915_OVERLAY_PUT_IMAGE, intel_overlay_put_image, DRM_MASTER|DRM_CONTROL_ALLOW),
+	DRM_IOCTL_DEF(DRM_I915_OVERLAY_ATTRS, intel_overlay_attrs, DRM_MASTER|DRM_CONTROL_ALLOW),
 };
 
 int i915_max_ioctl = DRM_ARRAY_SIZE(i915_ioctls);
diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index b93814c0d3e2..1b2e95455c05 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -33,7 +33,6 @@
 #include "i915_drm.h"
 #include "i915_drv.h"
 
-#include "drm_pciids.h"
 #include <linux/console.h>
 #include "drm_crtc_helper.h"
 
@@ -46,36 +45,162 @@ module_param_named(fbpercrtc, i915_fbpercrtc, int, 0400);
 unsigned int i915_powersave = 1;
 module_param_named(powersave, i915_powersave, int, 0400);
 
+unsigned int i915_lvds_downclock = 0;
+module_param_named(lvds_downclock, i915_lvds_downclock, int, 0400);
+
 static struct drm_driver driver;
+extern int intel_agp_enabled;
+
+#define INTEL_VGA_DEVICE(id, info) {		\
+	.class = PCI_CLASS_DISPLAY_VGA << 8,	\
+	.class_mask = 0xffff00,			\
+	.vendor = 0x8086,			\
+	.device = id,				\
+	.subvendor = PCI_ANY_ID,		\
+	.subdevice = PCI_ANY_ID,		\
+	.driver_data = (unsigned long) info }
+
+const static struct intel_device_info intel_i830_info = {
+	.is_i8xx = 1, .is_mobile = 1, .cursor_needs_physical = 1,
+};
+
+const static struct intel_device_info intel_845g_info = {
+	.is_i8xx = 1,
+};
+
+const static struct intel_device_info intel_i85x_info = {
+	.is_i8xx = 1, .is_mobile = 1, .cursor_needs_physical = 1,
+};
+
+const static struct intel_device_info intel_i865g_info = {
+	.is_i8xx = 1,
+};
+
+const static struct intel_device_info intel_i915g_info = {
+	.is_i915g = 1, .is_i9xx = 1, .cursor_needs_physical = 1,
+};
+const static struct intel_device_info intel_i915gm_info = {
+	.is_i9xx = 1,  .is_mobile = 1, .has_fbc = 1,
+	.cursor_needs_physical = 1,
+};
+const static struct intel_device_info intel_i945g_info = {
+	.is_i9xx = 1, .has_hotplug = 1, .cursor_needs_physical = 1,
+};
+const static struct intel_device_info intel_i945gm_info = {
+	.is_i945gm = 1, .is_i9xx = 1, .is_mobile = 1, .has_fbc = 1,
+	.has_hotplug = 1, .cursor_needs_physical = 1,
+};
 
-static struct pci_device_id pciidlist[] = {
-	i915_PCI_IDS
+const static struct intel_device_info intel_i965g_info = {
+	.is_i965g = 1, .is_i9xx = 1, .has_hotplug = 1,
+};
+
+const static struct intel_device_info intel_i965gm_info = {
+	.is_i965g = 1, .is_mobile = 1, .is_i965gm = 1, .is_i9xx = 1,
+	.is_mobile = 1, .has_fbc = 1, .has_rc6 = 1,
+	.has_hotplug = 1,
+};
+
+const static struct intel_device_info intel_g33_info = {
+	.is_g33 = 1, .is_i9xx = 1, .need_gfx_hws = 1,
+	.has_hotplug = 1,
+};
+
+const static struct intel_device_info intel_g45_info = {
+	.is_i965g = 1, .is_g4x = 1, .is_i9xx = 1, .need_gfx_hws = 1,
+	.has_pipe_cxsr = 1,
+	.has_hotplug = 1,
+};
+
+const static struct intel_device_info intel_gm45_info = {
+	.is_i965g = 1, .is_mobile = 1, .is_g4x = 1, .is_i9xx = 1,
+	.is_mobile = 1, .need_gfx_hws = 1, .has_fbc = 1, .has_rc6 = 1,
+	.has_pipe_cxsr = 1,
+	.has_hotplug = 1,
+};
+
+const static struct intel_device_info intel_pineview_info = {
+	.is_g33 = 1, .is_pineview = 1, .is_mobile = 1, .is_i9xx = 1,
+	.need_gfx_hws = 1,
+	.has_hotplug = 1,
+};
+
+const static struct intel_device_info intel_ironlake_d_info = {
+	.is_ironlake = 1, .is_i965g = 1, .is_i9xx = 1, .need_gfx_hws = 1,
+	.has_pipe_cxsr = 1,
+	.has_hotplug = 1,
+};
+
+const static struct intel_device_info intel_ironlake_m_info = {
+	.is_ironlake = 1, .is_mobile = 1, .is_i965g = 1, .is_i9xx = 1,
+	.need_gfx_hws = 1, .has_rc6 = 1,
+	.has_hotplug = 1,
+};
+
+const static struct intel_device_info intel_sandybridge_d_info = {
+	.is_i965g = 1, .is_i9xx = 1, .need_gfx_hws = 1,
+	.has_hotplug = 1,
+};
+
+const static struct intel_device_info intel_sandybridge_m_info = {
+	.is_i965g = 1, .is_mobile = 1, .is_i9xx = 1, .need_gfx_hws = 1,
+	.has_hotplug = 1,
+};
+
+const static struct pci_device_id pciidlist[] = {
+	INTEL_VGA_DEVICE(0x3577, &intel_i830_info),
+	INTEL_VGA_DEVICE(0x2562, &intel_845g_info),
+	INTEL_VGA_DEVICE(0x3582, &intel_i85x_info),
+	INTEL_VGA_DEVICE(0x35e8, &intel_i85x_info),
+	INTEL_VGA_DEVICE(0x2572, &intel_i865g_info),
+	INTEL_VGA_DEVICE(0x2582, &intel_i915g_info),
+	INTEL_VGA_DEVICE(0x258a, &intel_i915g_info),
+	INTEL_VGA_DEVICE(0x2592, &intel_i915gm_info),
+	INTEL_VGA_DEVICE(0x2772, &intel_i945g_info),
+	INTEL_VGA_DEVICE(0x27a2, &intel_i945gm_info),
+	INTEL_VGA_DEVICE(0x27ae, &intel_i945gm_info),
+	INTEL_VGA_DEVICE(0x2972, &intel_i965g_info),
+	INTEL_VGA_DEVICE(0x2982, &intel_i965g_info),
+	INTEL_VGA_DEVICE(0x2992, &intel_i965g_info),
+	INTEL_VGA_DEVICE(0x29a2, &intel_i965g_info),
+	INTEL_VGA_DEVICE(0x29b2, &intel_g33_info),
+	INTEL_VGA_DEVICE(0x29c2, &intel_g33_info),
+	INTEL_VGA_DEVICE(0x29d2, &intel_g33_info),
+	INTEL_VGA_DEVICE(0x2a02, &intel_i965gm_info),
+	INTEL_VGA_DEVICE(0x2a12, &intel_i965gm_info),
+	INTEL_VGA_DEVICE(0x2a42, &intel_gm45_info),
+	INTEL_VGA_DEVICE(0x2e02, &intel_g45_info),
+	INTEL_VGA_DEVICE(0x2e12, &intel_g45_info),
+	INTEL_VGA_DEVICE(0x2e22, &intel_g45_info),
+	INTEL_VGA_DEVICE(0x2e32, &intel_g45_info),
+	INTEL_VGA_DEVICE(0x2e42, &intel_g45_info),
+	INTEL_VGA_DEVICE(0xa001, &intel_pineview_info),
+	INTEL_VGA_DEVICE(0xa011, &intel_pineview_info),
+	INTEL_VGA_DEVICE(0x0042, &intel_ironlake_d_info),
+	INTEL_VGA_DEVICE(0x0046, &intel_ironlake_m_info),
+	INTEL_VGA_DEVICE(0x0102, &intel_sandybridge_d_info),
+	INTEL_VGA_DEVICE(0x0106, &intel_sandybridge_m_info),
+	{0, 0, 0}
 };
 
 #if defined(CONFIG_DRM_I915_KMS)
 MODULE_DEVICE_TABLE(pci, pciidlist);
 #endif
 
-static int i915_suspend(struct drm_device *dev, pm_message_t state)
+static int i915_drm_freeze(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
 
-	if (!dev || !dev_priv) {
-		DRM_ERROR("dev: %p, dev_priv: %p\n", dev, dev_priv);
-		DRM_ERROR("DRM not initialized, aborting suspend.\n");
-		return -ENODEV;
-	}
-
-	if (state.event == PM_EVENT_PRETHAW)
-		return 0;
-
 	pci_save_state(dev->pdev);
 
 	/* If KMS is active, we do the leavevt stuff here */
 	if (drm_core_check_feature(dev, DRIVER_MODESET)) {
-		if (i915_gem_idle(dev))
+		int error = i915_gem_idle(dev);
+		if (error) {
 			dev_err(&dev->pdev->dev,
-				"GEM idle failed, resume may fail\n");
+				"GEM idle failed, resume might fail\n");
+			return error;
+		}
 		drm_irq_uninstall(dev);
 	}
 
@@ -83,25 +208,42 @@ static int i915_suspend(struct drm_device *dev, pm_message_t state)
 
 	intel_opregion_free(dev, 1);
 
+	/* Modeset on resume, not lid events */
+	dev_priv->modeset_on_lid = 0;
+
+	return 0;
+}
+
+int i915_suspend(struct drm_device *dev, pm_message_t state)
+{
+	int error;
+
+	if (!dev || !dev->dev_private) {
+		DRM_ERROR("dev: %p\n", dev);
+		DRM_ERROR("DRM not initialized, aborting suspend.\n");
+		return -ENODEV;
+	}
+
+	if (state.event == PM_EVENT_PRETHAW)
+		return 0;
+
+	error = i915_drm_freeze(dev);
+	if (error)
+		return error;
+
 	if (state.event == PM_EVENT_SUSPEND) {
 		/* Shut down the device */
 		pci_disable_device(dev->pdev);
 		pci_set_power_state(dev->pdev, PCI_D3hot);
 	}
 
-	dev_priv->suspended = 1;
-
 	return 0;
 }
 
-static int i915_resume(struct drm_device *dev)
+static int i915_drm_thaw(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
-	int ret = 0;
-
-	if (pci_enable_device(dev->pdev))
-		return -1;
-	pci_set_master(dev->pdev);
+	int error = 0;
 
 	i915_restore_state(dev);
 
@@ -112,21 +254,28 @@ static int i915_resume(struct drm_device *dev)
 		mutex_lock(&dev->struct_mutex);
 		dev_priv->mm.suspended = 0;
 
-		ret = i915_gem_init_ringbuffer(dev);
-		if (ret != 0)
-			ret = -1;
+		error = i915_gem_init_ringbuffer(dev);
 		mutex_unlock(&dev->struct_mutex);
 
 		drm_irq_install(dev);
-	}
-	if (drm_core_check_feature(dev, DRIVER_MODESET)) {
+
 		/* Resume the modeset for every activated CRTC */
 		drm_helper_resume_force_mode(dev);
 	}
 
-	dev_priv->suspended = 0;
+	dev_priv->modeset_on_lid = 0;
+
+	return error;
+}
+
+int i915_resume(struct drm_device *dev)
+{
+	if (pci_enable_device(dev->pdev))
+		return -EIO;
 
-	return ret;
+	pci_set_master(dev->pdev);
+
+	return i915_drm_thaw(dev);
 }
 
 /**
@@ -267,22 +416,73 @@ i915_pci_remove(struct pci_dev *pdev)
 	drm_put_dev(dev);
 }
 
-static int
-i915_pci_suspend(struct pci_dev *pdev, pm_message_t state)
+static int i915_pm_suspend(struct device *dev)
 {
-	struct drm_device *dev = pci_get_drvdata(pdev);
+	struct pci_dev *pdev = to_pci_dev(dev);
+	struct drm_device *drm_dev = pci_get_drvdata(pdev);
+	int error;
 
-	return i915_suspend(dev, state);
+	if (!drm_dev || !drm_dev->dev_private) {
+		dev_err(dev, "DRM not initialized, aborting suspend.\n");
+		return -ENODEV;
+	}
+
+	error = i915_drm_freeze(drm_dev);
+	if (error)
+		return error;
+
+	pci_disable_device(pdev);
+	pci_set_power_state(pdev, PCI_D3hot);
+
+	return 0;
 }
 
-static int
-i915_pci_resume(struct pci_dev *pdev)
+static int i915_pm_resume(struct device *dev)
 {
-	struct drm_device *dev = pci_get_drvdata(pdev);
+	struct pci_dev *pdev = to_pci_dev(dev);
+	struct drm_device *drm_dev = pci_get_drvdata(pdev);
 
-	return i915_resume(dev);
+	return i915_resume(drm_dev);
 }
 
+static int i915_pm_freeze(struct device *dev)
+{
+	struct pci_dev *pdev = to_pci_dev(dev);
+	struct drm_device *drm_dev = pci_get_drvdata(pdev);
+
+	if (!drm_dev || !drm_dev->dev_private) {
+		dev_err(dev, "DRM not initialized, aborting suspend.\n");
+		return -ENODEV;
+	}
+
+	return i915_drm_freeze(drm_dev);
+}
+
+static int i915_pm_thaw(struct device *dev)
+{
+	struct pci_dev *pdev = to_pci_dev(dev);
+	struct drm_device *drm_dev = pci_get_drvdata(pdev);
+
+	return i915_drm_thaw(drm_dev);
+}
+
+static int i915_pm_poweroff(struct device *dev)
+{
+	struct pci_dev *pdev = to_pci_dev(dev);
+	struct drm_device *drm_dev = pci_get_drvdata(pdev);
+
+	return i915_drm_freeze(drm_dev);
+}
+
+const struct dev_pm_ops i915_pm_ops = {
+     .suspend = i915_pm_suspend,
+     .resume = i915_pm_resume,
+     .freeze = i915_pm_freeze,
+     .thaw = i915_pm_thaw,
+     .poweroff = i915_pm_poweroff,
+     .restore = i915_pm_resume,
+};
+
 static struct vm_operations_struct i915_gem_vm_ops = {
 	.fault = i915_gem_fault,
 	.open = drm_gem_vm_open,
@@ -302,8 +502,11 @@ static struct drm_driver driver = {
 	.lastclose = i915_driver_lastclose,
 	.preclose = i915_driver_preclose,
 	.postclose = i915_driver_postclose,
+
+	/* Used in place of i915_pm_ops for non-DRIVER_MODESET */
 	.suspend = i915_suspend,
 	.resume = i915_resume,
+
 	.device_is_agp = i915_driver_device_is_agp,
 	.enable_vblank = i915_enable_vblank,
 	.disable_vblank = i915_disable_vblank,
@@ -328,10 +531,11 @@ static struct drm_driver driver = {
 		 .owner = THIS_MODULE,
 		 .open = drm_open,
 		 .release = drm_release,
-		 .ioctl = drm_ioctl,
+		 .unlocked_ioctl = drm_ioctl,
 		 .mmap = drm_gem_mmap,
 		 .poll = drm_poll,
 		 .fasync = drm_fasync,
+		 .read = drm_read,
 #ifdef CONFIG_COMPAT
 		 .compat_ioctl = i915_compat_ioctl,
 #endif
@@ -342,10 +546,7 @@ static struct drm_driver driver = {
 		 .id_table = pciidlist,
 		 .probe = i915_pci_probe,
 		 .remove = i915_pci_remove,
-#ifdef CONFIG_PM
-		 .resume = i915_pci_resume,
-		 .suspend = i915_pci_suspend,
-#endif
+		 .driver.pm = &i915_pm_ops,
 	},
 
 	.name = DRIVER_NAME,
@@ -358,6 +559,11 @@ static struct drm_driver driver = {
 
 static int __init i915_init(void)
 {
+	if (!intel_agp_enabled) {
+		DRM_ERROR("drm/i915 can't work without intel_agp module!\n");
+		return -ENODEV;
+	}
+
 	driver.num_ioctls = i915_max_ioctl;
 
 	i915_gem_shrinker_init();
@@ -383,6 +589,11 @@ static int __init i915_init(void)
 		driver.driver_features &= ~DRIVER_MODESET;
 #endif
 
+	if (!(driver.driver_features & DRIVER_MODESET)) {
+		driver.suspend = i915_suspend;
+		driver.resume = i915_resume;
+	}
+
 	return drm_init(&driver);
 }
 
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index b24b2d145b75..979439cfb827 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -150,7 +150,27 @@ struct drm_i915_error_state {
 	u32 instps;
 	u32 instdone1;
 	u32 seqno;
+	u64 bbaddr;
 	struct timeval time;
+	struct drm_i915_error_object {
+		int page_count;
+		u32 gtt_offset;
+		u32 *pages[0];
+	} *ringbuffer, *batchbuffer[2];
+	struct drm_i915_error_buffer {
+		size_t size;
+		u32 name;
+		u32 seqno;
+		u32 gtt_offset;
+		u32 read_domains;
+		u32 write_domain;
+		u32 fence_reg;
+		s32 pinned:2;
+		u32 tiling:2;
+		u32 dirty:1;
+		u32 purgeable:1;
+	} *active_bo;
+	u32 active_bo_count;
 };
 
 struct drm_i915_display_funcs {
@@ -170,9 +190,41 @@ struct drm_i915_display_funcs {
 	/* clock gating init */
 };
 
+struct intel_overlay;
+
+struct intel_device_info {
+	u8 is_mobile : 1;
+	u8 is_i8xx : 1;
+	u8 is_i915g : 1;
+	u8 is_i9xx : 1;
+	u8 is_i945gm : 1;
+	u8 is_i965g : 1;
+	u8 is_i965gm : 1;
+	u8 is_g33 : 1;
+	u8 need_gfx_hws : 1;
+	u8 is_g4x : 1;
+	u8 is_pineview : 1;
+	u8 is_ironlake : 1;
+	u8 has_fbc : 1;
+	u8 has_rc6 : 1;
+	u8 has_pipe_cxsr : 1;
+	u8 has_hotplug : 1;
+	u8 cursor_needs_physical : 1;
+};
+
+enum no_fbc_reason {
+	FBC_STOLEN_TOO_SMALL, /* not enough space to hold compressed buffers */
+	FBC_UNSUPPORTED_MODE, /* interlace or doublescanned mode */
+	FBC_MODE_TOO_LARGE, /* mode too large for compression */
+	FBC_BAD_PLANE, /* fbc not supported on plane */
+	FBC_NOT_TILED, /* buffer not tiled */
+};
+
 typedef struct drm_i915_private {
 	struct drm_device *dev;
 
+	const struct intel_device_info *info;
+
 	int has_gem;
 
 	void __iomem *regs;
@@ -187,6 +239,7 @@ typedef struct drm_i915_private {
 	unsigned int status_gfx_addr;
 	drm_local_map_t hws_map;
 	struct drm_gem_object *hws_obj;
+	struct drm_gem_object *pwrctx;
 
 	struct resource mch_res;
 
@@ -202,14 +255,17 @@ typedef struct drm_i915_private {
 	spinlock_t user_irq_lock;
 	/** Refcount for i915_user_irq_get() versus i915_user_irq_put(). */
 	int user_irq_refcount;
+	u32 trace_irq_seqno;
 	/** Cached value of IMR to avoid reads in updating the bitfield */
 	u32 irq_mask_reg;
 	u32 pipestat[2];
-	/** splitted irq regs for graphics and display engine on IGDNG,
+	/** splitted irq regs for graphics and display engine on Ironlake,
 	    irq_mask_reg is still used for display irq. */
 	u32 gt_irq_mask_reg;
 	u32 gt_irq_enable_reg;
 	u32 de_irq_enable_reg;
+	u32 pch_irq_mask_reg;
+	u32 pch_irq_enable_reg;
 
 	u32 hotplug_supported_mask;
 	struct work_struct hotplug_work;
@@ -226,8 +282,6 @@ typedef struct drm_i915_private {
 	int hangcheck_count;
 	uint32_t last_acthd;
 
-	bool cursor_needs_physical;
-
 	struct drm_mm vram;
 
 	unsigned long cfb_size;
@@ -239,6 +293,9 @@ typedef struct drm_i915_private {
 
 	struct intel_opregion opregion;
 
+	/* overlay */
+	struct intel_overlay *overlay;
+
 	/* LVDS info */
 	int backlight_duty_cycle;  /* restore backlight to this value */
 	bool panel_wants_dither;
@@ -254,10 +311,11 @@ typedef struct drm_i915_private {
 	unsigned int lvds_use_ssc:1;
 	unsigned int edp_support:1;
 	int lvds_ssc_freq;
+	int edp_bpp;
 
 	struct notifier_block lid_notifier;
 
-	int crt_ddc_bus; /* -1 = unknown, else GPIO to use for CRT DDC */
+	int crt_ddc_bus; /* 0 = unknown, else GPIO to use for CRT DDC */
 	struct drm_i915_fence_reg fence_regs[16]; /* assume 965 */
 	int fence_reg_start; /* 4 if userland hasn't ioctl'd us yet */
 	int num_fence_regs; /* 8 on pre-965, 16 otherwise */
@@ -273,12 +331,11 @@ typedef struct drm_i915_private {
 	struct drm_i915_display_funcs display;
 
 	/* Register state */
-	bool suspended;
+	bool modeset_on_lid;
 	u8 saveLBB;
 	u32 saveDSPACNTR;
 	u32 saveDSPBCNTR;
 	u32 saveDSPARB;
-	u32 saveRENDERSTANDBY;
 	u32 saveHWS;
 	u32 savePIPEACONF;
 	u32 savePIPEBCONF;
@@ -295,6 +352,13 @@ typedef struct drm_i915_private {
 	u32 saveVBLANK_A;
 	u32 saveVSYNC_A;
 	u32 saveBCLRPAT_A;
+	u32 saveTRANSACONF;
+	u32 saveTRANS_HTOTAL_A;
+	u32 saveTRANS_HBLANK_A;
+	u32 saveTRANS_HSYNC_A;
+	u32 saveTRANS_VTOTAL_A;
+	u32 saveTRANS_VBLANK_A;
+	u32 saveTRANS_VSYNC_A;
 	u32 savePIPEASTAT;
 	u32 saveDSPASTRIDE;
 	u32 saveDSPASIZE;
@@ -303,8 +367,11 @@ typedef struct drm_i915_private {
 	u32 saveDSPASURF;
 	u32 saveDSPATILEOFF;
 	u32 savePFIT_PGM_RATIOS;
+	u32 saveBLC_HIST_CTL;
 	u32 saveBLC_PWM_CTL;
 	u32 saveBLC_PWM_CTL2;
+	u32 saveBLC_CPU_PWM_CTL;
+	u32 saveBLC_CPU_PWM_CTL2;
 	u32 saveFPB0;
 	u32 saveFPB1;
 	u32 saveDPLL_B;
@@ -316,6 +383,13 @@ typedef struct drm_i915_private {
 	u32 saveVBLANK_B;
 	u32 saveVSYNC_B;
 	u32 saveBCLRPAT_B;
+	u32 saveTRANSBCONF;
+	u32 saveTRANS_HTOTAL_B;
+	u32 saveTRANS_HBLANK_B;
+	u32 saveTRANS_HSYNC_B;
+	u32 saveTRANS_VTOTAL_B;
+	u32 saveTRANS_VBLANK_B;
+	u32 saveTRANS_VSYNC_B;
 	u32 savePIPEBSTAT;
 	u32 saveDSPBSTRIDE;
 	u32 saveDSPBSIZE;
@@ -341,6 +415,7 @@ typedef struct drm_i915_private {
 	u32 savePFIT_CONTROL;
 	u32 save_palette_a[256];
 	u32 save_palette_b[256];
+	u32 saveDPFC_CB_BASE;
 	u32 saveFBC_CFB_BASE;
 	u32 saveFBC_LL_BASE;
 	u32 saveFBC_CONTROL;
@@ -348,9 +423,13 @@ typedef struct drm_i915_private {
 	u32 saveIER;
 	u32 saveIIR;
 	u32 saveIMR;
+	u32 saveDEIER;
+	u32 saveDEIMR;
+	u32 saveGTIER;
+	u32 saveGTIMR;
+	u32 saveFDI_RXA_IMR;
+	u32 saveFDI_RXB_IMR;
 	u32 saveCACHE_MODE_0;
-	u32 saveD_STATE;
-	u32 saveDSPCLK_GATE_D;
 	u32 saveMI_ARB_STATE;
 	u32 saveSWF0[16];
 	u32 saveSWF1[16];
@@ -381,6 +460,27 @@ typedef struct drm_i915_private {
 	u32 savePIPEB_DP_LINK_M;
 	u32 savePIPEA_DP_LINK_N;
 	u32 savePIPEB_DP_LINK_N;
+	u32 saveFDI_RXA_CTL;
+	u32 saveFDI_TXA_CTL;
+	u32 saveFDI_RXB_CTL;
+	u32 saveFDI_TXB_CTL;
+	u32 savePFA_CTL_1;
+	u32 savePFB_CTL_1;
+	u32 savePFA_WIN_SZ;
+	u32 savePFB_WIN_SZ;
+	u32 savePFA_WIN_POS;
+	u32 savePFB_WIN_POS;
+	u32 savePCH_DREF_CONTROL;
+	u32 saveDISP_ARB_CTL;
+	u32 savePIPEA_DATA_M1;
+	u32 savePIPEA_DATA_N1;
+	u32 savePIPEA_LINK_M1;
+	u32 savePIPEA_LINK_N1;
+	u32 savePIPEB_DATA_M1;
+	u32 savePIPEB_DATA_N1;
+	u32 savePIPEB_LINK_M1;
+	u32 savePIPEB_LINK_N1;
+	u32 saveMCHBAR_RENDER_STANDBY;
 
 	struct {
 		struct drm_mm gtt_space;
@@ -422,6 +522,15 @@ typedef struct drm_i915_private {
 		struct list_head flushing_list;
 
 		/**
+		 * List of objects currently pending a GPU write flush.
+		 *
+		 * All elements on this list will belong to either the
+		 * active_list or flushing_list, last_rendering_seqno can
+		 * be used to differentiate between the two elements.
+		 */
+		struct list_head gpu_write_list;
+
+		/**
 		 * LRU list of objects which are not in the ringbuffer and
 		 * are ready to unbind, but are still in the GTT.
 		 *
@@ -491,14 +600,33 @@ typedef struct drm_i915_private {
 		struct drm_i915_gem_phys_object *phys_objs[I915_MAX_PHYS_OBJECT];
 	} mm;
 	struct sdvo_device_mapping sdvo_mappings[2];
+	/* indicate whether the LVDS_BORDER should be enabled or not */
+	unsigned int lvds_border_bits;
+
+	struct drm_crtc *plane_to_crtc_mapping[2];
+	struct drm_crtc *pipe_to_crtc_mapping[2];
+	wait_queue_head_t pending_flip_queue;
 
 	/* Reclocking support */
 	bool render_reclock_avail;
 	bool lvds_downclock_avail;
+	/* indicates the reduced downclock for LVDS*/
+	int lvds_downclock;
 	struct work_struct idle_work;
 	struct timer_list idle_timer;
 	bool busy;
 	u16 orig_clock;
+	int child_dev_num;
+	struct child_device_config *child_dev;
+	struct drm_connector *int_lvds_connector;
+
+	bool mchbar_need_disable;
+
+	u8 cur_delay;
+	u8 min_delay;
+	u8 max_delay;
+
+	enum no_fbc_reason no_fbc_reason;
 } drm_i915_private_t;
 
 /** driver private structure attached to each drm_gem_object */
@@ -510,6 +638,8 @@ struct drm_i915_gem_object {
 
 	/** This object's place on the active/flushing/inactive lists */
 	struct list_head list;
+	/** This object's place on GPU write list */
+	struct list_head gpu_write_list;
 
 	/** This object's place on the fenced object LRU */
 	struct list_head fence_list;
@@ -591,6 +721,13 @@ struct drm_i915_gem_object {
 	 * Advice: are the backing pages purgeable?
 	 */
 	int madv;
+
+	/**
+	 * Number of crtcs where this object is currently the fb, but
+	 * will be page flipped away on the next vblank.  When it
+	 * reaches 0, dev_priv->pending_flip_queue will be woken up.
+	 */
+	atomic_t pending_flip;
 };
 
 /**
@@ -634,7 +771,10 @@ extern struct drm_ioctl_desc i915_ioctls[];
 extern int i915_max_ioctl;
 extern unsigned int i915_fbpercrtc;
 extern unsigned int i915_powersave;
+extern unsigned int i915_lvds_downclock;
 
+extern int i915_suspend(struct drm_device *dev, pm_message_t state);
+extern int i915_resume(struct drm_device *dev);
 extern void i915_save_display(struct drm_device *dev);
 extern void i915_restore_display(struct drm_device *dev);
 extern int i915_master_create(struct drm_device *dev, struct drm_master *master);
@@ -660,11 +800,13 @@ extern int i965_reset(struct drm_device *dev, u8 flags);
 
 /* i915_irq.c */
 void i915_hangcheck_elapsed(unsigned long data);
+void i915_destroy_error_state(struct drm_device *dev);
 extern int i915_irq_emit(struct drm_device *dev, void *data,
 			 struct drm_file *file_priv);
 extern int i915_irq_wait(struct drm_device *dev, void *data,
 			 struct drm_file *file_priv);
 void i915_user_irq_get(struct drm_device *dev);
+void i915_trace_irq_get(struct drm_device *dev, u32 seqno);
 void i915_user_irq_put(struct drm_device *dev);
 extern void i915_enable_interrupt (struct drm_device *dev);
 
@@ -690,6 +832,8 @@ i915_enable_pipestat(drm_i915_private_t *dev_priv, int pipe, u32 mask);
 void
 i915_disable_pipestat(drm_i915_private_t *dev_priv, int pipe, u32 mask);
 
+void intel_enable_asle (struct drm_device *dev);
+
 
 /* i915_mem.c */
 extern int i915_mem_alloc(struct drm_device *dev, void *data,
@@ -722,6 +866,8 @@ int i915_gem_sw_finish_ioctl(struct drm_device *dev, void *data,
 			     struct drm_file *file_priv);
 int i915_gem_execbuffer(struct drm_device *dev, void *data,
 			struct drm_file *file_priv);
+int i915_gem_execbuffer2(struct drm_device *dev, void *data,
+			 struct drm_file *file_priv);
 int i915_gem_pin_ioctl(struct drm_device *dev, void *data,
 		       struct drm_file *file_priv);
 int i915_gem_unpin_ioctl(struct drm_device *dev, void *data,
@@ -765,17 +911,22 @@ void i915_gem_cleanup_ringbuffer(struct drm_device *dev);
 int i915_gem_do_init(struct drm_device *dev, unsigned long start,
 		     unsigned long end);
 int i915_gem_idle(struct drm_device *dev);
+uint32_t i915_add_request(struct drm_device *dev, struct drm_file *file_priv,
+			  uint32_t flush_domains);
+int i915_do_wait_request(struct drm_device *dev, uint32_t seqno, int interruptible);
 int i915_gem_fault(struct vm_area_struct *vma, struct vm_fault *vmf);
 int i915_gem_object_set_to_gtt_domain(struct drm_gem_object *obj,
 				      int write);
+int i915_gem_object_set_to_display_plane(struct drm_gem_object *obj);
 int i915_gem_attach_phys_object(struct drm_device *dev,
 				struct drm_gem_object *obj, int id);
 void i915_gem_detach_phys_object(struct drm_device *dev,
 				 struct drm_gem_object *obj);
 void i915_gem_free_all_phys_object(struct drm_device *dev);
-int i915_gem_object_get_pages(struct drm_gem_object *obj);
+int i915_gem_object_get_pages(struct drm_gem_object *obj, gfp_t gfpmask);
 void i915_gem_object_put_pages(struct drm_gem_object *obj);
 void i915_gem_release(struct drm_device * dev, struct drm_file *file_priv);
+void i915_gem_object_flush_write_domain(struct drm_gem_object *obj);
 
 void i915_gem_shrinker_init(void);
 void i915_gem_shrinker_exit(void);
@@ -784,6 +935,10 @@ void i915_gem_shrinker_exit(void);
 void i915_gem_detect_bit_6_swizzle(struct drm_device *dev);
 void i915_gem_object_do_bit_17_swizzle(struct drm_gem_object *obj);
 void i915_gem_object_save_bit_17_swizzle(struct drm_gem_object *obj);
+bool i915_tiling_ok(struct drm_device *dev, int stride, int size,
+		    int tiling_mode);
+bool i915_gem_object_fence_offset_ok(struct drm_gem_object *obj,
+				     int tiling_mode);
 
 /* i915_gem_debug.c */
 void i915_gem_dump_object(struct drm_gem_object *obj, int len,
@@ -815,11 +970,13 @@ extern int i915_restore_state(struct drm_device *dev);
 extern int intel_opregion_init(struct drm_device *dev, int resume);
 extern void intel_opregion_free(struct drm_device *dev, int suspend);
 extern void opregion_asle_intr(struct drm_device *dev);
+extern void ironlake_opregion_gse_intr(struct drm_device *dev);
 extern void opregion_enable_asle(struct drm_device *dev);
 #else
 static inline int intel_opregion_init(struct drm_device *dev, int resume) { return 0; }
 static inline void intel_opregion_free(struct drm_device *dev, int suspend) { return; }
 static inline void opregion_asle_intr(struct drm_device *dev) { return; }
+static inline void ironlake_opregion_gse_intr(struct drm_device *dev) { return; }
 static inline void opregion_enable_asle(struct drm_device *dev) { return; }
 #endif
 
@@ -904,82 +1061,76 @@ extern void g4x_disable_fbc(struct drm_device *dev);
 extern int i915_wrap_ring(struct drm_device * dev);
 extern int i915_wait_ring(struct drm_device * dev, int n, const char *caller);
 
-#define IS_I830(dev) ((dev)->pci_device == 0x3577)
-#define IS_845G(dev) ((dev)->pci_device == 0x2562)
-#define IS_I85X(dev) ((dev)->pci_device == 0x3582)
-#define IS_I855(dev) ((dev)->pci_device == 0x3582)
-#define IS_I865G(dev) ((dev)->pci_device == 0x2572)
-
-#define IS_I915G(dev) ((dev)->pci_device == 0x2582 || (dev)->pci_device == 0x258a)
-#define IS_I915GM(dev) ((dev)->pci_device == 0x2592)
-#define IS_I945G(dev) ((dev)->pci_device == 0x2772)
-#define IS_I945GM(dev) ((dev)->pci_device == 0x27A2 ||\
-		        (dev)->pci_device == 0x27AE)
-#define IS_I965G(dev) ((dev)->pci_device == 0x2972 || \
-		       (dev)->pci_device == 0x2982 || \
-		       (dev)->pci_device == 0x2992 || \
-		       (dev)->pci_device == 0x29A2 || \
-		       (dev)->pci_device == 0x2A02 || \
-		       (dev)->pci_device == 0x2A12 || \
-		       (dev)->pci_device == 0x2A42 || \
-		       (dev)->pci_device == 0x2E02 || \
-		       (dev)->pci_device == 0x2E12 || \
-		       (dev)->pci_device == 0x2E22 || \
-		       (dev)->pci_device == 0x2E32 || \
-		       (dev)->pci_device == 0x2E42 || \
-		       (dev)->pci_device == 0x0042 || \
-		       (dev)->pci_device == 0x0046)
-
-#define IS_I965GM(dev) ((dev)->pci_device == 0x2A02 || \
-			(dev)->pci_device == 0x2A12)
-
-#define IS_GM45(dev) ((dev)->pci_device == 0x2A42)
-
-#define IS_G4X(dev) ((dev)->pci_device == 0x2E02 || \
-		     (dev)->pci_device == 0x2E12 || \
-		     (dev)->pci_device == 0x2E22 || \
-		     (dev)->pci_device == 0x2E32 || \
-		     (dev)->pci_device == 0x2E42 || \
-		     IS_GM45(dev))
-
-#define IS_IGDG(dev) ((dev)->pci_device == 0xa001)
-#define IS_IGDGM(dev) ((dev)->pci_device == 0xa011)
-#define IS_IGD(dev) (IS_IGDG(dev) || IS_IGDGM(dev))
-
-#define IS_G33(dev)    ((dev)->pci_device == 0x29C2 ||	\
-			(dev)->pci_device == 0x29B2 ||	\
-			(dev)->pci_device == 0x29D2 ||  \
-			(IS_IGD(dev)))
-
-#define IS_IGDNG_D(dev)	((dev)->pci_device == 0x0042)
-#define IS_IGDNG_M(dev)	((dev)->pci_device == 0x0046)
-#define IS_IGDNG(dev)	(IS_IGDNG_D(dev) || IS_IGDNG_M(dev))
-
-#define IS_I9XX(dev) (IS_I915G(dev) || IS_I915GM(dev) || IS_I945G(dev) || \
-		      IS_I945GM(dev) || IS_I965G(dev) || IS_G33(dev) || \
-		      IS_IGDNG(dev))
-
-#define IS_MOBILE(dev) (IS_I830(dev) || IS_I85X(dev) || IS_I915GM(dev) || \
-			IS_I945GM(dev) || IS_I965GM(dev) || IS_GM45(dev) || \
-			IS_IGD(dev) || IS_IGDNG_M(dev))
-
-#define I915_NEED_GFX_HWS(dev) (IS_G33(dev) || IS_GM45(dev) || IS_G4X(dev) || \
-				IS_IGDNG(dev))
+#define INTEL_INFO(dev)	(((struct drm_i915_private *) (dev)->dev_private)->info)
+
+#define IS_I830(dev)		((dev)->pci_device == 0x3577)
+#define IS_845G(dev)		((dev)->pci_device == 0x2562)
+#define IS_I85X(dev)		((dev)->pci_device == 0x3582)
+#define IS_I865G(dev)		((dev)->pci_device == 0x2572)
+#define IS_GEN2(dev)		(INTEL_INFO(dev)->is_i8xx)
+#define IS_I915G(dev)		(INTEL_INFO(dev)->is_i915g)
+#define IS_I915GM(dev)		((dev)->pci_device == 0x2592)
+#define IS_I945G(dev)		((dev)->pci_device == 0x2772)
+#define IS_I945GM(dev)		(INTEL_INFO(dev)->is_i945gm)
+#define IS_I965G(dev)		(INTEL_INFO(dev)->is_i965g)
+#define IS_I965GM(dev)		(INTEL_INFO(dev)->is_i965gm)
+#define IS_GM45(dev)		((dev)->pci_device == 0x2A42)
+#define IS_G4X(dev)		(INTEL_INFO(dev)->is_g4x)
+#define IS_PINEVIEW_G(dev)	((dev)->pci_device == 0xa001)
+#define IS_PINEVIEW_M(dev)	((dev)->pci_device == 0xa011)
+#define IS_PINEVIEW(dev)	(INTEL_INFO(dev)->is_pineview)
+#define IS_G33(dev)		(INTEL_INFO(dev)->is_g33)
+#define IS_IRONLAKE_D(dev)	((dev)->pci_device == 0x0042)
+#define IS_IRONLAKE_M(dev)	((dev)->pci_device == 0x0046)
+#define IS_IRONLAKE(dev)	(INTEL_INFO(dev)->is_ironlake)
+#define IS_I9XX(dev)		(INTEL_INFO(dev)->is_i9xx)
+#define IS_MOBILE(dev)		(INTEL_INFO(dev)->is_mobile)
+
+#define IS_GEN3(dev)	(IS_I915G(dev) ||			\
+			 IS_I915GM(dev) ||			\
+			 IS_I945G(dev) ||			\
+			 IS_I945GM(dev) ||			\
+			 IS_G33(dev) || \
+			 IS_PINEVIEW(dev))
+#define IS_GEN4(dev)	((dev)->pci_device == 0x2972 ||		\
+			 (dev)->pci_device == 0x2982 ||		\
+			 (dev)->pci_device == 0x2992 ||		\
+			 (dev)->pci_device == 0x29A2 ||		\
+			 (dev)->pci_device == 0x2A02 ||		\
+			 (dev)->pci_device == 0x2A12 ||		\
+			 (dev)->pci_device == 0x2E02 ||		\
+			 (dev)->pci_device == 0x2E12 ||		\
+			 (dev)->pci_device == 0x2E22 ||		\
+			 (dev)->pci_device == 0x2E32 ||		\
+			 (dev)->pci_device == 0x2A42 ||		\
+			 (dev)->pci_device == 0x2E42)
+
+#define I915_NEED_GFX_HWS(dev)	(INTEL_INFO(dev)->need_gfx_hws)
+
+#define IS_GEN6(dev)	((dev)->pci_device == 0x0102)
+
 /* With the 945 and later, Y tiling got adjusted so that it was 32 128-byte
  * rows, which changed the alignment requirements and fence programming.
  */
 #define HAS_128_BYTE_Y_TILING(dev) (IS_I9XX(dev) && !(IS_I915G(dev) || \
 						      IS_I915GM(dev)))
-#define SUPPORTS_INTEGRATED_HDMI(dev)	(IS_G4X(dev) || IS_IGDNG(dev))
-#define SUPPORTS_INTEGRATED_DP(dev)	(IS_G4X(dev) || IS_IGDNG(dev))
-#define SUPPORTS_EDP(dev)		(IS_IGDNG_M(dev))
-#define I915_HAS_HOTPLUG(dev) (IS_I945G(dev) || IS_I945GM(dev) || IS_G33(dev) || IS_I965G(dev))
+#define SUPPORTS_DIGITAL_OUTPUTS(dev)	(IS_I9XX(dev) && !IS_PINEVIEW(dev))
+#define SUPPORTS_INTEGRATED_HDMI(dev)	(IS_G4X(dev) || IS_IRONLAKE(dev))
+#define SUPPORTS_INTEGRATED_DP(dev)	(IS_G4X(dev) || IS_IRONLAKE(dev))
+#define SUPPORTS_EDP(dev)		(IS_IRONLAKE_M(dev))
+#define SUPPORTS_TV(dev)		(IS_I9XX(dev) && IS_MOBILE(dev) && \
+					!IS_IRONLAKE(dev) && !IS_PINEVIEW(dev))
+#define I915_HAS_HOTPLUG(dev)		 (INTEL_INFO(dev)->has_hotplug)
 /* dsparb controlled by hw only */
-#define DSPARB_HWCONTROL(dev) (IS_G4X(dev) || IS_IGDNG(dev))
+#define DSPARB_HWCONTROL(dev) (IS_G4X(dev) || IS_IRONLAKE(dev))
+
+#define HAS_FW_BLC(dev) (IS_I9XX(dev) || IS_G4X(dev) || IS_IRONLAKE(dev))
+#define HAS_PIPE_CXSR(dev) (INTEL_INFO(dev)->has_pipe_cxsr)
+#define I915_HAS_FBC(dev) (INTEL_INFO(dev)->has_fbc)
+#define I915_HAS_RC6(dev) (INTEL_INFO(dev)->has_rc6)
 
-#define HAS_FW_BLC(dev) (IS_I9XX(dev) || IS_G4X(dev) || IS_IGDNG(dev))
-#define HAS_PIPE_CXSR(dev) (IS_G4X(dev) || IS_IGDNG(dev))
-#define I915_HAS_FBC(dev) (IS_MOBILE(dev) && (IS_I9XX(dev) || IS_I965G(dev)))
+#define HAS_PCH_SPLIT(dev) (IS_IRONLAKE(dev) ||	\
+			    IS_GEN6(dev))
 
 #define PRIMARY_RINGBUFFER_SIZE         (128*1024)
 
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 40727d4c2919..fba37e9f775d 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -128,9 +128,7 @@ i915_gem_create_ioctl(struct drm_device *dev, void *data,
 		return -ENOMEM;
 
 	ret = drm_gem_handle_create(file_priv, obj, &handle);
-	mutex_lock(&dev->struct_mutex);
-	drm_gem_object_handle_unreference(obj);
-	mutex_unlock(&dev->struct_mutex);
+	drm_gem_object_handle_unreference_unlocked(obj);
 
 	if (ret)
 		return ret;
@@ -277,7 +275,7 @@ i915_gem_shmem_pread_fast(struct drm_device *dev, struct drm_gem_object *obj,
 
 	mutex_lock(&dev->struct_mutex);
 
-	ret = i915_gem_object_get_pages(obj);
+	ret = i915_gem_object_get_pages(obj, 0);
 	if (ret != 0)
 		goto fail_unlock;
 
@@ -321,40 +319,24 @@ fail_unlock:
 	return ret;
 }
 
-static inline gfp_t
-i915_gem_object_get_page_gfp_mask (struct drm_gem_object *obj)
-{
-	return mapping_gfp_mask(obj->filp->f_path.dentry->d_inode->i_mapping);
-}
-
-static inline void
-i915_gem_object_set_page_gfp_mask (struct drm_gem_object *obj, gfp_t gfp)
-{
-	mapping_set_gfp_mask(obj->filp->f_path.dentry->d_inode->i_mapping, gfp);
-}
-
 static int
 i915_gem_object_get_pages_or_evict(struct drm_gem_object *obj)
 {
 	int ret;
 
-	ret = i915_gem_object_get_pages(obj);
+	ret = i915_gem_object_get_pages(obj, __GFP_NORETRY | __GFP_NOWARN);
 
 	/* If we've insufficient memory to map in the pages, attempt
 	 * to make some space by throwing out some old buffers.
 	 */
 	if (ret == -ENOMEM) {
 		struct drm_device *dev = obj->dev;
-		gfp_t gfp;
 
 		ret = i915_gem_evict_something(dev, obj->size);
 		if (ret)
 			return ret;
 
-		gfp = i915_gem_object_get_page_gfp_mask(obj);
-		i915_gem_object_set_page_gfp_mask(obj, gfp & ~__GFP_NORETRY);
-		ret = i915_gem_object_get_pages(obj);
-		i915_gem_object_set_page_gfp_mask (obj, gfp);
+		ret = i915_gem_object_get_pages(obj, 0);
 	}
 
 	return ret;
@@ -504,7 +486,7 @@ i915_gem_pread_ioctl(struct drm_device *dev, void *data,
 	 */
 	if (args->offset > obj->size || args->size > obj->size ||
 	    args->offset + args->size > obj->size) {
-		drm_gem_object_unreference(obj);
+		drm_gem_object_unreference_unlocked(obj);
 		return -EINVAL;
 	}
 
@@ -517,7 +499,7 @@ i915_gem_pread_ioctl(struct drm_device *dev, void *data,
 							file_priv);
 	}
 
-	drm_gem_object_unreference(obj);
+	drm_gem_object_unreference_unlocked(obj);
 
 	return ret;
 }
@@ -790,7 +772,7 @@ i915_gem_shmem_pwrite_fast(struct drm_device *dev, struct drm_gem_object *obj,
 
 	mutex_lock(&dev->struct_mutex);
 
-	ret = i915_gem_object_get_pages(obj);
+	ret = i915_gem_object_get_pages(obj, 0);
 	if (ret != 0)
 		goto fail_unlock;
 
@@ -977,7 +959,7 @@ i915_gem_pwrite_ioctl(struct drm_device *dev, void *data,
 	 */
 	if (args->offset > obj->size || args->size > obj->size ||
 	    args->offset + args->size > obj->size) {
-		drm_gem_object_unreference(obj);
+		drm_gem_object_unreference_unlocked(obj);
 		return -EINVAL;
 	}
 
@@ -1011,7 +993,7 @@ i915_gem_pwrite_ioctl(struct drm_device *dev, void *data,
 		DRM_INFO("pwrite failed %d\n", ret);
 #endif
 
-	drm_gem_object_unreference(obj);
+	drm_gem_object_unreference_unlocked(obj);
 
 	return ret;
 }
@@ -1154,9 +1136,7 @@ i915_gem_mmap_ioctl(struct drm_device *dev, void *data,
 		       PROT_READ | PROT_WRITE, MAP_SHARED,
 		       args->offset);
 	up_write(&current->mm->mmap_sem);
-	mutex_lock(&dev->struct_mutex);
-	drm_gem_object_unreference(obj);
-	mutex_unlock(&dev->struct_mutex);
+	drm_gem_object_unreference_unlocked(obj);
 	if (IS_ERR((void *)addr))
 		return addr;
 
@@ -1288,6 +1268,7 @@ i915_gem_create_mmap_offset(struct drm_gem_object *obj)
 	list->hash.key = list->file_offset_node->start;
 	if (drm_ht_insert_item(&mm->offset_hash, &list->hash)) {
 		DRM_ERROR("failed to add to map hash\n");
+		ret = -ENOMEM;
 		goto out_free_mm;
 	}
 
@@ -1309,7 +1290,7 @@ out_free_list:
  * i915_gem_release_mmap - remove physical page mappings
  * @obj: obj in question
  *
- * Preserve the reservation of the mmaping with the DRM core code, but
+ * Preserve the reservation of the mmapping with the DRM core code, but
  * relinquish ownership of the pages back to the system.
  *
  * It is vital that we remove the page mapping if we have mapped a tiled
@@ -1567,6 +1548,8 @@ i915_gem_object_move_to_inactive(struct drm_gem_object *obj)
 	else
 		list_move_tail(&obj_priv->list, &dev_priv->mm.inactive_list);
 
+	BUG_ON(!list_empty(&obj_priv->gpu_write_list));
+
 	obj_priv->last_rendering_seqno = 0;
 	if (obj_priv->active) {
 		obj_priv->active = 0;
@@ -1575,6 +1558,38 @@ i915_gem_object_move_to_inactive(struct drm_gem_object *obj)
 	i915_verify_inactive(dev, __FILE__, __LINE__);
 }
 
+static void
+i915_gem_process_flushing_list(struct drm_device *dev,
+			       uint32_t flush_domains, uint32_t seqno)
+{
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	struct drm_i915_gem_object *obj_priv, *next;
+
+	list_for_each_entry_safe(obj_priv, next,
+				 &dev_priv->mm.gpu_write_list,
+				 gpu_write_list) {
+		struct drm_gem_object *obj = obj_priv->obj;
+
+		if ((obj->write_domain & flush_domains) ==
+		    obj->write_domain) {
+			uint32_t old_write_domain = obj->write_domain;
+
+			obj->write_domain = 0;
+			list_del_init(&obj_priv->gpu_write_list);
+			i915_gem_object_move_to_active(obj, seqno);
+
+			/* update the fence lru list */
+			if (obj_priv->fence_reg != I915_FENCE_REG_NONE)
+				list_move_tail(&obj_priv->fence_list,
+						&dev_priv->mm.fence_list);
+
+			trace_i915_gem_object_change_domain(obj,
+							    obj->read_domains,
+							    old_write_domain);
+		}
+	}
+}
+
 /**
  * Creates a new sequence number, emitting a write of it to the status page
  * plus an interrupt, which will trigger i915_user_interrupt_handler.
@@ -1583,7 +1598,7 @@ i915_gem_object_move_to_inactive(struct drm_gem_object *obj)
  *
  * Returned sequence numbers are nonzero on success.
  */
-static uint32_t
+uint32_t
 i915_add_request(struct drm_device *dev, struct drm_file *file_priv,
 		 uint32_t flush_domains)
 {
@@ -1617,7 +1632,7 @@ i915_add_request(struct drm_device *dev, struct drm_file *file_priv,
 	OUT_RING(MI_USER_INTERRUPT);
 	ADVANCE_LP_RING();
 
-	DRM_DEBUG("%d\n", seqno);
+	DRM_DEBUG_DRIVER("%d\n", seqno);
 
 	request->seqno = seqno;
 	request->emitted_jiffies = jiffies;
@@ -1633,27 +1648,8 @@ i915_add_request(struct drm_device *dev, struct drm_file *file_priv,
 	/* Associate any objects on the flushing list matching the write
 	 * domain we're flushing with our flush.
 	 */
-	if (flush_domains != 0) {
-		struct drm_i915_gem_object *obj_priv, *next;
-
-		list_for_each_entry_safe(obj_priv, next,
-					 &dev_priv->mm.flushing_list, list) {
-			struct drm_gem_object *obj = obj_priv->obj;
-
-			if ((obj->write_domain & flush_domains) ==
-			    obj->write_domain) {
-				uint32_t old_write_domain = obj->write_domain;
-
-				obj->write_domain = 0;
-				i915_gem_object_move_to_active(obj, seqno);
-
-				trace_i915_gem_object_change_domain(obj,
-								    obj->read_domains,
-								    old_write_domain);
-			}
-		}
-
-	}
+	if (flush_domains != 0) 
+		i915_gem_process_flushing_list(dev, flush_domains, seqno);
 
 	if (!dev_priv->mm.suspended) {
 		mod_timer(&dev_priv->hangcheck_timer, jiffies + DRM_I915_HANGCHECK_PERIOD);
@@ -1770,7 +1766,7 @@ i915_gem_retire_requests(struct drm_device *dev)
 	drm_i915_private_t *dev_priv = dev->dev_private;
 	uint32_t seqno;
 
-	if (!dev_priv->hw_status_page)
+	if (!dev_priv->hw_status_page || list_empty(&dev_priv->mm.request_list))
 		return;
 
 	seqno = i915_get_gem_seqno(dev);
@@ -1794,6 +1790,12 @@ i915_gem_retire_requests(struct drm_device *dev)
 		} else
 			break;
 	}
+
+	if (unlikely (dev_priv->trace_irq_seqno &&
+		      i915_seqno_passed(dev_priv->trace_irq_seqno, seqno))) {
+		i915_user_irq_put(dev);
+		dev_priv->trace_irq_seqno = 0;
+	}
 }
 
 void
@@ -1814,12 +1816,8 @@ i915_gem_retire_work_handler(struct work_struct *work)
 	mutex_unlock(&dev->struct_mutex);
 }
 
-/**
- * Waits for a sequence number to be signaled, and cleans up the
- * request and object lists appropriately for that event.
- */
-static int
-i915_wait_request(struct drm_device *dev, uint32_t seqno)
+int
+i915_do_wait_request(struct drm_device *dev, uint32_t seqno, int interruptible)
 {
 	drm_i915_private_t *dev_priv = dev->dev_private;
 	u32 ier;
@@ -1831,7 +1829,7 @@ i915_wait_request(struct drm_device *dev, uint32_t seqno)
 		return -EIO;
 
 	if (!i915_seqno_passed(i915_get_gem_seqno(dev), seqno)) {
-		if (IS_IGDNG(dev))
+		if (HAS_PCH_SPLIT(dev))
 			ier = I915_READ(DEIER) | I915_READ(GTIER);
 		else
 			ier = I915_READ(IER);
@@ -1846,10 +1844,15 @@ i915_wait_request(struct drm_device *dev, uint32_t seqno)
 
 		dev_priv->mm.waiting_gem_seqno = seqno;
 		i915_user_irq_get(dev);
-		ret = wait_event_interruptible(dev_priv->irq_queue,
-					       i915_seqno_passed(i915_get_gem_seqno(dev),
-								 seqno) ||
-					       atomic_read(&dev_priv->mm.wedged));
+		if (interruptible)
+			ret = wait_event_interruptible(dev_priv->irq_queue,
+				i915_seqno_passed(i915_get_gem_seqno(dev), seqno) ||
+				atomic_read(&dev_priv->mm.wedged));
+		else
+			wait_event(dev_priv->irq_queue,
+				i915_seqno_passed(i915_get_gem_seqno(dev), seqno) ||
+				atomic_read(&dev_priv->mm.wedged));
+
 		i915_user_irq_put(dev);
 		dev_priv->mm.waiting_gem_seqno = 0;
 
@@ -1873,6 +1876,16 @@ i915_wait_request(struct drm_device *dev, uint32_t seqno)
 	return ret;
 }
 
+/**
+ * Waits for a sequence number to be signaled, and cleans up the
+ * request and object lists appropriately for that event.
+ */
+static int
+i915_wait_request(struct drm_device *dev, uint32_t seqno)
+{
+	return i915_do_wait_request(dev, seqno, 1);
+}
+
 static void
 i915_gem_flush(struct drm_device *dev,
 	       uint32_t invalidate_domains,
@@ -1941,7 +1954,7 @@ i915_gem_flush(struct drm_device *dev,
 #endif
 		BEGIN_LP_RING(2);
 		OUT_RING(cmd);
-		OUT_RING(0); /* noop */
+		OUT_RING(MI_NOOP);
 		ADVANCE_LP_RING();
 	}
 }
@@ -1985,6 +1998,7 @@ int
 i915_gem_object_unbind(struct drm_gem_object *obj)
 {
 	struct drm_device *dev = obj->dev;
+	drm_i915_private_t *dev_priv = dev->dev_private;
 	struct drm_i915_gem_object *obj_priv = obj->driver_private;
 	int ret = 0;
 
@@ -2003,9 +2017,6 @@ i915_gem_object_unbind(struct drm_gem_object *obj)
 	/* blow away mappings if mapped through GTT */
 	i915_gem_release_mmap(obj);
 
-	if (obj_priv->fence_reg != I915_FENCE_REG_NONE)
-		i915_gem_clear_fence_reg(obj);
-
 	/* Move the object to the CPU domain to ensure that
 	 * any possible CPU writes while it's not in the GTT
 	 * are flushed when we go to remap it. This will
@@ -2021,6 +2032,10 @@ i915_gem_object_unbind(struct drm_gem_object *obj)
 
 	BUG_ON(obj_priv->active);
 
+	/* release the fence reg _after_ flushing */
+	if (obj_priv->fence_reg != I915_FENCE_REG_NONE)
+		i915_gem_clear_fence_reg(obj);
+
 	if (obj_priv->agp_mem != NULL) {
 		drm_unbind_agp(obj_priv->agp_mem);
 		drm_free_agp(obj_priv->agp_mem, obj->size / PAGE_SIZE);
@@ -2039,8 +2054,10 @@ i915_gem_object_unbind(struct drm_gem_object *obj)
 	}
 
 	/* Remove ourselves from the LRU list if present. */
+	spin_lock(&dev_priv->mm.active_list_lock);
 	if (!list_empty(&obj_priv->list))
 		list_del_init(&obj_priv->list);
+	spin_unlock(&dev_priv->mm.active_list_lock);
 
 	if (i915_gem_object_is_purgeable(obj_priv))
 		i915_gem_object_truncate(obj);
@@ -2078,10 +2095,33 @@ i915_gem_find_inactive_object(struct drm_device *dev, int min_size)
 }
 
 static int
-i915_gem_evict_everything(struct drm_device *dev)
+i915_gpu_idle(struct drm_device *dev)
 {
 	drm_i915_private_t *dev_priv = dev->dev_private;
+	bool lists_empty;
 	uint32_t seqno;
+
+	spin_lock(&dev_priv->mm.active_list_lock);
+	lists_empty = list_empty(&dev_priv->mm.flushing_list) &&
+		      list_empty(&dev_priv->mm.active_list);
+	spin_unlock(&dev_priv->mm.active_list_lock);
+
+	if (lists_empty)
+		return 0;
+
+	/* Flush everything onto the inactive list. */
+	i915_gem_flush(dev, I915_GEM_GPU_DOMAINS, I915_GEM_GPU_DOMAINS);
+	seqno = i915_add_request(dev, NULL, I915_GEM_GPU_DOMAINS);
+	if (seqno == 0)
+		return -ENOMEM;
+
+	return i915_wait_request(dev, seqno);
+}
+
+static int
+i915_gem_evict_everything(struct drm_device *dev)
+{
+	drm_i915_private_t *dev_priv = dev->dev_private;
 	int ret;
 	bool lists_empty;
 
@@ -2095,15 +2135,12 @@ i915_gem_evict_everything(struct drm_device *dev)
 		return -ENOSPC;
 
 	/* Flush everything (on to the inactive lists) and evict */
-	i915_gem_flush(dev, I915_GEM_GPU_DOMAINS, I915_GEM_GPU_DOMAINS);
-	seqno = i915_add_request(dev, NULL, I915_GEM_GPU_DOMAINS);
-	if (seqno == 0)
-		return -ENOMEM;
-
-	ret = i915_wait_request(dev, seqno);
+	ret = i915_gpu_idle(dev);
 	if (ret)
 		return ret;
 
+	BUG_ON(!list_empty(&dev_priv->mm.flushing_list));
+
 	ret = i915_gem_evict_from_inactive_list(dev);
 	if (ret)
 		return ret;
@@ -2211,7 +2248,8 @@ i915_gem_evict_something(struct drm_device *dev, int min_size)
 }
 
 int
-i915_gem_object_get_pages(struct drm_gem_object *obj)
+i915_gem_object_get_pages(struct drm_gem_object *obj,
+			  gfp_t gfpmask)
 {
 	struct drm_i915_gem_object *obj_priv = obj->driver_private;
 	int page_count, i;
@@ -2237,7 +2275,10 @@ i915_gem_object_get_pages(struct drm_gem_object *obj)
 	inode = obj->filp->f_path.dentry->d_inode;
 	mapping = inode->i_mapping;
 	for (i = 0; i < page_count; i++) {
-		page = read_mapping_page(mapping, i, NULL);
+		page = read_cache_page_gfp(mapping, i,
+					   mapping_gfp_mask (mapping) |
+					   __GFP_COLD |
+					   gfpmask);
 		if (IS_ERR(page)) {
 			ret = PTR_ERR(page);
 			i915_gem_object_put_pages(obj);
@@ -2252,6 +2293,28 @@ i915_gem_object_get_pages(struct drm_gem_object *obj)
 	return 0;
 }
 
+static void sandybridge_write_fence_reg(struct drm_i915_fence_reg *reg)
+{
+	struct drm_gem_object *obj = reg->obj;
+	struct drm_device *dev = obj->dev;
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	struct drm_i915_gem_object *obj_priv = obj->driver_private;
+	int regnum = obj_priv->fence_reg;
+	uint64_t val;
+
+	val = (uint64_t)((obj_priv->gtt_offset + obj->size - 4096) &
+		    0xfffff000) << 32;
+	val |= obj_priv->gtt_offset & 0xfffff000;
+	val |= (uint64_t)((obj_priv->stride / 128) - 1) <<
+		SANDYBRIDGE_FENCE_PITCH_SHIFT;
+
+	if (obj_priv->tiling_mode == I915_TILING_Y)
+		val |= 1 << I965_FENCE_TILING_Y_SHIFT;
+	val |= I965_FENCE_REG_VALID;
+
+	I915_WRITE64(FENCE_REG_SANDYBRIDGE_0 + (regnum * 8), val);
+}
+
 static void i965_write_fence_reg(struct drm_i915_fence_reg *reg)
 {
 	struct drm_gem_object *obj = reg->obj;
@@ -2348,6 +2411,58 @@ static void i830_write_fence_reg(struct drm_i915_fence_reg *reg)
 	I915_WRITE(FENCE_REG_830_0 + (regnum * 4), val);
 }
 
+static int i915_find_fence_reg(struct drm_device *dev)
+{
+	struct drm_i915_fence_reg *reg = NULL;
+	struct drm_i915_gem_object *obj_priv = NULL;
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct drm_gem_object *obj = NULL;
+	int i, avail, ret;
+
+	/* First try to find a free reg */
+	avail = 0;
+	for (i = dev_priv->fence_reg_start; i < dev_priv->num_fence_regs; i++) {
+		reg = &dev_priv->fence_regs[i];
+		if (!reg->obj)
+			return i;
+
+		obj_priv = reg->obj->driver_private;
+		if (!obj_priv->pin_count)
+		    avail++;
+	}
+
+	if (avail == 0)
+		return -ENOSPC;
+
+	/* None available, try to steal one or wait for a user to finish */
+	i = I915_FENCE_REG_NONE;
+	list_for_each_entry(obj_priv, &dev_priv->mm.fence_list,
+			    fence_list) {
+		obj = obj_priv->obj;
+
+		if (obj_priv->pin_count)
+			continue;
+
+		/* found one! */
+		i = obj_priv->fence_reg;
+		break;
+	}
+
+	BUG_ON(i == I915_FENCE_REG_NONE);
+
+	/* We only have a reference on obj from the active list. put_fence_reg
+	 * might drop that one, causing a use-after-free in it. So hold a
+	 * private reference to obj like the other callers of put_fence_reg
+	 * (set_tiling ioctl) do. */
+	drm_gem_object_reference(obj);
+	ret = i915_gem_object_put_fence_reg(obj);
+	drm_gem_object_unreference(obj);
+	if (ret != 0)
+		return ret;
+
+	return i;
+}
+
 /**
  * i915_gem_object_get_fence_reg - set up a fence reg for an object
  * @obj: object to map through a fence reg
@@ -2368,8 +2483,7 @@ i915_gem_object_get_fence_reg(struct drm_gem_object *obj)
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct drm_i915_gem_object *obj_priv = obj->driver_private;
 	struct drm_i915_fence_reg *reg = NULL;
-	struct drm_i915_gem_object *old_obj_priv = NULL;
-	int i, ret, avail;
+	int ret;
 
 	/* Just update our place in the LRU if our fence is getting used. */
 	if (obj_priv->fence_reg != I915_FENCE_REG_NONE) {
@@ -2397,86 +2511,27 @@ i915_gem_object_get_fence_reg(struct drm_gem_object *obj)
 		break;
 	}
 
-	/* First try to find a free reg */
-	avail = 0;
-	for (i = dev_priv->fence_reg_start; i < dev_priv->num_fence_regs; i++) {
-		reg = &dev_priv->fence_regs[i];
-		if (!reg->obj)
-			break;
-
-		old_obj_priv = reg->obj->driver_private;
-		if (!old_obj_priv->pin_count)
-		    avail++;
-	}
-
-	/* None available, try to steal one or wait for a user to finish */
-	if (i == dev_priv->num_fence_regs) {
-		struct drm_gem_object *old_obj = NULL;
-
-		if (avail == 0)
-			return -ENOSPC;
-
-		list_for_each_entry(old_obj_priv, &dev_priv->mm.fence_list,
-				    fence_list) {
-			old_obj = old_obj_priv->obj;
-
-			if (old_obj_priv->pin_count)
-				continue;
-
-			/* Take a reference, as otherwise the wait_rendering
-			 * below may cause the object to get freed out from
-			 * under us.
-			 */
-			drm_gem_object_reference(old_obj);
-
-			/* i915 uses fences for GPU access to tiled buffers */
-			if (IS_I965G(dev) || !old_obj_priv->active)
-				break;
-
-			/* This brings the object to the head of the LRU if it
-			 * had been written to.  The only way this should
-			 * result in us waiting longer than the expected
-			 * optimal amount of time is if there was a
-			 * fence-using buffer later that was read-only.
-			 */
-			i915_gem_object_flush_gpu_write_domain(old_obj);
-			ret = i915_gem_object_wait_rendering(old_obj);
-			if (ret != 0) {
-				drm_gem_object_unreference(old_obj);
-				return ret;
-			}
-
-			break;
-		}
-
-		/*
-		 * Zap this virtual mapping so we can set up a fence again
-		 * for this object next time we need it.
-		 */
-		i915_gem_release_mmap(old_obj);
-
-		i = old_obj_priv->fence_reg;
-		reg = &dev_priv->fence_regs[i];
-
-		old_obj_priv->fence_reg = I915_FENCE_REG_NONE;
-		list_del_init(&old_obj_priv->fence_list);
-
-		drm_gem_object_unreference(old_obj);
-	}
+	ret = i915_find_fence_reg(dev);
+	if (ret < 0)
+		return ret;
 
-	obj_priv->fence_reg = i;
+	obj_priv->fence_reg = ret;
+	reg = &dev_priv->fence_regs[obj_priv->fence_reg];
 	list_add_tail(&obj_priv->fence_list, &dev_priv->mm.fence_list);
 
 	reg->obj = obj;
 
-	if (IS_I965G(dev))
+	if (IS_GEN6(dev))
+		sandybridge_write_fence_reg(reg);
+	else if (IS_I965G(dev))
 		i965_write_fence_reg(reg);
 	else if (IS_I9XX(dev))
 		i915_write_fence_reg(reg);
 	else
 		i830_write_fence_reg(reg);
 
-	trace_i915_gem_object_get_fence(obj, i, obj_priv->tiling_mode);
+	trace_i915_gem_object_get_fence(obj, obj_priv->fence_reg,
+			obj_priv->tiling_mode);
 
 	return 0;
 }
@@ -2495,9 +2550,12 @@ i915_gem_clear_fence_reg(struct drm_gem_object *obj)
 	drm_i915_private_t *dev_priv = dev->dev_private;
 	struct drm_i915_gem_object *obj_priv = obj->driver_private;
 
-	if (IS_I965G(dev))
+	if (IS_GEN6(dev)) {
+		I915_WRITE64(FENCE_REG_SANDYBRIDGE_0 +
+			     (obj_priv->fence_reg * 8), 0);
+	} else if (IS_I965G(dev)) {
 		I915_WRITE64(FENCE_REG_965_0 + (obj_priv->fence_reg * 8), 0);
-	else {
+	} else {
 		uint32_t fence_reg;
 
 		if (obj_priv->fence_reg < 8)
@@ -2531,6 +2589,12 @@ i915_gem_object_put_fence_reg(struct drm_gem_object *obj)
 	if (obj_priv->fence_reg == I915_FENCE_REG_NONE)
 		return 0;
 
+	/* If we've changed tiling, GTT-mappings of the object
+	 * need to re-fault to ensure that the correct fence register
+	 * setup is in place.
+	 */
+	i915_gem_release_mmap(obj);
+
 	/* On the i915, GPU access to tiled buffers is via a fence,
 	 * therefore we must wait for any outstanding access to complete
 	 * before clearing the fence.
@@ -2539,12 +2603,12 @@ i915_gem_object_put_fence_reg(struct drm_gem_object *obj)
 		int ret;
 
 		i915_gem_object_flush_gpu_write_domain(obj);
-		i915_gem_object_flush_gtt_write_domain(obj);
 		ret = i915_gem_object_wait_rendering(obj);
 		if (ret != 0)
 			return ret;
 	}
 
+	i915_gem_object_flush_gtt_write_domain(obj);
 	i915_gem_clear_fence_reg (obj);
 
 	return 0;
@@ -2560,12 +2624,9 @@ i915_gem_object_bind_to_gtt(struct drm_gem_object *obj, unsigned alignment)
 	drm_i915_private_t *dev_priv = dev->dev_private;
 	struct drm_i915_gem_object *obj_priv = obj->driver_private;
 	struct drm_mm_node *free_space;
-	bool retry_alloc = false;
+	gfp_t gfpmask =  __GFP_NORETRY | __GFP_NOWARN;
 	int ret;
 
-	if (dev_priv->mm.suspended)
-		return -EBUSY;
-
 	if (obj_priv->madv != I915_MADV_WILLNEED) {
 		DRM_ERROR("Attempting to bind a purgeable object\n");
 		return -EINVAL;
@@ -2607,15 +2668,7 @@ i915_gem_object_bind_to_gtt(struct drm_gem_object *obj, unsigned alignment)
 	DRM_INFO("Binding object of size %zd at 0x%08x\n",
 		 obj->size, obj_priv->gtt_offset);
 #endif
-	if (retry_alloc) {
-		i915_gem_object_set_page_gfp_mask (obj,
-						   i915_gem_object_get_page_gfp_mask (obj) & ~__GFP_NORETRY);
-	}
-	ret = i915_gem_object_get_pages(obj);
-	if (retry_alloc) {
-		i915_gem_object_set_page_gfp_mask (obj,
-						   i915_gem_object_get_page_gfp_mask (obj) | __GFP_NORETRY);
-	}
+	ret = i915_gem_object_get_pages(obj, gfpmask);
 	if (ret) {
 		drm_mm_put_block(obj_priv->gtt_space);
 		obj_priv->gtt_space = NULL;
@@ -2625,9 +2678,9 @@ i915_gem_object_bind_to_gtt(struct drm_gem_object *obj, unsigned alignment)
 			ret = i915_gem_evict_something(dev, obj->size);
 			if (ret) {
 				/* now try to shrink everyone else */
-				if (! retry_alloc) {
-				    retry_alloc = true;
-				    goto search_free;
+				if (gfpmask) {
+					gfpmask = 0;
+					goto search_free;
 				}
 
 				return ret;
@@ -2695,7 +2748,6 @@ static void
 i915_gem_object_flush_gpu_write_domain(struct drm_gem_object *obj)
 {
 	struct drm_device *dev = obj->dev;
-	uint32_t seqno;
 	uint32_t old_write_domain;
 
 	if ((obj->write_domain & I915_GEM_GPU_DOMAINS) == 0)
@@ -2704,9 +2756,8 @@ i915_gem_object_flush_gpu_write_domain(struct drm_gem_object *obj)
 	/* Queue the GPU write cache flushing we need. */
 	old_write_domain = obj->write_domain;
 	i915_gem_flush(dev, 0, obj->write_domain);
-	seqno = i915_add_request(dev, NULL, obj->write_domain);
-	obj->write_domain = 0;
-	i915_gem_object_move_to_active(obj, seqno);
+	(void) i915_add_request(dev, NULL, obj->write_domain);
+	BUG_ON(obj->write_domain);
 
 	trace_i915_gem_object_change_domain(obj,
 					    obj->read_domains,
@@ -2754,6 +2805,22 @@ i915_gem_object_flush_cpu_write_domain(struct drm_gem_object *obj)
 					    old_write_domain);
 }
 
+void
+i915_gem_object_flush_write_domain(struct drm_gem_object *obj)
+{
+	switch (obj->write_domain) {
+	case I915_GEM_DOMAIN_GTT:
+		i915_gem_object_flush_gtt_write_domain(obj);
+		break;
+	case I915_GEM_DOMAIN_CPU:
+		i915_gem_object_flush_cpu_write_domain(obj);
+		break;
+	default:
+		i915_gem_object_flush_gpu_write_domain(obj);
+		break;
+	}
+}
+
 /**
  * Moves a single object to the GTT read, and possibly write domain.
  *
@@ -2805,6 +2872,57 @@ i915_gem_object_set_to_gtt_domain(struct drm_gem_object *obj, int write)
 	return 0;
 }
 
+/*
+ * Prepare buffer for display plane. Use uninterruptible for possible flush
+ * wait, as in modesetting process we're not supposed to be interrupted.
+ */
+int
+i915_gem_object_set_to_display_plane(struct drm_gem_object *obj)
+{
+	struct drm_device *dev = obj->dev;
+	struct drm_i915_gem_object *obj_priv = obj->driver_private;
+	uint32_t old_write_domain, old_read_domains;
+	int ret;
+
+	/* Not valid to be called on unbound objects. */
+	if (obj_priv->gtt_space == NULL)
+		return -EINVAL;
+
+	i915_gem_object_flush_gpu_write_domain(obj);
+
+	/* Wait on any GPU rendering and flushing to occur. */
+	if (obj_priv->active) {
+#if WATCH_BUF
+		DRM_INFO("%s: object %p wait for seqno %08x\n",
+			  __func__, obj, obj_priv->last_rendering_seqno);
+#endif
+		ret = i915_do_wait_request(dev, obj_priv->last_rendering_seqno, 0);
+		if (ret != 0)
+			return ret;
+	}
+
+	old_write_domain = obj->write_domain;
+	old_read_domains = obj->read_domains;
+
+	obj->read_domains &= I915_GEM_DOMAIN_GTT;
+
+	i915_gem_object_flush_cpu_write_domain(obj);
+
+	/* It should now be out of any other write domains, and we can update
+	 * the domain values for our changes.
+	 */
+	BUG_ON((obj->write_domain & ~I915_GEM_DOMAIN_GTT) != 0);
+	obj->read_domains |= I915_GEM_DOMAIN_GTT;
+	obj->write_domain = I915_GEM_DOMAIN_GTT;
+	obj_priv->dirty = 1;
+
+	trace_i915_gem_object_change_domain(obj,
+					    old_read_domains,
+					    old_write_domain);
+
+	return 0;
+}
+
 /**
  * Moves a single object to the CPU read, and possibly write domain.
  *
@@ -3164,7 +3282,7 @@ i915_gem_object_set_cpu_read_domain_range(struct drm_gem_object *obj,
 static int
 i915_gem_object_pin_and_relocate(struct drm_gem_object *obj,
 				 struct drm_file *file_priv,
-				 struct drm_i915_gem_exec_object *entry,
+				 struct drm_i915_gem_exec_object2 *entry,
 				 struct drm_i915_gem_relocation_entry *relocs)
 {
 	struct drm_device *dev = obj->dev;
@@ -3172,12 +3290,36 @@ i915_gem_object_pin_and_relocate(struct drm_gem_object *obj,
 	struct drm_i915_gem_object *obj_priv = obj->driver_private;
 	int i, ret;
 	void __iomem *reloc_page;
+	bool need_fence;
+
+	need_fence = entry->flags & EXEC_OBJECT_NEEDS_FENCE &&
+	             obj_priv->tiling_mode != I915_TILING_NONE;
+
+	/* Check fence reg constraints and rebind if necessary */
+	if (need_fence && !i915_gem_object_fence_offset_ok(obj,
+	    obj_priv->tiling_mode))
+		i915_gem_object_unbind(obj);
 
 	/* Choose the GTT offset for our buffer and put it there. */
 	ret = i915_gem_object_pin(obj, (uint32_t) entry->alignment);
 	if (ret)
 		return ret;
 
+	/*
+	 * Pre-965 chips need a fence register set up in order to
+	 * properly handle blits to/from tiled surfaces.
+	 */
+	if (need_fence) {
+		ret = i915_gem_object_get_fence_reg(obj);
+		if (ret != 0) {
+			if (ret != -EBUSY && ret != -ERESTARTSYS)
+				DRM_ERROR("Failure to install fence: %d\n",
+					  ret);
+			i915_gem_object_unpin(obj);
+			return ret;
+		}
+	}
+
 	entry->offset = obj_priv->gtt_offset;
 
 	/* Apply the relocations, using the GTT aperture to avoid cache
@@ -3225,6 +3367,16 @@ i915_gem_object_pin_and_relocate(struct drm_gem_object *obj,
 		}
 
 		/* Validate that the target is in a valid r/w GPU domain */
+		if (reloc->write_domain & (reloc->write_domain - 1)) {
+			DRM_ERROR("reloc with multiple write domains: "
+				  "obj %p target %d offset %d "
+				  "read %08x write %08x",
+				  obj, reloc->target_handle,
+				  (int) reloc->offset,
+				  reloc->read_domains,
+				  reloc->write_domain);
+			return -EINVAL;
+		}
 		if (reloc->write_domain & I915_GEM_DOMAIN_CPU ||
 		    reloc->read_domains & I915_GEM_DOMAIN_CPU) {
 			DRM_ERROR("reloc with read/write CPU domains: "
@@ -3339,7 +3491,7 @@ i915_gem_object_pin_and_relocate(struct drm_gem_object *obj,
  */
 static int
 i915_dispatch_gem_execbuffer(struct drm_device *dev,
-			      struct drm_i915_gem_execbuffer *exec,
+			      struct drm_i915_gem_execbuffer2 *exec,
 			      struct drm_clip_rect *cliprects,
 			      uint64_t exec_offset)
 {
@@ -3352,7 +3504,7 @@ i915_dispatch_gem_execbuffer(struct drm_device *dev,
 	exec_start = (uint32_t) exec_offset + exec->batch_start_offset;
 	exec_len = (uint32_t) exec->batch_len;
 
-	trace_i915_gem_request_submit(dev, dev_priv->mm.next_gem_seqno);
+	trace_i915_gem_request_submit(dev, dev_priv->mm.next_gem_seqno + 1);
 
 	count = nbox ? nbox : 1;
 
@@ -3429,7 +3581,7 @@ i915_gem_ring_throttle(struct drm_device *dev, struct drm_file *file_priv)
 }
 
 static int
-i915_gem_get_relocs_from_user(struct drm_i915_gem_exec_object *exec_list,
+i915_gem_get_relocs_from_user(struct drm_i915_gem_exec_object2 *exec_list,
 			      uint32_t buffer_count,
 			      struct drm_i915_gem_relocation_entry **relocs)
 {
@@ -3444,8 +3596,10 @@ i915_gem_get_relocs_from_user(struct drm_i915_gem_exec_object *exec_list,
 	}
 
 	*relocs = drm_calloc_large(reloc_count, sizeof(**relocs));
-	if (*relocs == NULL)
+	if (*relocs == NULL) {
+		DRM_ERROR("failed to alloc relocs, count %d\n", reloc_count);
 		return -ENOMEM;
+	}
 
 	for (i = 0; i < buffer_count; i++) {
 		struct drm_i915_gem_relocation_entry __user *user_relocs;
@@ -3469,13 +3623,16 @@ i915_gem_get_relocs_from_user(struct drm_i915_gem_exec_object *exec_list,
 }
 
 static int
-i915_gem_put_relocs_to_user(struct drm_i915_gem_exec_object *exec_list,
+i915_gem_put_relocs_to_user(struct drm_i915_gem_exec_object2 *exec_list,
 			    uint32_t buffer_count,
 			    struct drm_i915_gem_relocation_entry *relocs)
 {
 	uint32_t reloc_count = 0, i;
 	int ret = 0;
 
+	if (relocs == NULL)
+	    return 0;
+
 	for (i = 0; i < buffer_count; i++) {
 		struct drm_i915_gem_relocation_entry __user *user_relocs;
 		int unwritten;
@@ -3502,7 +3659,7 @@ err:
 }
 
 static int
-i915_gem_check_execbuffer (struct drm_i915_gem_execbuffer *exec,
+i915_gem_check_execbuffer (struct drm_i915_gem_execbuffer2 *exec,
 			   uint64_t exec_offset)
 {
 	uint32_t exec_start, exec_len;
@@ -3519,22 +3676,57 @@ i915_gem_check_execbuffer (struct drm_i915_gem_execbuffer *exec,
 	return 0;
 }
 
+static int
+i915_gem_wait_for_pending_flip(struct drm_device *dev,
+			       struct drm_gem_object **object_list,
+			       int count)
+{
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	struct drm_i915_gem_object *obj_priv;
+	DEFINE_WAIT(wait);
+	int i, ret = 0;
+
+	for (;;) {
+		prepare_to_wait(&dev_priv->pending_flip_queue,
+				&wait, TASK_INTERRUPTIBLE);
+		for (i = 0; i < count; i++) {
+			obj_priv = object_list[i]->driver_private;
+			if (atomic_read(&obj_priv->pending_flip) > 0)
+				break;
+		}
+		if (i == count)
+			break;
+
+		if (!signal_pending(current)) {
+			mutex_unlock(&dev->struct_mutex);
+			schedule();
+			mutex_lock(&dev->struct_mutex);
+			continue;
+		}
+		ret = -ERESTARTSYS;
+		break;
+	}
+	finish_wait(&dev_priv->pending_flip_queue, &wait);
+
+	return ret;
+}
+
 int
-i915_gem_execbuffer(struct drm_device *dev, void *data,
-		    struct drm_file *file_priv)
+i915_gem_do_execbuffer(struct drm_device *dev, void *data,
+		       struct drm_file *file_priv,
+		       struct drm_i915_gem_execbuffer2 *args,
+		       struct drm_i915_gem_exec_object2 *exec_list)
 {
 	drm_i915_private_t *dev_priv = dev->dev_private;
-	struct drm_i915_gem_execbuffer *args = data;
-	struct drm_i915_gem_exec_object *exec_list = NULL;
 	struct drm_gem_object **object_list = NULL;
 	struct drm_gem_object *batch_obj;
 	struct drm_i915_gem_object *obj_priv;
 	struct drm_clip_rect *cliprects = NULL;
-	struct drm_i915_gem_relocation_entry *relocs;
-	int ret, ret2, i, pinned = 0;
+	struct drm_i915_gem_relocation_entry *relocs = NULL;
+	int ret = 0, ret2, i, pinned = 0;
 	uint64_t exec_offset;
 	uint32_t seqno, flush_domains, reloc_index;
-	int pin_tries;
+	int pin_tries, flips;
 
 #if WATCH_EXEC
 	DRM_INFO("buffers_ptr %d buffer_count %d len %08x\n",
@@ -3545,31 +3737,21 @@ i915_gem_execbuffer(struct drm_device *dev, void *data,
 		DRM_ERROR("execbuf with %d buffers\n", args->buffer_count);
 		return -EINVAL;
 	}
-	/* Copy in the exec list from userland */
-	exec_list = drm_calloc_large(sizeof(*exec_list), args->buffer_count);
-	object_list = drm_calloc_large(sizeof(*object_list), args->buffer_count);
-	if (exec_list == NULL || object_list == NULL) {
-		DRM_ERROR("Failed to allocate exec or object list "
-			  "for %d buffers\n",
+	object_list = drm_malloc_ab(sizeof(*object_list), args->buffer_count);
+	if (object_list == NULL) {
+		DRM_ERROR("Failed to allocate object list for %d buffers\n",
 			  args->buffer_count);
 		ret = -ENOMEM;
 		goto pre_mutex_err;
 	}
-	ret = copy_from_user(exec_list,
-			     (struct drm_i915_relocation_entry __user *)
-			     (uintptr_t) args->buffers_ptr,
-			     sizeof(*exec_list) * args->buffer_count);
-	if (ret != 0) {
-		DRM_ERROR("copy %d exec entries failed %d\n",
-			  args->buffer_count, ret);
-		goto pre_mutex_err;
-	}
 
 	if (args->num_cliprects != 0) {
 		cliprects = kcalloc(args->num_cliprects, sizeof(*cliprects),
 				    GFP_KERNEL);
-		if (cliprects == NULL)
+		if (cliprects == NULL) {
+			ret = -ENOMEM;
 			goto pre_mutex_err;
+		}
 
 		ret = copy_from_user(cliprects,
 				     (struct drm_clip_rect __user *)
@@ -3592,26 +3774,27 @@ i915_gem_execbuffer(struct drm_device *dev, void *data,
 	i915_verify_inactive(dev, __FILE__, __LINE__);
 
 	if (atomic_read(&dev_priv->mm.wedged)) {
-		DRM_ERROR("Execbuf while wedged\n");
 		mutex_unlock(&dev->struct_mutex);
 		ret = -EIO;
 		goto pre_mutex_err;
 	}
 
 	if (dev_priv->mm.suspended) {
-		DRM_ERROR("Execbuf while VT-switched.\n");
 		mutex_unlock(&dev->struct_mutex);
 		ret = -EBUSY;
 		goto pre_mutex_err;
 	}
 
 	/* Look up object handles */
+	flips = 0;
 	for (i = 0; i < args->buffer_count; i++) {
 		object_list[i] = drm_gem_object_lookup(dev, file_priv,
 						       exec_list[i].handle);
 		if (object_list[i] == NULL) {
 			DRM_ERROR("Invalid object handle %d at index %d\n",
 				   exec_list[i].handle, i);
+			/* prevent error path from reading uninitialized data */
+			args->buffer_count = i + 1;
 			ret = -EBADF;
 			goto err;
 		}
@@ -3620,10 +3803,20 @@ i915_gem_execbuffer(struct drm_device *dev, void *data,
 		if (obj_priv->in_execbuffer) {
 			DRM_ERROR("Object %p appears more than once in object list\n",
 				   object_list[i]);
+			/* prevent error path from reading uninitialized data */
+			args->buffer_count = i + 1;
 			ret = -EBADF;
 			goto err;
 		}
 		obj_priv->in_execbuffer = true;
+		flips += atomic_read(&obj_priv->pending_flip);
+	}
+
+	if (flips > 0) {
+		ret = i915_gem_wait_for_pending_flip(dev, object_list,
+						     args->buffer_count);
+		if (ret)
+			goto err;
 	}
 
 	/* Pin and relocate */
@@ -3725,16 +3918,23 @@ i915_gem_execbuffer(struct drm_device *dev, void *data,
 		i915_gem_flush(dev,
 			       dev->invalidate_domains,
 			       dev->flush_domains);
-		if (dev->flush_domains)
+		if (dev->flush_domains & I915_GEM_GPU_DOMAINS)
 			(void)i915_add_request(dev, file_priv,
 					       dev->flush_domains);
 	}
 
 	for (i = 0; i < args->buffer_count; i++) {
 		struct drm_gem_object *obj = object_list[i];
+		struct drm_i915_gem_object *obj_priv = obj->driver_private;
 		uint32_t old_write_domain = obj->write_domain;
 
 		obj->write_domain = obj->pending_write_domain;
+		if (obj->write_domain)
+			list_move_tail(&obj_priv->gpu_write_list,
+				       &dev_priv->mm.gpu_write_list);
+		else
+			list_del_init(&obj_priv->gpu_write_list);
+
 		trace_i915_gem_object_change_domain(obj,
 						    obj->read_domains,
 						    old_write_domain);
@@ -3808,8 +4008,101 @@ err:
 
 	mutex_unlock(&dev->struct_mutex);
 
+pre_mutex_err:
+	/* Copy the updated relocations out regardless of current error
+	 * state.  Failure to update the relocs would mean that the next
+	 * time userland calls execbuf, it would do so with presumed offset
+	 * state that didn't match the actual object state.
+	 */
+	ret2 = i915_gem_put_relocs_to_user(exec_list, args->buffer_count,
+					   relocs);
+	if (ret2 != 0) {
+		DRM_ERROR("Failed to copy relocations back out: %d\n", ret2);
+
+		if (ret == 0)
+			ret = ret2;
+	}
+
+	drm_free_large(object_list);
+	kfree(cliprects);
+
+	return ret;
+}
+
+/*
+ * Legacy execbuffer just creates an exec2 list from the original exec object
+ * list array and passes it to the real function.
+ */
+int
+i915_gem_execbuffer(struct drm_device *dev, void *data,
+		    struct drm_file *file_priv)
+{
+	struct drm_i915_gem_execbuffer *args = data;
+	struct drm_i915_gem_execbuffer2 exec2;
+	struct drm_i915_gem_exec_object *exec_list = NULL;
+	struct drm_i915_gem_exec_object2 *exec2_list = NULL;
+	int ret, i;
+
+#if WATCH_EXEC
+	DRM_INFO("buffers_ptr %d buffer_count %d len %08x\n",
+		  (int) args->buffers_ptr, args->buffer_count, args->batch_len);
+#endif
+
+	if (args->buffer_count < 1) {
+		DRM_ERROR("execbuf with %d buffers\n", args->buffer_count);
+		return -EINVAL;
+	}
+
+	/* Copy in the exec list from userland */
+	exec_list = drm_malloc_ab(sizeof(*exec_list), args->buffer_count);
+	exec2_list = drm_malloc_ab(sizeof(*exec2_list), args->buffer_count);
+	if (exec_list == NULL || exec2_list == NULL) {
+		DRM_ERROR("Failed to allocate exec list for %d buffers\n",
+			  args->buffer_count);
+		drm_free_large(exec_list);
+		drm_free_large(exec2_list);
+		return -ENOMEM;
+	}
+	ret = copy_from_user(exec_list,
+			     (struct drm_i915_relocation_entry __user *)
+			     (uintptr_t) args->buffers_ptr,
+			     sizeof(*exec_list) * args->buffer_count);
+	if (ret != 0) {
+		DRM_ERROR("copy %d exec entries failed %d\n",
+			  args->buffer_count, ret);
+		drm_free_large(exec_list);
+		drm_free_large(exec2_list);
+		return -EFAULT;
+	}
+
+	for (i = 0; i < args->buffer_count; i++) {
+		exec2_list[i].handle = exec_list[i].handle;
+		exec2_list[i].relocation_count = exec_list[i].relocation_count;
+		exec2_list[i].relocs_ptr = exec_list[i].relocs_ptr;
+		exec2_list[i].alignment = exec_list[i].alignment;
+		exec2_list[i].offset = exec_list[i].offset;
+		if (!IS_I965G(dev))
+			exec2_list[i].flags = EXEC_OBJECT_NEEDS_FENCE;
+		else
+			exec2_list[i].flags = 0;
+	}
+
+	exec2.buffers_ptr = args->buffers_ptr;
+	exec2.buffer_count = args->buffer_count;
+	exec2.batch_start_offset = args->batch_start_offset;
+	exec2.batch_len = args->batch_len;
+	exec2.DR1 = args->DR1;
+	exec2.DR4 = args->DR4;
+	exec2.num_cliprects = args->num_cliprects;
+	exec2.cliprects_ptr = args->cliprects_ptr;
+	exec2.flags = 0;
+
+	ret = i915_gem_do_execbuffer(dev, data, file_priv, &exec2, exec2_list);
 	if (!ret) {
 		/* Copy the new buffer offsets back to the user's exec list. */
+		for (i = 0; i < args->buffer_count; i++)
+			exec_list[i].offset = exec2_list[i].offset;
+		/* ... and back out to userspace */
 		ret = copy_to_user((struct drm_i915_relocation_entry __user *)
 				   (uintptr_t) args->buffers_ptr,
 				   exec_list,
@@ -3822,25 +4115,62 @@ err:
 		}
 	}
 
-	/* Copy the updated relocations out regardless of current error
-	 * state.  Failure to update the relocs would mean that the next
-	 * time userland calls execbuf, it would do so with presumed offset
-	 * state that didn't match the actual object state.
-	 */
-	ret2 = i915_gem_put_relocs_to_user(exec_list, args->buffer_count,
-					   relocs);
-	if (ret2 != 0) {
-		DRM_ERROR("Failed to copy relocations back out: %d\n", ret2);
+	drm_free_large(exec_list);
+	drm_free_large(exec2_list);
+	return ret;
+}
 
-		if (ret == 0)
-			ret = ret2;
+int
+i915_gem_execbuffer2(struct drm_device *dev, void *data,
+		     struct drm_file *file_priv)
+{
+	struct drm_i915_gem_execbuffer2 *args = data;
+	struct drm_i915_gem_exec_object2 *exec2_list = NULL;
+	int ret;
+
+#if WATCH_EXEC
+	DRM_INFO("buffers_ptr %d buffer_count %d len %08x\n",
+		  (int) args->buffers_ptr, args->buffer_count, args->batch_len);
+#endif
+
+	if (args->buffer_count < 1) {
+		DRM_ERROR("execbuf2 with %d buffers\n", args->buffer_count);
+		return -EINVAL;
 	}
 
-pre_mutex_err:
-	drm_free_large(object_list);
-	drm_free_large(exec_list);
-	kfree(cliprects);
+	exec2_list = drm_malloc_ab(sizeof(*exec2_list), args->buffer_count);
+	if (exec2_list == NULL) {
+		DRM_ERROR("Failed to allocate exec list for %d buffers\n",
+			  args->buffer_count);
+		return -ENOMEM;
+	}
+	ret = copy_from_user(exec2_list,
+			     (struct drm_i915_relocation_entry __user *)
+			     (uintptr_t) args->buffers_ptr,
+			     sizeof(*exec2_list) * args->buffer_count);
+	if (ret != 0) {
+		DRM_ERROR("copy %d exec entries failed %d\n",
+			  args->buffer_count, ret);
+		drm_free_large(exec2_list);
+		return -EFAULT;
+	}
 
+	ret = i915_gem_do_execbuffer(dev, data, file_priv, args, exec2_list);
+	if (!ret) {
+		/* Copy the new buffer offsets back to the user's exec list. */
+		ret = copy_to_user((struct drm_i915_relocation_entry __user *)
+				   (uintptr_t) args->buffers_ptr,
+				   exec2_list,
+				   sizeof(*exec2_list) * args->buffer_count);
+		if (ret) {
+			ret = -EFAULT;
+			DRM_ERROR("failed to copy %d exec entries "
+				  "back to user (%d)\n",
+				  args->buffer_count, ret);
+		}
+	}
+
+	drm_free_large(exec2_list);
 	return ret;
 }
 
@@ -3857,19 +4187,7 @@ i915_gem_object_pin(struct drm_gem_object *obj, uint32_t alignment)
 		if (ret)
 			return ret;
 	}
-	/*
-	 * Pre-965 chips need a fence register set up in order to
-	 * properly handle tiled surfaces.
-	 */
-	if (!IS_I965G(dev) && obj_priv->tiling_mode != I915_TILING_NONE) {
-		ret = i915_gem_object_get_fence_reg(obj);
-		if (ret != 0) {
-			if (ret != -EBUSY && ret != -ERESTARTSYS)
-				DRM_ERROR("Failure to install fence: %d\n",
-					  ret);
-			return ret;
-		}
-	}
+
 	obj_priv->pin_count++;
 
 	/* If the object is not active and not pending a flush,
@@ -4127,6 +4445,7 @@ int i915_gem_init_object(struct drm_gem_object *obj)
 	obj_priv->obj = obj;
 	obj_priv->fence_reg = I915_FENCE_REG_NONE;
 	INIT_LIST_HEAD(&obj_priv->list);
+	INIT_LIST_HEAD(&obj_priv->gpu_write_list);
 	INIT_LIST_HEAD(&obj_priv->fence_list);
 	obj_priv->madv = I915_MADV_WILLNEED;
 
@@ -4186,8 +4505,7 @@ int
 i915_gem_idle(struct drm_device *dev)
 {
 	drm_i915_private_t *dev_priv = dev->dev_private;
-	uint32_t seqno, cur_seqno, last_seqno;
-	int stuck, ret;
+	int ret;
 
 	mutex_lock(&dev->struct_mutex);
 
@@ -4196,115 +4514,36 @@ i915_gem_idle(struct drm_device *dev)
 		return 0;
 	}
 
-	/* Hack!  Don't let anybody do execbuf while we don't control the chip.
-	 * We need to replace this with a semaphore, or something.
-	 */
-	dev_priv->mm.suspended = 1;
-	del_timer(&dev_priv->hangcheck_timer);
-
-	/* Cancel the retire work handler, wait for it to finish if running
-	 */
-	mutex_unlock(&dev->struct_mutex);
-	cancel_delayed_work_sync(&dev_priv->mm.retire_work);
-	mutex_lock(&dev->struct_mutex);
-
-	i915_kernel_lost_context(dev);
-
-	/* Flush the GPU along with all non-CPU write domains
-	 */
-	i915_gem_flush(dev, I915_GEM_GPU_DOMAINS, I915_GEM_GPU_DOMAINS);
-	seqno = i915_add_request(dev, NULL, I915_GEM_GPU_DOMAINS);
-
-	if (seqno == 0) {
+	ret = i915_gpu_idle(dev);
+	if (ret) {
 		mutex_unlock(&dev->struct_mutex);
-		return -ENOMEM;
+		return ret;
 	}
 
-	dev_priv->mm.waiting_gem_seqno = seqno;
-	last_seqno = 0;
-	stuck = 0;
-	for (;;) {
-		cur_seqno = i915_get_gem_seqno(dev);
-		if (i915_seqno_passed(cur_seqno, seqno))
-			break;
-		if (last_seqno == cur_seqno) {
-			if (stuck++ > 100) {
-				DRM_ERROR("hardware wedged\n");
-				atomic_set(&dev_priv->mm.wedged, 1);
-				DRM_WAKEUP(&dev_priv->irq_queue);
-				break;
-			}
+	/* Under UMS, be paranoid and evict. */
+	if (!drm_core_check_feature(dev, DRIVER_MODESET)) {
+		ret = i915_gem_evict_from_inactive_list(dev);
+		if (ret) {
+			mutex_unlock(&dev->struct_mutex);
+			return ret;
 		}
-		msleep(10);
-		last_seqno = cur_seqno;
-	}
-	dev_priv->mm.waiting_gem_seqno = 0;
-
-	i915_gem_retire_requests(dev);
-
-	spin_lock(&dev_priv->mm.active_list_lock);
-	if (!atomic_read(&dev_priv->mm.wedged)) {
-		/* Active and flushing should now be empty as we've
-		 * waited for a sequence higher than any pending execbuffer
-		 */
-		WARN_ON(!list_empty(&dev_priv->mm.active_list));
-		WARN_ON(!list_empty(&dev_priv->mm.flushing_list));
-		/* Request should now be empty as we've also waited
-		 * for the last request in the list
-		 */
-		WARN_ON(!list_empty(&dev_priv->mm.request_list));
 	}
 
-	/* Empty the active and flushing lists to inactive.  If there's
-	 * anything left at this point, it means that we're wedged and
-	 * nothing good's going to happen by leaving them there.  So strip
-	 * the GPU domains and just stuff them onto inactive.
+	/* Hack!  Don't let anybody do execbuf while we don't control the chip.
+	 * We need to replace this with a semaphore, or something.
+	 * And not confound mm.suspended!
 	 */
-	while (!list_empty(&dev_priv->mm.active_list)) {
-		struct drm_gem_object *obj;
-		uint32_t old_write_domain;
-
-		obj = list_first_entry(&dev_priv->mm.active_list,
-				       struct drm_i915_gem_object,
-				       list)->obj;
-		old_write_domain = obj->write_domain;
-		obj->write_domain &= ~I915_GEM_GPU_DOMAINS;
-		i915_gem_object_move_to_inactive(obj);
-
-		trace_i915_gem_object_change_domain(obj,
-						    obj->read_domains,
-						    old_write_domain);
-	}
-	spin_unlock(&dev_priv->mm.active_list_lock);
-
-	while (!list_empty(&dev_priv->mm.flushing_list)) {
-		struct drm_gem_object *obj;
-		uint32_t old_write_domain;
-
-		obj = list_first_entry(&dev_priv->mm.flushing_list,
-				       struct drm_i915_gem_object,
-				       list)->obj;
-		old_write_domain = obj->write_domain;
-		obj->write_domain &= ~I915_GEM_GPU_DOMAINS;
-		i915_gem_object_move_to_inactive(obj);
-
-		trace_i915_gem_object_change_domain(obj,
-						    obj->read_domains,
-						    old_write_domain);
-	}
-
-
-	/* Move all inactive buffers out of the GTT. */
-	ret = i915_gem_evict_from_inactive_list(dev);
-	WARN_ON(!list_empty(&dev_priv->mm.inactive_list));
-	if (ret) {
-		mutex_unlock(&dev->struct_mutex);
-		return ret;
-	}
+	dev_priv->mm.suspended = 1;
+	del_timer(&dev_priv->hangcheck_timer);
 
+	i915_kernel_lost_context(dev);
 	i915_gem_cleanup_ringbuffer(dev);
+
 	mutex_unlock(&dev->struct_mutex);
 
+	/* Cancel the retire work handler, which should be idle now. */
+	cancel_delayed_work_sync(&dev_priv->mm.retire_work);
+
 	return 0;
 }
 
@@ -4348,9 +4587,14 @@ i915_gem_init_hws(struct drm_device *dev)
 	}
 	dev_priv->hws_obj = obj;
 	memset(dev_priv->hw_status_page, 0, PAGE_SIZE);
-	I915_WRITE(HWS_PGA, dev_priv->status_gfx_addr);
-	I915_READ(HWS_PGA); /* posting read */
-	DRM_DEBUG("hws offset: 0x%08x\n", dev_priv->status_gfx_addr);
+	if (IS_GEN6(dev)) {
+		I915_WRITE(HWS_PGA_GEN6, dev_priv->status_gfx_addr);
+		I915_READ(HWS_PGA_GEN6); /* posting read */
+	} else {
+		I915_WRITE(HWS_PGA, dev_priv->status_gfx_addr);
+		I915_READ(HWS_PGA); /* posting read */
+	}
+	DRM_DEBUG_DRIVER("hws offset: 0x%08x\n", dev_priv->status_gfx_addr);
 
 	return 0;
 }
@@ -4578,6 +4822,7 @@ i915_gem_load(struct drm_device *dev)
 	spin_lock_init(&dev_priv->mm.active_list_lock);
 	INIT_LIST_HEAD(&dev_priv->mm.active_list);
 	INIT_LIST_HEAD(&dev_priv->mm.flushing_list);
+	INIT_LIST_HEAD(&dev_priv->mm.gpu_write_list);
 	INIT_LIST_HEAD(&dev_priv->mm.inactive_list);
 	INIT_LIST_HEAD(&dev_priv->mm.request_list);
 	INIT_LIST_HEAD(&dev_priv->mm.fence_list);
@@ -4590,7 +4835,8 @@ i915_gem_load(struct drm_device *dev)
 	spin_unlock(&shrink_list_lock);
 
 	/* Old X drivers will take 0-2 for front, back, depth buffers */
-	dev_priv->fence_reg_start = 3;
+	if (!drm_core_check_feature(dev, DRIVER_MODESET))
+		dev_priv->fence_reg_start = 3;
 
 	if (IS_I965G(dev) || IS_I945G(dev) || IS_I945GM(dev) || IS_G33(dev))
 		dev_priv->num_fence_regs = 16;
@@ -4608,8 +4854,8 @@ i915_gem_load(struct drm_device *dev)
 			for (i = 0; i < 8; i++)
 				I915_WRITE(FENCE_REG_945_8 + (i * 4), 0);
 	}
-
 	i915_gem_detect_bit_6_swizzle(dev);
+	init_waitqueue_head(&dev_priv->pending_flip_queue);
 }
 
 /*
@@ -4632,7 +4878,7 @@ int i915_gem_init_phys_object(struct drm_device *dev,
 
 	phys_obj->id = id;
 
-	phys_obj->handle = drm_pci_alloc(dev, size, 0, 0xffffffff);
+	phys_obj->handle = drm_pci_alloc(dev, size, 0);
 	if (!phys_obj->handle) {
 		ret = -ENOMEM;
 		goto kfree_obj;
@@ -4690,7 +4936,7 @@ void i915_gem_detach_phys_object(struct drm_device *dev,
 	if (!obj_priv->phys_obj)
 		return;
 
-	ret = i915_gem_object_get_pages(obj);
+	ret = i915_gem_object_get_pages(obj, 0);
 	if (ret)
 		goto out;
 
@@ -4748,7 +4994,7 @@ i915_gem_attach_phys_object(struct drm_device *dev,
 	obj_priv->phys_obj = dev_priv->mm.phys_objs[id - 1];
 	obj_priv->phys_obj->cur_obj = obj;
 
-	ret = i915_gem_object_get_pages(obj);
+	ret = i915_gem_object_get_pages(obj, 0);
 	if (ret) {
 		DRM_ERROR("failed to get page list\n");
 		goto out;
@@ -4784,7 +5030,7 @@ i915_gem_phys_pwrite(struct drm_device *dev, struct drm_gem_object *obj,
 	user_data = (char __user *) (uintptr_t) args->data_ptr;
 	obj_addr = obj_priv->phys_obj->handle->vaddr + args->offset;
 
-	DRM_DEBUG("obj_addr %p, %lld\n", obj_addr, args->size);
+	DRM_DEBUG_DRIVER("obj_addr %p, %lld\n", obj_addr, args->size);
 	ret = copy_from_user(obj_addr, user_data, args->size);
 	if (ret)
 		return -EFAULT;
diff --git a/drivers/gpu/drm/i915/i915_gem_tiling.c b/drivers/gpu/drm/i915/i915_gem_tiling.c
index 200e398453ca..b5c55d88ff76 100644
--- a/drivers/gpu/drm/i915/i915_gem_tiling.c
+++ b/drivers/gpu/drm/i915/i915_gem_tiling.c
@@ -25,8 +25,6 @@
  *
  */
 
-#include <linux/acpi.h>
-#include <linux/pnp.h>
 #include "linux/string.h"
 #include "linux/bitops.h"
 #include "drmP.h"
@@ -83,120 +81,6 @@
  * to match what the GPU expects.
  */
 
-#define MCHBAR_I915 0x44
-#define MCHBAR_I965 0x48
-#define MCHBAR_SIZE (4*4096)
-
-#define DEVEN_REG 0x54
-#define   DEVEN_MCHBAR_EN (1 << 28)
-
-/* Allocate space for the MCH regs if needed, return nonzero on error */
-static int
-intel_alloc_mchbar_resource(struct drm_device *dev)
-{
-	drm_i915_private_t *dev_priv = dev->dev_private;
-	int reg = IS_I965G(dev) ? MCHBAR_I965 : MCHBAR_I915;
-	u32 temp_lo, temp_hi = 0;
-	u64 mchbar_addr;
-	int ret = 0;
-
-	if (IS_I965G(dev))
-		pci_read_config_dword(dev_priv->bridge_dev, reg + 4, &temp_hi);
-	pci_read_config_dword(dev_priv->bridge_dev, reg, &temp_lo);
-	mchbar_addr = ((u64)temp_hi << 32) | temp_lo;
-
-	/* If ACPI doesn't have it, assume we need to allocate it ourselves */
-#ifdef CONFIG_PNP
-	if (mchbar_addr &&
-	    pnp_range_reserved(mchbar_addr, mchbar_addr + MCHBAR_SIZE)) {
-		ret = 0;
-		goto out;
-	}
-#endif
-
-	/* Get some space for it */
-	ret = pci_bus_alloc_resource(dev_priv->bridge_dev->bus, &dev_priv->mch_res,
-				     MCHBAR_SIZE, MCHBAR_SIZE,
-				     PCIBIOS_MIN_MEM,
-				     0,   pcibios_align_resource,
-				     dev_priv->bridge_dev);
-	if (ret) {
-		DRM_DEBUG("failed bus alloc: %d\n", ret);
-		dev_priv->mch_res.start = 0;
-		goto out;
-	}
-
-	if (IS_I965G(dev))
-		pci_write_config_dword(dev_priv->bridge_dev, reg + 4,
-				       upper_32_bits(dev_priv->mch_res.start));
-
-	pci_write_config_dword(dev_priv->bridge_dev, reg,
-			       lower_32_bits(dev_priv->mch_res.start));
-out:
-	return ret;
-}
-
-/* Setup MCHBAR if possible, return true if we should disable it again */
-static bool
-intel_setup_mchbar(struct drm_device *dev)
-{
-	drm_i915_private_t *dev_priv = dev->dev_private;
-	int mchbar_reg = IS_I965G(dev) ? MCHBAR_I965 : MCHBAR_I915;
-	u32 temp;
-	bool need_disable = false, enabled;
-
-	if (IS_I915G(dev) || IS_I915GM(dev)) {
-		pci_read_config_dword(dev_priv->bridge_dev, DEVEN_REG, &temp);
-		enabled = !!(temp & DEVEN_MCHBAR_EN);
-	} else {
-		pci_read_config_dword(dev_priv->bridge_dev, mchbar_reg, &temp);
-		enabled = temp & 1;
-	}
-
-	/* If it's already enabled, don't have to do anything */
-	if (enabled)
-		goto out;
-
-	if (intel_alloc_mchbar_resource(dev))
-		goto out;
-
-	need_disable = true;
-
-	/* Space is allocated or reserved, so enable it. */
-	if (IS_I915G(dev) || IS_I915GM(dev)) {
-		pci_write_config_dword(dev_priv->bridge_dev, DEVEN_REG,
-				       temp | DEVEN_MCHBAR_EN);
-	} else {
-		pci_read_config_dword(dev_priv->bridge_dev, mchbar_reg, &temp);
-		pci_write_config_dword(dev_priv->bridge_dev, mchbar_reg, temp | 1);
-	}
-out:
-	return need_disable;
-}
-
-static void
-intel_teardown_mchbar(struct drm_device *dev, bool disable)
-{
-	drm_i915_private_t *dev_priv = dev->dev_private;
-	int mchbar_reg = IS_I965G(dev) ? MCHBAR_I965 : MCHBAR_I915;
-	u32 temp;
-
-	if (disable) {
-		if (IS_I915G(dev) || IS_I915GM(dev)) {
-			pci_read_config_dword(dev_priv->bridge_dev, DEVEN_REG, &temp);
-			temp &= ~DEVEN_MCHBAR_EN;
-			pci_write_config_dword(dev_priv->bridge_dev, DEVEN_REG, temp);
-		} else {
-			pci_read_config_dword(dev_priv->bridge_dev, mchbar_reg, &temp);
-			temp &= ~1;
-			pci_write_config_dword(dev_priv->bridge_dev, mchbar_reg, temp);
-		}
-	}
-
-	if (dev_priv->mch_res.start)
-		release_resource(&dev_priv->mch_res);
-}
-
 /**
  * Detects bit 6 swizzling of address lookup between IGD access and CPU
  * access through main memory.
@@ -207,10 +91,9 @@ i915_gem_detect_bit_6_swizzle(struct drm_device *dev)
 	drm_i915_private_t *dev_priv = dev->dev_private;
 	uint32_t swizzle_x = I915_BIT_6_SWIZZLE_UNKNOWN;
 	uint32_t swizzle_y = I915_BIT_6_SWIZZLE_UNKNOWN;
-	bool need_disable;
 
-	if (IS_IGDNG(dev)) {
-		/* On IGDNG whatever DRAM config, GPU always do
+	if (IS_IRONLAKE(dev) || IS_GEN6(dev)) {
+		/* On Ironlake whatever DRAM config, GPU always do
 		 * same swizzling setup.
 		 */
 		swizzle_x = I915_BIT_6_SWIZZLE_9_10;
@@ -224,9 +107,6 @@ i915_gem_detect_bit_6_swizzle(struct drm_device *dev)
 	} else if (IS_MOBILE(dev)) {
 		uint32_t dcc;
 
-		/* Try to make sure MCHBAR is enabled before poking at it */
-		need_disable = intel_setup_mchbar(dev);
-
 		/* On mobile 9xx chipsets, channel interleave by the CPU is
 		 * determined by DCC.  For single-channel, neither the CPU
 		 * nor the GPU do swizzling.  For dual channel interleaved,
@@ -266,8 +146,6 @@ i915_gem_detect_bit_6_swizzle(struct drm_device *dev)
 			swizzle_x = I915_BIT_6_SWIZZLE_UNKNOWN;
 			swizzle_y = I915_BIT_6_SWIZZLE_UNKNOWN;
 		}
-
-		intel_teardown_mchbar(dev, need_disable);
 	} else {
 		/* The 965, G33, and newer, have a very flexible memory
 		 * configuration.  It will enable dual-channel mode
@@ -302,37 +180,8 @@ i915_gem_detect_bit_6_swizzle(struct drm_device *dev)
 	dev_priv->mm.bit_6_swizzle_y = swizzle_y;
 }
 
-
-/**
- * Returns the size of the fence for a tiled object of the given size.
- */
-static int
-i915_get_fence_size(struct drm_device *dev, int size)
-{
-	int i;
-	int start;
-
-	if (IS_I965G(dev)) {
-		/* The 965 can have fences at any page boundary. */
-		return ALIGN(size, 4096);
-	} else {
-		/* Align the size to a power of two greater than the smallest
-		 * fence size.
-		 */
-		if (IS_I9XX(dev))
-			start = 1024 * 1024;
-		else
-			start = 512 * 1024;
-
-		for (i = start; i < size; i <<= 1)
-			;
-
-		return i;
-	}
-}
-
 /* Check pitch constriants for all chips & tiling formats */
-static bool
+bool
 i915_tiling_ok(struct drm_device *dev, int stride, int size, int tiling_mode)
 {
 	int tile_width;
@@ -384,16 +233,10 @@ i915_tiling_ok(struct drm_device *dev, int stride, int size, int tiling_mode)
 	if (stride & (stride - 1))
 		return false;
 
-	/* We don't 0handle the aperture area covered by the fence being bigger
-	 * than the object size.
-	 */
-	if (i915_get_fence_size(dev, size) != size)
-		return false;
-
 	return true;
 }
 
-static bool
+bool
 i915_gem_object_fence_offset_ok(struct drm_gem_object *obj, int tiling_mode)
 {
 	struct drm_device *dev = obj->dev;
@@ -440,9 +283,7 @@ i915_gem_set_tiling(struct drm_device *dev, void *data,
 	obj_priv = obj->driver_private;
 
 	if (!i915_tiling_ok(dev, args->stride, obj->size, args->tiling_mode)) {
-		mutex_lock(&dev->struct_mutex);
-		drm_gem_object_unreference(obj);
-		mutex_unlock(&dev->struct_mutex);
+		drm_gem_object_unreference_unlocked(obj);
 		return -EINVAL;
 	}
 
@@ -495,12 +336,6 @@ i915_gem_set_tiling(struct drm_device *dev, void *data,
 			goto err;
 		}
 
-		/* If we've changed tiling, GTT-mappings of the object
-		 * need to re-fault to ensure that the correct fence register
-		 * setup is in place.
-		 */
-		i915_gem_release_mmap(obj);
-
 		obj_priv->tiling_mode = args->tiling_mode;
 		obj_priv->stride = args->stride;
 	}
diff --git a/drivers/gpu/drm/i915/i915_ioc32.c b/drivers/gpu/drm/i915/i915_ioc32.c
index 1fe68a251b75..13b028994b2b 100644
--- a/drivers/gpu/drm/i915/i915_ioc32.c
+++ b/drivers/gpu/drm/i915/i915_ioc32.c
@@ -66,8 +66,7 @@ static int compat_i915_batchbuffer(struct file *file, unsigned int cmd,
 			  &batchbuffer->cliprects))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_I915_BATCHBUFFER,
+	return drm_ioctl(file, DRM_IOCTL_I915_BATCHBUFFER,
 			 (unsigned long)batchbuffer);
 }
 
@@ -102,8 +101,8 @@ static int compat_i915_cmdbuffer(struct file *file, unsigned int cmd,
 			  &cmdbuffer->cliprects))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_I915_CMDBUFFER, (unsigned long)cmdbuffer);
+	return drm_ioctl(file, DRM_IOCTL_I915_CMDBUFFER,
+			 (unsigned long)cmdbuffer);
 }
 
 typedef struct drm_i915_irq_emit32 {
@@ -125,8 +124,8 @@ static int compat_i915_irq_emit(struct file *file, unsigned int cmd,
 			  &request->irq_seq))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_I915_IRQ_EMIT, (unsigned long)request);
+	return drm_ioctl(file, DRM_IOCTL_I915_IRQ_EMIT,
+			 (unsigned long)request);
 }
 typedef struct drm_i915_getparam32 {
 	int param;
@@ -149,8 +148,8 @@ static int compat_i915_getparam(struct file *file, unsigned int cmd,
 			  &request->value))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_I915_GETPARAM, (unsigned long)request);
+	return drm_ioctl(file, DRM_IOCTL_I915_GETPARAM,
+			 (unsigned long)request);
 }
 
 typedef struct drm_i915_mem_alloc32 {
@@ -178,8 +177,8 @@ static int compat_i915_alloc(struct file *file, unsigned int cmd,
 			  &request->region_offset))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_I915_ALLOC, (unsigned long)request);
+	return drm_ioctl(file, DRM_IOCTL_I915_ALLOC,
+			 (unsigned long)request);
 }
 
 drm_ioctl_compat_t *i915_compat_ioctls[] = {
@@ -211,12 +210,10 @@ long i915_compat_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
 	if (nr < DRM_COMMAND_BASE + DRM_ARRAY_SIZE(i915_compat_ioctls))
 		fn = i915_compat_ioctls[nr - DRM_COMMAND_BASE];
 
-	lock_kernel();		/* XXX for now */
 	if (fn != NULL)
 		ret = (*fn) (filp, cmd, arg);
 	else
-		ret = drm_ioctl(filp->f_path.dentry->d_inode, filp, cmd, arg);
-	unlock_kernel();
+		ret = drm_ioctl(filp, cmd, arg);
 
 	return ret;
 }
diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c
index 4dfeec7cdd42..5388354da0d1 100644
--- a/drivers/gpu/drm/i915/i915_irq.c
+++ b/drivers/gpu/drm/i915/i915_irq.c
@@ -43,10 +43,13 @@
  * we leave them always unmasked in IMR and then control enabling them through
  * PIPESTAT alone.
  */
-#define I915_INTERRUPT_ENABLE_FIX (I915_ASLE_INTERRUPT |		 \
-				   I915_DISPLAY_PIPE_A_EVENT_INTERRUPT | \
-				   I915_DISPLAY_PIPE_B_EVENT_INTERRUPT | \
-				   I915_RENDER_COMMAND_PARSER_ERROR_INTERRUPT)
+#define I915_INTERRUPT_ENABLE_FIX			\
+	(I915_ASLE_INTERRUPT |				\
+	 I915_DISPLAY_PIPE_A_EVENT_INTERRUPT |		\
+	 I915_DISPLAY_PIPE_B_EVENT_INTERRUPT |		\
+	 I915_DISPLAY_PLANE_A_FLIP_PENDING_INTERRUPT |	\
+	 I915_DISPLAY_PLANE_B_FLIP_PENDING_INTERRUPT |	\
+	 I915_RENDER_COMMAND_PARSER_ERROR_INTERRUPT)
 
 /** Interrupts that we mask and unmask at runtime. */
 #define I915_INTERRUPT_ENABLE_VAR (I915_USER_INTERRUPT)
@@ -61,7 +64,7 @@
 					 DRM_I915_VBLANK_PIPE_B)
 
 void
-igdng_enable_graphics_irq(drm_i915_private_t *dev_priv, u32 mask)
+ironlake_enable_graphics_irq(drm_i915_private_t *dev_priv, u32 mask)
 {
 	if ((dev_priv->gt_irq_mask_reg & mask) != 0) {
 		dev_priv->gt_irq_mask_reg &= ~mask;
@@ -71,7 +74,7 @@ igdng_enable_graphics_irq(drm_i915_private_t *dev_priv, u32 mask)
 }
 
 static inline void
-igdng_disable_graphics_irq(drm_i915_private_t *dev_priv, u32 mask)
+ironlake_disable_graphics_irq(drm_i915_private_t *dev_priv, u32 mask)
 {
 	if ((dev_priv->gt_irq_mask_reg & mask) != mask) {
 		dev_priv->gt_irq_mask_reg |= mask;
@@ -82,7 +85,7 @@ igdng_disable_graphics_irq(drm_i915_private_t *dev_priv, u32 mask)
 
 /* For display hotplug interrupt */
 void
-igdng_enable_display_irq(drm_i915_private_t *dev_priv, u32 mask)
+ironlake_enable_display_irq(drm_i915_private_t *dev_priv, u32 mask)
 {
 	if ((dev_priv->irq_mask_reg & mask) != 0) {
 		dev_priv->irq_mask_reg &= ~mask;
@@ -92,7 +95,7 @@ igdng_enable_display_irq(drm_i915_private_t *dev_priv, u32 mask)
 }
 
 static inline void
-igdng_disable_display_irq(drm_i915_private_t *dev_priv, u32 mask)
+ironlake_disable_display_irq(drm_i915_private_t *dev_priv, u32 mask)
 {
 	if ((dev_priv->irq_mask_reg & mask) != mask) {
 		dev_priv->irq_mask_reg |= mask;
@@ -157,6 +160,20 @@ i915_disable_pipestat(drm_i915_private_t *dev_priv, int pipe, u32 mask)
 }
 
 /**
+ * intel_enable_asle - enable ASLE interrupt for OpRegion
+ */
+void intel_enable_asle (struct drm_device *dev)
+{
+	drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private;
+
+	if (HAS_PCH_SPLIT(dev))
+		ironlake_enable_display_irq(dev_priv, DE_GSE);
+	else
+		i915_enable_pipestat(dev_priv, 1,
+				     I915_LEGACY_BLC_EVENT_ENABLE);
+}
+
+/**
  * i915_pipe_enabled - check if a pipe is enabled
  * @dev: DRM device
  * @pipe: pipe to check
@@ -191,7 +208,8 @@ u32 i915_get_vblank_counter(struct drm_device *dev, int pipe)
 	low_frame = pipe ? PIPEBFRAMEPIXEL : PIPEAFRAMEPIXEL;
 
 	if (!i915_pipe_enabled(dev, pipe)) {
-		DRM_DEBUG("trying to get vblank count for disabled pipe %d\n", pipe);
+		DRM_DEBUG_DRIVER("trying to get vblank count for disabled "
+				"pipe %d\n", pipe);
 		return 0;
 	}
 
@@ -220,7 +238,8 @@ u32 gm45_get_vblank_counter(struct drm_device *dev, int pipe)
 	int reg = pipe ? PIPEB_FRMCOUNT_GM45 : PIPEA_FRMCOUNT_GM45;
 
 	if (!i915_pipe_enabled(dev, pipe)) {
-		DRM_DEBUG("trying to get vblank count for disabled pipe %d\n", pipe);
+		DRM_DEBUG_DRIVER("trying to get vblank count for disabled "
+					"pipe %d\n", pipe);
 		return 0;
 	}
 
@@ -250,46 +269,133 @@ static void i915_hotplug_work_func(struct work_struct *work)
 	drm_sysfs_hotplug_event(dev);
 }
 
-irqreturn_t igdng_irq_handler(struct drm_device *dev)
+static void i915_handle_rps_change(struct drm_device *dev)
+{
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	u32 busy_up, busy_down, max_avg, min_avg;
+	u16 rgvswctl;
+	u8 new_delay = dev_priv->cur_delay;
+
+	I915_WRITE(MEMINTRSTS, I915_READ(MEMINTRSTS) & ~MEMINT_EVAL_CHG);
+	busy_up = I915_READ(RCPREVBSYTUPAVG);
+	busy_down = I915_READ(RCPREVBSYTDNAVG);
+	max_avg = I915_READ(RCBMAXAVG);
+	min_avg = I915_READ(RCBMINAVG);
+
+	/* Handle RCS change request from hw */
+	if (busy_up > max_avg) {
+		if (dev_priv->cur_delay != dev_priv->max_delay)
+			new_delay = dev_priv->cur_delay - 1;
+		if (new_delay < dev_priv->max_delay)
+			new_delay = dev_priv->max_delay;
+	} else if (busy_down < min_avg) {
+		if (dev_priv->cur_delay != dev_priv->min_delay)
+			new_delay = dev_priv->cur_delay + 1;
+		if (new_delay > dev_priv->min_delay)
+			new_delay = dev_priv->min_delay;
+	}
+
+	DRM_DEBUG("rps change requested: %d -> %d\n",
+		  dev_priv->cur_delay, new_delay);
+
+	rgvswctl = I915_READ(MEMSWCTL);
+	if (rgvswctl & MEMCTL_CMD_STS) {
+		DRM_ERROR("gpu busy, RCS change rejected\n");
+		return; /* still busy with another command */
+	}
+
+	/* Program the new state */
+	rgvswctl = (MEMCTL_CMD_CHFREQ << MEMCTL_CMD_SHIFT) |
+		(new_delay << MEMCTL_FREQ_SHIFT) | MEMCTL_SFCAVM;
+	I915_WRITE(MEMSWCTL, rgvswctl);
+	POSTING_READ(MEMSWCTL);
+
+	rgvswctl |= MEMCTL_CMD_STS;
+	I915_WRITE(MEMSWCTL, rgvswctl);
+
+	dev_priv->cur_delay = new_delay;
+
+	DRM_DEBUG("rps changed\n");
+
+	return;
+}
+
+irqreturn_t ironlake_irq_handler(struct drm_device *dev)
 {
 	drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private;
 	int ret = IRQ_NONE;
-	u32 de_iir, gt_iir;
-	u32 new_de_iir, new_gt_iir;
+	u32 de_iir, gt_iir, de_ier, pch_iir;
 	struct drm_i915_master_private *master_priv;
 
+	/* disable master interrupt before clearing iir  */
+	de_ier = I915_READ(DEIER);
+	I915_WRITE(DEIER, de_ier & ~DE_MASTER_IRQ_CONTROL);
+	(void)I915_READ(DEIER);
+
 	de_iir = I915_READ(DEIIR);
 	gt_iir = I915_READ(GTIIR);
+	pch_iir = I915_READ(SDEIIR);
 
-	for (;;) {
-		if (de_iir == 0 && gt_iir == 0)
-			break;
+	if (de_iir == 0 && gt_iir == 0 && pch_iir == 0)
+		goto done;
 
-		ret = IRQ_HANDLED;
+	ret = IRQ_HANDLED;
 
-		I915_WRITE(DEIIR, de_iir);
-		new_de_iir = I915_READ(DEIIR);
-		I915_WRITE(GTIIR, gt_iir);
-		new_gt_iir = I915_READ(GTIIR);
+	if (dev->primary->master) {
+		master_priv = dev->primary->master->driver_priv;
+		if (master_priv->sarea_priv)
+			master_priv->sarea_priv->last_dispatch =
+				READ_BREADCRUMB(dev_priv);
+	}
 
-		if (dev->primary->master) {
-			master_priv = dev->primary->master->driver_priv;
-			if (master_priv->sarea_priv)
-				master_priv->sarea_priv->last_dispatch =
-					READ_BREADCRUMB(dev_priv);
-		}
+	if (gt_iir & GT_USER_INTERRUPT) {
+		u32 seqno = i915_get_gem_seqno(dev);
+		dev_priv->mm.irq_gem_seqno = seqno;
+		trace_i915_gem_request_complete(dev, seqno);
+		DRM_WAKEUP(&dev_priv->irq_queue);
+		dev_priv->hangcheck_count = 0;
+		mod_timer(&dev_priv->hangcheck_timer, jiffies + DRM_I915_HANGCHECK_PERIOD);
+	}
 
-		if (gt_iir & GT_USER_INTERRUPT) {
-			u32 seqno = i915_get_gem_seqno(dev);
-			dev_priv->mm.irq_gem_seqno = seqno;
-			trace_i915_gem_request_complete(dev, seqno);
-			DRM_WAKEUP(&dev_priv->irq_queue);
-		}
+	if (de_iir & DE_GSE)
+		ironlake_opregion_gse_intr(dev);
+
+	if (de_iir & DE_PLANEA_FLIP_DONE) {
+		intel_prepare_page_flip(dev, 0);
+		intel_finish_page_flip(dev, 0);
+	}
 
-		de_iir = new_de_iir;
-		gt_iir = new_gt_iir;
+	if (de_iir & DE_PLANEB_FLIP_DONE) {
+		intel_prepare_page_flip(dev, 1);
+		intel_finish_page_flip(dev, 1);
 	}
 
+	if (de_iir & DE_PIPEA_VBLANK)
+		drm_handle_vblank(dev, 0);
+
+	if (de_iir & DE_PIPEB_VBLANK)
+		drm_handle_vblank(dev, 1);
+
+	/* check event from PCH */
+	if ((de_iir & DE_PCH_EVENT) &&
+	    (pch_iir & SDE_HOTPLUG_MASK)) {
+		queue_work(dev_priv->wq, &dev_priv->hotplug_work);
+	}
+
+	if (de_iir & DE_PCU_EVENT) {
+		I915_WRITE(MEMINTRSTS, I915_READ(MEMINTRSTS));
+		i915_handle_rps_change(dev);
+	}
+
+	/* should clear PCH hotplug event before clear CPU irq */
+	I915_WRITE(SDEIIR, pch_iir);
+	I915_WRITE(GTIIR, gt_iir);
+	I915_WRITE(DEIIR, de_iir);
+
+done:
+	I915_WRITE(DEIER, de_ier);
+	(void)I915_READ(DEIER);
+
 	return ret;
 }
 
@@ -309,23 +415,138 @@ static void i915_error_work_func(struct work_struct *work)
 	char *reset_event[] = { "RESET=1", NULL };
 	char *reset_done_event[] = { "ERROR=0", NULL };
 
-	DRM_DEBUG("generating error event\n");
+	DRM_DEBUG_DRIVER("generating error event\n");
 	kobject_uevent_env(&dev->primary->kdev.kobj, KOBJ_CHANGE, error_event);
 
 	if (atomic_read(&dev_priv->mm.wedged)) {
 		if (IS_I965G(dev)) {
-			DRM_DEBUG("resetting chip\n");
+			DRM_DEBUG_DRIVER("resetting chip\n");
 			kobject_uevent_env(&dev->primary->kdev.kobj, KOBJ_CHANGE, reset_event);
 			if (!i965_reset(dev, GDRST_RENDER)) {
 				atomic_set(&dev_priv->mm.wedged, 0);
 				kobject_uevent_env(&dev->primary->kdev.kobj, KOBJ_CHANGE, reset_done_event);
 			}
 		} else {
-			printk("reboot required\n");
+			DRM_DEBUG_DRIVER("reboot required\n");
 		}
 	}
 }
 
+static struct drm_i915_error_object *
+i915_error_object_create(struct drm_device *dev,
+			 struct drm_gem_object *src)
+{
+	struct drm_i915_error_object *dst;
+	struct drm_i915_gem_object *src_priv;
+	int page, page_count;
+
+	if (src == NULL)
+		return NULL;
+
+	src_priv = src->driver_private;
+	if (src_priv->pages == NULL)
+		return NULL;
+
+	page_count = src->size / PAGE_SIZE;
+
+	dst = kmalloc(sizeof(*dst) + page_count * sizeof (u32 *), GFP_ATOMIC);
+	if (dst == NULL)
+		return NULL;
+
+	for (page = 0; page < page_count; page++) {
+		void *s, *d = kmalloc(PAGE_SIZE, GFP_ATOMIC);
+		if (d == NULL)
+			goto unwind;
+		s = kmap_atomic(src_priv->pages[page], KM_USER0);
+		memcpy(d, s, PAGE_SIZE);
+		kunmap_atomic(s, KM_USER0);
+		dst->pages[page] = d;
+	}
+	dst->page_count = page_count;
+	dst->gtt_offset = src_priv->gtt_offset;
+
+	return dst;
+
+unwind:
+	while (page--)
+		kfree(dst->pages[page]);
+	kfree(dst);
+	return NULL;
+}
+
+static void
+i915_error_object_free(struct drm_i915_error_object *obj)
+{
+	int page;
+
+	if (obj == NULL)
+		return;
+
+	for (page = 0; page < obj->page_count; page++)
+		kfree(obj->pages[page]);
+
+	kfree(obj);
+}
+
+static void
+i915_error_state_free(struct drm_device *dev,
+		      struct drm_i915_error_state *error)
+{
+	i915_error_object_free(error->batchbuffer[0]);
+	i915_error_object_free(error->batchbuffer[1]);
+	i915_error_object_free(error->ringbuffer);
+	kfree(error->active_bo);
+	kfree(error);
+}
+
+static u32
+i915_get_bbaddr(struct drm_device *dev, u32 *ring)
+{
+	u32 cmd;
+
+	if (IS_I830(dev) || IS_845G(dev))
+		cmd = MI_BATCH_BUFFER;
+	else if (IS_I965G(dev))
+		cmd = (MI_BATCH_BUFFER_START | (2 << 6) |
+		       MI_BATCH_NON_SECURE_I965);
+	else
+		cmd = (MI_BATCH_BUFFER_START | (2 << 6));
+
+	return ring[0] == cmd ? ring[1] : 0;
+}
+
+static u32
+i915_ringbuffer_last_batch(struct drm_device *dev)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	u32 head, bbaddr;
+	u32 *ring;
+
+	/* Locate the current position in the ringbuffer and walk back
+	 * to find the most recently dispatched batch buffer.
+	 */
+	bbaddr = 0;
+	head = I915_READ(PRB0_HEAD) & HEAD_ADDR;
+	ring = (u32 *)(dev_priv->ring.virtual_start + head);
+
+	while (--ring >= (u32 *)dev_priv->ring.virtual_start) {
+		bbaddr = i915_get_bbaddr(dev, ring);
+		if (bbaddr)
+			break;
+	}
+
+	if (bbaddr == 0) {
+		ring = (u32 *)(dev_priv->ring.virtual_start + dev_priv->ring.Size);
+		while (--ring >= (u32 *)dev_priv->ring.virtual_start) {
+			bbaddr = i915_get_bbaddr(dev, ring);
+			if (bbaddr)
+				break;
+		}
+	}
+
+	return bbaddr;
+}
+
 /**
  * i915_capture_error_state - capture an error record for later analysis
  * @dev: drm device
@@ -338,19 +559,26 @@ static void i915_error_work_func(struct work_struct *work)
 static void i915_capture_error_state(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct drm_i915_gem_object *obj_priv;
 	struct drm_i915_error_state *error;
+	struct drm_gem_object *batchbuffer[2];
 	unsigned long flags;
+	u32 bbaddr;
+	int count;
 
 	spin_lock_irqsave(&dev_priv->error_lock, flags);
-	if (dev_priv->first_error)
-		goto out;
+	error = dev_priv->first_error;
+	spin_unlock_irqrestore(&dev_priv->error_lock, flags);
+	if (error)
+		return;
 
 	error = kmalloc(sizeof(*error), GFP_ATOMIC);
 	if (!error) {
-		DRM_DEBUG("out ot memory, not capturing error state\n");
-		goto out;
+		DRM_DEBUG_DRIVER("out of memory, not capturing error state\n");
+		return;
 	}
 
+	error->seqno = i915_get_gem_seqno(dev);
 	error->eir = I915_READ(EIR);
 	error->pgtbl_er = I915_READ(PGTBL_ER);
 	error->pipeastat = I915_READ(PIPEASTAT);
@@ -361,6 +589,7 @@ static void i915_capture_error_state(struct drm_device *dev)
 		error->ipehr = I915_READ(IPEHR);
 		error->instdone = I915_READ(INSTDONE);
 		error->acthd = I915_READ(ACTHD);
+		error->bbaddr = 0;
 	} else {
 		error->ipeir = I915_READ(IPEIR_I965);
 		error->ipehr = I915_READ(IPEHR_I965);
@@ -368,14 +597,101 @@ static void i915_capture_error_state(struct drm_device *dev)
 		error->instps = I915_READ(INSTPS);
 		error->instdone1 = I915_READ(INSTDONE1);
 		error->acthd = I915_READ(ACTHD_I965);
+		error->bbaddr = I915_READ64(BB_ADDR);
 	}
 
-	do_gettimeofday(&error->time);
+	bbaddr = i915_ringbuffer_last_batch(dev);
+
+	/* Grab the current batchbuffer, most likely to have crashed. */
+	batchbuffer[0] = NULL;
+	batchbuffer[1] = NULL;
+	count = 0;
+	list_for_each_entry(obj_priv, &dev_priv->mm.active_list, list) {
+		struct drm_gem_object *obj = obj_priv->obj;
 
-	dev_priv->first_error = error;
+		if (batchbuffer[0] == NULL &&
+		    bbaddr >= obj_priv->gtt_offset &&
+		    bbaddr < obj_priv->gtt_offset + obj->size)
+			batchbuffer[0] = obj;
 
-out:
+		if (batchbuffer[1] == NULL &&
+		    error->acthd >= obj_priv->gtt_offset &&
+		    error->acthd < obj_priv->gtt_offset + obj->size &&
+		    batchbuffer[0] != obj)
+			batchbuffer[1] = obj;
+
+		count++;
+	}
+
+	/* We need to copy these to an anonymous buffer as the simplest
+	 * method to avoid being overwritten by userpace.
+	 */
+	error->batchbuffer[0] = i915_error_object_create(dev, batchbuffer[0]);
+	error->batchbuffer[1] = i915_error_object_create(dev, batchbuffer[1]);
+
+	/* Record the ringbuffer */
+	error->ringbuffer = i915_error_object_create(dev, dev_priv->ring.ring_obj);
+
+	/* Record buffers on the active list. */
+	error->active_bo = NULL;
+	error->active_bo_count = 0;
+
+	if (count)
+		error->active_bo = kmalloc(sizeof(*error->active_bo)*count,
+					   GFP_ATOMIC);
+
+	if (error->active_bo) {
+		int i = 0;
+		list_for_each_entry(obj_priv, &dev_priv->mm.active_list, list) {
+			struct drm_gem_object *obj = obj_priv->obj;
+
+			error->active_bo[i].size = obj->size;
+			error->active_bo[i].name = obj->name;
+			error->active_bo[i].seqno = obj_priv->last_rendering_seqno;
+			error->active_bo[i].gtt_offset = obj_priv->gtt_offset;
+			error->active_bo[i].read_domains = obj->read_domains;
+			error->active_bo[i].write_domain = obj->write_domain;
+			error->active_bo[i].fence_reg = obj_priv->fence_reg;
+			error->active_bo[i].pinned = 0;
+			if (obj_priv->pin_count > 0)
+				error->active_bo[i].pinned = 1;
+			if (obj_priv->user_pin_count > 0)
+				error->active_bo[i].pinned = -1;
+			error->active_bo[i].tiling = obj_priv->tiling_mode;
+			error->active_bo[i].dirty = obj_priv->dirty;
+			error->active_bo[i].purgeable = obj_priv->madv != I915_MADV_WILLNEED;
+
+			if (++i == count)
+				break;
+		}
+		error->active_bo_count = i;
+	}
+
+	do_gettimeofday(&error->time);
+
+	spin_lock_irqsave(&dev_priv->error_lock, flags);
+	if (dev_priv->first_error == NULL) {
+		dev_priv->first_error = error;
+		error = NULL;
+	}
 	spin_unlock_irqrestore(&dev_priv->error_lock, flags);
+
+	if (error)
+		i915_error_state_free(dev, error);
+}
+
+void i915_destroy_error_state(struct drm_device *dev)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct drm_i915_error_state *error;
+
+	spin_lock(&dev_priv->error_lock);
+	error = dev_priv->first_error;
+	dev_priv->first_error = NULL;
+	spin_unlock(&dev_priv->error_lock);
+
+	if (error)
+		i915_error_state_free(dev, error);
 }
 
 /**
@@ -504,7 +820,6 @@ static void i915_handle_error(struct drm_device *dev, bool wedged)
 		/*
 		 * Wakeup waiting processes so they don't hang
 		 */
-		printk("i915: Waking up sleeping processes\n");
 		DRM_WAKEUP(&dev_priv->irq_queue);
 	}
 
@@ -527,8 +842,8 @@ irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS)
 
 	atomic_inc(&dev_priv->irq_received);
 
-	if (IS_IGDNG(dev))
-		return igdng_irq_handler(dev);
+	if (HAS_PCH_SPLIT(dev))
+		return ironlake_irq_handler(dev);
 
 	iir = I915_READ(IIR);
 
@@ -560,14 +875,14 @@ irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS)
 		 */
 		if (pipea_stats & 0x8000ffff) {
 			if (pipea_stats &  PIPE_FIFO_UNDERRUN_STATUS)
-				DRM_DEBUG("pipe a underrun\n");
+				DRM_DEBUG_DRIVER("pipe a underrun\n");
 			I915_WRITE(PIPEASTAT, pipea_stats);
 			irq_received = 1;
 		}
 
 		if (pipeb_stats & 0x8000ffff) {
 			if (pipeb_stats &  PIPE_FIFO_UNDERRUN_STATUS)
-				DRM_DEBUG("pipe b underrun\n");
+				DRM_DEBUG_DRIVER("pipe b underrun\n");
 			I915_WRITE(PIPEBSTAT, pipeb_stats);
 			irq_received = 1;
 		}
@@ -583,7 +898,7 @@ irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS)
 		    (iir & I915_DISPLAY_PORT_INTERRUPT)) {
 			u32 hotplug_status = I915_READ(PORT_HOTPLUG_STAT);
 
-			DRM_DEBUG("hotplug event received, stat 0x%08x\n",
+			DRM_DEBUG_DRIVER("hotplug event received, stat 0x%08x\n",
 				  hotplug_status);
 			if (hotplug_status & dev_priv->hotplug_supported_mask)
 				queue_work(dev_priv->wq,
@@ -591,27 +906,6 @@ irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS)
 
 			I915_WRITE(PORT_HOTPLUG_STAT, hotplug_status);
 			I915_READ(PORT_HOTPLUG_STAT);
-
-			/* EOS interrupts occurs */
-			if (IS_IGD(dev) &&
-				(hotplug_status & CRT_EOS_INT_STATUS)) {
-				u32 temp;
-
-				DRM_DEBUG("EOS interrupt occurs\n");
-				/* status is already cleared */
-				temp = I915_READ(ADPA);
-				temp &= ~ADPA_DAC_ENABLE;
-				I915_WRITE(ADPA, temp);
-
-				temp = I915_READ(PORT_HOTPLUG_EN);
-				temp &= ~CRT_EOS_INT_EN;
-				I915_WRITE(PORT_HOTPLUG_EN, temp);
-
-				temp = I915_READ(PORT_HOTPLUG_STAT);
-				if (temp & CRT_EOS_INT_STATUS)
-					I915_WRITE(PORT_HOTPLUG_STAT,
-						CRT_EOS_INT_STATUS);
-			}
 		}
 
 		I915_WRITE(IIR, iir);
@@ -633,14 +927,22 @@ irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS)
 			mod_timer(&dev_priv->hangcheck_timer, jiffies + DRM_I915_HANGCHECK_PERIOD);
 		}
 
+		if (iir & I915_DISPLAY_PLANE_A_FLIP_PENDING_INTERRUPT)
+			intel_prepare_page_flip(dev, 0);
+
+		if (iir & I915_DISPLAY_PLANE_B_FLIP_PENDING_INTERRUPT)
+			intel_prepare_page_flip(dev, 1);
+
 		if (pipea_stats & vblank_status) {
 			vblank++;
 			drm_handle_vblank(dev, 0);
+			intel_finish_page_flip(dev, 0);
 		}
 
 		if (pipeb_stats & vblank_status) {
 			vblank++;
 			drm_handle_vblank(dev, 1);
+			intel_finish_page_flip(dev, 1);
 		}
 
 		if ((pipeb_stats & I915_LEGACY_BLC_EVENT_STATUS) ||
@@ -676,7 +978,7 @@ static int i915_emit_irq(struct drm_device * dev)
 
 	i915_kernel_lost_context(dev);
 
-	DRM_DEBUG("\n");
+	DRM_DEBUG_DRIVER("\n");
 
 	dev_priv->counter++;
 	if (dev_priv->counter > 0x7FFFFFFFUL)
@@ -701,8 +1003,8 @@ void i915_user_irq_get(struct drm_device *dev)
 
 	spin_lock_irqsave(&dev_priv->user_irq_lock, irqflags);
 	if (dev->irq_enabled && (++dev_priv->user_irq_refcount == 1)) {
-		if (IS_IGDNG(dev))
-			igdng_enable_graphics_irq(dev_priv, GT_USER_INTERRUPT);
+		if (HAS_PCH_SPLIT(dev))
+			ironlake_enable_graphics_irq(dev_priv, GT_USER_INTERRUPT);
 		else
 			i915_enable_irq(dev_priv, I915_USER_INTERRUPT);
 	}
@@ -717,21 +1019,31 @@ void i915_user_irq_put(struct drm_device *dev)
 	spin_lock_irqsave(&dev_priv->user_irq_lock, irqflags);
 	BUG_ON(dev->irq_enabled && dev_priv->user_irq_refcount <= 0);
 	if (dev->irq_enabled && (--dev_priv->user_irq_refcount == 0)) {
-		if (IS_IGDNG(dev))
-			igdng_disable_graphics_irq(dev_priv, GT_USER_INTERRUPT);
+		if (HAS_PCH_SPLIT(dev))
+			ironlake_disable_graphics_irq(dev_priv, GT_USER_INTERRUPT);
 		else
 			i915_disable_irq(dev_priv, I915_USER_INTERRUPT);
 	}
 	spin_unlock_irqrestore(&dev_priv->user_irq_lock, irqflags);
 }
 
+void i915_trace_irq_get(struct drm_device *dev, u32 seqno)
+{
+	drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private;
+
+	if (dev_priv->trace_irq_seqno == 0)
+		i915_user_irq_get(dev);
+
+	dev_priv->trace_irq_seqno = seqno;
+}
+
 static int i915_wait_irq(struct drm_device * dev, int irq_nr)
 {
 	drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private;
 	struct drm_i915_master_private *master_priv = dev->primary->master->driver_priv;
 	int ret = 0;
 
-	DRM_DEBUG("irq_nr=%d breadcrumb=%d\n", irq_nr,
+	DRM_DEBUG_DRIVER("irq_nr=%d breadcrumb=%d\n", irq_nr,
 		  READ_BREADCRUMB(dev_priv));
 
 	if (READ_BREADCRUMB(dev_priv) >= irq_nr) {
@@ -814,11 +1126,11 @@ int i915_enable_vblank(struct drm_device *dev, int pipe)
 	if (!(pipeconf & PIPEACONF_ENABLE))
 		return -EINVAL;
 
-	if (IS_IGDNG(dev))
-		return 0;
-
 	spin_lock_irqsave(&dev_priv->user_irq_lock, irqflags);
-	if (IS_I965G(dev))
+	if (HAS_PCH_SPLIT(dev))
+		ironlake_enable_display_irq(dev_priv, (pipe == 0) ? 
+					    DE_PIPEA_VBLANK: DE_PIPEB_VBLANK);
+	else if (IS_I965G(dev))
 		i915_enable_pipestat(dev_priv, pipe,
 				     PIPE_START_VBLANK_INTERRUPT_ENABLE);
 	else
@@ -836,13 +1148,14 @@ void i915_disable_vblank(struct drm_device *dev, int pipe)
 	drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private;
 	unsigned long irqflags;
 
-	if (IS_IGDNG(dev))
-		return;
-
 	spin_lock_irqsave(&dev_priv->user_irq_lock, irqflags);
-	i915_disable_pipestat(dev_priv, pipe,
-			      PIPE_VBLANK_INTERRUPT_ENABLE |
-			      PIPE_START_VBLANK_INTERRUPT_ENABLE);
+	if (HAS_PCH_SPLIT(dev))
+		ironlake_disable_display_irq(dev_priv, (pipe == 0) ? 
+					     DE_PIPEA_VBLANK: DE_PIPEB_VBLANK);
+	else
+		i915_disable_pipestat(dev_priv, pipe,
+				      PIPE_VBLANK_INTERRUPT_ENABLE |
+				      PIPE_START_VBLANK_INTERRUPT_ENABLE);
 	spin_unlock_irqrestore(&dev_priv->user_irq_lock, irqflags);
 }
 
@@ -850,7 +1163,7 @@ void i915_enable_interrupt (struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
 
-	if (!IS_IGDNG(dev))
+	if (!HAS_PCH_SPLIT(dev))
 		opregion_enable_asle(dev);
 	dev_priv->irq_enabled = 1;
 }
@@ -926,7 +1239,11 @@ void i915_hangcheck_elapsed(unsigned long data)
 	struct drm_device *dev = (struct drm_device *)data;
 	drm_i915_private_t *dev_priv = dev->dev_private;
 	uint32_t acthd;
-       
+
+	/* No reset support on this chip yet. */
+	if (IS_GEN6(dev))
+		return;
+
 	if (!IS_I965G(dev))
 		acthd = I915_READ(ACTHD);
 	else
@@ -958,7 +1275,7 @@ void i915_hangcheck_elapsed(unsigned long data)
 
 /* drm_dma.h hooks
 */
-static void igdng_irq_preinstall(struct drm_device *dev)
+static void ironlake_irq_preinstall(struct drm_device *dev)
 {
 	drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private;
 
@@ -974,17 +1291,25 @@ static void igdng_irq_preinstall(struct drm_device *dev)
 	I915_WRITE(GTIMR, 0xffffffff);
 	I915_WRITE(GTIER, 0x0);
 	(void) I915_READ(GTIER);
+
+	/* south display irq */
+	I915_WRITE(SDEIMR, 0xffffffff);
+	I915_WRITE(SDEIER, 0x0);
+	(void) I915_READ(SDEIER);
 }
 
-static int igdng_irq_postinstall(struct drm_device *dev)
+static int ironlake_irq_postinstall(struct drm_device *dev)
 {
 	drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private;
 	/* enable kind of interrupts always enabled */
-	u32 display_mask = DE_MASTER_IRQ_CONTROL /*| DE_PCH_EVENT */;
+	u32 display_mask = DE_MASTER_IRQ_CONTROL | DE_GSE | DE_PCH_EVENT |
+			   DE_PLANEA_FLIP_DONE | DE_PLANEB_FLIP_DONE;
 	u32 render_mask = GT_USER_INTERRUPT;
+	u32 hotplug_mask = SDE_CRT_HOTPLUG | SDE_PORTB_HOTPLUG |
+			   SDE_PORTC_HOTPLUG | SDE_PORTD_HOTPLUG;
 
 	dev_priv->irq_mask_reg = ~display_mask;
-	dev_priv->de_irq_enable_reg = display_mask;
+	dev_priv->de_irq_enable_reg = display_mask | DE_PIPEA_VBLANK | DE_PIPEB_VBLANK;
 
 	/* should always can generate irq */
 	I915_WRITE(DEIIR, I915_READ(DEIIR));
@@ -1001,6 +1326,21 @@ static int igdng_irq_postinstall(struct drm_device *dev)
 	I915_WRITE(GTIER, dev_priv->gt_irq_enable_reg);
 	(void) I915_READ(GTIER);
 
+	dev_priv->pch_irq_mask_reg = ~hotplug_mask;
+	dev_priv->pch_irq_enable_reg = hotplug_mask;
+
+	I915_WRITE(SDEIIR, I915_READ(SDEIIR));
+	I915_WRITE(SDEIMR, dev_priv->pch_irq_mask_reg);
+	I915_WRITE(SDEIER, dev_priv->pch_irq_enable_reg);
+	(void) I915_READ(SDEIER);
+
+	if (IS_IRONLAKE_M(dev)) {
+		/* Clear & enable PCU event interrupts */
+		I915_WRITE(DEIIR, DE_PCU_EVENT);
+		I915_WRITE(DEIER, I915_READ(DEIER) | DE_PCU_EVENT);
+		ironlake_enable_display_irq(dev_priv, DE_PCU_EVENT);
+	}
+
 	return 0;
 }
 
@@ -1013,8 +1353,8 @@ void i915_driver_irq_preinstall(struct drm_device * dev)
 	INIT_WORK(&dev_priv->hotplug_work, i915_hotplug_work_func);
 	INIT_WORK(&dev_priv->error_work, i915_error_work_func);
 
-	if (IS_IGDNG(dev)) {
-		igdng_irq_preinstall(dev);
+	if (HAS_PCH_SPLIT(dev)) {
+		ironlake_irq_preinstall(dev);
 		return;
 	}
 
@@ -1031,6 +1371,10 @@ void i915_driver_irq_preinstall(struct drm_device * dev)
 	(void) I915_READ(IER);
 }
 
+/*
+ * Must be called after intel_modeset_init or hotplug interrupts won't be
+ * enabled correctly.
+ */
 int i915_driver_irq_postinstall(struct drm_device *dev)
 {
 	drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private;
@@ -1041,8 +1385,8 @@ int i915_driver_irq_postinstall(struct drm_device *dev)
 
 	dev_priv->vblank_pipe = DRM_I915_VBLANK_PIPE_A | DRM_I915_VBLANK_PIPE_B;
 
-	if (IS_IGDNG(dev))
-		return igdng_irq_postinstall(dev);
+	if (HAS_PCH_SPLIT(dev))
+		return ironlake_irq_postinstall(dev);
 
 	/* Unmask the interrupts that we always want on. */
 	dev_priv->irq_mask_reg = ~I915_INTERRUPT_ENABLE_FIX;
@@ -1053,19 +1397,23 @@ int i915_driver_irq_postinstall(struct drm_device *dev)
 	if (I915_HAS_HOTPLUG(dev)) {
 		u32 hotplug_en = I915_READ(PORT_HOTPLUG_EN);
 
-		/* Leave other bits alone */
-		hotplug_en |= HOTPLUG_EN_MASK;
+		/* Note HDMI and DP share bits */
+		if (dev_priv->hotplug_supported_mask & HDMIB_HOTPLUG_INT_STATUS)
+			hotplug_en |= HDMIB_HOTPLUG_INT_EN;
+		if (dev_priv->hotplug_supported_mask & HDMIC_HOTPLUG_INT_STATUS)
+			hotplug_en |= HDMIC_HOTPLUG_INT_EN;
+		if (dev_priv->hotplug_supported_mask & HDMID_HOTPLUG_INT_STATUS)
+			hotplug_en |= HDMID_HOTPLUG_INT_EN;
+		if (dev_priv->hotplug_supported_mask & SDVOC_HOTPLUG_INT_STATUS)
+			hotplug_en |= SDVOC_HOTPLUG_INT_EN;
+		if (dev_priv->hotplug_supported_mask & SDVOB_HOTPLUG_INT_STATUS)
+			hotplug_en |= SDVOB_HOTPLUG_INT_EN;
+		if (dev_priv->hotplug_supported_mask & CRT_HOTPLUG_INT_STATUS)
+			hotplug_en |= CRT_HOTPLUG_INT_EN;
+		/* Ignore TV since it's buggy */
+
 		I915_WRITE(PORT_HOTPLUG_EN, hotplug_en);
 
-		dev_priv->hotplug_supported_mask = CRT_HOTPLUG_INT_STATUS |
-			TV_HOTPLUG_INT_STATUS | SDVOC_HOTPLUG_INT_STATUS |
-			SDVOB_HOTPLUG_INT_STATUS;
-		if (IS_G4X(dev)) {
-			dev_priv->hotplug_supported_mask |=
-				HDMIB_HOTPLUG_INT_STATUS |
-				HDMIC_HOTPLUG_INT_STATUS |
-				HDMID_HOTPLUG_INT_STATUS;
-		}
 		/* Enable in IER... */
 		enable_mask |= I915_DISPLAY_PORT_INTERRUPT;
 		/* and unmask in IMR */
@@ -1102,7 +1450,7 @@ int i915_driver_irq_postinstall(struct drm_device *dev)
 	return 0;
 }
 
-static void igdng_irq_uninstall(struct drm_device *dev)
+static void ironlake_irq_uninstall(struct drm_device *dev)
 {
 	drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private;
 	I915_WRITE(HWSTAM, 0xffffffff);
@@ -1125,8 +1473,8 @@ void i915_driver_irq_uninstall(struct drm_device * dev)
 
 	dev_priv->vblank_pipe = 0;
 
-	if (IS_IGDNG(dev)) {
-		igdng_irq_uninstall(dev);
+	if (HAS_PCH_SPLIT(dev)) {
+		ironlake_irq_uninstall(dev);
 		return;
 	}
 
diff --git a/drivers/gpu/drm/i915/i915_opregion.c b/drivers/gpu/drm/i915/i915_opregion.c
index 2d5193556d3f..7cc8410239cb 100644
--- a/drivers/gpu/drm/i915/i915_opregion.c
+++ b/drivers/gpu/drm/i915/i915_opregion.c
@@ -118,6 +118,10 @@ struct opregion_asle {
 #define ASLE_BACKLIGHT_FAIL    (2<<12)
 #define ASLE_PFIT_FAIL         (2<<14)
 #define ASLE_PWM_FREQ_FAIL     (2<<16)
+#define ASLE_ALS_ILLUM_FAILED	(1<<10)
+#define ASLE_BACKLIGHT_FAILED	(1<<12)
+#define ASLE_PFIT_FAILED	(1<<14)
+#define ASLE_PWM_FREQ_FAILED	(1<<16)
 
 /* ASLE backlight brightness to set */
 #define ASLE_BCLP_VALID                (1<<31)
@@ -163,7 +167,7 @@ static u32 asle_set_backlight(struct drm_device *dev, u32 bclp)
 	if (IS_I965G(dev) && (blc_pwm_ctl2 & BLM_COMBINATION_MODE))
 		pci_write_config_dword(dev->pdev, PCI_LBPC, bclp);
 	else {
-		if (IS_IGD(dev)) {
+		if (IS_PINEVIEW(dev)) {
 			blc_pwm_ctl &= ~(BACKLIGHT_DUTY_CYCLE_MASK - 1);
 			max_backlight = (blc_pwm_ctl & BACKLIGHT_MODULATION_FREQ_MASK) >> 
 					BACKLIGHT_MODULATION_FREQ_SHIFT;
@@ -224,7 +228,7 @@ void opregion_asle_intr(struct drm_device *dev)
 	asle_req = asle->aslc & ASLE_REQ_MSK;
 
 	if (!asle_req) {
-		DRM_DEBUG("non asle set request??\n");
+		DRM_DEBUG_DRIVER("non asle set request??\n");
 		return;
 	}
 
@@ -243,6 +247,73 @@ void opregion_asle_intr(struct drm_device *dev)
 	asle->aslc = asle_stat;
 }
 
+static u32 asle_set_backlight_ironlake(struct drm_device *dev, u32 bclp)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct opregion_asle *asle = dev_priv->opregion.asle;
+	u32 cpu_pwm_ctl, pch_pwm_ctl2;
+	u32 max_backlight, level;
+
+	if (!(bclp & ASLE_BCLP_VALID))
+		return ASLE_BACKLIGHT_FAILED;
+
+	bclp &= ASLE_BCLP_MSK;
+	if (bclp < 0 || bclp > 255)
+		return ASLE_BACKLIGHT_FAILED;
+
+	cpu_pwm_ctl = I915_READ(BLC_PWM_CPU_CTL);
+	pch_pwm_ctl2 = I915_READ(BLC_PWM_PCH_CTL2);
+	/* get the max PWM frequency */
+	max_backlight = (pch_pwm_ctl2 >> 16) & BACKLIGHT_DUTY_CYCLE_MASK;
+	/* calculate the expected PMW frequency */
+	level = (bclp * max_backlight) / 255;
+	/* reserve the high 16 bits */
+	cpu_pwm_ctl &= ~(BACKLIGHT_DUTY_CYCLE_MASK);
+	/* write the updated PWM frequency */
+	I915_WRITE(BLC_PWM_CPU_CTL, cpu_pwm_ctl | level);
+
+	asle->cblv = (bclp*0x64)/0xff | ASLE_CBLV_VALID;
+
+	return 0;
+}
+
+void ironlake_opregion_gse_intr(struct drm_device *dev)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct opregion_asle *asle = dev_priv->opregion.asle;
+	u32 asle_stat = 0;
+	u32 asle_req;
+
+	if (!asle)
+		return;
+
+	asle_req = asle->aslc & ASLE_REQ_MSK;
+
+	if (!asle_req) {
+		DRM_DEBUG_DRIVER("non asle set request??\n");
+		return;
+	}
+
+	if (asle_req & ASLE_SET_ALS_ILLUM) {
+		DRM_DEBUG_DRIVER("Illum is not supported\n");
+		asle_stat |= ASLE_ALS_ILLUM_FAILED;
+	}
+
+	if (asle_req & ASLE_SET_BACKLIGHT)
+		asle_stat |= asle_set_backlight_ironlake(dev, asle->bclp);
+
+	if (asle_req & ASLE_SET_PFIT) {
+		DRM_DEBUG_DRIVER("Pfit is not supported\n");
+		asle_stat |= ASLE_PFIT_FAILED;
+	}
+
+	if (asle_req & ASLE_SET_PWM_FREQ) {
+		DRM_DEBUG_DRIVER("PWM freq is not supported\n");
+		asle_stat |= ASLE_PWM_FREQ_FAILED;
+	}
+
+	asle->aslc = asle_stat;
+}
 #define ASLE_ALS_EN    (1<<0)
 #define ASLE_BLC_EN    (1<<1)
 #define ASLE_PFIT_EN   (1<<2)
@@ -258,8 +329,7 @@ void opregion_enable_asle(struct drm_device *dev)
 			unsigned long irqflags;
 
 			spin_lock_irqsave(&dev_priv->user_irq_lock, irqflags);
-			i915_enable_pipestat(dev_priv, 1,
-					     I915_LEGACY_BLC_EVENT_ENABLE);
+			intel_enable_asle(dev);
 			spin_unlock_irqrestore(&dev_priv->user_irq_lock,
 					       irqflags);
 		}
@@ -361,9 +431,9 @@ int intel_opregion_init(struct drm_device *dev, int resume)
 	int err = 0;
 
 	pci_read_config_dword(dev->pdev, PCI_ASLS, &asls);
-	DRM_DEBUG("graphic opregion physical addr: 0x%x\n", asls);
+	DRM_DEBUG_DRIVER("graphic opregion physical addr: 0x%x\n", asls);
 	if (asls == 0) {
-		DRM_DEBUG("ACPI OpRegion not supported!\n");
+		DRM_DEBUG_DRIVER("ACPI OpRegion not supported!\n");
 		return -ENOTSUPP;
 	}
 
@@ -373,30 +443,30 @@ int intel_opregion_init(struct drm_device *dev, int resume)
 
 	opregion->header = base;
 	if (memcmp(opregion->header->signature, OPREGION_SIGNATURE, 16)) {
-		DRM_DEBUG("opregion signature mismatch\n");
+		DRM_DEBUG_DRIVER("opregion signature mismatch\n");
 		err = -EINVAL;
 		goto err_out;
 	}
 
 	mboxes = opregion->header->mboxes;
 	if (mboxes & MBOX_ACPI) {
-		DRM_DEBUG("Public ACPI methods supported\n");
+		DRM_DEBUG_DRIVER("Public ACPI methods supported\n");
 		opregion->acpi = base + OPREGION_ACPI_OFFSET;
 		if (drm_core_check_feature(dev, DRIVER_MODESET))
 			intel_didl_outputs(dev);
 	} else {
-		DRM_DEBUG("Public ACPI methods not supported\n");
+		DRM_DEBUG_DRIVER("Public ACPI methods not supported\n");
 		err = -ENOTSUPP;
 		goto err_out;
 	}
 	opregion->enabled = 1;
 
 	if (mboxes & MBOX_SWSCI) {
-		DRM_DEBUG("SWSCI supported\n");
+		DRM_DEBUG_DRIVER("SWSCI supported\n");
 		opregion->swsci = base + OPREGION_SWSCI_OFFSET;
 	}
 	if (mboxes & MBOX_ASLE) {
-		DRM_DEBUG("ASLE supported\n");
+		DRM_DEBUG_DRIVER("ASLE supported\n");
 		opregion->asle = base + OPREGION_ASLE_OFFSET;
 		opregion_enable_asle(dev);
 	}
diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
index 0466ddbeba32..3d59862c7ccd 100644
--- a/drivers/gpu/drm/i915/i915_reg.h
+++ b/drivers/gpu/drm/i915/i915_reg.h
@@ -53,6 +53,25 @@
 #define INTEL_GMCH_GMS_STOLEN_224M	(0xc << 4)
 #define INTEL_GMCH_GMS_STOLEN_352M	(0xd << 4)
 
+#define SNB_GMCH_CTRL	0x50
+#define SNB_GMCH_GMS_STOLEN_MASK	0xF8
+#define SNB_GMCH_GMS_STOLEN_32M		(1 << 3)
+#define SNB_GMCH_GMS_STOLEN_64M		(2 << 3)
+#define SNB_GMCH_GMS_STOLEN_96M		(3 << 3)
+#define SNB_GMCH_GMS_STOLEN_128M	(4 << 3)
+#define SNB_GMCH_GMS_STOLEN_160M	(5 << 3)
+#define SNB_GMCH_GMS_STOLEN_192M	(6 << 3)
+#define SNB_GMCH_GMS_STOLEN_224M	(7 << 3)
+#define SNB_GMCH_GMS_STOLEN_256M	(8 << 3)
+#define SNB_GMCH_GMS_STOLEN_288M	(9 << 3)
+#define SNB_GMCH_GMS_STOLEN_320M	(0xa << 3)
+#define SNB_GMCH_GMS_STOLEN_352M	(0xb << 3)
+#define SNB_GMCH_GMS_STOLEN_384M	(0xc << 3)
+#define SNB_GMCH_GMS_STOLEN_416M	(0xd << 3)
+#define SNB_GMCH_GMS_STOLEN_448M	(0xe << 3)
+#define SNB_GMCH_GMS_STOLEN_480M	(0xf << 3)
+#define SNB_GMCH_GMS_STOLEN_512M	(0x10 << 3)
+
 /* PCI config space */
 
 #define HPLLCC	0xc0 /* 855 only */
@@ -61,6 +80,7 @@
 #define   GC_CLOCK_100_200		(1 << 0)
 #define   GC_CLOCK_100_133		(2 << 0)
 #define   GC_CLOCK_166_250		(3 << 0)
+#define GCFGC2	0xda
 #define GCFGC	0xf0 /* 915+ only */
 #define   GC_LOW_FREQUENCY_ENABLE	(1 << 7)
 #define   GC_DISPLAY_CLOCK_190_200_MHZ	(0 << 4)
@@ -140,6 +160,7 @@
 #define MI_NOOP			MI_INSTR(0, 0)
 #define MI_USER_INTERRUPT	MI_INSTR(0x02, 0)
 #define MI_WAIT_FOR_EVENT       MI_INSTR(0x03, 0)
+#define   MI_WAIT_FOR_OVERLAY_FLIP	(1<<16)
 #define   MI_WAIT_FOR_PLANE_B_FLIP      (1<<6)
 #define   MI_WAIT_FOR_PLANE_A_FLIP      (1<<2)
 #define   MI_WAIT_FOR_PLANE_A_SCANLINES (1<<1)
@@ -151,7 +172,13 @@
 #define   MI_END_SCENE		(1 << 4) /* flush binner and incr scene count */
 #define MI_BATCH_BUFFER_END	MI_INSTR(0x0a, 0)
 #define MI_REPORT_HEAD		MI_INSTR(0x07, 0)
+#define MI_OVERLAY_FLIP		MI_INSTR(0x11,0)
+#define   MI_OVERLAY_CONTINUE	(0x0<<21)
+#define   MI_OVERLAY_ON		(0x1<<21)
+#define   MI_OVERLAY_OFF	(0x2<<21)
 #define MI_LOAD_SCAN_LINES_INCL MI_INSTR(0x12, 0)
+#define MI_DISPLAY_FLIP		MI_INSTR(0x14, 2)
+#define   MI_DISPLAY_FLIP_PLANE(n) ((n) << 20)
 #define MI_STORE_DWORD_IMM	MI_INSTR(0x20, 1)
 #define   MI_MEM_VIRTUAL	(1 << 22) /* 965+ only */
 #define MI_STORE_DWORD_INDEX	MI_INSTR(0x21, 1)
@@ -227,6 +254,9 @@
 #define   I965_FENCE_REG_VALID		(1<<0)
 #define   I965_FENCE_MAX_PITCH_VAL	0x0400
 
+#define FENCE_REG_SANDYBRIDGE_0		0x100000
+#define   SANDYBRIDGE_FENCE_PITCH_SHIFT	32
+
 /*
  * Instruction and interrupt control regs
  */
@@ -258,8 +288,11 @@
 #define INSTDONE1	0x0207c /* 965+ only */
 #define ACTHD_I965	0x02074
 #define HWS_PGA		0x02080
+#define HWS_PGA_GEN6	0x04080
 #define HWS_ADDRESS_MASK	0xfffff000
 #define HWS_START_ADDRESS_SHIFT	4
+#define PWRCTXA		0x2088 /* 965GM+ only */
+#define   PWRCTX_EN	(1<<0)
 #define IPEIR		0x02088
 #define IPEHR		0x0208c
 #define INSTDONE	0x02090
@@ -273,7 +306,7 @@
 #define   I915_PIPE_CONTROL_NOTIFY_INTERRUPT		(1<<18)
 #define   I915_DISPLAY_PORT_INTERRUPT			(1<<17)
 #define   I915_RENDER_COMMAND_PARSER_ERROR_INTERRUPT	(1<<15)
-#define   I915_GMCH_THERMAL_SENSOR_EVENT_INTERRUPT	(1<<14)
+#define   I915_GMCH_THERMAL_SENSOR_EVENT_INTERRUPT	(1<<14) /* p-state */
 #define   I915_HWB_OOM_INTERRUPT			(1<<13)
 #define   I915_SYNC_STATUS_INTERRUPT			(1<<12)
 #define   I915_DISPLAY_PLANE_A_FLIP_PENDING_INTERRUPT	(1<<11)
@@ -297,11 +330,14 @@
 #define   I915_ERROR_MEMORY_REFRESH			(1<<1)
 #define   I915_ERROR_INSTRUCTION			(1<<0)
 #define INSTPM	        0x020c0
+#define   INSTPM_SELF_EN (1<<12) /* 915GM only */
 #define ACTHD	        0x020c8
 #define FW_BLC		0x020d8
 #define FW_BLC2	 	0x020dc
 #define FW_BLC_SELF	0x020e0 /* 915+ only */
-#define   FW_BLC_SELF_EN (1<<15)
+#define   FW_BLC_SELF_EN_MASK      (1<<31)
+#define   FW_BLC_SELF_FIFO_MASK    (1<<16) /* 945 only */
+#define   FW_BLC_SELF_EN           (1<<15) /* 945 only */
 #define MM_BURST_LENGTH     0x00700000
 #define MM_FIFO_WATERMARK   0x0001F000
 #define LM_BURST_LENGTH     0x00000700
@@ -315,6 +351,7 @@
 #define   CM0_COLOR_EVICT_DISABLE (1<<3)
 #define   CM0_DEPTH_WRITE_DISABLE (1<<1)
 #define   CM0_RC_OP_FLUSH_DISABLE (1<<0)
+#define BB_ADDR		0x02140 /* 8 bytes */
 #define GFX_FLSH_CNTL	0x02170 /* 915+ only */
 
 
@@ -329,6 +366,7 @@
 #define   FBC_CTL_PERIODIC	(1<<30)
 #define   FBC_CTL_INTERVAL_SHIFT (16)
 #define   FBC_CTL_UNCOMPRESSIBLE (1<<14)
+#define   FBC_C3_IDLE		(1<<13)
 #define   FBC_CTL_STRIDE_SHIFT	(5)
 #define   FBC_CTL_FENCENO	(1<<0)
 #define FBC_COMMAND		0x0320c
@@ -405,6 +443,13 @@
 # define GPIO_DATA_VAL_IN		(1 << 12)
 # define GPIO_DATA_PULLUP_DISABLE	(1 << 13)
 
+#define GMBUS0			0x5100
+#define GMBUS1			0x5104
+#define GMBUS2			0x5108
+#define GMBUS3			0x510c
+#define GMBUS4			0x5110
+#define GMBUS5			0x5120
+
 /*
  * Clock control & power management
  */
@@ -435,7 +480,7 @@
 #define   DPLLB_LVDS_P2_CLOCK_DIV_7	(1 << 24) /* i915 */
 #define   DPLL_P2_CLOCK_DIV_MASK	0x03000000 /* i915 */
 #define   DPLL_FPA01_P1_POST_DIV_MASK	0x00ff0000 /* i915 */
-#define   DPLL_FPA01_P1_POST_DIV_MASK_IGD	0x00ff8000 /* IGD */
+#define   DPLL_FPA01_P1_POST_DIV_MASK_PINEVIEW	0x00ff8000 /* Pineview */
 
 #define I915_FIFO_UNDERRUN_STATUS		(1UL<<31)
 #define I915_CRC_ERROR_ENABLE			(1UL<<29)
@@ -512,7 +557,7 @@
  */
 #define   DPLL_FPA01_P1_POST_DIV_MASK_I830_LVDS	0x003f0000
 #define   DPLL_FPA01_P1_POST_DIV_SHIFT	16
-#define   DPLL_FPA01_P1_POST_DIV_SHIFT_IGD 15
+#define   DPLL_FPA01_P1_POST_DIV_SHIFT_PINEVIEW 15
 /* i830, required in DVO non-gang */
 #define   PLL_P2_DIVIDE_BY_4		(1 << 23)
 #define   PLL_P1_DIVIDE_BY_TWO		(1 << 21) /* i830 */
@@ -522,7 +567,7 @@
 #define   PLLB_REF_INPUT_SPREADSPECTRUMIN (3 << 13)
 #define   PLL_REF_INPUT_MASK		(3 << 13)
 #define   PLL_LOAD_PULSE_PHASE_SHIFT		9
-/* IGDNG */
+/* Ironlake */
 # define PLL_REF_SDVO_HDMI_MULTIPLIER_SHIFT     9
 # define PLL_REF_SDVO_HDMI_MULTIPLIER_MASK      (7 << 9)
 # define PLL_REF_SDVO_HDMI_MULTIPLIER(x)	(((x)-1) << 9)
@@ -586,12 +631,12 @@
 #define FPB0	0x06048
 #define FPB1	0x0604c
 #define   FP_N_DIV_MASK		0x003f0000
-#define   FP_N_IGD_DIV_MASK	0x00ff0000
+#define   FP_N_PINEVIEW_DIV_MASK	0x00ff0000
 #define   FP_N_DIV_SHIFT		16
 #define   FP_M1_DIV_MASK	0x00003f00
 #define   FP_M1_DIV_SHIFT		 8
 #define   FP_M2_DIV_MASK	0x0000003f
-#define   FP_M2_IGD_DIV_MASK	0x000000ff
+#define   FP_M2_PINEVIEW_DIV_MASK	0x000000ff
 #define   FP_M2_DIV_SHIFT		 0
 #define DPLL_TEST	0x606c
 #define   DPLLB_TEST_SDVO_DIV_1		(0 << 22)
@@ -767,9 +812,144 @@
 #define CLKCFG_MEM_800					(3 << 4)
 #define CLKCFG_MEM_MASK					(7 << 4)
 
-/** GM965 GM45 render standby register */
-#define MCHBAR_RENDER_STANDBY	0x111B8
-
+#define CRSTANDVID		0x11100
+#define PXVFREQ_BASE		0x11110 /* P[0-15]VIDFREQ (0x1114c) (Ironlake) */
+#define   PXVFREQ_PX_MASK	0x7f000000
+#define   PXVFREQ_PX_SHIFT	24
+#define VIDFREQ_BASE		0x11110
+#define VIDFREQ1		0x11110 /* VIDFREQ1-4 (0x1111c) (Cantiga) */
+#define VIDFREQ2		0x11114
+#define VIDFREQ3		0x11118
+#define VIDFREQ4		0x1111c
+#define   VIDFREQ_P0_MASK	0x1f000000
+#define   VIDFREQ_P0_SHIFT	24
+#define   VIDFREQ_P0_CSCLK_MASK	0x00f00000
+#define   VIDFREQ_P0_CSCLK_SHIFT 20
+#define   VIDFREQ_P0_CRCLK_MASK	0x000f0000
+#define   VIDFREQ_P0_CRCLK_SHIFT 16
+#define   VIDFREQ_P1_MASK	0x00001f00
+#define   VIDFREQ_P1_SHIFT	8
+#define   VIDFREQ_P1_CSCLK_MASK	0x000000f0
+#define   VIDFREQ_P1_CSCLK_SHIFT 4
+#define   VIDFREQ_P1_CRCLK_MASK	0x0000000f
+#define INTTOEXT_BASE_ILK	0x11300
+#define INTTOEXT_BASE		0x11120 /* INTTOEXT1-8 (0x1113c) */
+#define   INTTOEXT_MAP3_SHIFT	24
+#define   INTTOEXT_MAP3_MASK	(0x1f << INTTOEXT_MAP3_SHIFT)
+#define   INTTOEXT_MAP2_SHIFT	16
+#define   INTTOEXT_MAP2_MASK	(0x1f << INTTOEXT_MAP2_SHIFT)
+#define   INTTOEXT_MAP1_SHIFT	8
+#define   INTTOEXT_MAP1_MASK	(0x1f << INTTOEXT_MAP1_SHIFT)
+#define   INTTOEXT_MAP0_SHIFT	0
+#define   INTTOEXT_MAP0_MASK	(0x1f << INTTOEXT_MAP0_SHIFT)
+#define MEMSWCTL		0x11170 /* Ironlake only */
+#define   MEMCTL_CMD_MASK	0xe000
+#define   MEMCTL_CMD_SHIFT	13
+#define   MEMCTL_CMD_RCLK_OFF	0
+#define   MEMCTL_CMD_RCLK_ON	1
+#define   MEMCTL_CMD_CHFREQ	2
+#define   MEMCTL_CMD_CHVID	3
+#define   MEMCTL_CMD_VMMOFF	4
+#define   MEMCTL_CMD_VMMON	5
+#define   MEMCTL_CMD_STS	(1<<12) /* write 1 triggers command, clears
+					   when command complete */
+#define   MEMCTL_FREQ_MASK	0x0f00 /* jitter, from 0-15 */
+#define   MEMCTL_FREQ_SHIFT	8
+#define   MEMCTL_SFCAVM		(1<<7)
+#define   MEMCTL_TGT_VID_MASK	0x007f
+#define MEMIHYST		0x1117c
+#define MEMINTREN		0x11180 /* 16 bits */
+#define   MEMINT_RSEXIT_EN	(1<<8)
+#define   MEMINT_CX_SUPR_EN	(1<<7)
+#define   MEMINT_CONT_BUSY_EN	(1<<6)
+#define   MEMINT_AVG_BUSY_EN	(1<<5)
+#define   MEMINT_EVAL_CHG_EN	(1<<4)
+#define   MEMINT_MON_IDLE_EN	(1<<3)
+#define   MEMINT_UP_EVAL_EN	(1<<2)
+#define   MEMINT_DOWN_EVAL_EN	(1<<1)
+#define   MEMINT_SW_CMD_EN	(1<<0)
+#define MEMINTRSTR		0x11182 /* 16 bits */
+#define   MEM_RSEXIT_MASK	0xc000
+#define   MEM_RSEXIT_SHIFT	14
+#define   MEM_CONT_BUSY_MASK	0x3000
+#define   MEM_CONT_BUSY_SHIFT	12
+#define   MEM_AVG_BUSY_MASK	0x0c00
+#define   MEM_AVG_BUSY_SHIFT	10
+#define   MEM_EVAL_CHG_MASK	0x0300
+#define   MEM_EVAL_BUSY_SHIFT	8
+#define   MEM_MON_IDLE_MASK	0x00c0
+#define   MEM_MON_IDLE_SHIFT	6
+#define   MEM_UP_EVAL_MASK	0x0030
+#define   MEM_UP_EVAL_SHIFT	4
+#define   MEM_DOWN_EVAL_MASK	0x000c
+#define   MEM_DOWN_EVAL_SHIFT	2
+#define   MEM_SW_CMD_MASK	0x0003
+#define   MEM_INT_STEER_GFX	0
+#define   MEM_INT_STEER_CMR	1
+#define   MEM_INT_STEER_SMI	2
+#define   MEM_INT_STEER_SCI	3
+#define MEMINTRSTS		0x11184
+#define   MEMINT_RSEXIT		(1<<7)
+#define   MEMINT_CONT_BUSY	(1<<6)
+#define   MEMINT_AVG_BUSY	(1<<5)
+#define   MEMINT_EVAL_CHG	(1<<4)
+#define   MEMINT_MON_IDLE	(1<<3)
+#define   MEMINT_UP_EVAL	(1<<2)
+#define   MEMINT_DOWN_EVAL	(1<<1)
+#define   MEMINT_SW_CMD		(1<<0)
+#define MEMMODECTL		0x11190
+#define   MEMMODE_BOOST_EN	(1<<31)
+#define   MEMMODE_BOOST_FREQ_MASK 0x0f000000 /* jitter for boost, 0-15 */
+#define   MEMMODE_BOOST_FREQ_SHIFT 24
+#define   MEMMODE_IDLE_MODE_MASK 0x00030000
+#define   MEMMODE_IDLE_MODE_SHIFT 16
+#define   MEMMODE_IDLE_MODE_EVAL 0
+#define   MEMMODE_IDLE_MODE_CONT 1
+#define   MEMMODE_HWIDLE_EN	(1<<15)
+#define   MEMMODE_SWMODE_EN	(1<<14)
+#define   MEMMODE_RCLK_GATE	(1<<13)
+#define   MEMMODE_HW_UPDATE	(1<<12)
+#define   MEMMODE_FSTART_MASK	0x00000f00 /* starting jitter, 0-15 */
+#define   MEMMODE_FSTART_SHIFT	8
+#define   MEMMODE_FMAX_MASK	0x000000f0 /* max jitter, 0-15 */
+#define   MEMMODE_FMAX_SHIFT	4
+#define   MEMMODE_FMIN_MASK	0x0000000f /* min jitter, 0-15 */
+#define RCBMAXAVG		0x1119c
+#define MEMSWCTL2		0x1119e /* Cantiga only */
+#define   SWMEMCMD_RENDER_OFF	(0 << 13)
+#define   SWMEMCMD_RENDER_ON	(1 << 13)
+#define   SWMEMCMD_SWFREQ	(2 << 13)
+#define   SWMEMCMD_TARVID	(3 << 13)
+#define   SWMEMCMD_VRM_OFF	(4 << 13)
+#define   SWMEMCMD_VRM_ON	(5 << 13)
+#define   CMDSTS		(1<<12)
+#define   SFCAVM		(1<<11)
+#define   SWFREQ_MASK		0x0380 /* P0-7 */
+#define   SWFREQ_SHIFT		7
+#define   TARVID_MASK		0x001f
+#define MEMSTAT_CTG		0x111a0
+#define RCBMINAVG		0x111a0
+#define RCUPEI			0x111b0
+#define RCDNEI			0x111b4
+#define MCHBAR_RENDER_STANDBY		0x111b8
+#define   RCX_SW_EXIT		(1<<23)
+#define   RSX_STATUS_MASK	0x00700000
+#define VIDCTL			0x111c0
+#define VIDSTS			0x111c8
+#define VIDSTART		0x111cc /* 8 bits */
+#define MEMSTAT_ILK			0x111f8
+#define   MEMSTAT_VID_MASK	0x7f00
+#define   MEMSTAT_VID_SHIFT	8
+#define   MEMSTAT_PSTATE_MASK	0x00f8
+#define   MEMSTAT_PSTATE_SHIFT  3
+#define   MEMSTAT_MON_ACTV	(1<<2)
+#define   MEMSTAT_SRC_CTL_MASK	0x0003
+#define   MEMSTAT_SRC_CTL_CORE	0
+#define   MEMSTAT_SRC_CTL_TRB	1
+#define   MEMSTAT_SRC_CTL_THM	2
+#define   MEMSTAT_SRC_CTL_STDBY 3
+#define RCPREVBSYTUPAVG		0x113b8
+#define RCPREVBSYTDNAVG		0x113bc
 #define PEG_BAND_GAP_DATA	0x14d68
 
 /*
@@ -844,7 +1024,6 @@
 #define   SDVOB_HOTPLUG_INT_EN			(1 << 26)
 #define   SDVOC_HOTPLUG_INT_EN			(1 << 25)
 #define   TV_HOTPLUG_INT_EN			(1 << 18)
-#define   CRT_EOS_INT_EN			(1 << 10)
 #define   CRT_HOTPLUG_INT_EN			(1 << 9)
 #define   CRT_HOTPLUG_FORCE_DETECT		(1 << 3)
 #define CRT_HOTPLUG_ACTIVATION_PERIOD_32	(0 << 8)
@@ -863,14 +1042,6 @@
 #define CRT_HOTPLUG_DETECT_VOLTAGE_475MV	(1 << 2)
 #define CRT_HOTPLUG_MASK			(0x3fc) /* Bits 9-2 */
 #define CRT_FORCE_HOTPLUG_MASK			0xfffffe1f
-#define HOTPLUG_EN_MASK (HDMIB_HOTPLUG_INT_EN | \
-			 HDMIC_HOTPLUG_INT_EN |	  \
-			 HDMID_HOTPLUG_INT_EN |	  \
-			 SDVOB_HOTPLUG_INT_EN |	  \
-			 SDVOC_HOTPLUG_INT_EN |	  \
-			 TV_HOTPLUG_INT_EN |	  \
-			 CRT_HOTPLUG_INT_EN)
-
 
 #define PORT_HOTPLUG_STAT	0x61114
 #define   HDMIB_HOTPLUG_INT_STATUS		(1 << 29)
@@ -879,7 +1050,6 @@
 #define   DPC_HOTPLUG_INT_STATUS		(1 << 28)
 #define   HDMID_HOTPLUG_INT_STATUS		(1 << 27)
 #define   DPD_HOTPLUG_INT_STATUS		(1 << 27)
-#define   CRT_EOS_INT_STATUS			(1 << 12)
 #define   CRT_HOTPLUG_INT_STATUS		(1 << 11)
 #define   TV_HOTPLUG_INT_STATUS			(1 << 10)
 #define   CRT_HOTPLUG_MONITOR_MASK		(3 << 8)
@@ -968,6 +1138,10 @@
 #define   LVDS_PORT_EN			(1 << 31)
 /* Selects pipe B for LVDS data.  Must be set on pre-965. */
 #define   LVDS_PIPEB_SELECT		(1 << 30)
+/* LVDS dithering flag on 965/g4x platform */
+#define   LVDS_ENABLE_DITHER		(1 << 25)
+/* Enable border for unscaled (or aspect-scaled) display */
+#define   LVDS_BORDER_ENABLE		(1 << 15)
 /*
  * Enables the A0-A2 data pairs and CLKA, containing 18 bits of color data per
  * pixel.
@@ -1078,6 +1252,8 @@
 #define   BACKLIGHT_DUTY_CYCLE_SHIFT		(0)
 #define   BACKLIGHT_DUTY_CYCLE_MASK		(0xffff)
 
+#define BLC_HIST_CTL		0x61260
+
 /* TV port control */
 #define TV_CTL			0x68000
 /** Enables the TV encoder */
@@ -1616,7 +1792,7 @@
 #define   DP_CLOCK_OUTPUT_ENABLE	(1 << 13)
 
 #define   DP_SCRAMBLING_DISABLE		(1 << 12)
-#define   DP_SCRAMBLING_DISABLE_IGDNG	(1 << 7)
+#define   DP_SCRAMBLING_DISABLE_IRONLAKE	(1 << 7)
 
 /** limit RGB values to avoid confusing TVs */
 #define   DP_COLOR_RANGE_16_235		(1 << 8)
@@ -1733,6 +1909,8 @@
 
 /* Display & cursor control */
 
+/* dithering flag on Ironlake */
+#define PIPE_ENABLE_DITHER	(1 << 4)
 /* Pipe A */
 #define PIPEADSL		0x70000
 #define PIPEACONF		0x70008
@@ -1780,6 +1958,11 @@
 #define   PIPE_START_VBLANK_INTERRUPT_STATUS	(1UL<<2) /* 965 or later */
 #define   PIPE_VBLANK_INTERRUPT_STATUS		(1UL<<1)
 #define   PIPE_OVERLAY_UPDATED_STATUS		(1UL<<0)
+#define   PIPE_BPC_MASK 			(7 << 5) /* Ironlake */
+#define   PIPE_8BPC				(0 << 5)
+#define   PIPE_10BPC				(1 << 5)
+#define   PIPE_6BPC				(2 << 5)
+#define   PIPE_12BPC				(3 << 5)
 
 #define DSPARB			0x70030
 #define   DSPARB_CSTART_MASK	(0x7f << 7)
@@ -1790,29 +1973,41 @@
 #define   DSPARB_AEND_SHIFT	0
 
 #define DSPFW1			0x70034
+#define   DSPFW_SR_SHIFT	23
+#define   DSPFW_CURSORB_SHIFT	16
+#define   DSPFW_PLANEB_SHIFT	8
 #define DSPFW2			0x70038
+#define   DSPFW_CURSORA_MASK	0x00003f00
+#define   DSPFW_CURSORA_SHIFT	8
 #define DSPFW3			0x7003c
-#define   IGD_SELF_REFRESH_EN	(1<<30)
+#define   DSPFW_HPLL_SR_EN	(1<<31)
+#define   DSPFW_CURSOR_SR_SHIFT	24
+#define   PINEVIEW_SELF_REFRESH_EN	(1<<30)
 
 /* FIFO watermark sizes etc */
+#define G4X_FIFO_LINE_SIZE	64
 #define I915_FIFO_LINE_SIZE	64
 #define I830_FIFO_LINE_SIZE	32
+
+#define G4X_FIFO_SIZE		127
 #define I945_FIFO_SIZE		127 /* 945 & 965 */
 #define I915_FIFO_SIZE		95
 #define I855GM_FIFO_SIZE	127 /* In cachelines */
 #define I830_FIFO_SIZE		95
+
+#define G4X_MAX_WM		0x3f
 #define I915_MAX_WM		0x3f
 
-#define IGD_DISPLAY_FIFO	512 /* in 64byte unit */
-#define IGD_FIFO_LINE_SIZE	64
-#define IGD_MAX_WM		0x1ff
-#define IGD_DFT_WM		0x3f
-#define IGD_DFT_HPLLOFF_WM	0
-#define IGD_GUARD_WM		10
-#define IGD_CURSOR_FIFO		64
-#define IGD_CURSOR_MAX_WM	0x3f
-#define IGD_CURSOR_DFT_WM	0
-#define IGD_CURSOR_GUARD_WM	5
+#define PINEVIEW_DISPLAY_FIFO	512 /* in 64byte unit */
+#define PINEVIEW_FIFO_LINE_SIZE	64
+#define PINEVIEW_MAX_WM		0x1ff
+#define PINEVIEW_DFT_WM		0x3f
+#define PINEVIEW_DFT_HPLLOFF_WM	0
+#define PINEVIEW_GUARD_WM		10
+#define PINEVIEW_CURSOR_FIFO		64
+#define PINEVIEW_CURSOR_MAX_WM	0x3f
+#define PINEVIEW_CURSOR_DFT_WM	0
+#define PINEVIEW_CURSOR_GUARD_WM	5
 
 /*
  * The two pipe frame counter registers are not synchronized, so
@@ -1886,6 +2081,7 @@
 #define   DISPPLANE_16BPP			(0x5<<26)
 #define   DISPPLANE_32BPP_NO_ALPHA		(0x6<<26)
 #define   DISPPLANE_32BPP			(0x7<<26)
+#define   DISPPLANE_32BPP_30BIT_NO_ALPHA	(0xa<<26)
 #define   DISPPLANE_STEREO_ENABLE		(1<<25)
 #define   DISPPLANE_STEREO_DISABLE		0
 #define   DISPPLANE_SEL_PIPE_MASK		(1<<24)
@@ -1897,7 +2093,7 @@
 #define   DISPPLANE_NO_LINE_DOUBLE		0
 #define   DISPPLANE_STEREO_POLARITY_FIRST	0
 #define   DISPPLANE_STEREO_POLARITY_SECOND	(1<<18)
-#define   DISPPLANE_TRICKLE_FEED_DISABLE	(1<<14) /* IGDNG */
+#define   DISPPLANE_TRICKLE_FEED_DISABLE	(1<<14) /* Ironlake */
 #define   DISPPLANE_TILED			(1<<10)
 #define DSPAADDR		0x70184
 #define DSPASTRIDE		0x70188
@@ -1950,7 +2146,7 @@
 # define VGA_2X_MODE				(1 << 30)
 # define VGA_PIPE_B_SELECT			(1 << 29)
 
-/* IGDNG */
+/* Ironlake */
 
 #define CPU_VGACNTRL	0x41000
 
@@ -2030,6 +2226,11 @@
 #define PFA_CTL_1               0x68080
 #define PFB_CTL_1               0x68880
 #define  PF_ENABLE              (1<<31)
+#define  PF_FILTER_MASK		(3<<23)
+#define  PF_FILTER_PROGRAMMED	(0<<23)
+#define  PF_FILTER_MED_3x3	(1<<23)
+#define  PF_FILTER_EDGE_ENHANCE	(2<<23)
+#define  PF_FILTER_EDGE_SOFTEN	(3<<23)
 #define PFA_WIN_SZ		0x68074
 #define PFB_WIN_SZ		0x68874
 #define PFA_WIN_POS		0x68070
@@ -2091,6 +2292,7 @@
 #define SDE_PORTC_HOTPLUG       (1 << 9)
 #define SDE_PORTB_HOTPLUG       (1 << 8)
 #define SDE_SDVOB_HOTPLUG       (1 << 6)
+#define SDE_HOTPLUG_MASK	(0xf << 8)
 
 #define SDEISR  0xc4000
 #define SDEIMR  0xc4004
@@ -2131,6 +2333,13 @@
 #define PCH_GPIOE               0xc5020
 #define PCH_GPIOF               0xc5024
 
+#define PCH_GMBUS0		0xc5100
+#define PCH_GMBUS1		0xc5104
+#define PCH_GMBUS2		0xc5108
+#define PCH_GMBUS3		0xc510c
+#define PCH_GMBUS4		0xc5110
+#define PCH_GMBUS5		0xc5120
+
 #define PCH_DPLL_A              0xc6014
 #define PCH_DPLL_B              0xc6018
 
@@ -2149,11 +2358,11 @@
 #define  DREF_CPU_SOURCE_OUTPUT_MASK		(3<<13)
 #define  DREF_SSC_SOURCE_DISABLE                (0<<11)
 #define  DREF_SSC_SOURCE_ENABLE                 (2<<11)
-#define  DREF_SSC_SOURCE_MASK			(2<<11)
+#define  DREF_SSC_SOURCE_MASK			(3<<11)
 #define  DREF_NONSPREAD_SOURCE_DISABLE          (0<<9)
 #define  DREF_NONSPREAD_CK505_ENABLE		(1<<9)
 #define  DREF_NONSPREAD_SOURCE_ENABLE           (2<<9)
-#define  DREF_NONSPREAD_SOURCE_MASK		(2<<9)
+#define  DREF_NONSPREAD_SOURCE_MASK		(3<<9)
 #define  DREF_SUPERSPREAD_SOURCE_DISABLE        (0<<7)
 #define  DREF_SUPERSPREAD_SOURCE_ENABLE         (2<<7)
 #define  DREF_SSC4_DOWNSPREAD                   (0<<6)
@@ -2266,7 +2475,7 @@
 #define  FDI_DP_PORT_WIDTH_X3           (2<<19)
 #define  FDI_DP_PORT_WIDTH_X4           (3<<19)
 #define  FDI_TX_ENHANCE_FRAME_ENABLE    (1<<18)
-/* IGDNG: hardwired to 1 */
+/* Ironlake: hardwired to 1 */
 #define  FDI_TX_PLL_ENABLE              (1<<14)
 /* both Tx and Rx */
 #define  FDI_SCRAMBLING_ENABLE          (0<<7)
diff --git a/drivers/gpu/drm/i915/i915_suspend.c b/drivers/gpu/drm/i915/i915_suspend.c
index bd6d8d91ca9f..ac0d1a73ac22 100644
--- a/drivers/gpu/drm/i915/i915_suspend.c
+++ b/drivers/gpu/drm/i915/i915_suspend.c
@@ -27,16 +27,20 @@
 #include "drmP.h"
 #include "drm.h"
 #include "i915_drm.h"
-#include "i915_drv.h"
+#include "intel_drv.h"
 
 static bool i915_pipe_enabled(struct drm_device *dev, enum pipe pipe)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
+	u32	dpll_reg;
 
-	if (pipe == PIPE_A)
-		return (I915_READ(DPLL_A) & DPLL_VCO_ENABLE);
-	else
-		return (I915_READ(DPLL_B) & DPLL_VCO_ENABLE);
+	if (IS_IRONLAKE(dev)) {
+		dpll_reg = (pipe == PIPE_A) ? PCH_DPLL_A: PCH_DPLL_B;
+	} else {
+		dpll_reg = (pipe == PIPE_A) ? DPLL_A: DPLL_B;
+	}
+
+	return (I915_READ(dpll_reg) & DPLL_VCO_ENABLE);
 }
 
 static void i915_save_palette(struct drm_device *dev, enum pipe pipe)
@@ -49,6 +53,9 @@ static void i915_save_palette(struct drm_device *dev, enum pipe pipe)
 	if (!i915_pipe_enabled(dev, pipe))
 		return;
 
+	if (IS_IRONLAKE(dev))
+		reg = (pipe == PIPE_A) ? LGC_PALETTE_A : LGC_PALETTE_B;
+
 	if (pipe == PIPE_A)
 		array = dev_priv->save_palette_a;
 	else
@@ -68,6 +75,9 @@ static void i915_restore_palette(struct drm_device *dev, enum pipe pipe)
 	if (!i915_pipe_enabled(dev, pipe))
 		return;
 
+	if (IS_IRONLAKE(dev))
+		reg = (pipe == PIPE_A) ? LGC_PALETTE_A : LGC_PALETTE_B;
+
 	if (pipe == PIPE_A)
 		array = dev_priv->save_palette_a;
 	else
@@ -229,13 +239,24 @@ static void i915_save_modeset_reg(struct drm_device *dev)
 	if (drm_core_check_feature(dev, DRIVER_MODESET))
 		return;
 
+	if (IS_IRONLAKE(dev)) {
+		dev_priv->savePCH_DREF_CONTROL = I915_READ(PCH_DREF_CONTROL);
+		dev_priv->saveDISP_ARB_CTL = I915_READ(DISP_ARB_CTL);
+	}
+
 	/* Pipe & plane A info */
 	dev_priv->savePIPEACONF = I915_READ(PIPEACONF);
 	dev_priv->savePIPEASRC = I915_READ(PIPEASRC);
-	dev_priv->saveFPA0 = I915_READ(FPA0);
-	dev_priv->saveFPA1 = I915_READ(FPA1);
-	dev_priv->saveDPLL_A = I915_READ(DPLL_A);
-	if (IS_I965G(dev))
+	if (IS_IRONLAKE(dev)) {
+		dev_priv->saveFPA0 = I915_READ(PCH_FPA0);
+		dev_priv->saveFPA1 = I915_READ(PCH_FPA1);
+		dev_priv->saveDPLL_A = I915_READ(PCH_DPLL_A);
+	} else {
+		dev_priv->saveFPA0 = I915_READ(FPA0);
+		dev_priv->saveFPA1 = I915_READ(FPA1);
+		dev_priv->saveDPLL_A = I915_READ(DPLL_A);
+	}
+	if (IS_I965G(dev) && !IS_IRONLAKE(dev))
 		dev_priv->saveDPLL_A_MD = I915_READ(DPLL_A_MD);
 	dev_priv->saveHTOTAL_A = I915_READ(HTOTAL_A);
 	dev_priv->saveHBLANK_A = I915_READ(HBLANK_A);
@@ -243,7 +264,30 @@ static void i915_save_modeset_reg(struct drm_device *dev)
 	dev_priv->saveVTOTAL_A = I915_READ(VTOTAL_A);
 	dev_priv->saveVBLANK_A = I915_READ(VBLANK_A);
 	dev_priv->saveVSYNC_A = I915_READ(VSYNC_A);
-	dev_priv->saveBCLRPAT_A = I915_READ(BCLRPAT_A);
+	if (!IS_IRONLAKE(dev))
+		dev_priv->saveBCLRPAT_A = I915_READ(BCLRPAT_A);
+
+	if (IS_IRONLAKE(dev)) {
+		dev_priv->savePIPEA_DATA_M1 = I915_READ(PIPEA_DATA_M1);
+		dev_priv->savePIPEA_DATA_N1 = I915_READ(PIPEA_DATA_N1);
+		dev_priv->savePIPEA_LINK_M1 = I915_READ(PIPEA_LINK_M1);
+		dev_priv->savePIPEA_LINK_N1 = I915_READ(PIPEA_LINK_N1);
+
+		dev_priv->saveFDI_TXA_CTL = I915_READ(FDI_TXA_CTL);
+		dev_priv->saveFDI_RXA_CTL = I915_READ(FDI_RXA_CTL);
+
+		dev_priv->savePFA_CTL_1 = I915_READ(PFA_CTL_1);
+		dev_priv->savePFA_WIN_SZ = I915_READ(PFA_WIN_SZ);
+		dev_priv->savePFA_WIN_POS = I915_READ(PFA_WIN_POS);
+
+		dev_priv->saveTRANSACONF = I915_READ(TRANSACONF);
+		dev_priv->saveTRANS_HTOTAL_A = I915_READ(TRANS_HTOTAL_A);
+		dev_priv->saveTRANS_HBLANK_A = I915_READ(TRANS_HBLANK_A);
+		dev_priv->saveTRANS_HSYNC_A = I915_READ(TRANS_HSYNC_A);
+		dev_priv->saveTRANS_VTOTAL_A = I915_READ(TRANS_VTOTAL_A);
+		dev_priv->saveTRANS_VBLANK_A = I915_READ(TRANS_VBLANK_A);
+		dev_priv->saveTRANS_VSYNC_A = I915_READ(TRANS_VSYNC_A);
+	}
 
 	dev_priv->saveDSPACNTR = I915_READ(DSPACNTR);
 	dev_priv->saveDSPASTRIDE = I915_READ(DSPASTRIDE);
@@ -260,10 +304,16 @@ static void i915_save_modeset_reg(struct drm_device *dev)
 	/* Pipe & plane B info */
 	dev_priv->savePIPEBCONF = I915_READ(PIPEBCONF);
 	dev_priv->savePIPEBSRC = I915_READ(PIPEBSRC);
-	dev_priv->saveFPB0 = I915_READ(FPB0);
-	dev_priv->saveFPB1 = I915_READ(FPB1);
-	dev_priv->saveDPLL_B = I915_READ(DPLL_B);
-	if (IS_I965G(dev))
+	if (IS_IRONLAKE(dev)) {
+		dev_priv->saveFPB0 = I915_READ(PCH_FPB0);
+		dev_priv->saveFPB1 = I915_READ(PCH_FPB1);
+		dev_priv->saveDPLL_B = I915_READ(PCH_DPLL_B);
+	} else {
+		dev_priv->saveFPB0 = I915_READ(FPB0);
+		dev_priv->saveFPB1 = I915_READ(FPB1);
+		dev_priv->saveDPLL_B = I915_READ(DPLL_B);
+	}
+	if (IS_I965G(dev) && !IS_IRONLAKE(dev))
 		dev_priv->saveDPLL_B_MD = I915_READ(DPLL_B_MD);
 	dev_priv->saveHTOTAL_B = I915_READ(HTOTAL_B);
 	dev_priv->saveHBLANK_B = I915_READ(HBLANK_B);
@@ -271,7 +321,30 @@ static void i915_save_modeset_reg(struct drm_device *dev)
 	dev_priv->saveVTOTAL_B = I915_READ(VTOTAL_B);
 	dev_priv->saveVBLANK_B = I915_READ(VBLANK_B);
 	dev_priv->saveVSYNC_B = I915_READ(VSYNC_B);
-	dev_priv->saveBCLRPAT_A = I915_READ(BCLRPAT_A);
+	if (!IS_IRONLAKE(dev))
+		dev_priv->saveBCLRPAT_B = I915_READ(BCLRPAT_B);
+
+	if (IS_IRONLAKE(dev)) {
+		dev_priv->savePIPEB_DATA_M1 = I915_READ(PIPEB_DATA_M1);
+		dev_priv->savePIPEB_DATA_N1 = I915_READ(PIPEB_DATA_N1);
+		dev_priv->savePIPEB_LINK_M1 = I915_READ(PIPEB_LINK_M1);
+		dev_priv->savePIPEB_LINK_N1 = I915_READ(PIPEB_LINK_N1);
+
+		dev_priv->saveFDI_TXB_CTL = I915_READ(FDI_TXB_CTL);
+		dev_priv->saveFDI_RXB_CTL = I915_READ(FDI_RXB_CTL);
+
+		dev_priv->savePFB_CTL_1 = I915_READ(PFB_CTL_1);
+		dev_priv->savePFB_WIN_SZ = I915_READ(PFB_WIN_SZ);
+		dev_priv->savePFB_WIN_POS = I915_READ(PFB_WIN_POS);
+
+		dev_priv->saveTRANSBCONF = I915_READ(TRANSBCONF);
+		dev_priv->saveTRANS_HTOTAL_B = I915_READ(TRANS_HTOTAL_B);
+		dev_priv->saveTRANS_HBLANK_B = I915_READ(TRANS_HBLANK_B);
+		dev_priv->saveTRANS_HSYNC_B = I915_READ(TRANS_HSYNC_B);
+		dev_priv->saveTRANS_VTOTAL_B = I915_READ(TRANS_VTOTAL_B);
+		dev_priv->saveTRANS_VBLANK_B = I915_READ(TRANS_VBLANK_B);
+		dev_priv->saveTRANS_VSYNC_B = I915_READ(TRANS_VSYNC_B);
+	}
 
 	dev_priv->saveDSPBCNTR = I915_READ(DSPBCNTR);
 	dev_priv->saveDSPBSTRIDE = I915_READ(DSPBSTRIDE);
@@ -290,23 +363,46 @@ static void i915_save_modeset_reg(struct drm_device *dev)
 static void i915_restore_modeset_reg(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
+	int dpll_a_reg, fpa0_reg, fpa1_reg;
+	int dpll_b_reg, fpb0_reg, fpb1_reg;
 
 	if (drm_core_check_feature(dev, DRIVER_MODESET))
 		return;
 
+	if (IS_IRONLAKE(dev)) {
+		dpll_a_reg = PCH_DPLL_A;
+		dpll_b_reg = PCH_DPLL_B;
+		fpa0_reg = PCH_FPA0;
+		fpb0_reg = PCH_FPB0;
+		fpa1_reg = PCH_FPA1;
+		fpb1_reg = PCH_FPB1;
+	} else {
+		dpll_a_reg = DPLL_A;
+		dpll_b_reg = DPLL_B;
+		fpa0_reg = FPA0;
+		fpb0_reg = FPB0;
+		fpa1_reg = FPA1;
+		fpb1_reg = FPB1;
+	}
+
+	if (IS_IRONLAKE(dev)) {
+		I915_WRITE(PCH_DREF_CONTROL, dev_priv->savePCH_DREF_CONTROL);
+		I915_WRITE(DISP_ARB_CTL, dev_priv->saveDISP_ARB_CTL);
+	}
+
 	/* Pipe & plane A info */
 	/* Prime the clock */
 	if (dev_priv->saveDPLL_A & DPLL_VCO_ENABLE) {
-		I915_WRITE(DPLL_A, dev_priv->saveDPLL_A &
+		I915_WRITE(dpll_a_reg, dev_priv->saveDPLL_A &
 			   ~DPLL_VCO_ENABLE);
 		DRM_UDELAY(150);
 	}
-	I915_WRITE(FPA0, dev_priv->saveFPA0);
-	I915_WRITE(FPA1, dev_priv->saveFPA1);
+	I915_WRITE(fpa0_reg, dev_priv->saveFPA0);
+	I915_WRITE(fpa1_reg, dev_priv->saveFPA1);
 	/* Actually enable it */
-	I915_WRITE(DPLL_A, dev_priv->saveDPLL_A);
+	I915_WRITE(dpll_a_reg, dev_priv->saveDPLL_A);
 	DRM_UDELAY(150);
-	if (IS_I965G(dev))
+	if (IS_I965G(dev) && !IS_IRONLAKE(dev))
 		I915_WRITE(DPLL_A_MD, dev_priv->saveDPLL_A_MD);
 	DRM_UDELAY(150);
 
@@ -317,7 +413,30 @@ static void i915_restore_modeset_reg(struct drm_device *dev)
 	I915_WRITE(VTOTAL_A, dev_priv->saveVTOTAL_A);
 	I915_WRITE(VBLANK_A, dev_priv->saveVBLANK_A);
 	I915_WRITE(VSYNC_A, dev_priv->saveVSYNC_A);
-	I915_WRITE(BCLRPAT_A, dev_priv->saveBCLRPAT_A);
+	if (!IS_IRONLAKE(dev))
+		I915_WRITE(BCLRPAT_A, dev_priv->saveBCLRPAT_A);
+
+	if (IS_IRONLAKE(dev)) {
+		I915_WRITE(PIPEA_DATA_M1, dev_priv->savePIPEA_DATA_M1);
+		I915_WRITE(PIPEA_DATA_N1, dev_priv->savePIPEA_DATA_N1);
+		I915_WRITE(PIPEA_LINK_M1, dev_priv->savePIPEA_LINK_M1);
+		I915_WRITE(PIPEA_LINK_N1, dev_priv->savePIPEA_LINK_N1);
+
+		I915_WRITE(FDI_RXA_CTL, dev_priv->saveFDI_RXA_CTL);
+		I915_WRITE(FDI_TXA_CTL, dev_priv->saveFDI_TXA_CTL);
+
+		I915_WRITE(PFA_CTL_1, dev_priv->savePFA_CTL_1);
+		I915_WRITE(PFA_WIN_SZ, dev_priv->savePFA_WIN_SZ);
+		I915_WRITE(PFA_WIN_POS, dev_priv->savePFA_WIN_POS);
+
+		I915_WRITE(TRANSACONF, dev_priv->saveTRANSACONF);
+		I915_WRITE(TRANS_HTOTAL_A, dev_priv->saveTRANS_HTOTAL_A);
+		I915_WRITE(TRANS_HBLANK_A, dev_priv->saveTRANS_HBLANK_A);
+		I915_WRITE(TRANS_HSYNC_A, dev_priv->saveTRANS_HSYNC_A);
+		I915_WRITE(TRANS_VTOTAL_A, dev_priv->saveTRANS_VTOTAL_A);
+		I915_WRITE(TRANS_VBLANK_A, dev_priv->saveTRANS_VBLANK_A);
+		I915_WRITE(TRANS_VSYNC_A, dev_priv->saveTRANS_VSYNC_A);
+	}
 
 	/* Restore plane info */
 	I915_WRITE(DSPASIZE, dev_priv->saveDSPASIZE);
@@ -339,16 +458,16 @@ static void i915_restore_modeset_reg(struct drm_device *dev)
 
 	/* Pipe & plane B info */
 	if (dev_priv->saveDPLL_B & DPLL_VCO_ENABLE) {
-		I915_WRITE(DPLL_B, dev_priv->saveDPLL_B &
+		I915_WRITE(dpll_b_reg, dev_priv->saveDPLL_B &
 			   ~DPLL_VCO_ENABLE);
 		DRM_UDELAY(150);
 	}
-	I915_WRITE(FPB0, dev_priv->saveFPB0);
-	I915_WRITE(FPB1, dev_priv->saveFPB1);
+	I915_WRITE(fpb0_reg, dev_priv->saveFPB0);
+	I915_WRITE(fpb1_reg, dev_priv->saveFPB1);
 	/* Actually enable it */
-	I915_WRITE(DPLL_B, dev_priv->saveDPLL_B);
+	I915_WRITE(dpll_b_reg, dev_priv->saveDPLL_B);
 	DRM_UDELAY(150);
-	if (IS_I965G(dev))
+	if (IS_I965G(dev) && !IS_IRONLAKE(dev))
 		I915_WRITE(DPLL_B_MD, dev_priv->saveDPLL_B_MD);
 	DRM_UDELAY(150);
 
@@ -359,7 +478,30 @@ static void i915_restore_modeset_reg(struct drm_device *dev)
 	I915_WRITE(VTOTAL_B, dev_priv->saveVTOTAL_B);
 	I915_WRITE(VBLANK_B, dev_priv->saveVBLANK_B);
 	I915_WRITE(VSYNC_B, dev_priv->saveVSYNC_B);
-	I915_WRITE(BCLRPAT_B, dev_priv->saveBCLRPAT_B);
+	if (!IS_IRONLAKE(dev))
+		I915_WRITE(BCLRPAT_B, dev_priv->saveBCLRPAT_B);
+
+	if (IS_IRONLAKE(dev)) {
+		I915_WRITE(PIPEB_DATA_M1, dev_priv->savePIPEB_DATA_M1);
+		I915_WRITE(PIPEB_DATA_N1, dev_priv->savePIPEB_DATA_N1);
+		I915_WRITE(PIPEB_LINK_M1, dev_priv->savePIPEB_LINK_M1);
+		I915_WRITE(PIPEB_LINK_N1, dev_priv->savePIPEB_LINK_N1);
+
+		I915_WRITE(FDI_RXB_CTL, dev_priv->saveFDI_RXB_CTL);
+		I915_WRITE(FDI_TXB_CTL, dev_priv->saveFDI_TXB_CTL);
+
+		I915_WRITE(PFB_CTL_1, dev_priv->savePFB_CTL_1);
+		I915_WRITE(PFB_WIN_SZ, dev_priv->savePFB_WIN_SZ);
+		I915_WRITE(PFB_WIN_POS, dev_priv->savePFB_WIN_POS);
+
+		I915_WRITE(TRANSBCONF, dev_priv->saveTRANSBCONF);
+		I915_WRITE(TRANS_HTOTAL_B, dev_priv->saveTRANS_HTOTAL_B);
+		I915_WRITE(TRANS_HBLANK_B, dev_priv->saveTRANS_HBLANK_B);
+		I915_WRITE(TRANS_HSYNC_B, dev_priv->saveTRANS_HSYNC_B);
+		I915_WRITE(TRANS_VTOTAL_B, dev_priv->saveTRANS_VTOTAL_B);
+		I915_WRITE(TRANS_VBLANK_B, dev_priv->saveTRANS_VBLANK_B);
+		I915_WRITE(TRANS_VSYNC_B, dev_priv->saveTRANS_VSYNC_B);
+	}
 
 	/* Restore plane info */
 	I915_WRITE(DSPBSIZE, dev_priv->saveDSPBSIZE);
@@ -404,21 +546,43 @@ void i915_save_display(struct drm_device *dev)
 		dev_priv->saveCURSIZE = I915_READ(CURSIZE);
 
 	/* CRT state */
-	dev_priv->saveADPA = I915_READ(ADPA);
+	if (IS_IRONLAKE(dev)) {
+		dev_priv->saveADPA = I915_READ(PCH_ADPA);
+	} else {
+		dev_priv->saveADPA = I915_READ(ADPA);
+	}
 
 	/* LVDS state */
-	dev_priv->savePP_CONTROL = I915_READ(PP_CONTROL);
-	dev_priv->savePFIT_PGM_RATIOS = I915_READ(PFIT_PGM_RATIOS);
-	dev_priv->saveBLC_PWM_CTL = I915_READ(BLC_PWM_CTL);
-	if (IS_I965G(dev))
-		dev_priv->saveBLC_PWM_CTL2 = I915_READ(BLC_PWM_CTL2);
-	if (IS_MOBILE(dev) && !IS_I830(dev))
-		dev_priv->saveLVDS = I915_READ(LVDS);
-	if (!IS_I830(dev) && !IS_845G(dev))
+	if (IS_IRONLAKE(dev)) {
+		dev_priv->savePP_CONTROL = I915_READ(PCH_PP_CONTROL);
+		dev_priv->saveBLC_PWM_CTL = I915_READ(BLC_PWM_PCH_CTL1);
+		dev_priv->saveBLC_PWM_CTL2 = I915_READ(BLC_PWM_PCH_CTL2);
+		dev_priv->saveBLC_CPU_PWM_CTL = I915_READ(BLC_PWM_CPU_CTL);
+		dev_priv->saveBLC_CPU_PWM_CTL2 = I915_READ(BLC_PWM_CPU_CTL2);
+		dev_priv->saveLVDS = I915_READ(PCH_LVDS);
+	} else {
+		dev_priv->savePP_CONTROL = I915_READ(PP_CONTROL);
+		dev_priv->savePFIT_PGM_RATIOS = I915_READ(PFIT_PGM_RATIOS);
+		dev_priv->saveBLC_PWM_CTL = I915_READ(BLC_PWM_CTL);
+		dev_priv->saveBLC_HIST_CTL = I915_READ(BLC_HIST_CTL);
+		if (IS_I965G(dev))
+			dev_priv->saveBLC_PWM_CTL2 = I915_READ(BLC_PWM_CTL2);
+		if (IS_MOBILE(dev) && !IS_I830(dev))
+			dev_priv->saveLVDS = I915_READ(LVDS);
+	}
+
+	if (!IS_I830(dev) && !IS_845G(dev) && !IS_IRONLAKE(dev))
 		dev_priv->savePFIT_CONTROL = I915_READ(PFIT_CONTROL);
-	dev_priv->savePP_ON_DELAYS = I915_READ(PP_ON_DELAYS);
-	dev_priv->savePP_OFF_DELAYS = I915_READ(PP_OFF_DELAYS);
-	dev_priv->savePP_DIVISOR = I915_READ(PP_DIVISOR);
+
+	if (IS_IRONLAKE(dev)) {
+		dev_priv->savePP_ON_DELAYS = I915_READ(PCH_PP_ON_DELAYS);
+		dev_priv->savePP_OFF_DELAYS = I915_READ(PCH_PP_OFF_DELAYS);
+		dev_priv->savePP_DIVISOR = I915_READ(PCH_PP_DIVISOR);
+	} else {
+		dev_priv->savePP_ON_DELAYS = I915_READ(PP_ON_DELAYS);
+		dev_priv->savePP_OFF_DELAYS = I915_READ(PP_OFF_DELAYS);
+		dev_priv->savePP_DIVISOR = I915_READ(PP_DIVISOR);
+	}
 
 	/* Display Port state */
 	if (SUPPORTS_INTEGRATED_DP(dev)) {
@@ -437,16 +601,23 @@ void i915_save_display(struct drm_device *dev)
 	/* FIXME: save TV & SDVO state */
 
 	/* FBC state */
-	dev_priv->saveFBC_CFB_BASE = I915_READ(FBC_CFB_BASE);
-	dev_priv->saveFBC_LL_BASE = I915_READ(FBC_LL_BASE);
-	dev_priv->saveFBC_CONTROL2 = I915_READ(FBC_CONTROL2);
-	dev_priv->saveFBC_CONTROL = I915_READ(FBC_CONTROL);
+	if (IS_GM45(dev)) {
+		dev_priv->saveDPFC_CB_BASE = I915_READ(DPFC_CB_BASE);
+	} else {
+		dev_priv->saveFBC_CFB_BASE = I915_READ(FBC_CFB_BASE);
+		dev_priv->saveFBC_LL_BASE = I915_READ(FBC_LL_BASE);
+		dev_priv->saveFBC_CONTROL2 = I915_READ(FBC_CONTROL2);
+		dev_priv->saveFBC_CONTROL = I915_READ(FBC_CONTROL);
+	}
 
 	/* VGA state */
 	dev_priv->saveVGA0 = I915_READ(VGA0);
 	dev_priv->saveVGA1 = I915_READ(VGA1);
 	dev_priv->saveVGA_PD = I915_READ(VGA_PD);
-	dev_priv->saveVGACNTRL = I915_READ(VGACNTRL);
+	if (IS_IRONLAKE(dev))
+		dev_priv->saveVGACNTRL = I915_READ(CPU_VGACNTRL);
+	else
+		dev_priv->saveVGACNTRL = I915_READ(VGACNTRL);
 
 	i915_save_vga(dev);
 }
@@ -485,22 +656,43 @@ void i915_restore_display(struct drm_device *dev)
 		I915_WRITE(CURSIZE, dev_priv->saveCURSIZE);
 
 	/* CRT state */
-	I915_WRITE(ADPA, dev_priv->saveADPA);
+	if (IS_IRONLAKE(dev))
+		I915_WRITE(PCH_ADPA, dev_priv->saveADPA);
+	else
+		I915_WRITE(ADPA, dev_priv->saveADPA);
 
 	/* LVDS state */
-	if (IS_I965G(dev))
+	if (IS_I965G(dev) && !IS_IRONLAKE(dev))
 		I915_WRITE(BLC_PWM_CTL2, dev_priv->saveBLC_PWM_CTL2);
-	if (IS_MOBILE(dev) && !IS_I830(dev))
+
+	if (IS_IRONLAKE(dev)) {
+		I915_WRITE(PCH_LVDS, dev_priv->saveLVDS);
+	} else if (IS_MOBILE(dev) && !IS_I830(dev))
 		I915_WRITE(LVDS, dev_priv->saveLVDS);
-	if (!IS_I830(dev) && !IS_845G(dev))
+
+	if (!IS_I830(dev) && !IS_845G(dev) && !IS_IRONLAKE(dev))
 		I915_WRITE(PFIT_CONTROL, dev_priv->savePFIT_CONTROL);
 
-	I915_WRITE(PFIT_PGM_RATIOS, dev_priv->savePFIT_PGM_RATIOS);
-	I915_WRITE(BLC_PWM_CTL, dev_priv->saveBLC_PWM_CTL);
-	I915_WRITE(PP_ON_DELAYS, dev_priv->savePP_ON_DELAYS);
-	I915_WRITE(PP_OFF_DELAYS, dev_priv->savePP_OFF_DELAYS);
-	I915_WRITE(PP_DIVISOR, dev_priv->savePP_DIVISOR);
-	I915_WRITE(PP_CONTROL, dev_priv->savePP_CONTROL);
+	if (IS_IRONLAKE(dev)) {
+		I915_WRITE(BLC_PWM_PCH_CTL1, dev_priv->saveBLC_PWM_CTL);
+		I915_WRITE(BLC_PWM_PCH_CTL2, dev_priv->saveBLC_PWM_CTL2);
+		I915_WRITE(BLC_PWM_CPU_CTL, dev_priv->saveBLC_CPU_PWM_CTL);
+		I915_WRITE(BLC_PWM_CPU_CTL2, dev_priv->saveBLC_CPU_PWM_CTL2);
+		I915_WRITE(PCH_PP_ON_DELAYS, dev_priv->savePP_ON_DELAYS);
+		I915_WRITE(PCH_PP_OFF_DELAYS, dev_priv->savePP_OFF_DELAYS);
+		I915_WRITE(PCH_PP_DIVISOR, dev_priv->savePP_DIVISOR);
+		I915_WRITE(PCH_PP_CONTROL, dev_priv->savePP_CONTROL);
+		I915_WRITE(MCHBAR_RENDER_STANDBY,
+			   dev_priv->saveMCHBAR_RENDER_STANDBY);
+	} else {
+		I915_WRITE(PFIT_PGM_RATIOS, dev_priv->savePFIT_PGM_RATIOS);
+		I915_WRITE(BLC_PWM_CTL, dev_priv->saveBLC_PWM_CTL);
+		I915_WRITE(BLC_HIST_CTL, dev_priv->saveBLC_HIST_CTL);
+		I915_WRITE(PP_ON_DELAYS, dev_priv->savePP_ON_DELAYS);
+		I915_WRITE(PP_OFF_DELAYS, dev_priv->savePP_OFF_DELAYS);
+		I915_WRITE(PP_DIVISOR, dev_priv->savePP_DIVISOR);
+		I915_WRITE(PP_CONTROL, dev_priv->savePP_CONTROL);
+	}
 
 	/* Display Port state */
 	if (SUPPORTS_INTEGRATED_DP(dev)) {
@@ -511,13 +703,22 @@ void i915_restore_display(struct drm_device *dev)
 	/* FIXME: restore TV & SDVO state */
 
 	/* FBC info */
-	I915_WRITE(FBC_CFB_BASE, dev_priv->saveFBC_CFB_BASE);
-	I915_WRITE(FBC_LL_BASE, dev_priv->saveFBC_LL_BASE);
-	I915_WRITE(FBC_CONTROL2, dev_priv->saveFBC_CONTROL2);
-	I915_WRITE(FBC_CONTROL, dev_priv->saveFBC_CONTROL);
+	if (IS_GM45(dev)) {
+		g4x_disable_fbc(dev);
+		I915_WRITE(DPFC_CB_BASE, dev_priv->saveDPFC_CB_BASE);
+	} else {
+		i8xx_disable_fbc(dev);
+		I915_WRITE(FBC_CFB_BASE, dev_priv->saveFBC_CFB_BASE);
+		I915_WRITE(FBC_LL_BASE, dev_priv->saveFBC_LL_BASE);
+		I915_WRITE(FBC_CONTROL2, dev_priv->saveFBC_CONTROL2);
+		I915_WRITE(FBC_CONTROL, dev_priv->saveFBC_CONTROL);
+	}
 
 	/* VGA state */
-	I915_WRITE(VGACNTRL, dev_priv->saveVGACNTRL);
+	if (IS_IRONLAKE(dev))
+		I915_WRITE(CPU_VGACNTRL, dev_priv->saveVGACNTRL);
+	else
+		I915_WRITE(VGACNTRL, dev_priv->saveVGACNTRL);
 	I915_WRITE(VGA0, dev_priv->saveVGA0);
 	I915_WRITE(VGA1, dev_priv->saveVGA1);
 	I915_WRITE(VGA_PD, dev_priv->saveVGA_PD);
@@ -533,22 +734,28 @@ int i915_save_state(struct drm_device *dev)
 
 	pci_read_config_byte(dev->pdev, LBB, &dev_priv->saveLBB);
 
-	/* Render Standby */
-	if (IS_I965G(dev) && IS_MOBILE(dev))
-		dev_priv->saveRENDERSTANDBY = I915_READ(MCHBAR_RENDER_STANDBY);
-
 	/* Hardware status page */
 	dev_priv->saveHWS = I915_READ(HWS_PGA);
 
 	i915_save_display(dev);
 
 	/* Interrupt state */
-	dev_priv->saveIER = I915_READ(IER);
-	dev_priv->saveIMR = I915_READ(IMR);
+	if (IS_IRONLAKE(dev)) {
+		dev_priv->saveDEIER = I915_READ(DEIER);
+		dev_priv->saveDEIMR = I915_READ(DEIMR);
+		dev_priv->saveGTIER = I915_READ(GTIER);
+		dev_priv->saveGTIMR = I915_READ(GTIMR);
+		dev_priv->saveFDI_RXA_IMR = I915_READ(FDI_RXA_IMR);
+		dev_priv->saveFDI_RXB_IMR = I915_READ(FDI_RXB_IMR);
+		dev_priv->saveMCHBAR_RENDER_STANDBY =
+			I915_READ(MCHBAR_RENDER_STANDBY);
+	} else {
+		dev_priv->saveIER = I915_READ(IER);
+		dev_priv->saveIMR = I915_READ(IMR);
+	}
 
-	/* Clock gating state */
-	dev_priv->saveD_STATE = I915_READ(D_STATE);
-	dev_priv->saveDSPCLK_GATE_D = I915_READ(DSPCLK_GATE_D); /* Not sure about this */
+	if (IS_IRONLAKE_M(dev))
+		ironlake_disable_drps(dev);
 
 	/* Cache mode state */
 	dev_priv->saveCACHE_MODE_0 = I915_READ(CACHE_MODE_0);
@@ -587,10 +794,6 @@ int i915_restore_state(struct drm_device *dev)
 
 	pci_write_config_byte(dev->pdev, LBB, dev_priv->saveLBB);
 
-	/* Render Standby */
-	if (IS_I965G(dev) && IS_MOBILE(dev))
-		I915_WRITE(MCHBAR_RENDER_STANDBY, dev_priv->saveRENDERSTANDBY);
-
 	/* Hardware status page */
 	I915_WRITE(HWS_PGA, dev_priv->saveHWS);
 
@@ -609,12 +812,23 @@ int i915_restore_state(struct drm_device *dev)
 	i915_restore_display(dev);
 
 	/* Interrupt state */
-	I915_WRITE (IER, dev_priv->saveIER);
-	I915_WRITE (IMR,  dev_priv->saveIMR);
+	if (IS_IRONLAKE(dev)) {
+		I915_WRITE(DEIER, dev_priv->saveDEIER);
+		I915_WRITE(DEIMR, dev_priv->saveDEIMR);
+		I915_WRITE(GTIER, dev_priv->saveGTIER);
+		I915_WRITE(GTIMR, dev_priv->saveGTIMR);
+		I915_WRITE(FDI_RXA_IMR, dev_priv->saveFDI_RXA_IMR);
+		I915_WRITE(FDI_RXB_IMR, dev_priv->saveFDI_RXB_IMR);
+	} else {
+		I915_WRITE (IER, dev_priv->saveIER);
+		I915_WRITE (IMR,  dev_priv->saveIMR);
+	}
 
 	/* Clock gating state */
-	I915_WRITE (D_STATE, dev_priv->saveD_STATE);
-	I915_WRITE (DSPCLK_GATE_D, dev_priv->saveDSPCLK_GATE_D);
+	intel_init_clock_gating(dev);
+
+	if (IS_IRONLAKE_M(dev))
+		ironlake_enable_drps(dev);
 
 	/* Cache mode state */
 	I915_WRITE (CACHE_MODE_0, dev_priv->saveCACHE_MODE_0 | 0xffff0000);
@@ -629,6 +843,9 @@ int i915_restore_state(struct drm_device *dev)
 	for (i = 0; i < 3; i++)
 		I915_WRITE(SWF30 + (i << 2), dev_priv->saveSWF2[i]);
 
+	/* I2C state */
+	intel_i2c_reset_gmbus(dev);
+
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/i915/i915_trace.h b/drivers/gpu/drm/i915/i915_trace.h
index 5567a40816f3..01840d9bc38f 100644
--- a/drivers/gpu/drm/i915/i915_trace.h
+++ b/drivers/gpu/drm/i915/i915_trace.h
@@ -158,16 +158,17 @@ TRACE_EVENT(i915_gem_request_submit,
 	    TP_ARGS(dev, seqno),
 
 	    TP_STRUCT__entry(
-			     __field(struct drm_device *, dev)
+			     __field(u32, dev)
 			     __field(u32, seqno)
 			     ),
 
 	    TP_fast_assign(
-			   __entry->dev = dev;
+			   __entry->dev = dev->primary->index;
 			   __entry->seqno = seqno;
+			   i915_trace_irq_get(dev, seqno);
 			   ),
 
-	    TP_printk("dev=%p, seqno=%u", __entry->dev, __entry->seqno)
+	    TP_printk("dev=%u, seqno=%u", __entry->dev, __entry->seqno)
 );
 
 TRACE_EVENT(i915_gem_request_flush,
@@ -178,20 +179,20 @@ TRACE_EVENT(i915_gem_request_flush,
 	    TP_ARGS(dev, seqno, flush_domains, invalidate_domains),
 
 	    TP_STRUCT__entry(
-			     __field(struct drm_device *, dev)
+			     __field(u32, dev)
 			     __field(u32, seqno)
 			     __field(u32, flush_domains)
 			     __field(u32, invalidate_domains)
 			     ),
 
 	    TP_fast_assign(
-			   __entry->dev = dev;
+			   __entry->dev = dev->primary->index;
 			   __entry->seqno = seqno;
 			   __entry->flush_domains = flush_domains;
 			   __entry->invalidate_domains = invalidate_domains;
 			   ),
 
-	    TP_printk("dev=%p, seqno=%u, flush=%04x, invalidate=%04x",
+	    TP_printk("dev=%u, seqno=%u, flush=%04x, invalidate=%04x",
 		      __entry->dev, __entry->seqno,
 		      __entry->flush_domains, __entry->invalidate_domains)
 );
@@ -204,16 +205,16 @@ TRACE_EVENT(i915_gem_request_complete,
 	    TP_ARGS(dev, seqno),
 
 	    TP_STRUCT__entry(
-			     __field(struct drm_device *, dev)
+			     __field(u32, dev)
 			     __field(u32, seqno)
 			     ),
 
 	    TP_fast_assign(
-			   __entry->dev = dev;
+			   __entry->dev = dev->primary->index;
 			   __entry->seqno = seqno;
 			   ),
 
-	    TP_printk("dev=%p, seqno=%u", __entry->dev, __entry->seqno)
+	    TP_printk("dev=%u, seqno=%u", __entry->dev, __entry->seqno)
 );
 
 TRACE_EVENT(i915_gem_request_retire,
@@ -223,16 +224,16 @@ TRACE_EVENT(i915_gem_request_retire,
 	    TP_ARGS(dev, seqno),
 
 	    TP_STRUCT__entry(
-			     __field(struct drm_device *, dev)
+			     __field(u32, dev)
 			     __field(u32, seqno)
 			     ),
 
 	    TP_fast_assign(
-			   __entry->dev = dev;
+			   __entry->dev = dev->primary->index;
 			   __entry->seqno = seqno;
 			   ),
 
-	    TP_printk("dev=%p, seqno=%u", __entry->dev, __entry->seqno)
+	    TP_printk("dev=%u, seqno=%u", __entry->dev, __entry->seqno)
 );
 
 TRACE_EVENT(i915_gem_request_wait_begin,
@@ -242,16 +243,16 @@ TRACE_EVENT(i915_gem_request_wait_begin,
 	    TP_ARGS(dev, seqno),
 
 	    TP_STRUCT__entry(
-			     __field(struct drm_device *, dev)
+			     __field(u32, dev)
 			     __field(u32, seqno)
 			     ),
 
 	    TP_fast_assign(
-			   __entry->dev = dev;
+			   __entry->dev = dev->primary->index;
 			   __entry->seqno = seqno;
 			   ),
 
-	    TP_printk("dev=%p, seqno=%u", __entry->dev, __entry->seqno)
+	    TP_printk("dev=%u, seqno=%u", __entry->dev, __entry->seqno)
 );
 
 TRACE_EVENT(i915_gem_request_wait_end,
@@ -261,16 +262,16 @@ TRACE_EVENT(i915_gem_request_wait_end,
 	    TP_ARGS(dev, seqno),
 
 	    TP_STRUCT__entry(
-			     __field(struct drm_device *, dev)
+			     __field(u32, dev)
 			     __field(u32, seqno)
 			     ),
 
 	    TP_fast_assign(
-			   __entry->dev = dev;
+			   __entry->dev = dev->primary->index;
 			   __entry->seqno = seqno;
 			   ),
 
-	    TP_printk("dev=%p, seqno=%u", __entry->dev, __entry->seqno)
+	    TP_printk("dev=%u, seqno=%u", __entry->dev, __entry->seqno)
 );
 
 TRACE_EVENT(i915_ring_wait_begin,
@@ -280,14 +281,14 @@ TRACE_EVENT(i915_ring_wait_begin,
 	    TP_ARGS(dev),
 
 	    TP_STRUCT__entry(
-			     __field(struct drm_device *, dev)
+			     __field(u32, dev)
 			     ),
 
 	    TP_fast_assign(
-			   __entry->dev = dev;
+			   __entry->dev = dev->primary->index;
 			   ),
 
-	    TP_printk("dev=%p", __entry->dev)
+	    TP_printk("dev=%u", __entry->dev)
 );
 
 TRACE_EVENT(i915_ring_wait_end,
@@ -297,14 +298,14 @@ TRACE_EVENT(i915_ring_wait_end,
 	    TP_ARGS(dev),
 
 	    TP_STRUCT__entry(
-			     __field(struct drm_device *, dev)
+			     __field(u32, dev)
 			     ),
 
 	    TP_fast_assign(
-			   __entry->dev = dev;
+			   __entry->dev = dev->primary->index;
 			   ),
 
-	    TP_printk("dev=%p", __entry->dev)
+	    TP_printk("dev=%u", __entry->dev)
 );
 
 #endif /* _I915_TRACE_H_ */
diff --git a/drivers/gpu/drm/i915/intel_bios.c b/drivers/gpu/drm/i915/intel_bios.c
index 4337414846b6..70c9d4ba7042 100644
--- a/drivers/gpu/drm/i915/intel_bios.c
+++ b/drivers/gpu/drm/i915/intel_bios.c
@@ -33,6 +33,8 @@
 #define	SLAVE_ADDR1	0x70
 #define	SLAVE_ADDR2	0x72
 
+static int panel_type;
+
 static void *
 find_section(struct bdb_header *bdb, int section_id)
 {
@@ -114,6 +116,8 @@ parse_lfp_panel_data(struct drm_i915_private *dev_priv,
 	struct lvds_dvo_timing *dvo_timing;
 	struct drm_display_mode *panel_fixed_mode;
 	int lfp_data_size, dvo_timing_offset;
+	int i, temp_downclock;
+	struct drm_display_mode *temp_mode;
 
 	/* Defaults if we can't find VBT info */
 	dev_priv->lvds_dither = 0;
@@ -126,6 +130,7 @@ parse_lfp_panel_data(struct drm_i915_private *dev_priv,
 	dev_priv->lvds_dither = lvds_options->pixel_dither;
 	if (lvds_options->panel_type == 0xff)
 		return;
+	panel_type = lvds_options->panel_type;
 
 	lvds_lfp_data = find_section(bdb, BDB_LVDS_LFP_DATA);
 	if (!lvds_lfp_data)
@@ -159,9 +164,50 @@ parse_lfp_panel_data(struct drm_i915_private *dev_priv,
 
 	dev_priv->lfp_lvds_vbt_mode = panel_fixed_mode;
 
-	DRM_DEBUG("Found panel mode in BIOS VBT tables:\n");
+	DRM_DEBUG_KMS("Found panel mode in BIOS VBT tables:\n");
 	drm_mode_debug_printmodeline(panel_fixed_mode);
 
+	temp_mode = kzalloc(sizeof(*temp_mode), GFP_KERNEL);
+	temp_downclock = panel_fixed_mode->clock;
+	/*
+	 * enumerate the LVDS panel timing info entry in VBT to check whether
+	 * the LVDS downclock is found.
+	 */
+	for (i = 0; i < 16; i++) {
+		entry = (struct bdb_lvds_lfp_data_entry *)
+			((uint8_t *)lvds_lfp_data->data + (lfp_data_size * i));
+		dvo_timing = (struct lvds_dvo_timing *)
+			((unsigned char *)entry + dvo_timing_offset);
+
+		fill_detail_timing_data(temp_mode, dvo_timing);
+
+		if (temp_mode->hdisplay == panel_fixed_mode->hdisplay &&
+		temp_mode->hsync_start == panel_fixed_mode->hsync_start &&
+		temp_mode->hsync_end == panel_fixed_mode->hsync_end &&
+		temp_mode->htotal == panel_fixed_mode->htotal &&
+		temp_mode->vdisplay == panel_fixed_mode->vdisplay &&
+		temp_mode->vsync_start == panel_fixed_mode->vsync_start &&
+		temp_mode->vsync_end == panel_fixed_mode->vsync_end &&
+		temp_mode->vtotal == panel_fixed_mode->vtotal &&
+		temp_mode->clock < temp_downclock) {
+			/*
+			 * downclock is already found. But we expect
+			 * to find the lower downclock.
+			 */
+			temp_downclock = temp_mode->clock;
+		}
+		/* clear it to zero */
+		memset(temp_mode, 0, sizeof(*temp_mode));
+	}
+	kfree(temp_mode);
+	if (temp_downclock < panel_fixed_mode->clock &&
+	    i915_lvds_downclock) {
+		dev_priv->lvds_downclock_avail = 1;
+		dev_priv->lvds_downclock = temp_downclock;
+		DRM_DEBUG_KMS("LVDS downclock is found in VBT. ",
+				"Normal Clock %dKHz, downclock %dKHz\n",
+				temp_downclock, panel_fixed_mode->clock);
+	}
 	return;
 }
 
@@ -201,6 +247,7 @@ static void
 parse_general_features(struct drm_i915_private *dev_priv,
 		       struct bdb_header *bdb)
 {
+	struct drm_device *dev = dev_priv->dev;
 	struct bdb_general_features *general;
 
 	/* Set sensible defaults in case we can't find the general block */
@@ -217,7 +264,7 @@ parse_general_features(struct drm_i915_private *dev_priv,
 			if (IS_I85X(dev_priv->dev))
 				dev_priv->lvds_ssc_freq =
 					general->ssc_freq ? 66 : 48;
-			else if (IS_IGDNG(dev_priv->dev))
+			else if (IS_IRONLAKE(dev_priv->dev) || IS_GEN6(dev))
 				dev_priv->lvds_ssc_freq =
 					general->ssc_freq ? 100 : 120;
 			else
@@ -241,22 +288,18 @@ parse_general_definitions(struct drm_i915_private *dev_priv,
 		GPIOF,
 	};
 
-	/* Set sensible defaults in case we can't find the general block
-	   or it is the wrong chipset */
-	dev_priv->crt_ddc_bus = -1;
-
 	general = find_section(bdb, BDB_GENERAL_DEFINITIONS);
 	if (general) {
 		u16 block_size = get_blocksize(general);
 		if (block_size >= sizeof(*general)) {
 			int bus_pin = general->crt_ddc_gmbus_pin;
-			DRM_DEBUG("crt_ddc_bus_pin: %d\n", bus_pin);
+			DRM_DEBUG_KMS("crt_ddc_bus_pin: %d\n", bus_pin);
 			if ((bus_pin >= 1) && (bus_pin <= 6)) {
 				dev_priv->crt_ddc_bus =
 					crt_bus_map_table[bus_pin-1];
 			}
 		} else {
-			DRM_DEBUG("BDB_GD too small (%d). Invalid.\n",
+			DRM_DEBUG_KMS("BDB_GD too small (%d). Invalid.\n",
 				  block_size);
 		}
 	}
@@ -274,7 +317,7 @@ parse_sdvo_device_mapping(struct drm_i915_private *dev_priv,
 
 	p_defs = find_section(bdb, BDB_GENERAL_DEFINITIONS);
 	if (!p_defs) {
-		DRM_DEBUG("No general definition block is found\n");
+		DRM_DEBUG_KMS("No general definition block is found\n");
 		return;
 	}
 	/* judge whether the size of child device meets the requirements.
@@ -284,7 +327,7 @@ parse_sdvo_device_mapping(struct drm_i915_private *dev_priv,
 	 */
 	if (p_defs->child_dev_size != sizeof(*p_child)) {
 		/* different child dev size . Ignore it */
-		DRM_DEBUG("different child size is found. Invalid.\n");
+		DRM_DEBUG_KMS("different child size is found. Invalid.\n");
 		return;
 	}
 	/* get the block size of general definitions */
@@ -310,11 +353,11 @@ parse_sdvo_device_mapping(struct drm_i915_private *dev_priv,
 		if (p_child->dvo_port != DEVICE_PORT_DVOB &&
 			p_child->dvo_port != DEVICE_PORT_DVOC) {
 			/* skip the incorrect SDVO port */
-			DRM_DEBUG("Incorrect SDVO port. Skip it \n");
+			DRM_DEBUG_KMS("Incorrect SDVO port. Skip it \n");
 			continue;
 		}
-		DRM_DEBUG("the SDVO device with slave addr %2x is found on "
-				"%s port\n",
+		DRM_DEBUG_KMS("the SDVO device with slave addr %2x is found on"
+				" %s port\n",
 				p_child->slave_addr,
 				(p_child->dvo_port == DEVICE_PORT_DVOB) ?
 					"SDVOB" : "SDVOC");
@@ -325,21 +368,21 @@ parse_sdvo_device_mapping(struct drm_i915_private *dev_priv,
 			p_mapping->dvo_wiring = p_child->dvo_wiring;
 			p_mapping->initialized = 1;
 		} else {
-			DRM_DEBUG("Maybe one SDVO port is shared by "
+			DRM_DEBUG_KMS("Maybe one SDVO port is shared by "
 					 "two SDVO device.\n");
 		}
 		if (p_child->slave2_addr) {
 			/* Maybe this is a SDVO device with multiple inputs */
 			/* And the mapping info is not added */
-			DRM_DEBUG("there exists the slave2_addr. Maybe this "
-				"is a SDVO device with multiple inputs.\n");
+			DRM_DEBUG_KMS("there exists the slave2_addr. Maybe this"
+				" is a SDVO device with multiple inputs.\n");
 		}
 		count++;
 	}
 
 	if (!count) {
 		/* No SDVO device info is found */
-		DRM_DEBUG("No SDVO device info is found in VBT\n");
+		DRM_DEBUG_KMS("No SDVO device info is found in VBT\n");
 	}
 	return;
 }
@@ -351,23 +394,113 @@ parse_driver_features(struct drm_i915_private *dev_priv,
 	struct drm_device *dev = dev_priv->dev;
 	struct bdb_driver_features *driver;
 
-	/* set default for chips without eDP */
-	if (!SUPPORTS_EDP(dev)) {
-		dev_priv->edp_support = 0;
-		return;
-	}
-
 	driver = find_section(bdb, BDB_DRIVER_FEATURES);
 	if (!driver)
 		return;
 
-	if (driver->lvds_config == BDB_DRIVER_FEATURE_EDP)
+	if (driver && SUPPORTS_EDP(dev) &&
+	    driver->lvds_config == BDB_DRIVER_FEATURE_EDP) {
 		dev_priv->edp_support = 1;
+	} else {
+		dev_priv->edp_support = 0;
+	}
 
-	if (driver->dual_frequency)
+	if (driver && driver->dual_frequency)
 		dev_priv->render_reclock_avail = true;
 }
 
+static void
+parse_edp(struct drm_i915_private *dev_priv, struct bdb_header *bdb)
+{
+	struct bdb_edp *edp;
+
+	edp = find_section(bdb, BDB_EDP);
+	if (!edp) {
+		if (SUPPORTS_EDP(dev_priv->dev) && dev_priv->edp_support) {
+			DRM_DEBUG_KMS("No eDP BDB found but eDP panel supported,\
+				       assume 18bpp panel color depth.\n");
+			dev_priv->edp_bpp = 18;
+		}
+		return;
+	}
+
+	switch ((edp->color_depth >> (panel_type * 2)) & 3) {
+	case EDP_18BPP:
+		dev_priv->edp_bpp = 18;
+		break;
+	case EDP_24BPP:
+		dev_priv->edp_bpp = 24;
+		break;
+	case EDP_30BPP:
+		dev_priv->edp_bpp = 30;
+		break;
+	}
+}
+
+static void
+parse_device_mapping(struct drm_i915_private *dev_priv,
+		       struct bdb_header *bdb)
+{
+	struct bdb_general_definitions *p_defs;
+	struct child_device_config *p_child, *child_dev_ptr;
+	int i, child_device_num, count;
+	u16	block_size;
+
+	p_defs = find_section(bdb, BDB_GENERAL_DEFINITIONS);
+	if (!p_defs) {
+		DRM_DEBUG_KMS("No general definition block is found\n");
+		return;
+	}
+	/* judge whether the size of child device meets the requirements.
+	 * If the child device size obtained from general definition block
+	 * is different with sizeof(struct child_device_config), skip the
+	 * parsing of sdvo device info
+	 */
+	if (p_defs->child_dev_size != sizeof(*p_child)) {
+		/* different child dev size . Ignore it */
+		DRM_DEBUG_KMS("different child size is found. Invalid.\n");
+		return;
+	}
+	/* get the block size of general definitions */
+	block_size = get_blocksize(p_defs);
+	/* get the number of child device */
+	child_device_num = (block_size - sizeof(*p_defs)) /
+				sizeof(*p_child);
+	count = 0;
+	/* get the number of child device that is present */
+	for (i = 0; i < child_device_num; i++) {
+		p_child = &(p_defs->devices[i]);
+		if (!p_child->device_type) {
+			/* skip the device block if device type is invalid */
+			continue;
+		}
+		count++;
+	}
+	if (!count) {
+		DRM_DEBUG_KMS("no child dev is parsed from VBT \n");
+		return;
+	}
+	dev_priv->child_dev = kzalloc(sizeof(*p_child) * count, GFP_KERNEL);
+	if (!dev_priv->child_dev) {
+		DRM_DEBUG_KMS("No memory space for child device\n");
+		return;
+	}
+
+	dev_priv->child_dev_num = count;
+	count = 0;
+	for (i = 0; i < child_device_num; i++) {
+		p_child = &(p_defs->devices[i]);
+		if (!p_child->device_type) {
+			/* skip the device block if device type is invalid */
+			continue;
+		}
+		child_dev_ptr = dev_priv->child_dev + count;
+		count++;
+		memcpy((void *)child_dev_ptr, (void *)p_child,
+					sizeof(*p_child));
+	}
+	return;
+}
 /**
  * intel_init_bios - initialize VBIOS settings & find VBT
  * @dev: DRM device
@@ -419,7 +552,9 @@ intel_init_bios(struct drm_device *dev)
 	parse_lfp_panel_data(dev_priv, bdb);
 	parse_sdvo_panel_data(dev_priv, bdb);
 	parse_sdvo_device_mapping(dev_priv, bdb);
+	parse_device_mapping(dev_priv, bdb);
 	parse_driver_features(dev_priv, bdb);
+	parse_edp(dev_priv, bdb);
 
 	pci_unmap_rom(pdev, bios);
 
diff --git a/drivers/gpu/drm/i915/intel_bios.h b/drivers/gpu/drm/i915/intel_bios.h
index 0f8e5f69ac7a..4c18514f6f80 100644
--- a/drivers/gpu/drm/i915/intel_bios.h
+++ b/drivers/gpu/drm/i915/intel_bios.h
@@ -98,6 +98,7 @@ struct vbios_data {
 #define BDB_SDVO_LVDS_PNP_IDS	 24
 #define BDB_SDVO_LVDS_POWER_SEQ	 25
 #define BDB_TV_OPTIONS		 26
+#define BDB_EDP			 27
 #define BDB_LVDS_OPTIONS	 40
 #define BDB_LVDS_LFP_DATA_PTRS	 41
 #define BDB_LVDS_LFP_DATA	 42
@@ -426,6 +427,45 @@ struct bdb_driver_features {
 	u8 custom_vbt_version;
 } __attribute__((packed));
 
+#define EDP_18BPP	0
+#define EDP_24BPP	1
+#define EDP_30BPP	2
+#define EDP_RATE_1_62	0
+#define EDP_RATE_2_7	1
+#define EDP_LANE_1	0
+#define EDP_LANE_2	1
+#define EDP_LANE_4	3
+#define EDP_PREEMPHASIS_NONE	0
+#define EDP_PREEMPHASIS_3_5dB	1
+#define EDP_PREEMPHASIS_6dB	2
+#define EDP_PREEMPHASIS_9_5dB	3
+#define EDP_VSWING_0_4V		0
+#define EDP_VSWING_0_6V		1
+#define EDP_VSWING_0_8V		2
+#define EDP_VSWING_1_2V		3
+
+struct edp_power_seq {
+	u16 t3;
+	u16 t7;
+	u16 t9;
+	u16 t10;
+	u16 t12;
+} __attribute__ ((packed));
+
+struct edp_link_params {
+	u8 rate:4;
+	u8 lanes:4;
+	u8 preemphasis:4;
+	u8 vswing:4;
+} __attribute__ ((packed));
+
+struct bdb_edp {
+	struct edp_power_seq power_seqs[16];
+	u32 color_depth;
+	u32 sdrrs_msa_timing_delay;
+	struct edp_link_params link_params[16];
+} __attribute__ ((packed));
+
 bool intel_init_bios(struct drm_device *dev);
 
 /*
@@ -549,4 +589,21 @@ bool intel_init_bios(struct drm_device *dev);
 #define   SWF14_APM_STANDBY	0x1
 #define   SWF14_APM_RESTORE	0x0
 
+/* Add the device class for LFP, TV, HDMI */
+#define	 DEVICE_TYPE_INT_LFP	0x1022
+#define	 DEVICE_TYPE_INT_TV	0x1009
+#define	 DEVICE_TYPE_HDMI	0x60D2
+#define	 DEVICE_TYPE_DP		0x68C6
+#define	 DEVICE_TYPE_eDP	0x78C6
+
+/* define the DVO port for HDMI output type */
+#define		DVO_B		1
+#define		DVO_C		2
+#define		DVO_D		3
+
+/* define the PORT for DP output type */
+#define		PORT_IDPB	7
+#define		PORT_IDPC	8
+#define		PORT_IDPD	9
+
 #endif /* _I830_BIOS_H_ */
diff --git a/drivers/gpu/drm/i915/intel_crt.c b/drivers/gpu/drm/i915/intel_crt.c
index 212e22740fc1..fccf07470c8f 100644
--- a/drivers/gpu/drm/i915/intel_crt.c
+++ b/drivers/gpu/drm/i915/intel_crt.c
@@ -39,7 +39,7 @@ static void intel_crt_dpms(struct drm_encoder *encoder, int mode)
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	u32 temp, reg;
 
-	if (IS_IGDNG(dev))
+	if (HAS_PCH_SPLIT(dev))
 		reg = PCH_ADPA;
 	else
 		reg = ADPA;
@@ -64,34 +64,6 @@ static void intel_crt_dpms(struct drm_encoder *encoder, int mode)
 	}
 
 	I915_WRITE(reg, temp);
-
-	if (IS_IGD(dev)) {
-		if (mode == DRM_MODE_DPMS_OFF) {
-			/* turn off DAC */
-			temp = I915_READ(PORT_HOTPLUG_EN);
-			temp &= ~CRT_EOS_INT_EN;
-			I915_WRITE(PORT_HOTPLUG_EN, temp);
-
-			temp = I915_READ(PORT_HOTPLUG_STAT);
-			if (temp & CRT_EOS_INT_STATUS)
-				I915_WRITE(PORT_HOTPLUG_STAT,
-					CRT_EOS_INT_STATUS);
-		} else {
-			/* turn on DAC. EOS interrupt must be enabled after DAC
-			 * is enabled, so it sounds not good to enable it in
-			 * i915_driver_irq_postinstall()
-			 * wait 12.5ms after DAC is enabled
-			 */
-			msleep(13);
-			temp = I915_READ(PORT_HOTPLUG_STAT);
-			if (temp & CRT_EOS_INT_STATUS)
-				I915_WRITE(PORT_HOTPLUG_STAT,
-					CRT_EOS_INT_STATUS);
-			temp = I915_READ(PORT_HOTPLUG_EN);
-			temp |= CRT_EOS_INT_EN;
-			I915_WRITE(PORT_HOTPLUG_EN, temp);
-		}
-	}
 }
 
 static int intel_crt_mode_valid(struct drm_connector *connector,
@@ -141,7 +113,7 @@ static void intel_crt_mode_set(struct drm_encoder *encoder,
 	else
 		dpll_md_reg = DPLL_B_MD;
 
-	if (IS_IGDNG(dev))
+	if (HAS_PCH_SPLIT(dev))
 		adpa_reg = PCH_ADPA;
 	else
 		adpa_reg = ADPA;
@@ -150,7 +122,7 @@ static void intel_crt_mode_set(struct drm_encoder *encoder,
 	 * Disable separate mode multiplier used when cloning SDVO to CRT
 	 * XXX this needs to be adjusted when we really are cloning
 	 */
-	if (IS_I965G(dev) && !IS_IGDNG(dev)) {
+	if (IS_I965G(dev) && !HAS_PCH_SPLIT(dev)) {
 		dpll_md = I915_READ(dpll_md_reg);
 		I915_WRITE(dpll_md_reg,
 			   dpll_md & ~DPLL_MD_UDI_MULTIPLIER_MASK);
@@ -164,18 +136,18 @@ static void intel_crt_mode_set(struct drm_encoder *encoder,
 
 	if (intel_crtc->pipe == 0) {
 		adpa |= ADPA_PIPE_A_SELECT;
-		if (!IS_IGDNG(dev))
+		if (!HAS_PCH_SPLIT(dev))
 			I915_WRITE(BCLRPAT_A, 0);
 	} else {
 		adpa |= ADPA_PIPE_B_SELECT;
-		if (!IS_IGDNG(dev))
+		if (!HAS_PCH_SPLIT(dev))
 			I915_WRITE(BCLRPAT_B, 0);
 	}
 
 	I915_WRITE(adpa_reg, adpa);
 }
 
-static bool intel_igdng_crt_detect_hotplug(struct drm_connector *connector)
+static bool intel_ironlake_crt_detect_hotplug(struct drm_connector *connector)
 {
 	struct drm_device *dev = connector->dev;
 	struct drm_i915_private *dev_priv = dev->dev_private;
@@ -185,6 +157,9 @@ static bool intel_igdng_crt_detect_hotplug(struct drm_connector *connector)
 	adpa = I915_READ(PCH_ADPA);
 
 	adpa &= ~ADPA_CRT_HOTPLUG_MASK;
+	/* disable HPD first */
+	I915_WRITE(PCH_ADPA, adpa);
+	(void)I915_READ(PCH_ADPA);
 
 	adpa |= (ADPA_CRT_HOTPLUG_PERIOD_128 |
 			ADPA_CRT_HOTPLUG_WARMUP_10MS |
@@ -194,7 +169,7 @@ static bool intel_igdng_crt_detect_hotplug(struct drm_connector *connector)
 			ADPA_CRT_HOTPLUG_ENABLE |
 			ADPA_CRT_HOTPLUG_FORCE_TRIGGER);
 
-	DRM_DEBUG("pch crt adpa 0x%x", adpa);
+	DRM_DEBUG_KMS("pch crt adpa 0x%x", adpa);
 	I915_WRITE(PCH_ADPA, adpa);
 
 	while ((I915_READ(PCH_ADPA) & ADPA_CRT_HOTPLUG_FORCE_TRIGGER) != 0)
@@ -227,8 +202,8 @@ static bool intel_crt_detect_hotplug(struct drm_connector *connector)
 	u32 hotplug_en;
 	int i, tries = 0;
 
-	if (IS_IGDNG(dev))
-		return intel_igdng_crt_detect_hotplug(connector);
+	if (HAS_PCH_SPLIT(dev))
+		return intel_ironlake_crt_detect_hotplug(connector);
 
 	/*
 	 * On 4 series desktop, CRT detect sequence need to be done twice
@@ -262,8 +237,8 @@ static bool intel_crt_detect_hotplug(struct drm_connector *connector)
 		} while (time_after(timeout, jiffies));
 	}
 
-	if ((I915_READ(PORT_HOTPLUG_STAT) & CRT_HOTPLUG_MONITOR_MASK) ==
-	    CRT_HOTPLUG_MONITOR_COLOR)
+	if ((I915_READ(PORT_HOTPLUG_STAT) & CRT_HOTPLUG_MONITOR_MASK) !=
+	    CRT_HOTPLUG_MONITOR_NONE)
 		return true;
 
 	return false;
@@ -549,12 +524,12 @@ void intel_crt_init(struct drm_device *dev)
 					  &intel_output->enc);
 
 	/* Set up the DDC bus. */
-	if (IS_IGDNG(dev))
+	if (HAS_PCH_SPLIT(dev))
 		i2c_reg = PCH_GPIOA;
 	else {
 		i2c_reg = GPIOA;
 		/* Use VBT information for CRT DDC if available */
-		if (dev_priv->crt_ddc_bus != -1)
+		if (dev_priv->crt_ddc_bus != 0)
 			i2c_reg = dev_priv->crt_ddc_bus;
 	}
 	intel_output->ddc_bus = intel_i2c_create(dev, i2c_reg, "CRTDDC_A");
@@ -576,4 +551,6 @@ void intel_crt_init(struct drm_device *dev)
 	drm_connector_helper_add(connector, &intel_crt_connector_helper_funcs);
 
 	drm_sysfs_connector_add(connector);
+
+	dev_priv->hotplug_supported_mask |= CRT_HOTPLUG_INT_STATUS;
 }
diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
index 93ff6c03733e..9cd6de5f9906 100644
--- a/drivers/gpu/drm/i915/intel_display.c
+++ b/drivers/gpu/drm/i915/intel_display.c
@@ -32,7 +32,7 @@
 #include "intel_drv.h"
 #include "i915_drm.h"
 #include "i915_drv.h"
-#include "intel_dp.h"
+#include "drm_dp_helper.h"
 
 #include "drm_crtc_helper.h"
 
@@ -70,8 +70,6 @@ struct intel_limit {
     intel_p2_t	    p2;
     bool (* find_pll)(const intel_limit_t *, struct drm_crtc *,
 		      int, int, intel_clock_t *);
-    bool (* find_reduced_pll)(const intel_limit_t *, struct drm_crtc *,
-			      int, int, intel_clock_t *);
 };
 
 #define I8XX_DOT_MIN		  25000
@@ -102,32 +100,32 @@ struct intel_limit {
 #define I9XX_DOT_MAX		 400000
 #define I9XX_VCO_MIN		1400000
 #define I9XX_VCO_MAX		2800000
-#define IGD_VCO_MIN		1700000
-#define IGD_VCO_MAX		3500000
+#define PINEVIEW_VCO_MIN		1700000
+#define PINEVIEW_VCO_MAX		3500000
 #define I9XX_N_MIN		      1
 #define I9XX_N_MAX		      6
-/* IGD's Ncounter is a ring counter */
-#define IGD_N_MIN		      3
-#define IGD_N_MAX		      6
+/* Pineview's Ncounter is a ring counter */
+#define PINEVIEW_N_MIN		      3
+#define PINEVIEW_N_MAX		      6
 #define I9XX_M_MIN		     70
 #define I9XX_M_MAX		    120
-#define IGD_M_MIN		      2
-#define IGD_M_MAX		    256
+#define PINEVIEW_M_MIN		      2
+#define PINEVIEW_M_MAX		    256
 #define I9XX_M1_MIN		     10
 #define I9XX_M1_MAX		     22
 #define I9XX_M2_MIN		      5
 #define I9XX_M2_MAX		      9
-/* IGD M1 is reserved, and must be 0 */
-#define IGD_M1_MIN		      0
-#define IGD_M1_MAX		      0
-#define IGD_M2_MIN		      0
-#define IGD_M2_MAX		      254
+/* Pineview M1 is reserved, and must be 0 */
+#define PINEVIEW_M1_MIN		      0
+#define PINEVIEW_M1_MAX		      0
+#define PINEVIEW_M2_MIN		      0
+#define PINEVIEW_M2_MAX		      254
 #define I9XX_P_SDVO_DAC_MIN	      5
 #define I9XX_P_SDVO_DAC_MAX	     80
 #define I9XX_P_LVDS_MIN		      7
 #define I9XX_P_LVDS_MAX		     98
-#define IGD_P_LVDS_MIN		      7
-#define IGD_P_LVDS_MAX		     112
+#define PINEVIEW_P_LVDS_MIN		      7
+#define PINEVIEW_P_LVDS_MAX		     112
 #define I9XX_P1_MIN		      1
 #define I9XX_P1_MAX		      8
 #define I9XX_P2_SDVO_DAC_SLOW		     10
@@ -234,53 +232,108 @@ struct intel_limit {
 #define G4X_P2_DISPLAY_PORT_FAST           10
 #define G4X_P2_DISPLAY_PORT_LIMIT          0
 
-/* IGDNG */
+/* Ironlake / Sandybridge */
 /* as we calculate clock using (register_value + 2) for
    N/M1/M2, so here the range value for them is (actual_value-2).
  */
-#define IGDNG_DOT_MIN         25000
-#define IGDNG_DOT_MAX         350000
-#define IGDNG_VCO_MIN         1760000
-#define IGDNG_VCO_MAX         3510000
-#define IGDNG_N_MIN           1
-#define IGDNG_N_MAX           5
-#define IGDNG_M_MIN           79
-#define IGDNG_M_MAX           118
-#define IGDNG_M1_MIN          12
-#define IGDNG_M1_MAX          23
-#define IGDNG_M2_MIN          5
-#define IGDNG_M2_MAX          9
-#define IGDNG_P_SDVO_DAC_MIN  5
-#define IGDNG_P_SDVO_DAC_MAX  80
-#define IGDNG_P_LVDS_MIN      28
-#define IGDNG_P_LVDS_MAX      112
-#define IGDNG_P1_MIN          1
-#define IGDNG_P1_MAX          8
-#define IGDNG_P2_SDVO_DAC_SLOW 10
-#define IGDNG_P2_SDVO_DAC_FAST 5
-#define IGDNG_P2_LVDS_SLOW    14 /* single channel */
-#define IGDNG_P2_LVDS_FAST    7  /* double channel */
-#define IGDNG_P2_DOT_LIMIT    225000 /* 225Mhz */
+#define IRONLAKE_DOT_MIN         25000
+#define IRONLAKE_DOT_MAX         350000
+#define IRONLAKE_VCO_MIN         1760000
+#define IRONLAKE_VCO_MAX         3510000
+#define IRONLAKE_M1_MIN          12
+#define IRONLAKE_M1_MAX          22
+#define IRONLAKE_M2_MIN          5
+#define IRONLAKE_M2_MAX          9
+#define IRONLAKE_P2_DOT_LIMIT    225000 /* 225Mhz */
+
+/* We have parameter ranges for different type of outputs. */
+
+/* DAC & HDMI Refclk 120Mhz */
+#define IRONLAKE_DAC_N_MIN	1
+#define IRONLAKE_DAC_N_MAX	5
+#define IRONLAKE_DAC_M_MIN	79
+#define IRONLAKE_DAC_M_MAX	127
+#define IRONLAKE_DAC_P_MIN	5
+#define IRONLAKE_DAC_P_MAX	80
+#define IRONLAKE_DAC_P1_MIN	1
+#define IRONLAKE_DAC_P1_MAX	8
+#define IRONLAKE_DAC_P2_SLOW	10
+#define IRONLAKE_DAC_P2_FAST	5
+
+/* LVDS single-channel 120Mhz refclk */
+#define IRONLAKE_LVDS_S_N_MIN	1
+#define IRONLAKE_LVDS_S_N_MAX	3
+#define IRONLAKE_LVDS_S_M_MIN	79
+#define IRONLAKE_LVDS_S_M_MAX	118
+#define IRONLAKE_LVDS_S_P_MIN	28
+#define IRONLAKE_LVDS_S_P_MAX	112
+#define IRONLAKE_LVDS_S_P1_MIN	2
+#define IRONLAKE_LVDS_S_P1_MAX	8
+#define IRONLAKE_LVDS_S_P2_SLOW	14
+#define IRONLAKE_LVDS_S_P2_FAST	14
+
+/* LVDS dual-channel 120Mhz refclk */
+#define IRONLAKE_LVDS_D_N_MIN	1
+#define IRONLAKE_LVDS_D_N_MAX	3
+#define IRONLAKE_LVDS_D_M_MIN	79
+#define IRONLAKE_LVDS_D_M_MAX	127
+#define IRONLAKE_LVDS_D_P_MIN	14
+#define IRONLAKE_LVDS_D_P_MAX	56
+#define IRONLAKE_LVDS_D_P1_MIN	2
+#define IRONLAKE_LVDS_D_P1_MAX	8
+#define IRONLAKE_LVDS_D_P2_SLOW	7
+#define IRONLAKE_LVDS_D_P2_FAST	7
+
+/* LVDS single-channel 100Mhz refclk */
+#define IRONLAKE_LVDS_S_SSC_N_MIN	1
+#define IRONLAKE_LVDS_S_SSC_N_MAX	2
+#define IRONLAKE_LVDS_S_SSC_M_MIN	79
+#define IRONLAKE_LVDS_S_SSC_M_MAX	126
+#define IRONLAKE_LVDS_S_SSC_P_MIN	28
+#define IRONLAKE_LVDS_S_SSC_P_MAX	112
+#define IRONLAKE_LVDS_S_SSC_P1_MIN	2
+#define IRONLAKE_LVDS_S_SSC_P1_MAX	8
+#define IRONLAKE_LVDS_S_SSC_P2_SLOW	14
+#define IRONLAKE_LVDS_S_SSC_P2_FAST	14
+
+/* LVDS dual-channel 100Mhz refclk */
+#define IRONLAKE_LVDS_D_SSC_N_MIN	1
+#define IRONLAKE_LVDS_D_SSC_N_MAX	3
+#define IRONLAKE_LVDS_D_SSC_M_MIN	79
+#define IRONLAKE_LVDS_D_SSC_M_MAX	126
+#define IRONLAKE_LVDS_D_SSC_P_MIN	14
+#define IRONLAKE_LVDS_D_SSC_P_MAX	42
+#define IRONLAKE_LVDS_D_SSC_P1_MIN	2
+#define IRONLAKE_LVDS_D_SSC_P1_MAX	6
+#define IRONLAKE_LVDS_D_SSC_P2_SLOW	7
+#define IRONLAKE_LVDS_D_SSC_P2_FAST	7
+
+/* DisplayPort */
+#define IRONLAKE_DP_N_MIN		1
+#define IRONLAKE_DP_N_MAX		2
+#define IRONLAKE_DP_M_MIN		81
+#define IRONLAKE_DP_M_MAX		90
+#define IRONLAKE_DP_P_MIN		10
+#define IRONLAKE_DP_P_MAX		20
+#define IRONLAKE_DP_P2_FAST		10
+#define IRONLAKE_DP_P2_SLOW		10
+#define IRONLAKE_DP_P2_LIMIT		0
+#define IRONLAKE_DP_P1_MIN		1
+#define IRONLAKE_DP_P1_MAX		2
 
 static bool
 intel_find_best_PLL(const intel_limit_t *limit, struct drm_crtc *crtc,
 		    int target, int refclk, intel_clock_t *best_clock);
 static bool
-intel_find_best_reduced_PLL(const intel_limit_t *limit, struct drm_crtc *crtc,
-			    int target, int refclk, intel_clock_t *best_clock);
-static bool
 intel_g4x_find_best_PLL(const intel_limit_t *limit, struct drm_crtc *crtc,
 			int target, int refclk, intel_clock_t *best_clock);
-static bool
-intel_igdng_find_best_PLL(const intel_limit_t *limit, struct drm_crtc *crtc,
-			int target, int refclk, intel_clock_t *best_clock);
 
 static bool
 intel_find_pll_g4x_dp(const intel_limit_t *, struct drm_crtc *crtc,
 		      int target, int refclk, intel_clock_t *best_clock);
 static bool
-intel_find_pll_igdng_dp(const intel_limit_t *, struct drm_crtc *crtc,
-		      int target, int refclk, intel_clock_t *best_clock);
+intel_find_pll_ironlake_dp(const intel_limit_t *, struct drm_crtc *crtc,
+			   int target, int refclk, intel_clock_t *best_clock);
 
 static const intel_limit_t intel_limits_i8xx_dvo = {
         .dot = { .min = I8XX_DOT_MIN,		.max = I8XX_DOT_MAX },
@@ -294,7 +347,6 @@ static const intel_limit_t intel_limits_i8xx_dvo = {
 	.p2  = { .dot_limit = I8XX_P2_SLOW_LIMIT,
 		 .p2_slow = I8XX_P2_SLOW,	.p2_fast = I8XX_P2_FAST },
 	.find_pll = intel_find_best_PLL,
-	.find_reduced_pll = intel_find_best_reduced_PLL,
 };
 
 static const intel_limit_t intel_limits_i8xx_lvds = {
@@ -309,7 +361,6 @@ static const intel_limit_t intel_limits_i8xx_lvds = {
 	.p2  = { .dot_limit = I8XX_P2_SLOW_LIMIT,
 		 .p2_slow = I8XX_P2_LVDS_SLOW,	.p2_fast = I8XX_P2_LVDS_FAST },
 	.find_pll = intel_find_best_PLL,
-	.find_reduced_pll = intel_find_best_reduced_PLL,
 };
 	
 static const intel_limit_t intel_limits_i9xx_sdvo = {
@@ -324,7 +375,6 @@ static const intel_limit_t intel_limits_i9xx_sdvo = {
 	.p2  = { .dot_limit = I9XX_P2_SDVO_DAC_SLOW_LIMIT,
 		 .p2_slow = I9XX_P2_SDVO_DAC_SLOW,	.p2_fast = I9XX_P2_SDVO_DAC_FAST },
 	.find_pll = intel_find_best_PLL,
-	.find_reduced_pll = intel_find_best_reduced_PLL,
 };
 
 static const intel_limit_t intel_limits_i9xx_lvds = {
@@ -342,7 +392,6 @@ static const intel_limit_t intel_limits_i9xx_lvds = {
 	.p2  = { .dot_limit = I9XX_P2_LVDS_SLOW_LIMIT,
 		 .p2_slow = I9XX_P2_LVDS_SLOW,	.p2_fast = I9XX_P2_LVDS_FAST },
 	.find_pll = intel_find_best_PLL,
-	.find_reduced_pll = intel_find_best_reduced_PLL,
 };
 
     /* below parameter and function is for G4X Chipset Family*/
@@ -360,7 +409,6 @@ static const intel_limit_t intel_limits_g4x_sdvo = {
 		 .p2_fast = G4X_P2_SDVO_FAST
 	},
 	.find_pll = intel_g4x_find_best_PLL,
-	.find_reduced_pll = intel_g4x_find_best_PLL,
 };
 
 static const intel_limit_t intel_limits_g4x_hdmi = {
@@ -377,7 +425,6 @@ static const intel_limit_t intel_limits_g4x_hdmi = {
 		 .p2_fast = G4X_P2_HDMI_DAC_FAST
 	},
 	.find_pll = intel_g4x_find_best_PLL,
-	.find_reduced_pll = intel_g4x_find_best_PLL,
 };
 
 static const intel_limit_t intel_limits_g4x_single_channel_lvds = {
@@ -402,7 +449,6 @@ static const intel_limit_t intel_limits_g4x_single_channel_lvds = {
 		 .p2_fast = G4X_P2_SINGLE_CHANNEL_LVDS_FAST
 	},
 	.find_pll = intel_g4x_find_best_PLL,
-	.find_reduced_pll = intel_g4x_find_best_PLL,
 };
 
 static const intel_limit_t intel_limits_g4x_dual_channel_lvds = {
@@ -427,7 +473,6 @@ static const intel_limit_t intel_limits_g4x_dual_channel_lvds = {
 		 .p2_fast = G4X_P2_DUAL_CHANNEL_LVDS_FAST
 	},
 	.find_pll = intel_g4x_find_best_PLL,
-	.find_reduced_pll = intel_g4x_find_best_PLL,
 };
 
 static const intel_limit_t intel_limits_g4x_display_port = {
@@ -453,74 +498,162 @@ static const intel_limit_t intel_limits_g4x_display_port = {
         .find_pll = intel_find_pll_g4x_dp,
 };
 
-static const intel_limit_t intel_limits_igd_sdvo = {
+static const intel_limit_t intel_limits_pineview_sdvo = {
         .dot = { .min = I9XX_DOT_MIN,		.max = I9XX_DOT_MAX},
-        .vco = { .min = IGD_VCO_MIN,		.max = IGD_VCO_MAX },
-        .n   = { .min = IGD_N_MIN,		.max = IGD_N_MAX },
-        .m   = { .min = IGD_M_MIN,		.max = IGD_M_MAX },
-        .m1  = { .min = IGD_M1_MIN,		.max = IGD_M1_MAX },
-        .m2  = { .min = IGD_M2_MIN,		.max = IGD_M2_MAX },
+        .vco = { .min = PINEVIEW_VCO_MIN,		.max = PINEVIEW_VCO_MAX },
+        .n   = { .min = PINEVIEW_N_MIN,		.max = PINEVIEW_N_MAX },
+        .m   = { .min = PINEVIEW_M_MIN,		.max = PINEVIEW_M_MAX },
+        .m1  = { .min = PINEVIEW_M1_MIN,		.max = PINEVIEW_M1_MAX },
+        .m2  = { .min = PINEVIEW_M2_MIN,		.max = PINEVIEW_M2_MAX },
         .p   = { .min = I9XX_P_SDVO_DAC_MIN,    .max = I9XX_P_SDVO_DAC_MAX },
         .p1  = { .min = I9XX_P1_MIN,		.max = I9XX_P1_MAX },
 	.p2  = { .dot_limit = I9XX_P2_SDVO_DAC_SLOW_LIMIT,
 		 .p2_slow = I9XX_P2_SDVO_DAC_SLOW,	.p2_fast = I9XX_P2_SDVO_DAC_FAST },
 	.find_pll = intel_find_best_PLL,
-	.find_reduced_pll = intel_find_best_reduced_PLL,
 };
 
-static const intel_limit_t intel_limits_igd_lvds = {
+static const intel_limit_t intel_limits_pineview_lvds = {
         .dot = { .min = I9XX_DOT_MIN,		.max = I9XX_DOT_MAX },
-        .vco = { .min = IGD_VCO_MIN,		.max = IGD_VCO_MAX },
-        .n   = { .min = IGD_N_MIN,		.max = IGD_N_MAX },
-        .m   = { .min = IGD_M_MIN,		.max = IGD_M_MAX },
-        .m1  = { .min = IGD_M1_MIN,		.max = IGD_M1_MAX },
-        .m2  = { .min = IGD_M2_MIN,		.max = IGD_M2_MAX },
-        .p   = { .min = IGD_P_LVDS_MIN,	.max = IGD_P_LVDS_MAX },
+        .vco = { .min = PINEVIEW_VCO_MIN,		.max = PINEVIEW_VCO_MAX },
+        .n   = { .min = PINEVIEW_N_MIN,		.max = PINEVIEW_N_MAX },
+        .m   = { .min = PINEVIEW_M_MIN,		.max = PINEVIEW_M_MAX },
+        .m1  = { .min = PINEVIEW_M1_MIN,		.max = PINEVIEW_M1_MAX },
+        .m2  = { .min = PINEVIEW_M2_MIN,		.max = PINEVIEW_M2_MAX },
+        .p   = { .min = PINEVIEW_P_LVDS_MIN,	.max = PINEVIEW_P_LVDS_MAX },
         .p1  = { .min = I9XX_P1_MIN,		.max = I9XX_P1_MAX },
-	/* IGD only supports single-channel mode. */
+	/* Pineview only supports single-channel mode. */
 	.p2  = { .dot_limit = I9XX_P2_LVDS_SLOW_LIMIT,
 		 .p2_slow = I9XX_P2_LVDS_SLOW,	.p2_fast = I9XX_P2_LVDS_SLOW },
 	.find_pll = intel_find_best_PLL,
-	.find_reduced_pll = intel_find_best_reduced_PLL,
 };
 
-static const intel_limit_t intel_limits_igdng_sdvo = {
-	.dot = { .min = IGDNG_DOT_MIN,          .max = IGDNG_DOT_MAX },
-	.vco = { .min = IGDNG_VCO_MIN,          .max = IGDNG_VCO_MAX },
-	.n   = { .min = IGDNG_N_MIN,            .max = IGDNG_N_MAX },
-	.m   = { .min = IGDNG_M_MIN,            .max = IGDNG_M_MAX },
-	.m1  = { .min = IGDNG_M1_MIN,           .max = IGDNG_M1_MAX },
-	.m2  = { .min = IGDNG_M2_MIN,           .max = IGDNG_M2_MAX },
-	.p   = { .min = IGDNG_P_SDVO_DAC_MIN,   .max = IGDNG_P_SDVO_DAC_MAX },
-	.p1  = { .min = IGDNG_P1_MIN,           .max = IGDNG_P1_MAX },
-	.p2  = { .dot_limit = IGDNG_P2_DOT_LIMIT,
-		 .p2_slow = IGDNG_P2_SDVO_DAC_SLOW,
-		 .p2_fast = IGDNG_P2_SDVO_DAC_FAST },
-	.find_pll = intel_igdng_find_best_PLL,
+static const intel_limit_t intel_limits_ironlake_dac = {
+	.dot = { .min = IRONLAKE_DOT_MIN,          .max = IRONLAKE_DOT_MAX },
+	.vco = { .min = IRONLAKE_VCO_MIN,          .max = IRONLAKE_VCO_MAX },
+	.n   = { .min = IRONLAKE_DAC_N_MIN,        .max = IRONLAKE_DAC_N_MAX },
+	.m   = { .min = IRONLAKE_DAC_M_MIN,        .max = IRONLAKE_DAC_M_MAX },
+	.m1  = { .min = IRONLAKE_M1_MIN,           .max = IRONLAKE_M1_MAX },
+	.m2  = { .min = IRONLAKE_M2_MIN,           .max = IRONLAKE_M2_MAX },
+	.p   = { .min = IRONLAKE_DAC_P_MIN,	   .max = IRONLAKE_DAC_P_MAX },
+	.p1  = { .min = IRONLAKE_DAC_P1_MIN,       .max = IRONLAKE_DAC_P1_MAX },
+	.p2  = { .dot_limit = IRONLAKE_P2_DOT_LIMIT,
+		 .p2_slow = IRONLAKE_DAC_P2_SLOW,
+		 .p2_fast = IRONLAKE_DAC_P2_FAST },
+	.find_pll = intel_g4x_find_best_PLL,
+};
+
+static const intel_limit_t intel_limits_ironlake_single_lvds = {
+	.dot = { .min = IRONLAKE_DOT_MIN,          .max = IRONLAKE_DOT_MAX },
+	.vco = { .min = IRONLAKE_VCO_MIN,          .max = IRONLAKE_VCO_MAX },
+	.n   = { .min = IRONLAKE_LVDS_S_N_MIN,     .max = IRONLAKE_LVDS_S_N_MAX },
+	.m   = { .min = IRONLAKE_LVDS_S_M_MIN,     .max = IRONLAKE_LVDS_S_M_MAX },
+	.m1  = { .min = IRONLAKE_M1_MIN,           .max = IRONLAKE_M1_MAX },
+	.m2  = { .min = IRONLAKE_M2_MIN,           .max = IRONLAKE_M2_MAX },
+	.p   = { .min = IRONLAKE_LVDS_S_P_MIN,     .max = IRONLAKE_LVDS_S_P_MAX },
+	.p1  = { .min = IRONLAKE_LVDS_S_P1_MIN,    .max = IRONLAKE_LVDS_S_P1_MAX },
+	.p2  = { .dot_limit = IRONLAKE_P2_DOT_LIMIT,
+		 .p2_slow = IRONLAKE_LVDS_S_P2_SLOW,
+		 .p2_fast = IRONLAKE_LVDS_S_P2_FAST },
+	.find_pll = intel_g4x_find_best_PLL,
+};
+
+static const intel_limit_t intel_limits_ironlake_dual_lvds = {
+	.dot = { .min = IRONLAKE_DOT_MIN,          .max = IRONLAKE_DOT_MAX },
+	.vco = { .min = IRONLAKE_VCO_MIN,          .max = IRONLAKE_VCO_MAX },
+	.n   = { .min = IRONLAKE_LVDS_D_N_MIN,     .max = IRONLAKE_LVDS_D_N_MAX },
+	.m   = { .min = IRONLAKE_LVDS_D_M_MIN,     .max = IRONLAKE_LVDS_D_M_MAX },
+	.m1  = { .min = IRONLAKE_M1_MIN,           .max = IRONLAKE_M1_MAX },
+	.m2  = { .min = IRONLAKE_M2_MIN,           .max = IRONLAKE_M2_MAX },
+	.p   = { .min = IRONLAKE_LVDS_D_P_MIN,     .max = IRONLAKE_LVDS_D_P_MAX },
+	.p1  = { .min = IRONLAKE_LVDS_D_P1_MIN,    .max = IRONLAKE_LVDS_D_P1_MAX },
+	.p2  = { .dot_limit = IRONLAKE_P2_DOT_LIMIT,
+		 .p2_slow = IRONLAKE_LVDS_D_P2_SLOW,
+		 .p2_fast = IRONLAKE_LVDS_D_P2_FAST },
+	.find_pll = intel_g4x_find_best_PLL,
 };
 
-static const intel_limit_t intel_limits_igdng_lvds = {
-	.dot = { .min = IGDNG_DOT_MIN,          .max = IGDNG_DOT_MAX },
-	.vco = { .min = IGDNG_VCO_MIN,          .max = IGDNG_VCO_MAX },
-	.n   = { .min = IGDNG_N_MIN,            .max = IGDNG_N_MAX },
-	.m   = { .min = IGDNG_M_MIN,            .max = IGDNG_M_MAX },
-	.m1  = { .min = IGDNG_M1_MIN,           .max = IGDNG_M1_MAX },
-	.m2  = { .min = IGDNG_M2_MIN,           .max = IGDNG_M2_MAX },
-	.p   = { .min = IGDNG_P_LVDS_MIN,       .max = IGDNG_P_LVDS_MAX },
-	.p1  = { .min = IGDNG_P1_MIN,           .max = IGDNG_P1_MAX },
-	.p2  = { .dot_limit = IGDNG_P2_DOT_LIMIT,
-		 .p2_slow = IGDNG_P2_LVDS_SLOW,
-		 .p2_fast = IGDNG_P2_LVDS_FAST },
-	.find_pll = intel_igdng_find_best_PLL,
+static const intel_limit_t intel_limits_ironlake_single_lvds_100m = {
+	.dot = { .min = IRONLAKE_DOT_MIN,          .max = IRONLAKE_DOT_MAX },
+	.vco = { .min = IRONLAKE_VCO_MIN,          .max = IRONLAKE_VCO_MAX },
+	.n   = { .min = IRONLAKE_LVDS_S_SSC_N_MIN, .max = IRONLAKE_LVDS_S_SSC_N_MAX },
+	.m   = { .min = IRONLAKE_LVDS_S_SSC_M_MIN, .max = IRONLAKE_LVDS_S_SSC_M_MAX },
+	.m1  = { .min = IRONLAKE_M1_MIN,           .max = IRONLAKE_M1_MAX },
+	.m2  = { .min = IRONLAKE_M2_MIN,           .max = IRONLAKE_M2_MAX },
+	.p   = { .min = IRONLAKE_LVDS_S_SSC_P_MIN, .max = IRONLAKE_LVDS_S_SSC_P_MAX },
+	.p1  = { .min = IRONLAKE_LVDS_S_SSC_P1_MIN,.max = IRONLAKE_LVDS_S_SSC_P1_MAX },
+	.p2  = { .dot_limit = IRONLAKE_P2_DOT_LIMIT,
+		 .p2_slow = IRONLAKE_LVDS_S_SSC_P2_SLOW,
+		 .p2_fast = IRONLAKE_LVDS_S_SSC_P2_FAST },
+	.find_pll = intel_g4x_find_best_PLL,
 };
 
-static const intel_limit_t *intel_igdng_limit(struct drm_crtc *crtc)
+static const intel_limit_t intel_limits_ironlake_dual_lvds_100m = {
+	.dot = { .min = IRONLAKE_DOT_MIN,          .max = IRONLAKE_DOT_MAX },
+	.vco = { .min = IRONLAKE_VCO_MIN,          .max = IRONLAKE_VCO_MAX },
+	.n   = { .min = IRONLAKE_LVDS_D_SSC_N_MIN, .max = IRONLAKE_LVDS_D_SSC_N_MAX },
+	.m   = { .min = IRONLAKE_LVDS_D_SSC_M_MIN, .max = IRONLAKE_LVDS_D_SSC_M_MAX },
+	.m1  = { .min = IRONLAKE_M1_MIN,           .max = IRONLAKE_M1_MAX },
+	.m2  = { .min = IRONLAKE_M2_MIN,           .max = IRONLAKE_M2_MAX },
+	.p   = { .min = IRONLAKE_LVDS_D_SSC_P_MIN, .max = IRONLAKE_LVDS_D_SSC_P_MAX },
+	.p1  = { .min = IRONLAKE_LVDS_D_SSC_P1_MIN,.max = IRONLAKE_LVDS_D_SSC_P1_MAX },
+	.p2  = { .dot_limit = IRONLAKE_P2_DOT_LIMIT,
+		 .p2_slow = IRONLAKE_LVDS_D_SSC_P2_SLOW,
+		 .p2_fast = IRONLAKE_LVDS_D_SSC_P2_FAST },
+	.find_pll = intel_g4x_find_best_PLL,
+};
+
+static const intel_limit_t intel_limits_ironlake_display_port = {
+        .dot = { .min = IRONLAKE_DOT_MIN,
+                 .max = IRONLAKE_DOT_MAX },
+        .vco = { .min = IRONLAKE_VCO_MIN,
+                 .max = IRONLAKE_VCO_MAX},
+        .n   = { .min = IRONLAKE_DP_N_MIN,
+                 .max = IRONLAKE_DP_N_MAX },
+        .m   = { .min = IRONLAKE_DP_M_MIN,
+                 .max = IRONLAKE_DP_M_MAX },
+        .m1  = { .min = IRONLAKE_M1_MIN,
+                 .max = IRONLAKE_M1_MAX },
+        .m2  = { .min = IRONLAKE_M2_MIN,
+                 .max = IRONLAKE_M2_MAX },
+        .p   = { .min = IRONLAKE_DP_P_MIN,
+                 .max = IRONLAKE_DP_P_MAX },
+        .p1  = { .min = IRONLAKE_DP_P1_MIN,
+                 .max = IRONLAKE_DP_P1_MAX},
+        .p2  = { .dot_limit = IRONLAKE_DP_P2_LIMIT,
+                 .p2_slow = IRONLAKE_DP_P2_SLOW,
+                 .p2_fast = IRONLAKE_DP_P2_FAST },
+        .find_pll = intel_find_pll_ironlake_dp,
+};
+
+static const intel_limit_t *intel_ironlake_limit(struct drm_crtc *crtc)
 {
+	struct drm_device *dev = crtc->dev;
+	struct drm_i915_private *dev_priv = dev->dev_private;
 	const intel_limit_t *limit;
-	if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS))
-		limit = &intel_limits_igdng_lvds;
+	int refclk = 120;
+
+	if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS)) {
+		if (dev_priv->lvds_use_ssc && dev_priv->lvds_ssc_freq == 100)
+			refclk = 100;
+
+		if ((I915_READ(PCH_LVDS) & LVDS_CLKB_POWER_MASK) ==
+		    LVDS_CLKB_POWER_UP) {
+			/* LVDS dual channel */
+			if (refclk == 100)
+				limit = &intel_limits_ironlake_dual_lvds_100m;
+			else
+				limit = &intel_limits_ironlake_dual_lvds;
+		} else {
+			if (refclk == 100)
+				limit = &intel_limits_ironlake_single_lvds_100m;
+			else
+				limit = &intel_limits_ironlake_single_lvds;
+		}
+	} else if (intel_pipe_has_type(crtc, INTEL_OUTPUT_DISPLAYPORT) ||
+			HAS_eDP)
+		limit = &intel_limits_ironlake_display_port;
 	else
-		limit = &intel_limits_igdng_sdvo;
+		limit = &intel_limits_ironlake_dac;
 
 	return limit;
 }
@@ -557,20 +690,20 @@ static const intel_limit_t *intel_limit(struct drm_crtc *crtc)
 	struct drm_device *dev = crtc->dev;
 	const intel_limit_t *limit;
 
-	if (IS_IGDNG(dev))
-		limit = intel_igdng_limit(crtc);
+	if (HAS_PCH_SPLIT(dev))
+		limit = intel_ironlake_limit(crtc);
 	else if (IS_G4X(dev)) {
 		limit = intel_g4x_limit(crtc);
-	} else if (IS_I9XX(dev) && !IS_IGD(dev)) {
+	} else if (IS_I9XX(dev) && !IS_PINEVIEW(dev)) {
 		if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS))
 			limit = &intel_limits_i9xx_lvds;
 		else
 			limit = &intel_limits_i9xx_sdvo;
-	} else if (IS_IGD(dev)) {
+	} else if (IS_PINEVIEW(dev)) {
 		if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS))
-			limit = &intel_limits_igd_lvds;
+			limit = &intel_limits_pineview_lvds;
 		else
-			limit = &intel_limits_igd_sdvo;
+			limit = &intel_limits_pineview_sdvo;
 	} else {
 		if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS))
 			limit = &intel_limits_i8xx_lvds;
@@ -580,8 +713,8 @@ static const intel_limit_t *intel_limit(struct drm_crtc *crtc)
 	return limit;
 }
 
-/* m1 is reserved as 0 in IGD, n is a ring counter */
-static void igd_clock(int refclk, intel_clock_t *clock)
+/* m1 is reserved as 0 in Pineview, n is a ring counter */
+static void pineview_clock(int refclk, intel_clock_t *clock)
 {
 	clock->m = clock->m2 + 2;
 	clock->p = clock->p1 * clock->p2;
@@ -591,8 +724,8 @@ static void igd_clock(int refclk, intel_clock_t *clock)
 
 static void intel_clock(struct drm_device *dev, int refclk, intel_clock_t *clock)
 {
-	if (IS_IGD(dev)) {
-		igd_clock(refclk, clock);
+	if (IS_PINEVIEW(dev)) {
+		pineview_clock(refclk, clock);
 		return;
 	}
 	clock->m = 5 * (clock->m1 + 2) + (clock->m2 + 2);
@@ -657,7 +790,7 @@ static bool intel_PLL_is_valid(struct drm_crtc *crtc, intel_clock_t *clock)
 		INTELPllInvalid ("m2 out of range\n");
 	if (clock->m1  < limit->m1.min  || limit->m1.max  < clock->m1)
 		INTELPllInvalid ("m1 out of range\n");
-	if (clock->m1 <= clock->m2 && !IS_IGD(dev))
+	if (clock->m1 <= clock->m2 && !IS_PINEVIEW(dev))
 		INTELPllInvalid ("m1 <= m2\n");
 	if (clock->m   < limit->m.min   || limit->m.max   < clock->m)
 		INTELPllInvalid ("m out of range\n");
@@ -706,16 +839,17 @@ intel_find_best_PLL(const intel_limit_t *limit, struct drm_crtc *crtc,
 
 	memset (best_clock, 0, sizeof (*best_clock));
 
-	for (clock.p1 = limit->p1.max; clock.p1 >= limit->p1.min; clock.p1--) {
-		for (clock.m1 = limit->m1.min; clock.m1 <= limit->m1.max;
-		     clock.m1++) {
-			for (clock.m2 = limit->m2.min;
-			     clock.m2 <= limit->m2.max; clock.m2++) {
-				/* m1 is always 0 in IGD */
-				if (clock.m2 >= clock.m1 && !IS_IGD(dev))
-					break;
-				for (clock.n = limit->n.min;
-				     clock.n <= limit->n.max; clock.n++) {
+	for (clock.m1 = limit->m1.min; clock.m1 <= limit->m1.max;
+	     clock.m1++) {
+		for (clock.m2 = limit->m2.min;
+		     clock.m2 <= limit->m2.max; clock.m2++) {
+			/* m1 is always 0 in Pineview */
+			if (clock.m2 >= clock.m1 && !IS_PINEVIEW(dev))
+				break;
+			for (clock.n = limit->n.min;
+			     clock.n <= limit->n.max; clock.n++) {
+				for (clock.p1 = limit->p1.min;
+					clock.p1 <= limit->p1.max; clock.p1++) {
 					int this_err;
 
 					intel_clock(dev, refclk, &clock);
@@ -736,46 +870,6 @@ intel_find_best_PLL(const intel_limit_t *limit, struct drm_crtc *crtc,
 	return (err != target);
 }
 
-
-static bool
-intel_find_best_reduced_PLL(const intel_limit_t *limit, struct drm_crtc *crtc,
-			    int target, int refclk, intel_clock_t *best_clock)
-
-{
-	struct drm_device *dev = crtc->dev;
-	intel_clock_t clock;
-	int err = target;
-	bool found = false;
-
-	memcpy(&clock, best_clock, sizeof(intel_clock_t));
-
-	for (clock.m1 = limit->m1.min; clock.m1 <= limit->m1.max; clock.m1++) {
-		for (clock.m2 = limit->m2.min; clock.m2 <= limit->m2.max; clock.m2++) {
-			/* m1 is always 0 in IGD */
-			if (clock.m2 >= clock.m1 && !IS_IGD(dev))
-				break;
-			for (clock.n = limit->n.min; clock.n <= limit->n.max;
-			     clock.n++) {
-				int this_err;
-
-				intel_clock(dev, refclk, &clock);
-
-				if (!intel_PLL_is_valid(crtc, &clock))
-					continue;
-
-				this_err = abs(clock.dot - target);
-				if (this_err < err) {
-					*best_clock = clock;
-					err = this_err;
-					found = true;
-				}
-			}
-		}
-	}
-
-	return found;
-}
-
 static bool
 intel_g4x_find_best_PLL(const intel_limit_t *limit, struct drm_crtc *crtc,
 			int target, int refclk, intel_clock_t *best_clock)
@@ -790,7 +884,13 @@ intel_g4x_find_best_PLL(const intel_limit_t *limit, struct drm_crtc *crtc,
 	found = false;
 
 	if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS)) {
-		if ((I915_READ(LVDS) & LVDS_CLKB_POWER_MASK) ==
+		int lvds_reg;
+
+		if (HAS_PCH_SPLIT(dev))
+			lvds_reg = PCH_LVDS;
+		else
+			lvds_reg = LVDS;
+		if ((I915_READ(lvds_reg) & LVDS_CLKB_POWER_MASK) ==
 		    LVDS_CLKB_POWER_UP)
 			clock.p2 = limit->p2.p2_fast;
 		else
@@ -833,11 +933,16 @@ intel_g4x_find_best_PLL(const intel_limit_t *limit, struct drm_crtc *crtc,
 }
 
 static bool
-intel_find_pll_igdng_dp(const intel_limit_t *limit, struct drm_crtc *crtc,
-		      int target, int refclk, intel_clock_t *best_clock)
+intel_find_pll_ironlake_dp(const intel_limit_t *limit, struct drm_crtc *crtc,
+			   int target, int refclk, intel_clock_t *best_clock)
 {
 	struct drm_device *dev = crtc->dev;
 	intel_clock_t clock;
+
+	/* return directly when it is eDP */
+	if (HAS_eDP)
+		return true;
+
 	if (target < 200000) {
 		clock.n = 1;
 		clock.p1 = 2;
@@ -856,71 +961,6 @@ intel_find_pll_igdng_dp(const intel_limit_t *limit, struct drm_crtc *crtc,
 	return true;
 }
 
-static bool
-intel_igdng_find_best_PLL(const intel_limit_t *limit, struct drm_crtc *crtc,
-			int target, int refclk, intel_clock_t *best_clock)
-{
-	struct drm_device *dev = crtc->dev;
-	struct drm_i915_private *dev_priv = dev->dev_private;
-	intel_clock_t clock;
-	int max_n;
-	bool found;
-	int err_most = 47;
-	found = false;
-
-	/* eDP has only 2 clock choice, no n/m/p setting */
-	if (HAS_eDP)
-		return true;
-
-	if (intel_pipe_has_type(crtc, INTEL_OUTPUT_DISPLAYPORT))
-		return intel_find_pll_igdng_dp(limit, crtc, target,
-					       refclk, best_clock);
-
-	if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS)) {
-		if ((I915_READ(PCH_LVDS) & LVDS_CLKB_POWER_MASK) ==
-		    LVDS_CLKB_POWER_UP)
-			clock.p2 = limit->p2.p2_fast;
-		else
-			clock.p2 = limit->p2.p2_slow;
-	} else {
-		if (target < limit->p2.dot_limit)
-			clock.p2 = limit->p2.p2_slow;
-		else
-			clock.p2 = limit->p2.p2_fast;
-	}
-
-	memset(best_clock, 0, sizeof(*best_clock));
-	max_n = limit->n.max;
-	for (clock.p1 = limit->p1.max; clock.p1 >= limit->p1.min; clock.p1--) {
-		/* based on hardware requriment prefer smaller n to precision */
-		for (clock.n = limit->n.min; clock.n <= max_n; clock.n++) {
-			/* based on hardware requirment prefere larger m1,m2 */
-			for (clock.m1 = limit->m1.max;
-			     clock.m1 >= limit->m1.min; clock.m1--) {
-				for (clock.m2 = limit->m2.max;
-				     clock.m2 >= limit->m2.min; clock.m2--) {
-					int this_err;
-
-					intel_clock(dev, refclk, &clock);
-					if (!intel_PLL_is_valid(crtc, &clock))
-						continue;
-					this_err = abs((10000 - (target*10000/clock.dot)));
-					if (this_err < err_most) {
-						*best_clock = clock;
-						err_most = this_err;
-						max_n = clock.n;
-						found = true;
-						/* found on first matching */
-						goto out;
-					}
-				}
-			}
-		}
-	}
-out:
-	return found;
-}
-
 /* DisplayPort has only two frequencies, 162MHz and 270MHz */
 static bool
 intel_find_pll_g4x_dp(const intel_limit_t *limit, struct drm_crtc *crtc,
@@ -943,6 +983,7 @@ intel_find_pll_g4x_dp(const intel_limit_t *limit, struct drm_crtc *crtc,
     clock.m = 5 * (clock.m1 + 2) + (clock.m2 + 2);
     clock.p = (clock.p1 * clock.p2);
     clock.dot = 96000 * clock.m / (clock.n + 2) / clock.p;
+    clock.vco = 0;
     memcpy(best_clock, &clock, sizeof(intel_clock_t));
     return true;
 }
@@ -951,7 +992,7 @@ void
 intel_wait_for_vblank(struct drm_device *dev)
 {
 	/* Wait for 20ms, i.e. one cycle at 50hz. */
-	mdelay(20);
+	msleep(20);
 }
 
 /* Parameters have changed, update FBC info */
@@ -990,13 +1031,15 @@ static void i8xx_enable_fbc(struct drm_crtc *crtc, unsigned long interval)
 
 	/* enable it... */
 	fbc_ctl = FBC_CTL_EN | FBC_CTL_PERIODIC;
+	if (IS_I945GM(dev))
+		fbc_ctl |= FBC_C3_IDLE; /* 945 needs special SR handling */
 	fbc_ctl |= (dev_priv->cfb_pitch & 0xff) << FBC_CTL_STRIDE_SHIFT;
 	fbc_ctl |= (interval & 0x2fff) << FBC_CTL_INTERVAL_SHIFT;
 	if (obj_priv->tiling_mode != I915_TILING_NONE)
 		fbc_ctl |= dev_priv->cfb_fence;
 	I915_WRITE(FBC_CONTROL, fbc_ctl);
 
-	DRM_DEBUG("enabled FBC, pitch %ld, yoff %d, plane %d, ",
+	DRM_DEBUG_KMS("enabled FBC, pitch %ld, yoff %d, plane %d, ",
 		  dev_priv->cfb_pitch, crtc->y, dev_priv->cfb_plane);
 }
 
@@ -1019,7 +1062,7 @@ void i8xx_disable_fbc(struct drm_device *dev)
 
 	intel_wait_for_vblank(dev);
 
-	DRM_DEBUG("disabled FBC\n");
+	DRM_DEBUG_KMS("disabled FBC\n");
 }
 
 static bool i8xx_fbc_enabled(struct drm_crtc *crtc)
@@ -1064,7 +1107,7 @@ static void g4x_enable_fbc(struct drm_crtc *crtc, unsigned long interval)
 	/* enable it... */
 	I915_WRITE(DPFC_CONTROL, I915_READ(DPFC_CONTROL) | DPFC_CTL_EN);
 
-	DRM_DEBUG("enabled fbc on plane %d\n", intel_crtc->plane);
+	DRM_DEBUG_KMS("enabled fbc on plane %d\n", intel_crtc->plane);
 }
 
 void g4x_disable_fbc(struct drm_device *dev)
@@ -1078,7 +1121,7 @@ void g4x_disable_fbc(struct drm_device *dev)
 	I915_WRITE(DPFC_CONTROL, dpfc_ctl);
 	intel_wait_for_vblank(dev);
 
-	DRM_DEBUG("disabled FBC\n");
+	DRM_DEBUG_KMS("disabled FBC\n");
 }
 
 static bool g4x_fbc_enabled(struct drm_crtc *crtc)
@@ -1143,25 +1186,32 @@ static void intel_update_fbc(struct drm_crtc *crtc,
 	 *   - going to an unsupported config (interlace, pixel multiply, etc.)
 	 */
 	if (intel_fb->obj->size > dev_priv->cfb_size) {
-		DRM_DEBUG("framebuffer too large, disabling compression\n");
+		DRM_DEBUG_KMS("framebuffer too large, disabling "
+				"compression\n");
+		dev_priv->no_fbc_reason = FBC_STOLEN_TOO_SMALL;
 		goto out_disable;
 	}
 	if ((mode->flags & DRM_MODE_FLAG_INTERLACE) ||
 	    (mode->flags & DRM_MODE_FLAG_DBLSCAN)) {
-		DRM_DEBUG("mode incompatible with compression, disabling\n");
+		DRM_DEBUG_KMS("mode incompatible with compression, "
+				"disabling\n");
+		dev_priv->no_fbc_reason = FBC_UNSUPPORTED_MODE;
 		goto out_disable;
 	}
 	if ((mode->hdisplay > 2048) ||
 	    (mode->vdisplay > 1536)) {
-		DRM_DEBUG("mode too large for compression, disabling\n");
+		DRM_DEBUG_KMS("mode too large for compression, disabling\n");
+		dev_priv->no_fbc_reason = FBC_MODE_TOO_LARGE;
 		goto out_disable;
 	}
 	if ((IS_I915GM(dev) || IS_I945GM(dev)) && plane != 0) {
-		DRM_DEBUG("plane not 0, disabling compression\n");
+		DRM_DEBUG_KMS("plane not 0, disabling compression\n");
+		dev_priv->no_fbc_reason = FBC_BAD_PLANE;
 		goto out_disable;
 	}
 	if (obj_priv->tiling_mode != I915_TILING_X) {
-		DRM_DEBUG("framebuffer not tiled, disabling compression\n");
+		DRM_DEBUG_KMS("framebuffer not tiled, disabling compression\n");
+		dev_priv->no_fbc_reason = FBC_NOT_TILED;
 		goto out_disable;
 	}
 
@@ -1183,13 +1233,57 @@ static void intel_update_fbc(struct drm_crtc *crtc,
 	return;
 
 out_disable:
-	DRM_DEBUG("unsupported config, disabling FBC\n");
+	DRM_DEBUG_KMS("unsupported config, disabling FBC\n");
 	/* Multiple disables should be harmless */
 	if (dev_priv->display.fbc_enabled(crtc))
 		dev_priv->display.disable_fbc(dev);
 }
 
 static int
+intel_pin_and_fence_fb_obj(struct drm_device *dev, struct drm_gem_object *obj)
+{
+	struct drm_i915_gem_object *obj_priv = obj->driver_private;
+	u32 alignment;
+	int ret;
+
+	switch (obj_priv->tiling_mode) {
+	case I915_TILING_NONE:
+		alignment = 64 * 1024;
+		break;
+	case I915_TILING_X:
+		/* pin() will align the object as required by fence */
+		alignment = 0;
+		break;
+	case I915_TILING_Y:
+		/* FIXME: Is this true? */
+		DRM_ERROR("Y tiled not allowed for scan out buffers\n");
+		return -EINVAL;
+	default:
+		BUG();
+	}
+
+	ret = i915_gem_object_pin(obj, alignment);
+	if (ret != 0)
+		return ret;
+
+	/* Install a fence for tiled scan-out. Pre-i965 always needs a
+	 * fence, whereas 965+ only requires a fence if using
+	 * framebuffer compression.  For simplicity, we always install
+	 * a fence as the cost is not that onerous.
+	 */
+	if (obj_priv->fence_reg == I915_FENCE_REG_NONE &&
+	    obj_priv->tiling_mode != I915_TILING_NONE) {
+		ret = i915_gem_object_get_fence_reg(obj);
+		if (ret != 0) {
+			i915_gem_object_unpin(obj);
+			return ret;
+		}
+	}
+
+	return 0;
+}
+
+static int
 intel_pipe_set_base(struct drm_crtc *crtc, int x, int y,
 		    struct drm_framebuffer *old_fb)
 {
@@ -1208,12 +1302,12 @@ intel_pipe_set_base(struct drm_crtc *crtc, int x, int y,
 	int dspstride = (plane == 0) ? DSPASTRIDE : DSPBSTRIDE;
 	int dsptileoff = (plane == 0 ? DSPATILEOFF : DSPBTILEOFF);
 	int dspcntr_reg = (plane == 0) ? DSPACNTR : DSPBCNTR;
-	u32 dspcntr, alignment;
+	u32 dspcntr;
 	int ret;
 
 	/* no fb bound */
 	if (!crtc->fb) {
-		DRM_DEBUG("No FB bound\n");
+		DRM_DEBUG_KMS("No FB bound\n");
 		return 0;
 	}
 
@@ -1230,48 +1324,20 @@ intel_pipe_set_base(struct drm_crtc *crtc, int x, int y,
 	obj = intel_fb->obj;
 	obj_priv = obj->driver_private;
 
-	switch (obj_priv->tiling_mode) {
-	case I915_TILING_NONE:
-		alignment = 64 * 1024;
-		break;
-	case I915_TILING_X:
-		/* pin() will align the object as required by fence */
-		alignment = 0;
-		break;
-	case I915_TILING_Y:
-		/* FIXME: Is this true? */
-		DRM_ERROR("Y tiled not allowed for scan out buffers\n");
-		return -EINVAL;
-	default:
-		BUG();
-	}
-
 	mutex_lock(&dev->struct_mutex);
-	ret = i915_gem_object_pin(obj, alignment);
+	ret = intel_pin_and_fence_fb_obj(dev, obj);
 	if (ret != 0) {
 		mutex_unlock(&dev->struct_mutex);
 		return ret;
 	}
 
-	ret = i915_gem_object_set_to_gtt_domain(obj, 1);
+	ret = i915_gem_object_set_to_display_plane(obj);
 	if (ret != 0) {
 		i915_gem_object_unpin(obj);
 		mutex_unlock(&dev->struct_mutex);
 		return ret;
 	}
 
-	/* Pre-i965 needs to install a fence for tiled scan-out */
-	if (!IS_I965G(dev) &&
-	    obj_priv->fence_reg == I915_FENCE_REG_NONE &&
-	    obj_priv->tiling_mode != I915_TILING_NONE) {
-		ret = i915_gem_object_get_fence_reg(obj);
-		if (ret != 0) {
-			i915_gem_object_unpin(obj);
-			mutex_unlock(&dev->struct_mutex);
-			return ret;
-		}
-	}
-
 	dspcntr = I915_READ(dspcntr_reg);
 	/* Mask out pixel format bits in case we change it */
 	dspcntr &= ~DISPPLANE_PIXFORMAT_MASK;
@@ -1287,7 +1353,10 @@ intel_pipe_set_base(struct drm_crtc *crtc, int x, int y,
 		break;
 	case 24:
 	case 32:
-		dspcntr |= DISPPLANE_32BPP_NO_ALPHA;
+		if (crtc->fb->depth == 30)
+			dspcntr |= DISPPLANE_32BPP_30BIT_NO_ALPHA;
+		else
+			dspcntr |= DISPPLANE_32BPP_NO_ALPHA;
 		break;
 	default:
 		DRM_ERROR("Unknown color depth\n");
@@ -1302,7 +1371,7 @@ intel_pipe_set_base(struct drm_crtc *crtc, int x, int y,
 			dspcntr &= ~DISPPLANE_TILED;
 	}
 
-	if (IS_IGDNG(dev))
+	if (HAS_PCH_SPLIT(dev))
 		/* must disable */
 		dspcntr |= DISPPLANE_TRICKLE_FEED_DISABLE;
 
@@ -1311,7 +1380,7 @@ intel_pipe_set_base(struct drm_crtc *crtc, int x, int y,
 	Start = obj_priv->gtt_offset;
 	Offset = y * crtc->fb->pitch + x * (crtc->fb->bits_per_pixel / 8);
 
-	DRM_DEBUG("Writing base %08lX %08lX %d %d\n", Start, Offset, x, y);
+	DRM_DEBUG_KMS("Writing base %08lX %08lX %d %d\n", Start, Offset, x, y);
 	I915_WRITE(dspstride, crtc->fb->pitch);
 	if (IS_I965G(dev)) {
 		I915_WRITE(dspbase, Offset);
@@ -1363,7 +1432,7 @@ static void i915_disable_vga (struct drm_device *dev)
 	u8 sr1;
 	u32 vga_reg;
 
-	if (IS_IGDNG(dev))
+	if (HAS_PCH_SPLIT(dev))
 		vga_reg = CPU_VGACNTRL;
 	else
 		vga_reg = VGACNTRL;
@@ -1379,19 +1448,19 @@ static void i915_disable_vga (struct drm_device *dev)
 	I915_WRITE(vga_reg, VGA_DISP_DISABLE);
 }
 
-static void igdng_disable_pll_edp (struct drm_crtc *crtc)
+static void ironlake_disable_pll_edp (struct drm_crtc *crtc)
 {
 	struct drm_device *dev = crtc->dev;
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	u32 dpa_ctl;
 
-	DRM_DEBUG("\n");
+	DRM_DEBUG_KMS("\n");
 	dpa_ctl = I915_READ(DP_A);
 	dpa_ctl &= ~DP_PLL_ENABLE;
 	I915_WRITE(DP_A, dpa_ctl);
 }
 
-static void igdng_enable_pll_edp (struct drm_crtc *crtc)
+static void ironlake_enable_pll_edp (struct drm_crtc *crtc)
 {
 	struct drm_device *dev = crtc->dev;
 	struct drm_i915_private *dev_priv = dev->dev_private;
@@ -1404,13 +1473,13 @@ static void igdng_enable_pll_edp (struct drm_crtc *crtc)
 }
 
 
-static void igdng_set_pll_edp (struct drm_crtc *crtc, int clock)
+static void ironlake_set_pll_edp (struct drm_crtc *crtc, int clock)
 {
 	struct drm_device *dev = crtc->dev;
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	u32 dpa_ctl;
 
-	DRM_DEBUG("eDP PLL enable for clock %d\n", clock);
+	DRM_DEBUG_KMS("eDP PLL enable for clock %d\n", clock);
 	dpa_ctl = I915_READ(DP_A);
 	dpa_ctl &= ~DP_PLL_FREQ_MASK;
 
@@ -1440,7 +1509,7 @@ static void igdng_set_pll_edp (struct drm_crtc *crtc, int clock)
 	udelay(500);
 }
 
-static void igdng_crtc_dpms(struct drm_crtc *crtc, int mode)
+static void ironlake_crtc_dpms(struct drm_crtc *crtc, int mode)
 {
 	struct drm_device *dev = crtc->dev;
 	struct drm_i915_private *dev_priv = dev->dev_private;
@@ -1473,6 +1542,10 @@ static void igdng_crtc_dpms(struct drm_crtc *crtc, int mode)
 	int trans_vsync_reg = (pipe == 0) ? TRANS_VSYNC_A : TRANS_VSYNC_B;
 	u32 temp;
 	int tries = 5, j, n;
+	u32 pipe_bpc;
+
+	temp = I915_READ(pipeconf_reg);
+	pipe_bpc = temp & PIPE_BPC_MASK;
 
 	/* XXX: When our outputs are all unaware of DPMS modes other than off
 	 * and on, we should map those modes to DRM_MODE_DPMS_OFF in the CRTC.
@@ -1481,10 +1554,19 @@ static void igdng_crtc_dpms(struct drm_crtc *crtc, int mode)
 	case DRM_MODE_DPMS_ON:
 	case DRM_MODE_DPMS_STANDBY:
 	case DRM_MODE_DPMS_SUSPEND:
-		DRM_DEBUG("crtc %d dpms on\n", pipe);
+		DRM_DEBUG_KMS("crtc %d dpms on\n", pipe);
+
+		if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS)) {
+			temp = I915_READ(PCH_LVDS);
+			if ((temp & LVDS_PORT_EN) == 0) {
+				I915_WRITE(PCH_LVDS, temp | LVDS_PORT_EN);
+				POSTING_READ(PCH_LVDS);
+			}
+		}
+
 		if (HAS_eDP) {
 			/* enable eDP PLL */
-			igdng_enable_pll_edp(crtc);
+			ironlake_enable_pll_edp(crtc);
 		} else {
 			/* enable PCH DPLL */
 			temp = I915_READ(pch_dpll_reg);
@@ -1495,13 +1577,19 @@ static void igdng_crtc_dpms(struct drm_crtc *crtc, int mode)
 
 			/* enable PCH FDI RX PLL, wait warmup plus DMI latency */
 			temp = I915_READ(fdi_rx_reg);
+			/*
+			 * make the BPC in FDI Rx be consistent with that in
+			 * pipeconf reg.
+			 */
+			temp &= ~(0x7 << 16);
+			temp |= (pipe_bpc << 11);
 			I915_WRITE(fdi_rx_reg, temp | FDI_RX_PLL_ENABLE |
 					FDI_SEL_PCDCLK |
 					FDI_DP_PORT_WIDTH_X4); /* default 4 lanes */
 			I915_READ(fdi_rx_reg);
 			udelay(200);
 
-			/* Enable CPU FDI TX PLL, always on for IGDNG */
+			/* Enable CPU FDI TX PLL, always on for Ironlake */
 			temp = I915_READ(fdi_tx_reg);
 			if ((temp & FDI_TX_PLL_ENABLE) == 0) {
 				I915_WRITE(fdi_tx_reg, temp | FDI_TX_PLL_ENABLE);
@@ -1513,7 +1601,7 @@ static void igdng_crtc_dpms(struct drm_crtc *crtc, int mode)
 		/* Enable panel fitting for LVDS */
 		if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS)) {
 			temp = I915_READ(pf_ctl_reg);
-			I915_WRITE(pf_ctl_reg, temp | PF_ENABLE);
+			I915_WRITE(pf_ctl_reg, temp | PF_ENABLE | PF_FILTER_MED_3x3);
 
 			/* currently full aspect */
 			I915_WRITE(pf_win_pos, 0);
@@ -1568,12 +1656,13 @@ static void igdng_crtc_dpms(struct drm_crtc *crtc, int mode)
 			udelay(150);
 
 			temp = I915_READ(fdi_rx_iir_reg);
-			DRM_DEBUG("FDI_RX_IIR 0x%x\n", temp);
+			DRM_DEBUG_KMS("FDI_RX_IIR 0x%x\n", temp);
 
 			if ((temp & FDI_RX_BIT_LOCK) == 0) {
 				for (j = 0; j < tries; j++) {
 					temp = I915_READ(fdi_rx_iir_reg);
-					DRM_DEBUG("FDI_RX_IIR 0x%x\n", temp);
+					DRM_DEBUG_KMS("FDI_RX_IIR 0x%x\n",
+								temp);
 					if (temp & FDI_RX_BIT_LOCK)
 						break;
 					udelay(200);
@@ -1582,11 +1671,11 @@ static void igdng_crtc_dpms(struct drm_crtc *crtc, int mode)
 					I915_WRITE(fdi_rx_iir_reg,
 							temp | FDI_RX_BIT_LOCK);
 				else
-					DRM_DEBUG("train 1 fail\n");
+					DRM_DEBUG_KMS("train 1 fail\n");
 			} else {
 				I915_WRITE(fdi_rx_iir_reg,
 						temp | FDI_RX_BIT_LOCK);
-				DRM_DEBUG("train 1 ok 2!\n");
+				DRM_DEBUG_KMS("train 1 ok 2!\n");
 			}
 			temp = I915_READ(fdi_tx_reg);
 			temp &= ~FDI_LINK_TRAIN_NONE;
@@ -1601,12 +1690,13 @@ static void igdng_crtc_dpms(struct drm_crtc *crtc, int mode)
 			udelay(150);
 
 			temp = I915_READ(fdi_rx_iir_reg);
-			DRM_DEBUG("FDI_RX_IIR 0x%x\n", temp);
+			DRM_DEBUG_KMS("FDI_RX_IIR 0x%x\n", temp);
 
 			if ((temp & FDI_RX_SYMBOL_LOCK) == 0) {
 				for (j = 0; j < tries; j++) {
 					temp = I915_READ(fdi_rx_iir_reg);
-					DRM_DEBUG("FDI_RX_IIR 0x%x\n", temp);
+					DRM_DEBUG_KMS("FDI_RX_IIR 0x%x\n",
+								temp);
 					if (temp & FDI_RX_SYMBOL_LOCK)
 						break;
 					udelay(200);
@@ -1614,15 +1704,15 @@ static void igdng_crtc_dpms(struct drm_crtc *crtc, int mode)
 				if (j != tries) {
 					I915_WRITE(fdi_rx_iir_reg,
 							temp | FDI_RX_SYMBOL_LOCK);
-					DRM_DEBUG("train 2 ok 1!\n");
+					DRM_DEBUG_KMS("train 2 ok 1!\n");
 				} else
-					DRM_DEBUG("train 2 fail\n");
+					DRM_DEBUG_KMS("train 2 fail\n");
 			} else {
 				I915_WRITE(fdi_rx_iir_reg,
 						temp | FDI_RX_SYMBOL_LOCK);
-				DRM_DEBUG("train 2 ok 2!\n");
+				DRM_DEBUG_KMS("train 2 ok 2!\n");
 			}
-			DRM_DEBUG("train done\n");
+			DRM_DEBUG_KMS("train done\n");
 
 			/* set transcoder timing */
 			I915_WRITE(trans_htot_reg, I915_READ(cpu_htot_reg));
@@ -1635,6 +1725,12 @@ static void igdng_crtc_dpms(struct drm_crtc *crtc, int mode)
 
 			/* enable PCH transcoder */
 			temp = I915_READ(transconf_reg);
+			/*
+			 * make the BPC in transcoder be consistent with
+			 * that in pipeconf reg.
+			 */
+			temp &= ~PIPE_BPC_MASK;
+			temp |= pipe_bpc;
 			I915_WRITE(transconf_reg, temp | TRANS_ENABLE);
 			I915_READ(transconf_reg);
 
@@ -1664,10 +1760,9 @@ static void igdng_crtc_dpms(struct drm_crtc *crtc, int mode)
 
 	break;
 	case DRM_MODE_DPMS_OFF:
-		DRM_DEBUG("crtc %d dpms off\n", pipe);
-
-		i915_disable_vga(dev);
+		DRM_DEBUG_KMS("crtc %d dpms off\n", pipe);
 
+		drm_vblank_off(dev, pipe);
 		/* Disable display plane */
 		temp = I915_READ(dspcntr_reg);
 		if ((temp & DISPLAY_PLANE_ENABLE) != 0) {
@@ -1677,6 +1772,8 @@ static void igdng_crtc_dpms(struct drm_crtc *crtc, int mode)
 			I915_READ(dspbase_reg);
 		}
 
+		i915_disable_vga(dev);
+
 		/* disable cpu pipe, disable after all planes disabled */
 		temp = I915_READ(pipeconf_reg);
 		if ((temp & PIPEACONF_ENABLE) != 0) {
@@ -1690,16 +1787,23 @@ static void igdng_crtc_dpms(struct drm_crtc *crtc, int mode)
 					udelay(500);
 					continue;
 				} else {
-					DRM_DEBUG("pipe %d off delay\n", pipe);
+					DRM_DEBUG_KMS("pipe %d off delay\n",
+								pipe);
 					break;
 				}
 			}
 		} else
-			DRM_DEBUG("crtc %d is disabled\n", pipe);
+			DRM_DEBUG_KMS("crtc %d is disabled\n", pipe);
 
-		if (HAS_eDP) {
-			igdng_disable_pll_edp(crtc);
+		udelay(100);
+
+		/* Disable PF */
+		temp = I915_READ(pf_ctl_reg);
+		if ((temp & PF_ENABLE) != 0) {
+			I915_WRITE(pf_ctl_reg, temp & ~PF_ENABLE);
+			I915_READ(pf_ctl_reg);
 		}
+		I915_WRITE(pf_win_size, 0);
 
 		/* disable CPU FDI tx and PCH FDI rx */
 		temp = I915_READ(fdi_tx_reg);
@@ -1707,6 +1811,9 @@ static void igdng_crtc_dpms(struct drm_crtc *crtc, int mode)
 		I915_READ(fdi_tx_reg);
 
 		temp = I915_READ(fdi_rx_reg);
+		/* BPC in FDI rx is consistent with that in pipeconf */
+		temp &= ~(0x07 << 16);
+		temp |= (pipe_bpc << 11);
 		I915_WRITE(fdi_rx_reg, temp & ~FDI_RX_ENABLE);
 		I915_READ(fdi_rx_reg);
 
@@ -1725,6 +1832,13 @@ static void igdng_crtc_dpms(struct drm_crtc *crtc, int mode)
 
 		udelay(100);
 
+		if (intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS)) {
+			temp = I915_READ(PCH_LVDS);
+			I915_WRITE(PCH_LVDS, temp & ~LVDS_PORT_EN);
+			I915_READ(PCH_LVDS);
+			udelay(100);
+		}
+
 		/* disable PCH transcoder */
 		temp = I915_READ(transconf_reg);
 		if ((temp & TRANS_ENABLE) != 0) {
@@ -1738,11 +1852,19 @@ static void igdng_crtc_dpms(struct drm_crtc *crtc, int mode)
 					udelay(500);
 					continue;
 				} else {
-					DRM_DEBUG("transcoder %d off delay\n", pipe);
+					DRM_DEBUG_KMS("transcoder %d off "
+							"delay\n", pipe);
 					break;
 				}
 			}
 		}
+		temp = I915_READ(transconf_reg);
+		/* BPC in transcoder is consistent with that in pipeconf */
+		temp &= ~PIPE_BPC_MASK;
+		temp |= pipe_bpc;
+		I915_WRITE(transconf_reg, temp);
+		I915_READ(transconf_reg);
+		udelay(100);
 
 		/* disable PCH DPLL */
 		temp = I915_READ(pch_dpll_reg);
@@ -1751,14 +1873,20 @@ static void igdng_crtc_dpms(struct drm_crtc *crtc, int mode)
 			I915_READ(pch_dpll_reg);
 		}
 
-		temp = I915_READ(fdi_rx_reg);
-		if ((temp & FDI_RX_PLL_ENABLE) != 0) {
-			temp &= ~FDI_SEL_PCDCLK;
-			temp &= ~FDI_RX_PLL_ENABLE;
-			I915_WRITE(fdi_rx_reg, temp);
-			I915_READ(fdi_rx_reg);
+		if (HAS_eDP) {
+			ironlake_disable_pll_edp(crtc);
 		}
 
+		temp = I915_READ(fdi_rx_reg);
+		temp &= ~FDI_SEL_PCDCLK;
+		I915_WRITE(fdi_rx_reg, temp);
+		I915_READ(fdi_rx_reg);
+
+		temp = I915_READ(fdi_rx_reg);
+		temp &= ~FDI_RX_PLL_ENABLE;
+		I915_WRITE(fdi_rx_reg, temp);
+		I915_READ(fdi_rx_reg);
+
 		/* Disable CPU FDI TX PLL */
 		temp = I915_READ(fdi_tx_reg);
 		if ((temp & FDI_TX_PLL_ENABLE) != 0) {
@@ -1767,20 +1895,43 @@ static void igdng_crtc_dpms(struct drm_crtc *crtc, int mode)
 			udelay(100);
 		}
 
-		/* Disable PF */
-		temp = I915_READ(pf_ctl_reg);
-		if ((temp & PF_ENABLE) != 0) {
-			I915_WRITE(pf_ctl_reg, temp & ~PF_ENABLE);
-			I915_READ(pf_ctl_reg);
-		}
-		I915_WRITE(pf_win_size, 0);
-
 		/* Wait for the clocks to turn off. */
-		udelay(150);
+		udelay(100);
 		break;
 	}
 }
 
+static void intel_crtc_dpms_overlay(struct intel_crtc *intel_crtc, bool enable)
+{
+	struct intel_overlay *overlay;
+	int ret;
+
+	if (!enable && intel_crtc->overlay) {
+		overlay = intel_crtc->overlay;
+		mutex_lock(&overlay->dev->struct_mutex);
+		for (;;) {
+			ret = intel_overlay_switch_off(overlay);
+			if (ret == 0)
+				break;
+
+			ret = intel_overlay_recover_from_interrupt(overlay, 0);
+			if (ret != 0) {
+				/* overlay doesn't react anymore. Usually
+				 * results in a black screen and an unkillable
+				 * X server. */
+				BUG();
+				overlay->hw_wedged = HW_WEDGED;
+				break;
+			}
+		}
+		mutex_unlock(&overlay->dev->struct_mutex);
+	}
+	/* Let userspace switch the overlay on again. In most cases userspace
+	 * has to recompute where to put it anyway. */
+
+	return;
+}
+
 static void i9xx_crtc_dpms(struct drm_crtc *crtc, int mode)
 {
 	struct drm_device *dev = crtc->dev;
@@ -1801,6 +1952,8 @@ static void i9xx_crtc_dpms(struct drm_crtc *crtc, int mode)
 	case DRM_MODE_DPMS_ON:
 	case DRM_MODE_DPMS_STANDBY:
 	case DRM_MODE_DPMS_SUSPEND:
+		intel_update_watermarks(dev);
+
 		/* Enable the DPLL */
 		temp = I915_READ(dpll_reg);
 		if ((temp & DPLL_VCO_ENABLE) == 0) {
@@ -1837,13 +1990,14 @@ static void i9xx_crtc_dpms(struct drm_crtc *crtc, int mode)
 			intel_update_fbc(crtc, &crtc->mode);
 
 		/* Give the overlay scaler a chance to enable if it's on this pipe */
-		//intel_crtc_dpms_video(crtc, true); TODO
-		intel_update_watermarks(dev);
+		intel_crtc_dpms_overlay(intel_crtc, true);
 	break;
 	case DRM_MODE_DPMS_OFF:
 		intel_update_watermarks(dev);
+
 		/* Give the overlay scaler a chance to disable if it's on this pipe */
-		//intel_crtc_dpms_video(crtc, FALSE); TODO
+		intel_crtc_dpms_overlay(intel_crtc, false);
+		drm_vblank_off(dev, pipe);
 
 		if (dev_priv->cfb_plane == plane &&
 		    dev_priv->display.disable_fbc)
@@ -1962,7 +2116,7 @@ static bool intel_crtc_mode_fixup(struct drm_crtc *crtc,
 				  struct drm_display_mode *adjusted_mode)
 {
 	struct drm_device *dev = crtc->dev;
-	if (IS_IGDNG(dev)) {
+	if (HAS_PCH_SPLIT(dev)) {
 		/* FDI link clock is fixed at 2.7G */
 		if (mode->clock * 3 > 27000 * 4)
 			return MODE_CLOCK_HIGH;
@@ -2038,7 +2192,7 @@ static int i830_get_display_clock_speed(struct drm_device *dev)
  * Return the pipe currently connected to the panel fitter,
  * or -1 if the panel fitter is not present or not in use
  */
-static int intel_panel_fitter_pipe (struct drm_device *dev)
+int intel_panel_fitter_pipe (struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	u32  pfit_control;
@@ -2082,9 +2236,8 @@ fdi_reduce_ratio(u32 *num, u32 *den)
 #define LINK_N 0x80000
 
 static void
-igdng_compute_m_n(int bytes_per_pixel, int nlanes,
-		int pixel_clock, int link_clock,
-		struct fdi_m_n *m_n)
+ironlake_compute_m_n(int bits_per_pixel, int nlanes, int pixel_clock,
+		     int link_clock, struct fdi_m_n *m_n)
 {
 	u64 temp;
 
@@ -2092,7 +2245,8 @@ igdng_compute_m_n(int bytes_per_pixel, int nlanes,
 
 	temp = (u64) DATA_N * pixel_clock;
 	temp = div_u64(temp, link_clock);
-	m_n->gmch_m = div_u64(temp * bytes_per_pixel, nlanes);
+	m_n->gmch_m = div_u64(temp * bits_per_pixel, nlanes);
+	m_n->gmch_m >>= 3; /* convert to bytes_per_pixel */
 	m_n->gmch_n = DATA_N;
 	fdi_reduce_ratio(&m_n->gmch_m, &m_n->gmch_n);
 
@@ -2111,34 +2265,41 @@ struct intel_watermark_params {
 	unsigned long cacheline_size;
 };
 
-/* IGD has different values for various configs */
-static struct intel_watermark_params igd_display_wm = {
-	IGD_DISPLAY_FIFO,
-	IGD_MAX_WM,
-	IGD_DFT_WM,
-	IGD_GUARD_WM,
-	IGD_FIFO_LINE_SIZE
+/* Pineview has different values for various configs */
+static struct intel_watermark_params pineview_display_wm = {
+	PINEVIEW_DISPLAY_FIFO,
+	PINEVIEW_MAX_WM,
+	PINEVIEW_DFT_WM,
+	PINEVIEW_GUARD_WM,
+	PINEVIEW_FIFO_LINE_SIZE
 };
-static struct intel_watermark_params igd_display_hplloff_wm = {
-	IGD_DISPLAY_FIFO,
-	IGD_MAX_WM,
-	IGD_DFT_HPLLOFF_WM,
-	IGD_GUARD_WM,
-	IGD_FIFO_LINE_SIZE
+static struct intel_watermark_params pineview_display_hplloff_wm = {
+	PINEVIEW_DISPLAY_FIFO,
+	PINEVIEW_MAX_WM,
+	PINEVIEW_DFT_HPLLOFF_WM,
+	PINEVIEW_GUARD_WM,
+	PINEVIEW_FIFO_LINE_SIZE
 };
-static struct intel_watermark_params igd_cursor_wm = {
-	IGD_CURSOR_FIFO,
-	IGD_CURSOR_MAX_WM,
-	IGD_CURSOR_DFT_WM,
-	IGD_CURSOR_GUARD_WM,
-	IGD_FIFO_LINE_SIZE,
+static struct intel_watermark_params pineview_cursor_wm = {
+	PINEVIEW_CURSOR_FIFO,
+	PINEVIEW_CURSOR_MAX_WM,
+	PINEVIEW_CURSOR_DFT_WM,
+	PINEVIEW_CURSOR_GUARD_WM,
+	PINEVIEW_FIFO_LINE_SIZE,
 };
-static struct intel_watermark_params igd_cursor_hplloff_wm = {
-	IGD_CURSOR_FIFO,
-	IGD_CURSOR_MAX_WM,
-	IGD_CURSOR_DFT_WM,
-	IGD_CURSOR_GUARD_WM,
-	IGD_FIFO_LINE_SIZE
+static struct intel_watermark_params pineview_cursor_hplloff_wm = {
+	PINEVIEW_CURSOR_FIFO,
+	PINEVIEW_CURSOR_MAX_WM,
+	PINEVIEW_CURSOR_DFT_WM,
+	PINEVIEW_CURSOR_GUARD_WM,
+	PINEVIEW_FIFO_LINE_SIZE
+};
+static struct intel_watermark_params g4x_wm_info = {
+	G4X_FIFO_SIZE,
+	G4X_MAX_WM,
+	G4X_MAX_WM,
+	2,
+	G4X_FIFO_LINE_SIZE,
 };
 static struct intel_watermark_params i945_wm_info = {
 	I945_FIFO_SIZE,
@@ -2204,11 +2365,11 @@ static unsigned long intel_calculate_wm(unsigned long clock_in_khz,
 		1000;
 	entries_required /= wm->cacheline_size;
 
-	DRM_DEBUG("FIFO entries required for mode: %d\n", entries_required);
+	DRM_DEBUG_KMS("FIFO entries required for mode: %d\n", entries_required);
 
 	wm_size = wm->fifo_size - (entries_required + wm->guard_size);
 
-	DRM_DEBUG("FIFO watermark level: %d\n", wm_size);
+	DRM_DEBUG_KMS("FIFO watermark level: %d\n", wm_size);
 
 	/* Don't promote wm_size to unsigned... */
 	if (wm_size > (long)wm->max_wm)
@@ -2270,50 +2431,50 @@ static struct cxsr_latency *intel_get_cxsr_latency(int is_desktop, int fsb,
 			return latency;
 	}
 
-	DRM_DEBUG("Unknown FSB/MEM found, disable CxSR\n");
+	DRM_DEBUG_KMS("Unknown FSB/MEM found, disable CxSR\n");
 
 	return NULL;
 }
 
-static void igd_disable_cxsr(struct drm_device *dev)
+static void pineview_disable_cxsr(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	u32 reg;
 
 	/* deactivate cxsr */
 	reg = I915_READ(DSPFW3);
-	reg &= ~(IGD_SELF_REFRESH_EN);
+	reg &= ~(PINEVIEW_SELF_REFRESH_EN);
 	I915_WRITE(DSPFW3, reg);
 	DRM_INFO("Big FIFO is disabled\n");
 }
 
-static void igd_enable_cxsr(struct drm_device *dev, unsigned long clock,
-			    int pixel_size)
+static void pineview_enable_cxsr(struct drm_device *dev, unsigned long clock,
+				 int pixel_size)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	u32 reg;
 	unsigned long wm;
 	struct cxsr_latency *latency;
 
-	latency = intel_get_cxsr_latency(IS_IGDG(dev), dev_priv->fsb_freq,
+	latency = intel_get_cxsr_latency(IS_PINEVIEW_G(dev), dev_priv->fsb_freq,
 		dev_priv->mem_freq);
 	if (!latency) {
-		DRM_DEBUG("Unknown FSB/MEM found, disable CxSR\n");
-		igd_disable_cxsr(dev);
+		DRM_DEBUG_KMS("Unknown FSB/MEM found, disable CxSR\n");
+		pineview_disable_cxsr(dev);
 		return;
 	}
 
 	/* Display SR */
-	wm = intel_calculate_wm(clock, &igd_display_wm, pixel_size,
+	wm = intel_calculate_wm(clock, &pineview_display_wm, pixel_size,
 				latency->display_sr);
 	reg = I915_READ(DSPFW1);
 	reg &= 0x7fffff;
 	reg |= wm << 23;
 	I915_WRITE(DSPFW1, reg);
-	DRM_DEBUG("DSPFW1 register is %x\n", reg);
+	DRM_DEBUG_KMS("DSPFW1 register is %x\n", reg);
 
 	/* cursor SR */
-	wm = intel_calculate_wm(clock, &igd_cursor_wm, pixel_size,
+	wm = intel_calculate_wm(clock, &pineview_cursor_wm, pixel_size,
 				latency->cursor_sr);
 	reg = I915_READ(DSPFW3);
 	reg &= ~(0x3f << 24);
@@ -2321,7 +2482,7 @@ static void igd_enable_cxsr(struct drm_device *dev, unsigned long clock,
 	I915_WRITE(DSPFW3, reg);
 
 	/* Display HPLL off SR */
-	wm = intel_calculate_wm(clock, &igd_display_hplloff_wm,
+	wm = intel_calculate_wm(clock, &pineview_display_hplloff_wm,
 		latency->display_hpll_disable, I915_FIFO_LINE_SIZE);
 	reg = I915_READ(DSPFW3);
 	reg &= 0xfffffe00;
@@ -2329,17 +2490,17 @@ static void igd_enable_cxsr(struct drm_device *dev, unsigned long clock,
 	I915_WRITE(DSPFW3, reg);
 
 	/* cursor HPLL off SR */
-	wm = intel_calculate_wm(clock, &igd_cursor_hplloff_wm, pixel_size,
+	wm = intel_calculate_wm(clock, &pineview_cursor_hplloff_wm, pixel_size,
 				latency->cursor_hpll_disable);
 	reg = I915_READ(DSPFW3);
 	reg &= ~(0x3f << 16);
 	reg |= (wm & 0x3f) << 16;
 	I915_WRITE(DSPFW3, reg);
-	DRM_DEBUG("DSPFW3 register is %x\n", reg);
+	DRM_DEBUG_KMS("DSPFW3 register is %x\n", reg);
 
 	/* activate cxsr */
 	reg = I915_READ(DSPFW3);
-	reg |= IGD_SELF_REFRESH_EN;
+	reg |= PINEVIEW_SELF_REFRESH_EN;
 	I915_WRITE(DSPFW3, reg);
 
 	DRM_INFO("Big FIFO is enabled\n");
@@ -2361,7 +2522,7 @@ static void igd_enable_cxsr(struct drm_device *dev, unsigned long clock,
  * A value of 5us seems to be a good balance; safe for very low end
  * platforms but not overly aggressive on lower latency configs.
  */
-const static int latency_ns = 5000;
+static const int latency_ns = 5000;
 
 static int i9xx_get_fifo_size(struct drm_device *dev, int plane)
 {
@@ -2375,8 +2536,8 @@ static int i9xx_get_fifo_size(struct drm_device *dev, int plane)
 		size = ((dsparb >> DSPARB_CSTART_SHIFT) & 0x7f) -
 			(dsparb & 0x7f);
 
-	DRM_DEBUG("FIFO size - (0x%08x) %s: %d\n", dsparb, plane ? "B" : "A",
-		  size);
+	DRM_DEBUG_KMS("FIFO size - (0x%08x) %s: %d\n", dsparb,
+			plane ? "B" : "A", size);
 
 	return size;
 }
@@ -2394,8 +2555,8 @@ static int i85x_get_fifo_size(struct drm_device *dev, int plane)
 			(dsparb & 0x1ff);
 	size >>= 1; /* Convert to cachelines */
 
-	DRM_DEBUG("FIFO size - (0x%08x) %s: %d\n", dsparb, plane ? "B" : "A",
-		  size);
+	DRM_DEBUG_KMS("FIFO size - (0x%08x) %s: %d\n", dsparb,
+			plane ? "B" : "A", size);
 
 	return size;
 }
@@ -2409,7 +2570,8 @@ static int i845_get_fifo_size(struct drm_device *dev, int plane)
 	size = dsparb & 0x7f;
 	size >>= 2; /* Convert to cachelines */
 
-	DRM_DEBUG("FIFO size - (0x%08x) %s: %d\n", dsparb, plane ? "B" : "A",
+	DRM_DEBUG_KMS("FIFO size - (0x%08x) %s: %d\n", dsparb,
+			plane ? "B" : "A",
 		  size);
 
 	return size;
@@ -2424,34 +2586,123 @@ static int i830_get_fifo_size(struct drm_device *dev, int plane)
 	size = dsparb & 0x7f;
 	size >>= 1; /* Convert to cachelines */
 
-	DRM_DEBUG("FIFO size - (0x%08x) %s: %d\n", dsparb, plane ? "B" : "A",
-		  size);
+	DRM_DEBUG_KMS("FIFO size - (0x%08x) %s: %d\n", dsparb,
+			plane ? "B" : "A", size);
 
 	return size;
 }
 
-static void g4x_update_wm(struct drm_device *dev, int unused, int unused2,
-			  int unused3, int unused4)
+static void g4x_update_wm(struct drm_device *dev,  int planea_clock,
+			  int planeb_clock, int sr_hdisplay, int pixel_size)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
-	u32 fw_blc_self = I915_READ(FW_BLC_SELF);
+	int total_size, cacheline_size;
+	int planea_wm, planeb_wm, cursora_wm, cursorb_wm, cursor_sr;
+	struct intel_watermark_params planea_params, planeb_params;
+	unsigned long line_time_us;
+	int sr_clock, sr_entries = 0, entries_required;
 
-	if (i915_powersave)
-		fw_blc_self |= FW_BLC_SELF_EN;
-	else
-		fw_blc_self &= ~FW_BLC_SELF_EN;
-	I915_WRITE(FW_BLC_SELF, fw_blc_self);
+	/* Create copies of the base settings for each pipe */
+	planea_params = planeb_params = g4x_wm_info;
+
+	/* Grab a couple of global values before we overwrite them */
+	total_size = planea_params.fifo_size;
+	cacheline_size = planea_params.cacheline_size;
+
+	/*
+	 * Note: we need to make sure we don't overflow for various clock &
+	 * latency values.
+	 * clocks go from a few thousand to several hundred thousand.
+	 * latency is usually a few thousand
+	 */
+	entries_required = ((planea_clock / 1000) * pixel_size * latency_ns) /
+		1000;
+	entries_required /= G4X_FIFO_LINE_SIZE;
+	planea_wm = entries_required + planea_params.guard_size;
+
+	entries_required = ((planeb_clock / 1000) * pixel_size * latency_ns) /
+		1000;
+	entries_required /= G4X_FIFO_LINE_SIZE;
+	planeb_wm = entries_required + planeb_params.guard_size;
+
+	cursora_wm = cursorb_wm = 16;
+	cursor_sr = 32;
+
+	DRM_DEBUG("FIFO watermarks - A: %d, B: %d\n", planea_wm, planeb_wm);
+
+	/* Calc sr entries for one plane configs */
+	if (sr_hdisplay && (!planea_clock || !planeb_clock)) {
+		/* self-refresh has much higher latency */
+		static const int sr_latency_ns = 12000;
+
+		sr_clock = planea_clock ? planea_clock : planeb_clock;
+		line_time_us = ((sr_hdisplay * 1000) / sr_clock);
+
+		/* Use ns/us then divide to preserve precision */
+		sr_entries = (((sr_latency_ns / line_time_us) + 1) *
+			      pixel_size * sr_hdisplay) / 1000;
+		sr_entries = roundup(sr_entries / cacheline_size, 1);
+		DRM_DEBUG("self-refresh entries: %d\n", sr_entries);
+		I915_WRITE(FW_BLC_SELF, FW_BLC_SELF_EN);
+	} else {
+		/* Turn off self refresh if both pipes are enabled */
+		I915_WRITE(FW_BLC_SELF, I915_READ(FW_BLC_SELF)
+					& ~FW_BLC_SELF_EN);
+	}
+
+	DRM_DEBUG("Setting FIFO watermarks - A: %d, B: %d, SR %d\n",
+		  planea_wm, planeb_wm, sr_entries);
+
+	planea_wm &= 0x3f;
+	planeb_wm &= 0x3f;
+
+	I915_WRITE(DSPFW1, (sr_entries << DSPFW_SR_SHIFT) |
+		   (cursorb_wm << DSPFW_CURSORB_SHIFT) |
+		   (planeb_wm << DSPFW_PLANEB_SHIFT) | planea_wm);
+	I915_WRITE(DSPFW2, (I915_READ(DSPFW2) & DSPFW_CURSORA_MASK) |
+		   (cursora_wm << DSPFW_CURSORA_SHIFT));
+	/* HPLL off in SR has some issues on G4x... disable it */
+	I915_WRITE(DSPFW3, (I915_READ(DSPFW3) & ~DSPFW_HPLL_SR_EN) |
+		   (cursor_sr << DSPFW_CURSOR_SR_SHIFT));
 }
 
-static void i965_update_wm(struct drm_device *dev, int unused, int unused2,
-			   int unused3, int unused4)
+static void i965_update_wm(struct drm_device *dev, int planea_clock,
+			   int planeb_clock, int sr_hdisplay, int pixel_size)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
+	unsigned long line_time_us;
+	int sr_clock, sr_entries, srwm = 1;
+
+	/* Calc sr entries for one plane configs */
+	if (sr_hdisplay && (!planea_clock || !planeb_clock)) {
+		/* self-refresh has much higher latency */
+		static const int sr_latency_ns = 12000;
+
+		sr_clock = planea_clock ? planea_clock : planeb_clock;
+		line_time_us = ((sr_hdisplay * 1000) / sr_clock);
+
+		/* Use ns/us then divide to preserve precision */
+		sr_entries = (((sr_latency_ns / line_time_us) + 1) *
+			      pixel_size * sr_hdisplay) / 1000;
+		sr_entries = roundup(sr_entries / I915_FIFO_LINE_SIZE, 1);
+		DRM_DEBUG("self-refresh entries: %d\n", sr_entries);
+		srwm = I945_FIFO_SIZE - sr_entries;
+		if (srwm < 0)
+			srwm = 1;
+		srwm &= 0x3f;
+		I915_WRITE(FW_BLC_SELF, FW_BLC_SELF_EN);
+	} else {
+		/* Turn off self refresh if both pipes are enabled */
+		I915_WRITE(FW_BLC_SELF, I915_READ(FW_BLC_SELF)
+					& ~FW_BLC_SELF_EN);
+	}
 
-	DRM_DEBUG("Setting FIFO watermarks - A: 8, B: 8, C: 8, SR 8\n");
+	DRM_DEBUG_KMS("Setting FIFO watermarks - A: 8, B: 8, C: 8, SR %d\n",
+		      srwm);
 
 	/* 965 has limitations... */
-	I915_WRITE(DSPFW1, (8 << 16) | (8 << 8) | (8 << 0));
+	I915_WRITE(DSPFW1, (srwm << DSPFW_SR_SHIFT) | (8 << 16) | (8 << 8) |
+		   (8 << 0));
 	I915_WRITE(DSPFW2, (8 << 8) | (8 << 0));
 }
 
@@ -2487,7 +2738,7 @@ static void i9xx_update_wm(struct drm_device *dev, int planea_clock,
 				       pixel_size, latency_ns);
 	planeb_wm = intel_calculate_wm(planeb_clock, &planeb_params,
 				       pixel_size, latency_ns);
-	DRM_DEBUG("FIFO watermarks - A: %d, B: %d\n", planea_wm, planeb_wm);
+	DRM_DEBUG_KMS("FIFO watermarks - A: %d, B: %d\n", planea_wm, planeb_wm);
 
 	/*
 	 * Overlay gets an aggressive default since video jitter is bad.
@@ -2498,7 +2749,7 @@ static void i9xx_update_wm(struct drm_device *dev, int planea_clock,
 	if (HAS_FW_BLC(dev) && sr_hdisplay &&
 	    (!planea_clock || !planeb_clock)) {
 		/* self-refresh has much higher latency */
-		const static int sr_latency_ns = 6000;
+		static const int sr_latency_ns = 6000;
 
 		sr_clock = planea_clock ? planea_clock : planeb_clock;
 		line_time_us = ((sr_hdisplay * 1000) / sr_clock);
@@ -2507,14 +2758,29 @@ static void i9xx_update_wm(struct drm_device *dev, int planea_clock,
 		sr_entries = (((sr_latency_ns / line_time_us) + 1) *
 			      pixel_size * sr_hdisplay) / 1000;
 		sr_entries = roundup(sr_entries / cacheline_size, 1);
-		DRM_DEBUG("self-refresh entries: %d\n", sr_entries);
+		DRM_DEBUG_KMS("self-refresh entries: %d\n", sr_entries);
 		srwm = total_size - sr_entries;
 		if (srwm < 0)
 			srwm = 1;
-		I915_WRITE(FW_BLC_SELF, FW_BLC_SELF_EN | (srwm & 0x3f));
+
+		if (IS_I945G(dev) || IS_I945GM(dev))
+			I915_WRITE(FW_BLC_SELF, FW_BLC_SELF_FIFO_MASK | (srwm & 0xff));
+		else if (IS_I915GM(dev)) {
+			/* 915M has a smaller SRWM field */
+			I915_WRITE(FW_BLC_SELF, srwm & 0x3f);
+			I915_WRITE(INSTPM, I915_READ(INSTPM) | INSTPM_SELF_EN);
+		}
+	} else {
+		/* Turn off self refresh if both pipes are enabled */
+		if (IS_I945G(dev) || IS_I945GM(dev)) {
+			I915_WRITE(FW_BLC_SELF, I915_READ(FW_BLC_SELF)
+				   & ~FW_BLC_SELF_EN);
+		} else if (IS_I915GM(dev)) {
+			I915_WRITE(INSTPM, I915_READ(INSTPM) & ~INSTPM_SELF_EN);
+		}
 	}
 
-	DRM_DEBUG("Setting FIFO watermarks - A: %d, B: %d, C: %d, SR %d\n",
+	DRM_DEBUG_KMS("Setting FIFO watermarks - A: %d, B: %d, C: %d, SR %d\n",
 		  planea_wm, planeb_wm, cwm, srwm);
 
 	fwater_lo = ((planeb_wm & 0x3f) << 16) | (planea_wm & 0x3f);
@@ -2541,7 +2807,7 @@ static void i830_update_wm(struct drm_device *dev, int planea_clock, int unused,
 				       pixel_size, latency_ns);
 	fwater_lo |= (3<<8) | planea_wm;
 
-	DRM_DEBUG("Setting FIFO watermarks - A: %d\n", planea_wm);
+	DRM_DEBUG_KMS("Setting FIFO watermarks - A: %d\n", planea_wm);
 
 	I915_WRITE(FW_BLC, fwater_lo);
 }
@@ -2586,17 +2852,20 @@ static void intel_update_watermarks(struct drm_device *dev)
 	unsigned long planea_clock = 0, planeb_clock = 0, sr_clock = 0;
 	int enabled = 0, pixel_size = 0;
 
+	if (!dev_priv->display.update_wm)
+		return;
+
 	/* Get the clock config from both planes */
 	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
 		intel_crtc = to_intel_crtc(crtc);
 		if (crtc->enabled) {
 			enabled++;
 			if (intel_crtc->plane == 0) {
-				DRM_DEBUG("plane A (pipe %d) clock: %d\n",
+				DRM_DEBUG_KMS("plane A (pipe %d) clock: %d\n",
 					  intel_crtc->pipe, crtc->mode.clock);
 				planea_clock = crtc->mode.clock;
 			} else {
-				DRM_DEBUG("plane B (pipe %d) clock: %d\n",
+				DRM_DEBUG_KMS("plane B (pipe %d) clock: %d\n",
 					  intel_crtc->pipe, crtc->mode.clock);
 				planeb_clock = crtc->mode.clock;
 			}
@@ -2613,10 +2882,10 @@ static void intel_update_watermarks(struct drm_device *dev)
 		return;
 
 	/* Single plane configs can enable self refresh */
-	if (enabled == 1 && IS_IGD(dev))
-		igd_enable_cxsr(dev, sr_clock, pixel_size);
-	else if (IS_IGD(dev))
-		igd_disable_cxsr(dev);
+	if (enabled == 1 && IS_PINEVIEW(dev))
+		pineview_enable_cxsr(dev, sr_clock, pixel_size);
+	else if (IS_PINEVIEW(dev))
+		pineview_disable_cxsr(dev);
 
 	dev_priv->display.update_wm(dev, planea_clock, planeb_clock,
 				    sr_hdisplay, pixel_size);
@@ -2710,10 +2979,11 @@ static int intel_crtc_mode_set(struct drm_crtc *crtc,
 
 	if (is_lvds && dev_priv->lvds_use_ssc && num_outputs < 2) {
 		refclk = dev_priv->lvds_ssc_freq * 1000;
-		DRM_DEBUG("using SSC reference clock of %d MHz\n", refclk / 1000);
+		DRM_DEBUG_KMS("using SSC reference clock of %d MHz\n",
+					refclk / 1000);
 	} else if (IS_I9XX(dev)) {
 		refclk = 96000;
-		if (IS_IGDNG(dev))
+		if (HAS_PCH_SPLIT(dev))
 			refclk = 120000; /* 120Mhz refclk */
 	} else {
 		refclk = 48000;
@@ -2733,14 +3003,23 @@ static int intel_crtc_mode_set(struct drm_crtc *crtc,
 		return -EINVAL;
 	}
 
-	if (limit->find_reduced_pll && dev_priv->lvds_downclock_avail) {
-		memcpy(&reduced_clock, &clock, sizeof(intel_clock_t));
-		has_reduced_clock = limit->find_reduced_pll(limit, crtc,
-							    (adjusted_mode->clock*3/4),
+	if (is_lvds && dev_priv->lvds_downclock_avail) {
+		has_reduced_clock = limit->find_pll(limit, crtc,
+							    dev_priv->lvds_downclock,
 							    refclk,
 							    &reduced_clock);
+		if (has_reduced_clock && (clock.p != reduced_clock.p)) {
+			/*
+			 * If the different P is found, it means that we can't
+			 * switch the display clock by using the FP0/FP1.
+			 * In such case we will disable the LVDS downclock
+			 * feature.
+			 */
+			DRM_DEBUG_KMS("Different P is found for "
+						"LVDS clock/downclock\n");
+			has_reduced_clock = 0;
+		}
 	}
-
 	/* SDVO TV has fixed PLL values depend on its clock range,
 	   this mirrors vbios setting. */
 	if (is_sdvo && is_tv) {
@@ -2762,8 +3041,8 @@ static int intel_crtc_mode_set(struct drm_crtc *crtc,
 	}
 
 	/* FDI link */
-	if (IS_IGDNG(dev)) {
-		int lane, link_bw;
+	if (HAS_PCH_SPLIT(dev)) {
+		int lane, link_bw, bpp;
 		/* eDP doesn't require FDI link, so just set DP M/N
 		   according to current link config */
 		if (is_edp) {
@@ -2782,11 +3061,99 @@ static int intel_crtc_mode_set(struct drm_crtc *crtc,
 			lane = 4;
 			link_bw = 270000;
 		}
-		igdng_compute_m_n(3, lane, target_clock,
-				  link_bw, &m_n);
+
+		/* determine panel color depth */
+		temp = I915_READ(pipeconf_reg);
+		temp &= ~PIPE_BPC_MASK;
+		if (is_lvds) {
+			int lvds_reg = I915_READ(PCH_LVDS);
+			/* the BPC will be 6 if it is 18-bit LVDS panel */
+			if ((lvds_reg & LVDS_A3_POWER_MASK) == LVDS_A3_POWER_UP)
+				temp |= PIPE_8BPC;
+			else
+				temp |= PIPE_6BPC;
+		} else if (is_edp) {
+			switch (dev_priv->edp_bpp/3) {
+			case 8:
+				temp |= PIPE_8BPC;
+				break;
+			case 10:
+				temp |= PIPE_10BPC;
+				break;
+			case 6:
+				temp |= PIPE_6BPC;
+				break;
+			case 12:
+				temp |= PIPE_12BPC;
+				break;
+			}
+		} else
+			temp |= PIPE_8BPC;
+		I915_WRITE(pipeconf_reg, temp);
+		I915_READ(pipeconf_reg);
+
+		switch (temp & PIPE_BPC_MASK) {
+		case PIPE_8BPC:
+			bpp = 24;
+			break;
+		case PIPE_10BPC:
+			bpp = 30;
+			break;
+		case PIPE_6BPC:
+			bpp = 18;
+			break;
+		case PIPE_12BPC:
+			bpp = 36;
+			break;
+		default:
+			DRM_ERROR("unknown pipe bpc value\n");
+			bpp = 24;
+		}
+
+		ironlake_compute_m_n(bpp, lane, target_clock, link_bw, &m_n);
+	}
+
+	/* Ironlake: try to setup display ref clock before DPLL
+	 * enabling. This is only under driver's control after
+	 * PCH B stepping, previous chipset stepping should be
+	 * ignoring this setting.
+	 */
+	if (HAS_PCH_SPLIT(dev)) {
+		temp = I915_READ(PCH_DREF_CONTROL);
+		/* Always enable nonspread source */
+		temp &= ~DREF_NONSPREAD_SOURCE_MASK;
+		temp |= DREF_NONSPREAD_SOURCE_ENABLE;
+		I915_WRITE(PCH_DREF_CONTROL, temp);
+		POSTING_READ(PCH_DREF_CONTROL);
+
+		temp &= ~DREF_SSC_SOURCE_MASK;
+		temp |= DREF_SSC_SOURCE_ENABLE;
+		I915_WRITE(PCH_DREF_CONTROL, temp);
+		POSTING_READ(PCH_DREF_CONTROL);
+
+		udelay(200);
+
+		if (is_edp) {
+			if (dev_priv->lvds_use_ssc) {
+				temp |= DREF_SSC1_ENABLE;
+				I915_WRITE(PCH_DREF_CONTROL, temp);
+				POSTING_READ(PCH_DREF_CONTROL);
+
+				udelay(200);
+
+				temp &= ~DREF_CPU_SOURCE_OUTPUT_MASK;
+				temp |= DREF_CPU_SOURCE_OUTPUT_DOWNSPREAD;
+				I915_WRITE(PCH_DREF_CONTROL, temp);
+				POSTING_READ(PCH_DREF_CONTROL);
+			} else {
+				temp |= DREF_CPU_SOURCE_OUTPUT_NONSPREAD;
+				I915_WRITE(PCH_DREF_CONTROL, temp);
+				POSTING_READ(PCH_DREF_CONTROL);
+			}
+		}
 	}
 
-	if (IS_IGD(dev)) {
+	if (IS_PINEVIEW(dev)) {
 		fp = (1 << clock.n) << 16 | clock.m1 << 8 | clock.m2;
 		if (has_reduced_clock)
 			fp2 = (1 << reduced_clock.n) << 16 |
@@ -2798,7 +3165,7 @@ static int intel_crtc_mode_set(struct drm_crtc *crtc,
 				reduced_clock.m2;
 	}
 
-	if (!IS_IGDNG(dev))
+	if (!HAS_PCH_SPLIT(dev))
 		dpll = DPLL_VGA_MODE_DIS;
 
 	if (IS_I9XX(dev)) {
@@ -2811,19 +3178,19 @@ static int intel_crtc_mode_set(struct drm_crtc *crtc,
 			sdvo_pixel_multiply = adjusted_mode->clock / mode->clock;
 			if (IS_I945G(dev) || IS_I945GM(dev) || IS_G33(dev))
 				dpll |= (sdvo_pixel_multiply - 1) << SDVO_MULTIPLIER_SHIFT_HIRES;
-			else if (IS_IGDNG(dev))
+			else if (HAS_PCH_SPLIT(dev))
 				dpll |= (sdvo_pixel_multiply - 1) << PLL_REF_SDVO_HDMI_MULTIPLIER_SHIFT;
 		}
 		if (is_dp)
 			dpll |= DPLL_DVO_HIGH_SPEED;
 
 		/* compute bitmask from p1 value */
-		if (IS_IGD(dev))
-			dpll |= (1 << (clock.p1 - 1)) << DPLL_FPA01_P1_POST_DIV_SHIFT_IGD;
+		if (IS_PINEVIEW(dev))
+			dpll |= (1 << (clock.p1 - 1)) << DPLL_FPA01_P1_POST_DIV_SHIFT_PINEVIEW;
 		else {
 			dpll |= (1 << (clock.p1 - 1)) << DPLL_FPA01_P1_POST_DIV_SHIFT;
 			/* also FPA1 */
-			if (IS_IGDNG(dev))
+			if (HAS_PCH_SPLIT(dev))
 				dpll |= (1 << (clock.p1 - 1)) << DPLL_FPA1_P1_POST_DIV_SHIFT;
 			if (IS_G4X(dev) && has_reduced_clock)
 				dpll |= (1 << (reduced_clock.p1 - 1)) << DPLL_FPA1_P1_POST_DIV_SHIFT;
@@ -2842,7 +3209,7 @@ static int intel_crtc_mode_set(struct drm_crtc *crtc,
 			dpll |= DPLLB_LVDS_P2_CLOCK_DIV_14;
 			break;
 		}
-		if (IS_I965G(dev) && !IS_IGDNG(dev))
+		if (IS_I965G(dev) && !HAS_PCH_SPLIT(dev))
 			dpll |= (6 << PLL_LOAD_PULSE_PHASE_SHIFT);
 	} else {
 		if (is_lvds) {
@@ -2874,9 +3241,9 @@ static int intel_crtc_mode_set(struct drm_crtc *crtc,
 	/* Set up the display plane register */
 	dspcntr = DISPPLANE_GAMMA_ENABLE;
 
-	/* IGDNG's plane is forced to pipe, bit 24 is to
+	/* Ironlake's plane is forced to pipe, bit 24 is to
 	   enable color space conversion */
-	if (!IS_IGDNG(dev)) {
+	if (!HAS_PCH_SPLIT(dev)) {
 		if (pipe == 0)
 			dspcntr &= ~DISPPLANE_SEL_PIPE_MASK;
 		else
@@ -2903,20 +3270,20 @@ static int intel_crtc_mode_set(struct drm_crtc *crtc,
 
 
 	/* Disable the panel fitter if it was on our pipe */
-	if (!IS_IGDNG(dev) && intel_panel_fitter_pipe(dev) == pipe)
+	if (!HAS_PCH_SPLIT(dev) && intel_panel_fitter_pipe(dev) == pipe)
 		I915_WRITE(PFIT_CONTROL, 0);
 
-	DRM_DEBUG("Mode for pipe %c:\n", pipe == 0 ? 'A' : 'B');
+	DRM_DEBUG_KMS("Mode for pipe %c:\n", pipe == 0 ? 'A' : 'B');
 	drm_mode_debug_printmodeline(mode);
 
-	/* assign to IGDNG registers */
-	if (IS_IGDNG(dev)) {
+	/* assign to Ironlake registers */
+	if (HAS_PCH_SPLIT(dev)) {
 		fp_reg = pch_fp_reg;
 		dpll_reg = pch_dpll_reg;
 	}
 
 	if (is_edp) {
-		igdng_disable_pll_edp(crtc);
+		ironlake_disable_pll_edp(crtc);
 	} else if ((dpll & DPLL_VCO_ENABLE)) {
 		I915_WRITE(fp_reg, fp);
 		I915_WRITE(dpll_reg, dpll & ~DPLL_VCO_ENABLE);
@@ -2931,11 +3298,13 @@ static int intel_crtc_mode_set(struct drm_crtc *crtc,
 	if (is_lvds) {
 		u32 lvds;
 
-		if (IS_IGDNG(dev))
+		if (HAS_PCH_SPLIT(dev))
 			lvds_reg = PCH_LVDS;
 
 		lvds = I915_READ(lvds_reg);
 		lvds |= LVDS_PORT_EN | LVDS_A0A2_CLKA_POWER_UP | LVDS_PIPEB_SELECT;
+		/* set the corresponsding LVDS_BORDER bit */
+		lvds |= dev_priv->lvds_border_bits;
 		/* Set the B0-B3 data pairs corresponding to whether we're going to
 		 * set the DPLLs for dual-channel mode or not.
 		 */
@@ -2948,7 +3317,20 @@ static int intel_crtc_mode_set(struct drm_crtc *crtc,
 		 * appropriately here, but we need to look more thoroughly into how
 		 * panels behave in the two modes.
 		 */
-
+		/* set the dithering flag */
+		if (IS_I965G(dev)) {
+			if (dev_priv->lvds_dither) {
+				if (HAS_PCH_SPLIT(dev))
+					pipeconf |= PIPE_ENABLE_DITHER;
+				else
+					lvds |= LVDS_ENABLE_DITHER;
+			} else {
+				if (HAS_PCH_SPLIT(dev))
+					pipeconf &= ~PIPE_ENABLE_DITHER;
+				else
+					lvds &= ~LVDS_ENABLE_DITHER;
+			}
+		}
 		I915_WRITE(lvds_reg, lvds);
 		I915_READ(lvds_reg);
 	}
@@ -2962,7 +3344,7 @@ static int intel_crtc_mode_set(struct drm_crtc *crtc,
 		/* Wait for the clocks to stabilize. */
 		udelay(150);
 
-		if (IS_I965G(dev) && !IS_IGDNG(dev)) {
+		if (IS_I965G(dev) && !HAS_PCH_SPLIT(dev)) {
 			if (is_sdvo) {
 				sdvo_pixel_multiply = adjusted_mode->clock / mode->clock;
 				I915_WRITE(dpll_md_reg, (0 << DPLL_MD_UDI_DIVIDER_SHIFT) |
@@ -2982,14 +3364,14 @@ static int intel_crtc_mode_set(struct drm_crtc *crtc,
 		I915_WRITE(fp_reg + 4, fp2);
 		intel_crtc->lowfreq_avail = true;
 		if (HAS_PIPE_CXSR(dev)) {
-			DRM_DEBUG("enabling CxSR downclocking\n");
+			DRM_DEBUG_KMS("enabling CxSR downclocking\n");
 			pipeconf |= PIPECONF_CXSR_DOWNCLOCK;
 		}
 	} else {
 		I915_WRITE(fp_reg + 4, fp);
 		intel_crtc->lowfreq_avail = false;
 		if (HAS_PIPE_CXSR(dev)) {
-			DRM_DEBUG("disabling CxSR downclocking\n");
+			DRM_DEBUG_KMS("disabling CxSR downclocking\n");
 			pipeconf &= ~PIPECONF_CXSR_DOWNCLOCK;
 		}
 	}
@@ -3009,21 +3391,21 @@ static int intel_crtc_mode_set(struct drm_crtc *crtc,
 	/* pipesrc and dspsize control the size that is scaled from, which should
 	 * always be the user's requested size.
 	 */
-	if (!IS_IGDNG(dev)) {
+	if (!HAS_PCH_SPLIT(dev)) {
 		I915_WRITE(dspsize_reg, ((mode->vdisplay - 1) << 16) |
 				(mode->hdisplay - 1));
 		I915_WRITE(dsppos_reg, 0);
 	}
 	I915_WRITE(pipesrc_reg, ((mode->hdisplay - 1) << 16) | (mode->vdisplay - 1));
 
-	if (IS_IGDNG(dev)) {
+	if (HAS_PCH_SPLIT(dev)) {
 		I915_WRITE(data_m1_reg, TU_SIZE(m_n.tu) | m_n.gmch_m);
 		I915_WRITE(data_n1_reg, TU_SIZE(m_n.tu) | m_n.gmch_n);
 		I915_WRITE(link_m1_reg, m_n.link_m);
 		I915_WRITE(link_n1_reg, m_n.link_n);
 
 		if (is_edp) {
-			igdng_set_pll_edp(crtc, adjusted_mode->clock);
+			ironlake_set_pll_edp(crtc, adjusted_mode->clock);
 		} else {
 			/* enable FDI RX PLL too */
 			temp = I915_READ(fdi_rx_reg);
@@ -3037,7 +3419,7 @@ static int intel_crtc_mode_set(struct drm_crtc *crtc,
 
 	intel_wait_for_vblank(dev);
 
-	if (IS_IGDNG(dev)) {
+	if (IS_IRONLAKE(dev)) {
 		/* enable address swizzle for tiling buffer */
 		temp = I915_READ(DISP_ARB_CTL);
 		I915_WRITE(DISP_ARB_CTL, temp | DISP_TILE_SURFACE_SWIZZLING);
@@ -3071,8 +3453,8 @@ void intel_crtc_load_lut(struct drm_crtc *crtc)
 	if (!crtc->enabled)
 		return;
 
-	/* use legacy palette for IGDNG */
-	if (IS_IGDNG(dev))
+	/* use legacy palette for Ironlake */
+	if (HAS_PCH_SPLIT(dev))
 		palreg = (intel_crtc->pipe == 0) ? LGC_PALETTE_A :
 						   LGC_PALETTE_B;
 
@@ -3095,18 +3477,17 @@ static int intel_crtc_cursor_set(struct drm_crtc *crtc,
 	struct drm_gem_object *bo;
 	struct drm_i915_gem_object *obj_priv;
 	int pipe = intel_crtc->pipe;
-	int plane = intel_crtc->plane;
 	uint32_t control = (pipe == 0) ? CURACNTR : CURBCNTR;
 	uint32_t base = (pipe == 0) ? CURABASE : CURBBASE;
 	uint32_t temp = I915_READ(control);
 	size_t addr;
 	int ret;
 
-	DRM_DEBUG("\n");
+	DRM_DEBUG_KMS("\n");
 
 	/* if we want to turn off the cursor ignore width and height */
 	if (!handle) {
-		DRM_DEBUG("cursor off\n");
+		DRM_DEBUG_KMS("cursor off\n");
 		if (IS_MOBILE(dev) || IS_I9XX(dev)) {
 			temp &= ~(CURSOR_MODE | MCURSOR_GAMMA_ENABLE);
 			temp |= CURSOR_MODE_DISABLE;
@@ -3139,7 +3520,7 @@ static int intel_crtc_cursor_set(struct drm_crtc *crtc,
 
 	/* we only need to pin inside GTT if cursor is non-phy */
 	mutex_lock(&dev->struct_mutex);
-	if (!dev_priv->cursor_needs_physical) {
+	if (!dev_priv->info->cursor_needs_physical) {
 		ret = i915_gem_object_pin(bo, PAGE_SIZE);
 		if (ret) {
 			DRM_ERROR("failed to pin cursor bo\n");
@@ -3174,7 +3555,7 @@ static int intel_crtc_cursor_set(struct drm_crtc *crtc,
 	I915_WRITE(base, addr);
 
 	if (intel_crtc->cursor_bo) {
-		if (dev_priv->cursor_needs_physical) {
+		if (dev_priv->info->cursor_needs_physical) {
 			if (intel_crtc->cursor_bo != bo)
 				i915_gem_detach_phys_object(dev, intel_crtc->cursor_bo);
 		} else
@@ -3182,20 +3563,16 @@ static int intel_crtc_cursor_set(struct drm_crtc *crtc,
 		drm_gem_object_unreference(intel_crtc->cursor_bo);
 	}
 
-	if ((IS_I965G(dev) || plane == 0))
-		intel_update_fbc(crtc, &crtc->mode);
-
 	mutex_unlock(&dev->struct_mutex);
 
 	intel_crtc->cursor_addr = addr;
 	intel_crtc->cursor_bo = bo;
 
 	return 0;
-fail:
-	mutex_lock(&dev->struct_mutex);
 fail_locked:
-	drm_gem_object_unreference(bo);
 	mutex_unlock(&dev->struct_mutex);
+fail:
+	drm_gem_object_unreference_unlocked(bo);
 	return ret;
 }
 
@@ -3244,6 +3621,16 @@ void intel_crtc_fb_gamma_set(struct drm_crtc *crtc, u16 red, u16 green,
 	intel_crtc->lut_b[regno] = blue >> 8;
 }
 
+void intel_crtc_fb_gamma_get(struct drm_crtc *crtc, u16 *red, u16 *green,
+			     u16 *blue, int regno)
+{
+	struct intel_crtc *intel_crtc = to_intel_crtc(crtc);
+
+	*red = intel_crtc->lut_r[regno] << 8;
+	*green = intel_crtc->lut_g[regno] << 8;
+	*blue = intel_crtc->lut_b[regno] << 8;
+}
+
 static void intel_crtc_gamma_set(struct drm_crtc *crtc, u16 *red, u16 *green,
 				 u16 *blue, uint32_t size)
 {
@@ -3407,18 +3794,18 @@ static int intel_crtc_clock_get(struct drm_device *dev, struct drm_crtc *crtc)
 		fp = I915_READ((pipe == 0) ? FPA1 : FPB1);
 
 	clock.m1 = (fp & FP_M1_DIV_MASK) >> FP_M1_DIV_SHIFT;
-	if (IS_IGD(dev)) {
-		clock.n = ffs((fp & FP_N_IGD_DIV_MASK) >> FP_N_DIV_SHIFT) - 1;
-		clock.m2 = (fp & FP_M2_IGD_DIV_MASK) >> FP_M2_DIV_SHIFT;
+	if (IS_PINEVIEW(dev)) {
+		clock.n = ffs((fp & FP_N_PINEVIEW_DIV_MASK) >> FP_N_DIV_SHIFT) - 1;
+		clock.m2 = (fp & FP_M2_PINEVIEW_DIV_MASK) >> FP_M2_DIV_SHIFT;
 	} else {
 		clock.n = (fp & FP_N_DIV_MASK) >> FP_N_DIV_SHIFT;
 		clock.m2 = (fp & FP_M2_DIV_MASK) >> FP_M2_DIV_SHIFT;
 	}
 
 	if (IS_I9XX(dev)) {
-		if (IS_IGD(dev))
-			clock.p1 = ffs((dpll & DPLL_FPA01_P1_POST_DIV_MASK_IGD) >>
-				DPLL_FPA01_P1_POST_DIV_SHIFT_IGD);
+		if (IS_PINEVIEW(dev))
+			clock.p1 = ffs((dpll & DPLL_FPA01_P1_POST_DIV_MASK_PINEVIEW) >>
+				DPLL_FPA01_P1_POST_DIV_SHIFT_PINEVIEW);
 		else
 			clock.p1 = ffs((dpll & DPLL_FPA01_P1_POST_DIV_MASK) >>
 			       DPLL_FPA01_P1_POST_DIV_SHIFT);
@@ -3433,7 +3820,7 @@ static int intel_crtc_clock_get(struct drm_device *dev, struct drm_crtc *crtc)
 				7 : 14;
 			break;
 		default:
-			DRM_DEBUG("Unknown DPLL mode %08x in programmed "
+			DRM_DEBUG_KMS("Unknown DPLL mode %08x in programmed "
 				  "mode\n", (int)(dpll & DPLL_MODE_MASK));
 			return 0;
 		}
@@ -3519,132 +3906,13 @@ static void intel_gpu_idle_timer(unsigned long arg)
 	struct drm_device *dev = (struct drm_device *)arg;
 	drm_i915_private_t *dev_priv = dev->dev_private;
 
-	DRM_DEBUG("idle timer fired, downclocking\n");
+	DRM_DEBUG_DRIVER("idle timer fired, downclocking\n");
 
 	dev_priv->busy = false;
 
 	queue_work(dev_priv->wq, &dev_priv->idle_work);
 }
 
-void intel_increase_renderclock(struct drm_device *dev, bool schedule)
-{
-	drm_i915_private_t *dev_priv = dev->dev_private;
-
-	if (IS_IGDNG(dev))
-		return;
-
-	if (!dev_priv->render_reclock_avail) {
-		DRM_DEBUG("not reclocking render clock\n");
-		return;
-	}
-
-	/* Restore render clock frequency to original value */
-	if (IS_G4X(dev) || IS_I9XX(dev))
-		pci_write_config_word(dev->pdev, GCFGC, dev_priv->orig_clock);
-	else if (IS_I85X(dev))
-		pci_write_config_word(dev->pdev, HPLLCC, dev_priv->orig_clock);
-	DRM_DEBUG("increasing render clock frequency\n");
-
-	/* Schedule downclock */
-	if (schedule)
-		mod_timer(&dev_priv->idle_timer, jiffies +
-			  msecs_to_jiffies(GPU_IDLE_TIMEOUT));
-}
-
-void intel_decrease_renderclock(struct drm_device *dev)
-{
-	drm_i915_private_t *dev_priv = dev->dev_private;
-
-	if (IS_IGDNG(dev))
-		return;
-
-	if (!dev_priv->render_reclock_avail) {
-		DRM_DEBUG("not reclocking render clock\n");
-		return;
-	}
-
-	if (IS_G4X(dev)) {
-		u16 gcfgc;
-
-		/* Adjust render clock... */
-		pci_read_config_word(dev->pdev, GCFGC, &gcfgc);
-
-		/* Down to minimum... */
-		gcfgc &= ~GM45_GC_RENDER_CLOCK_MASK;
-		gcfgc |= GM45_GC_RENDER_CLOCK_266_MHZ;
-
-		pci_write_config_word(dev->pdev, GCFGC, gcfgc);
-	} else if (IS_I965G(dev)) {
-		u16 gcfgc;
-
-		/* Adjust render clock... */
-		pci_read_config_word(dev->pdev, GCFGC, &gcfgc);
-
-		/* Down to minimum... */
-		gcfgc &= ~I965_GC_RENDER_CLOCK_MASK;
-		gcfgc |= I965_GC_RENDER_CLOCK_267_MHZ;
-
-		pci_write_config_word(dev->pdev, GCFGC, gcfgc);
-	} else if (IS_I945G(dev) || IS_I945GM(dev)) {
-		u16 gcfgc;
-
-		/* Adjust render clock... */
-		pci_read_config_word(dev->pdev, GCFGC, &gcfgc);
-
-		/* Down to minimum... */
-		gcfgc &= ~I945_GC_RENDER_CLOCK_MASK;
-		gcfgc |= I945_GC_RENDER_CLOCK_166_MHZ;
-
-		pci_write_config_word(dev->pdev, GCFGC, gcfgc);
-	} else if (IS_I915G(dev)) {
-		u16 gcfgc;
-
-		/* Adjust render clock... */
-		pci_read_config_word(dev->pdev, GCFGC, &gcfgc);
-
-		/* Down to minimum... */
-		gcfgc &= ~I915_GC_RENDER_CLOCK_MASK;
-		gcfgc |= I915_GC_RENDER_CLOCK_166_MHZ;
-
-		pci_write_config_word(dev->pdev, GCFGC, gcfgc);
-	} else if (IS_I85X(dev)) {
-		u16 hpllcc;
-
-		/* Adjust render clock... */
-		pci_read_config_word(dev->pdev, HPLLCC, &hpllcc);
-
-		/* Up to maximum... */
-		hpllcc &= ~GC_CLOCK_CONTROL_MASK;
-		hpllcc |= GC_CLOCK_133_200;
-
-		pci_write_config_word(dev->pdev, HPLLCC, hpllcc);
-	}
-	DRM_DEBUG("decreasing render clock frequency\n");
-}
-
-/* Note that no increase function is needed for this - increase_renderclock()
- *  will also rewrite these bits
- */
-void intel_decrease_displayclock(struct drm_device *dev)
-{
-	if (IS_IGDNG(dev))
-		return;
-
-	if (IS_I945G(dev) || IS_I945GM(dev) || IS_I915G(dev) ||
-	    IS_I915GM(dev)) {
-		u16 gcfgc;
-
-		/* Adjust render clock... */
-		pci_read_config_word(dev->pdev, GCFGC, &gcfgc);
-
-		/* Down to minimum... */
-		gcfgc &= ~0xf0;
-		gcfgc |= 0x80;
-
-		pci_write_config_word(dev->pdev, GCFGC, gcfgc);
-	}
-}
-
 #define CRTC_IDLE_TIMEOUT 1000 /* ms */
 
 static void intel_crtc_idle_timer(unsigned long arg)
@@ -3653,7 +3921,7 @@ static void intel_crtc_idle_timer(unsigned long arg)
 	struct drm_crtc *crtc = &intel_crtc->base;
 	drm_i915_private_t *dev_priv = crtc->dev->dev_private;
 
-	DRM_DEBUG("idle timer fired, downclocking\n");
+	DRM_DEBUG_DRIVER("idle timer fired, downclocking\n");
 
 	intel_crtc->busy = false;
 
@@ -3669,14 +3937,14 @@ static void intel_increase_pllclock(struct drm_crtc *crtc, bool schedule)
 	int dpll_reg = (pipe == 0) ? DPLL_A : DPLL_B;
 	int dpll = I915_READ(dpll_reg);
 
-	if (IS_IGDNG(dev))
+	if (HAS_PCH_SPLIT(dev))
 		return;
 
 	if (!dev_priv->lvds_downclock_avail)
 		return;
 
 	if (!HAS_PIPE_CXSR(dev) && (dpll & DISPLAY_RATE_SELECT_FPA1)) {
-		DRM_DEBUG("upclocking LVDS\n");
+		DRM_DEBUG_DRIVER("upclocking LVDS\n");
 
 		/* Unlock panel regs */
 		I915_WRITE(PP_CONTROL, I915_READ(PP_CONTROL) | (0xabcd << 16));
@@ -3687,7 +3955,7 @@ static void intel_increase_pllclock(struct drm_crtc *crtc, bool schedule)
 		intel_wait_for_vblank(dev);
 		dpll = I915_READ(dpll_reg);
 		if (dpll & DISPLAY_RATE_SELECT_FPA1)
-			DRM_DEBUG("failed to upclock LVDS!\n");
+			DRM_DEBUG_DRIVER("failed to upclock LVDS!\n");
 
 		/* ...and lock them again */
 		I915_WRITE(PP_CONTROL, I915_READ(PP_CONTROL) & 0x3);
@@ -3708,7 +3976,7 @@ static void intel_decrease_pllclock(struct drm_crtc *crtc)
 	int dpll_reg = (pipe == 0) ? DPLL_A : DPLL_B;
 	int dpll = I915_READ(dpll_reg);
 
-	if (IS_IGDNG(dev))
+	if (HAS_PCH_SPLIT(dev))
 		return;
 
 	if (!dev_priv->lvds_downclock_avail)
@@ -3719,7 +3987,7 @@ static void intel_decrease_pllclock(struct drm_crtc *crtc)
 	 * the manual case.
 	 */
 	if (!HAS_PIPE_CXSR(dev) && intel_crtc->lowfreq_avail) {
-		DRM_DEBUG("downclocking LVDS\n");
+		DRM_DEBUG_DRIVER("downclocking LVDS\n");
 
 		/* Unlock panel regs */
 		I915_WRITE(PP_CONTROL, I915_READ(PP_CONTROL) | (0xabcd << 16));
@@ -3730,7 +3998,7 @@ static void intel_decrease_pllclock(struct drm_crtc *crtc)
 		intel_wait_for_vblank(dev);
 		dpll = I915_READ(dpll_reg);
 		if (!(dpll & DISPLAY_RATE_SELECT_FPA1))
-			DRM_DEBUG("failed to downclock LVDS!\n");
+			DRM_DEBUG_DRIVER("failed to downclock LVDS!\n");
 
 		/* ...and lock them again */
 		I915_WRITE(PP_CONTROL, I915_READ(PP_CONTROL) & 0x3);
@@ -3758,10 +4026,9 @@ static void intel_idle_update(struct work_struct *work)
 
 	mutex_lock(&dev->struct_mutex);
 
-	/* GPU isn't processing, downclock it. */
-	if (!dev_priv->busy) {
-		intel_decrease_renderclock(dev);
-		intel_decrease_displayclock(dev);
+	if (IS_I945G(dev) || IS_I945GM(dev)) {
+		DRM_DEBUG_DRIVER("enable memory self refresh on 945\n");
+		I915_WRITE(FW_BLC_SELF, FW_BLC_SELF_EN_MASK | FW_BLC_SELF_EN);
 	}
 
 	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
@@ -3797,8 +4064,19 @@ void intel_mark_busy(struct drm_device *dev, struct drm_gem_object *obj)
 	if (!drm_core_check_feature(dev, DRIVER_MODESET))
 		return;
 
-	dev_priv->busy = true;
-	intel_increase_renderclock(dev, true);
+	if (!dev_priv->busy) {
+		if (IS_I945G(dev) || IS_I945GM(dev)) {
+			u32 fw_blc_self;
+
+			DRM_DEBUG_DRIVER("disable memory self refresh on 945\n");
+			fw_blc_self = I915_READ(FW_BLC_SELF);
+			fw_blc_self &= ~FW_BLC_SELF_EN;
+			I915_WRITE(FW_BLC_SELF, fw_blc_self | FW_BLC_SELF_EN_MASK);
+		}
+		dev_priv->busy = true;
+	} else
+		mod_timer(&dev_priv->idle_timer, jiffies +
+			  msecs_to_jiffies(GPU_IDLE_TIMEOUT));
 
 	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
 		if (!crtc->fb)
@@ -3808,6 +4086,14 @@ void intel_mark_busy(struct drm_device *dev, struct drm_gem_object *obj)
 		intel_fb = to_intel_framebuffer(crtc->fb);
 		if (intel_fb->obj == obj) {
 			if (!intel_crtc->busy) {
+				if (IS_I945G(dev) || IS_I945GM(dev)) {
+					u32 fw_blc_self;
+
+					DRM_DEBUG_DRIVER("disable memory self refresh on 945\n");
+					fw_blc_self = I915_READ(FW_BLC_SELF);
+					fw_blc_self &= ~FW_BLC_SELF_EN;
+					I915_WRITE(FW_BLC_SELF, fw_blc_self | FW_BLC_SELF_EN_MASK);
+				}
 				/* Non-busy -> busy, upclock */
 				intel_increase_pllclock(crtc, true);
 				intel_crtc->busy = true;
@@ -3828,6 +4114,180 @@ static void intel_crtc_destroy(struct drm_crtc *crtc)
 	kfree(intel_crtc);
 }
 
+struct intel_unpin_work {
+	struct work_struct work;
+	struct drm_device *dev;
+	struct drm_gem_object *old_fb_obj;
+	struct drm_gem_object *pending_flip_obj;
+	struct drm_pending_vblank_event *event;
+	int pending;
+};
+
+static void intel_unpin_work_fn(struct work_struct *__work)
+{
+	struct intel_unpin_work *work =
+		container_of(__work, struct intel_unpin_work, work);
+
+	mutex_lock(&work->dev->struct_mutex);
+	i915_gem_object_unpin(work->old_fb_obj);
+	drm_gem_object_unreference(work->pending_flip_obj);
+	drm_gem_object_unreference(work->old_fb_obj);
+	mutex_unlock(&work->dev->struct_mutex);
+	kfree(work);
+}
+
+void intel_finish_page_flip(struct drm_device *dev, int pipe)
+{
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	struct drm_crtc *crtc = dev_priv->pipe_to_crtc_mapping[pipe];
+	struct intel_crtc *intel_crtc = to_intel_crtc(crtc);
+	struct intel_unpin_work *work;
+	struct drm_i915_gem_object *obj_priv;
+	struct drm_pending_vblank_event *e;
+	struct timeval now;
+	unsigned long flags;
+
+	/* Ignore early vblank irqs */
+	if (intel_crtc == NULL)
+		return;
+
+	spin_lock_irqsave(&dev->event_lock, flags);
+	work = intel_crtc->unpin_work;
+	if (work == NULL || !work->pending) {
+		if (work && !work->pending) {
+			obj_priv = work->pending_flip_obj->driver_private;
+			DRM_DEBUG_DRIVER("flip finish: %p (%d) not pending?\n",
+					 obj_priv,
+					 atomic_read(&obj_priv->pending_flip));
+		}
+		spin_unlock_irqrestore(&dev->event_lock, flags);
+		return;
+	}
+
+	intel_crtc->unpin_work = NULL;
+	drm_vblank_put(dev, intel_crtc->pipe);
+
+	if (work->event) {
+		e = work->event;
+		do_gettimeofday(&now);
+		e->event.sequence = drm_vblank_count(dev, intel_crtc->pipe);
+		e->event.tv_sec = now.tv_sec;
+		e->event.tv_usec = now.tv_usec;
+		list_add_tail(&e->base.link,
+			      &e->base.file_priv->event_list);
+		wake_up_interruptible(&e->base.file_priv->event_wait);
+	}
+
+	spin_unlock_irqrestore(&dev->event_lock, flags);
+
+	obj_priv = work->pending_flip_obj->driver_private;
+
+	/* Initial scanout buffer will have a 0 pending flip count */
+	if ((atomic_read(&obj_priv->pending_flip) == 0) ||
+	    atomic_dec_and_test(&obj_priv->pending_flip))
+		DRM_WAKEUP(&dev_priv->pending_flip_queue);
+	schedule_work(&work->work);
+}
+
+void intel_prepare_page_flip(struct drm_device *dev, int plane)
+{
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	struct intel_crtc *intel_crtc =
+		to_intel_crtc(dev_priv->plane_to_crtc_mapping[plane]);
+	unsigned long flags;
+
+	spin_lock_irqsave(&dev->event_lock, flags);
+	if (intel_crtc->unpin_work) {
+		intel_crtc->unpin_work->pending = 1;
+	} else {
+		DRM_DEBUG_DRIVER("preparing flip with no unpin work?\n");
+	}
+	spin_unlock_irqrestore(&dev->event_lock, flags);
+}
+
+static int intel_crtc_page_flip(struct drm_crtc *crtc,
+				struct drm_framebuffer *fb,
+				struct drm_pending_vblank_event *event)
+{
+	struct drm_device *dev = crtc->dev;
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct intel_framebuffer *intel_fb;
+	struct drm_i915_gem_object *obj_priv;
+	struct drm_gem_object *obj;
+	struct intel_crtc *intel_crtc = to_intel_crtc(crtc);
+	struct intel_unpin_work *work;
+	unsigned long flags;
+	int pipesrc_reg = (intel_crtc->pipe == 0) ? PIPEASRC : PIPEBSRC;
+	int ret, pipesrc;
+	RING_LOCALS;
+
+	work = kzalloc(sizeof *work, GFP_KERNEL);
+	if (work == NULL)
+		return -ENOMEM;
+
+	mutex_lock(&dev->struct_mutex);
+
+	work->event = event;
+	work->dev = crtc->dev;
+	intel_fb = to_intel_framebuffer(crtc->fb);
+	work->old_fb_obj = intel_fb->obj;
+	INIT_WORK(&work->work, intel_unpin_work_fn);
+
+	/* We borrow the event spin lock for protecting unpin_work */
+	spin_lock_irqsave(&dev->event_lock, flags);
+	if (intel_crtc->unpin_work) {
+		DRM_DEBUG_DRIVER("flip queue: crtc already busy\n");
+		spin_unlock_irqrestore(&dev->event_lock, flags);
+		kfree(work);
+		mutex_unlock(&dev->struct_mutex);
+		return -EBUSY;
+	}
+	intel_crtc->unpin_work = work;
+	spin_unlock_irqrestore(&dev->event_lock, flags);
+
+	intel_fb = to_intel_framebuffer(fb);
+	obj = intel_fb->obj;
+
+	ret = intel_pin_and_fence_fb_obj(dev, obj);
+	if (ret != 0) {
+		DRM_DEBUG_DRIVER("flip queue: %p pin & fence failed\n",
+			  obj->driver_private);
+		kfree(work);
+		intel_crtc->unpin_work = NULL;
+		mutex_unlock(&dev->struct_mutex);
+		return ret;
+	}
+
+	/* Reference the objects for the scheduled work. */
+	drm_gem_object_reference(work->old_fb_obj);
+	drm_gem_object_reference(obj);
+
+	crtc->fb = fb;
+	i915_gem_object_flush_write_domain(obj);
+	drm_vblank_get(dev, intel_crtc->pipe);
+	obj_priv = obj->driver_private;
+	atomic_inc(&obj_priv->pending_flip);
+	work->pending_flip_obj = obj;
+
+	BEGIN_LP_RING(4);
+	OUT_RING(MI_DISPLAY_FLIP |
+		 MI_DISPLAY_FLIP_PLANE(intel_crtc->plane));
+	OUT_RING(fb->pitch);
+	if (IS_I965G(dev)) {
+		OUT_RING(obj_priv->gtt_offset | obj_priv->tiling_mode);
+		pipesrc = I915_READ(pipesrc_reg); 
+		OUT_RING(pipesrc & 0x0fff0fff);
+	} else {
+		OUT_RING(obj_priv->gtt_offset);
+		OUT_RING(MI_NOOP);
+	}
+	ADVANCE_LP_RING();
+
+	mutex_unlock(&dev->struct_mutex);
+
+	return 0;
+}
+
 static const struct drm_crtc_helper_funcs intel_helper_funcs = {
 	.dpms = intel_crtc_dpms,
 	.mode_fixup = intel_crtc_mode_fixup,
@@ -3835,6 +4295,7 @@ static const struct drm_crtc_helper_funcs intel_helper_funcs = {
 	.mode_set_base = intel_pipe_set_base,
 	.prepare = intel_crtc_prepare,
 	.commit = intel_crtc_commit,
+	.load_lut = intel_crtc_load_lut,
 };
 
 static const struct drm_crtc_funcs intel_crtc_funcs = {
@@ -3843,11 +4304,13 @@ static const struct drm_crtc_funcs intel_crtc_funcs = {
 	.gamma_set = intel_crtc_gamma_set,
 	.set_config = drm_crtc_helper_set_config,
 	.destroy = intel_crtc_destroy,
+	.page_flip = intel_crtc_page_flip,
 };
 
 
 static void intel_crtc_init(struct drm_device *dev, int pipe)
 {
+	drm_i915_private_t *dev_priv = dev->dev_private;
 	struct intel_crtc *intel_crtc;
 	int i;
 
@@ -3870,10 +4333,15 @@ static void intel_crtc_init(struct drm_device *dev, int pipe)
 	intel_crtc->pipe = pipe;
 	intel_crtc->plane = pipe;
 	if (IS_MOBILE(dev) && (IS_I9XX(dev) && !IS_I965G(dev))) {
-		DRM_DEBUG("swapping pipes & planes for FBC\n");
+		DRM_DEBUG_KMS("swapping pipes & planes for FBC\n");
 		intel_crtc->plane = ((pipe == 0) ? 1 : 0);
 	}
 
+	BUG_ON(pipe >= ARRAY_SIZE(dev_priv->plane_to_crtc_mapping) ||
+	       dev_priv->plane_to_crtc_mapping[intel_crtc->plane] != NULL);
+	dev_priv->plane_to_crtc_mapping[intel_crtc->plane] = &intel_crtc->base;
+	dev_priv->pipe_to_crtc_mapping[intel_crtc->pipe] = &intel_crtc->base;
+
 	intel_crtc->cursor_addr = 0;
 	intel_crtc->dpms_mode = DRM_MODE_DPMS_OFF;
 	drm_crtc_helper_add(&intel_crtc->base, &intel_helper_funcs);
@@ -3950,7 +4418,7 @@ static void intel_setup_outputs(struct drm_device *dev)
 	if (IS_MOBILE(dev) && !IS_I830(dev))
 		intel_lvds_init(dev);
 
-	if (IS_IGDNG(dev)) {
+	if (HAS_PCH_SPLIT(dev)) {
 		int found;
 
 		if (IS_MOBILE(dev) && (I915_READ(DP_A) & DP_DETECTED))
@@ -3978,37 +4446,51 @@ static void intel_setup_outputs(struct drm_device *dev)
 		if (I915_READ(PCH_DP_D) & DP_DETECTED)
 			intel_dp_init(dev, PCH_DP_D);
 
-	} else if (IS_I9XX(dev)) {
+	} else if (SUPPORTS_DIGITAL_OUTPUTS(dev)) {
 		bool found = false;
 
 		if (I915_READ(SDVOB) & SDVO_DETECTED) {
+			DRM_DEBUG_KMS("probing SDVOB\n");
 			found = intel_sdvo_init(dev, SDVOB);
-			if (!found && SUPPORTS_INTEGRATED_HDMI(dev))
+			if (!found && SUPPORTS_INTEGRATED_HDMI(dev)) {
+				DRM_DEBUG_KMS("probing HDMI on SDVOB\n");
 				intel_hdmi_init(dev, SDVOB);
+			}
 
-			if (!found && SUPPORTS_INTEGRATED_DP(dev))
+			if (!found && SUPPORTS_INTEGRATED_DP(dev)) {
+				DRM_DEBUG_KMS("probing DP_B\n");
 				intel_dp_init(dev, DP_B);
+			}
 		}
 
 		/* Before G4X SDVOC doesn't have its own detect register */
 
-		if (I915_READ(SDVOB) & SDVO_DETECTED)
+		if (I915_READ(SDVOB) & SDVO_DETECTED) {
+			DRM_DEBUG_KMS("probing SDVOC\n");
 			found = intel_sdvo_init(dev, SDVOC);
+		}
 
 		if (!found && (I915_READ(SDVOC) & SDVO_DETECTED)) {
 
-			if (SUPPORTS_INTEGRATED_HDMI(dev))
+			if (SUPPORTS_INTEGRATED_HDMI(dev)) {
+				DRM_DEBUG_KMS("probing HDMI on SDVOC\n");
 				intel_hdmi_init(dev, SDVOC);
-			if (SUPPORTS_INTEGRATED_DP(dev))
+			}
+			if (SUPPORTS_INTEGRATED_DP(dev)) {
+				DRM_DEBUG_KMS("probing DP_C\n");
 				intel_dp_init(dev, DP_C);
+			}
 		}
 
-		if (SUPPORTS_INTEGRATED_DP(dev) && (I915_READ(DP_D) & DP_DETECTED))
+		if (SUPPORTS_INTEGRATED_DP(dev) &&
+		    (I915_READ(DP_D) & DP_DETECTED)) {
+			DRM_DEBUG_KMS("probing DP_D\n");
 			intel_dp_init(dev, DP_D);
-	} else
+		}
+	} else if (IS_GEN2(dev))
 		intel_dvo_init(dev);
 
-	if (IS_I9XX(dev) && IS_MOBILE(dev) && !IS_IGDNG(dev))
+	if (SUPPORTS_TV(dev))
 		intel_tv_init(dev);
 
 	list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
@@ -4030,9 +4512,7 @@ static void intel_user_framebuffer_destroy(struct drm_framebuffer *fb)
 		intelfb_remove(dev, fb);
 
 	drm_framebuffer_cleanup(fb);
-	mutex_lock(&dev->struct_mutex);
-	drm_gem_object_unreference(intel_fb->obj);
-	mutex_unlock(&dev->struct_mutex);
+	drm_gem_object_unreference_unlocked(intel_fb->obj);
 
 	kfree(intel_fb);
 }
@@ -4095,9 +4575,7 @@ intel_user_framebuffer_create(struct drm_device *dev,
 
 	ret = intel_framebuffer_create(dev, mode_cmd, &fb, obj);
 	if (ret) {
-		mutex_lock(&dev->struct_mutex);
-		drm_gem_object_unreference(obj);
-		mutex_unlock(&dev->struct_mutex);
+		drm_gem_object_unreference_unlocked(obj);
 		return NULL;
 	}
 
@@ -4109,6 +4587,127 @@ static const struct drm_mode_config_funcs intel_mode_funcs = {
 	.fb_changed = intelfb_probe,
 };
 
+static struct drm_gem_object *
+intel_alloc_power_context(struct drm_device *dev)
+{
+	struct drm_gem_object *pwrctx;
+	int ret;
+
+	pwrctx = drm_gem_object_alloc(dev, 4096);
+	if (!pwrctx) {
+		DRM_DEBUG("failed to alloc power context, RC6 disabled\n");
+		return NULL;
+	}
+
+	mutex_lock(&dev->struct_mutex);
+	ret = i915_gem_object_pin(pwrctx, 4096);
+	if (ret) {
+		DRM_ERROR("failed to pin power context: %d\n", ret);
+		goto err_unref;
+	}
+
+	ret = i915_gem_object_set_to_gtt_domain(pwrctx, 1);
+	if (ret) {
+		DRM_ERROR("failed to set-domain on power context: %d\n", ret);
+		goto err_unpin;
+	}
+	mutex_unlock(&dev->struct_mutex);
+
+	return pwrctx;
+
+err_unpin:
+	i915_gem_object_unpin(pwrctx);
+err_unref:
+	drm_gem_object_unreference(pwrctx);
+	mutex_unlock(&dev->struct_mutex);
+	return NULL;
+}
+
+void ironlake_enable_drps(struct drm_device *dev)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	u32 rgvmodectl = I915_READ(MEMMODECTL), rgvswctl;
+	u8 fmax, fmin, fstart, vstart;
+	int i = 0;
+
+	/* 100ms RC evaluation intervals */
+	I915_WRITE(RCUPEI, 100000);
+	I915_WRITE(RCDNEI, 100000);
+
+	/* Set max/min thresholds to 90ms and 80ms respectively */
+	I915_WRITE(RCBMAXAVG, 90000);
+	I915_WRITE(RCBMINAVG, 80000);
+
+	I915_WRITE(MEMIHYST, 1);
+
+	/* Set up min, max, and cur for interrupt handling */
+	fmax = (rgvmodectl & MEMMODE_FMAX_MASK) >> MEMMODE_FMAX_SHIFT;
+	fmin = (rgvmodectl & MEMMODE_FMIN_MASK);
+	fstart = (rgvmodectl & MEMMODE_FSTART_MASK) >>
+		MEMMODE_FSTART_SHIFT;
+	vstart = (I915_READ(PXVFREQ_BASE + (fstart * 4)) & PXVFREQ_PX_MASK) >>
+		PXVFREQ_PX_SHIFT;
+
+	dev_priv->max_delay = fstart; /* can't go to fmax w/o IPS */
+	dev_priv->min_delay = fmin;
+	dev_priv->cur_delay = fstart;
+
+	I915_WRITE(MEMINTREN, MEMINT_CX_SUPR_EN | MEMINT_EVAL_CHG_EN);
+
+	/*
+	 * Interrupts will be enabled in ironlake_irq_postinstall
+	 */
+
+	I915_WRITE(VIDSTART, vstart);
+	POSTING_READ(VIDSTART);
+
+	rgvmodectl |= MEMMODE_SWMODE_EN;
+	I915_WRITE(MEMMODECTL, rgvmodectl);
+
+	while (I915_READ(MEMSWCTL) & MEMCTL_CMD_STS) {
+		if (i++ > 100) {
+			DRM_ERROR("stuck trying to change perf mode\n");
+			break;
+		}
+		msleep(1);
+	}
+	msleep(1);
+
+	rgvswctl = (MEMCTL_CMD_CHFREQ << MEMCTL_CMD_SHIFT) |
+		(fstart << MEMCTL_FREQ_SHIFT) | MEMCTL_SFCAVM;
+	I915_WRITE(MEMSWCTL, rgvswctl);
+	POSTING_READ(MEMSWCTL);
+
+	rgvswctl |= MEMCTL_CMD_STS;
+	I915_WRITE(MEMSWCTL, rgvswctl);
+}
+
+void ironlake_disable_drps(struct drm_device *dev)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	u32 rgvswctl;
+	u8 fstart;
+
+	/* Ack interrupts, disable EFC interrupt */
+	I915_WRITE(MEMINTREN, I915_READ(MEMINTREN) & ~MEMINT_EVAL_CHG_EN);
+	I915_WRITE(MEMINTRSTS, MEMINT_EVAL_CHG);
+	I915_WRITE(DEIER, I915_READ(DEIER) & ~DE_PCU_EVENT);
+	I915_WRITE(DEIIR, DE_PCU_EVENT);
+	I915_WRITE(DEIMR, I915_READ(DEIMR) | DE_PCU_EVENT);
+
+	/* Go back to the starting frequency */
+	fstart = (I915_READ(MEMMODECTL) & MEMMODE_FSTART_MASK) >>
+		MEMMODE_FSTART_SHIFT;
+	rgvswctl = (MEMCTL_CMD_CHFREQ << MEMCTL_CMD_SHIFT) |
+		(fstart << MEMCTL_FREQ_SHIFT) | MEMCTL_SFCAVM;
+	I915_WRITE(MEMSWCTL, rgvswctl);
+	msleep(1);
+	rgvswctl |= MEMCTL_CMD_STS;
+	I915_WRITE(MEMSWCTL, rgvswctl);
+	msleep(1);
+
+}
+
 void intel_init_clock_gating(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
@@ -4117,7 +4716,9 @@ void intel_init_clock_gating(struct drm_device *dev)
 	 * Disable clock gating reported to work incorrectly according to the
 	 * specs, but enable as much else as we can.
 	 */
-	if (IS_G4X(dev)) {
+	if (HAS_PCH_SPLIT(dev)) {
+		return;
+	} else if (IS_G4X(dev)) {
 		uint32_t dspclk_gate;
 		I915_WRITE(RENCLK_GATE_D1, 0);
 		I915_WRITE(RENCLK_GATE_D2, VF_UNIT_CLOCK_GATE_DISABLE |
@@ -4149,11 +4750,37 @@ void intel_init_clock_gating(struct drm_device *dev)
 		dstate |= DSTATE_PLL_D3_OFF | DSTATE_GFX_CLOCK_GATING |
 			DSTATE_DOT_CLOCK_GATING;
 		I915_WRITE(D_STATE, dstate);
-	} else if (IS_I855(dev) || IS_I865G(dev)) {
+	} else if (IS_I85X(dev) || IS_I865G(dev)) {
 		I915_WRITE(RENCLK_GATE_D1, SV_CLOCK_GATE_DISABLE);
 	} else if (IS_I830(dev)) {
 		I915_WRITE(DSPCLK_GATE_D, OVRUNIT_CLOCK_GATE_DISABLE);
 	}
+
+	/*
+	 * GPU can automatically power down the render unit if given a page
+	 * to save state.
+	 */
+	if (I915_HAS_RC6(dev) && drm_core_check_feature(dev, DRIVER_MODESET)) {
+		struct drm_i915_gem_object *obj_priv = NULL;
+
+		if (dev_priv->pwrctx) {
+			obj_priv = dev_priv->pwrctx->driver_private;
+		} else {
+			struct drm_gem_object *pwrctx;
+
+			pwrctx = intel_alloc_power_context(dev);
+			if (pwrctx) {
+				dev_priv->pwrctx = pwrctx;
+				obj_priv = pwrctx->driver_private;
+			}
+		}
+
+		if (obj_priv) {
+			I915_WRITE(PWRCTXA, obj_priv->gtt_offset | PWRCTX_EN);
+			I915_WRITE(MCHBAR_RENDER_STANDBY,
+				   I915_READ(MCHBAR_RENDER_STANDBY) & ~RCX_SW_EXIT);
+		}
+	}
 }
 
 /* Set up chip specific display functions */
@@ -4162,8 +4789,8 @@ static void intel_init_display(struct drm_device *dev)
 	struct drm_i915_private *dev_priv = dev->dev_private;
 
 	/* We always want a DPMS function */
-	if (IS_IGDNG(dev))
-		dev_priv->display.dpms = igdng_crtc_dpms;
+	if (HAS_PCH_SPLIT(dev))
+		dev_priv->display.dpms = ironlake_crtc_dpms;
 	else
 		dev_priv->display.dpms = i9xx_crtc_dpms;
 
@@ -4182,13 +4809,13 @@ static void intel_init_display(struct drm_device *dev)
 	}
 
 	/* Returns the core display clock speed */
-	if (IS_I945G(dev))
+	if (IS_I945G(dev) || (IS_G33(dev) && ! IS_PINEVIEW_M(dev)))
 		dev_priv->display.get_display_clock_speed =
 			i945_get_display_clock_speed;
 	else if (IS_I915G(dev))
 		dev_priv->display.get_display_clock_speed =
 			i915_get_display_clock_speed;
-	else if (IS_I945GM(dev) || IS_845G(dev) || IS_IGDGM(dev))
+	else if (IS_I945GM(dev) || IS_845G(dev) || IS_PINEVIEW_M(dev))
 		dev_priv->display.get_display_clock_speed =
 			i9xx_misc_get_display_clock_speed;
 	else if (IS_I915GM(dev))
@@ -4197,7 +4824,7 @@ static void intel_init_display(struct drm_device *dev)
 	else if (IS_I865G(dev))
 		dev_priv->display.get_display_clock_speed =
 			i865_get_display_clock_speed;
-	else if (IS_I855(dev))
+	else if (IS_I85X(dev))
 		dev_priv->display.get_display_clock_speed =
 			i855_get_display_clock_speed;
 	else /* 852, 830 */
@@ -4205,7 +4832,9 @@ static void intel_init_display(struct drm_device *dev)
 			i830_get_display_clock_speed;
 
 	/* For FIFO watermark updates */
-	if (IS_G4X(dev))
+	if (HAS_PCH_SPLIT(dev))
+		dev_priv->display.update_wm = NULL;
+	else if (IS_G4X(dev))
 		dev_priv->display.update_wm = g4x_update_wm;
 	else if (IS_I965G(dev))
 		dev_priv->display.update_wm = i965_update_wm;
@@ -4259,14 +4888,9 @@ void intel_modeset_init(struct drm_device *dev)
 		num_pipe = 2;
 	else
 		num_pipe = 1;
-	DRM_DEBUG("%d display pipe%s available.\n",
+	DRM_DEBUG_KMS("%d display pipe%s available.\n",
 		  num_pipe, num_pipe > 1 ? "s" : "");
 
-	if (IS_I85X(dev))
-		pci_read_config_word(dev->pdev, HPLLCC, &dev_priv->orig_clock);
-	else if (IS_I9XX(dev) || IS_G4X(dev))
-		pci_read_config_word(dev->pdev, GCFGC, &dev_priv->orig_clock);
-
 	for (i = 0; i < num_pipe; i++) {
 		intel_crtc_init(dev, i);
 	}
@@ -4275,9 +4899,21 @@ void intel_modeset_init(struct drm_device *dev)
 
 	intel_init_clock_gating(dev);
 
+	if (IS_IRONLAKE_M(dev))
+		ironlake_enable_drps(dev);
+
 	INIT_WORK(&dev_priv->idle_work, intel_idle_update);
 	setup_timer(&dev_priv->idle_timer, intel_gpu_idle_timer,
 		    (unsigned long)dev);
+
+	intel_setup_overlay(dev);
+
+	if (IS_PINEVIEW(dev) && !intel_get_cxsr_latency(IS_PINEVIEW_G(dev),
+							dev_priv->fsb_freq,
+							dev_priv->mem_freq))
+		DRM_INFO("failed to find known CxSR latency "
+			 "(found fsb freq %d, mem freq %d), disabling CxSR\n",
+			 dev_priv->fsb_freq, dev_priv->mem_freq);
 }
 
 void intel_modeset_cleanup(struct drm_device *dev)
@@ -4298,14 +4934,26 @@ void intel_modeset_cleanup(struct drm_device *dev)
 		del_timer_sync(&intel_crtc->idle_timer);
 	}
 
-	intel_increase_renderclock(dev, false);
 	del_timer_sync(&dev_priv->idle_timer);
 
-	mutex_unlock(&dev->struct_mutex);
-
 	if (dev_priv->display.disable_fbc)
 		dev_priv->display.disable_fbc(dev);
 
+	if (dev_priv->pwrctx) {
+		struct drm_i915_gem_object *obj_priv;
+
+		obj_priv = dev_priv->pwrctx->driver_private;
+		I915_WRITE(PWRCTXA, obj_priv->gtt_offset &~ PWRCTX_EN);
+		I915_READ(PWRCTXA);
+		i915_gem_object_unpin(dev_priv->pwrctx);
+		drm_gem_object_unreference(dev_priv->pwrctx);
+	}
+
+	if (IS_IRONLAKE_M(dev))
+		ironlake_disable_drps(dev);
+
+	mutex_unlock(&dev->struct_mutex);
+
 	drm_mode_config_cleanup(dev);
 }
 
diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c
index f4856a510476..3ef3a0d0edd0 100644
--- a/drivers/gpu/drm/i915/intel_dp.c
+++ b/drivers/gpu/drm/i915/intel_dp.c
@@ -33,7 +33,8 @@
 #include "intel_drv.h"
 #include "i915_drm.h"
 #include "i915_drv.h"
-#include "intel_dp.h"
+#include "drm_dp_helper.h"
+
 
 #define DP_LINK_STATUS_SIZE	6
 #define DP_LINK_CHECK_TIMEOUT	(10 * 1000)
@@ -124,9 +125,15 @@ intel_dp_link_clock(uint8_t link_bw)
 
 /* I think this is a fiction */
 static int
-intel_dp_link_required(int pixel_clock)
+intel_dp_link_required(struct drm_device *dev,
+		       struct intel_output *intel_output, int pixel_clock)
 {
-	return pixel_clock * 3;
+	struct drm_i915_private *dev_priv = dev->dev_private;
+
+	if (IS_eDP(intel_output))
+		return (pixel_clock * dev_priv->edp_bpp) / 8;
+	else
+		return pixel_clock * 3;
 }
 
 static int
@@ -137,7 +144,8 @@ intel_dp_mode_valid(struct drm_connector *connector,
 	int max_link_clock = intel_dp_link_clock(intel_dp_max_link_bw(intel_output));
 	int max_lanes = intel_dp_max_lane_count(intel_output);
 
-	if (intel_dp_link_required(mode->clock) > max_link_clock * max_lanes)
+	if (intel_dp_link_required(connector->dev, intel_output, mode->clock)
+			> max_link_clock * max_lanes)
 		return MODE_CLOCK_HIGH;
 
 	if (mode->clock < 10000)
@@ -223,8 +231,8 @@ intel_dp_aux_ch(struct intel_output *intel_output,
 	 */
 	if (IS_eDP(intel_output))
 		aux_clock_divider = 225; /* eDP input clock at 450Mhz */
-	else if (IS_IGDNG(dev))
-		aux_clock_divider = 62; /* IGDNG: input clock fixed at 125Mhz */
+	else if (HAS_PCH_SPLIT(dev))
+		aux_clock_divider = 62; /* IRL input clock fixed at 125Mhz */
 	else
 		aux_clock_divider = intel_hrawclk(dev) / 2;
 
@@ -282,7 +290,7 @@ intel_dp_aux_ch(struct intel_output *intel_output,
 	/* Timeouts occur when the device isn't connected, so they're
 	 * "normal" -- don't fill the kernel log with these */
 	if (status & DP_AUX_CH_CTL_TIME_OUT_ERROR) {
-		DRM_DEBUG("dp_aux_ch timeout status 0x%08x\n", status);
+		DRM_DEBUG_KMS("dp_aux_ch timeout status 0x%08x\n", status);
 		return -ETIMEDOUT;
 	}
 
@@ -382,17 +390,77 @@ intel_dp_aux_native_read(struct intel_output *intel_output,
 }
 
 static int
-intel_dp_i2c_aux_ch(struct i2c_adapter *adapter,
-		    uint8_t *send, int send_bytes,
-		    uint8_t *recv, int recv_bytes)
+intel_dp_i2c_aux_ch(struct i2c_adapter *adapter, int mode,
+		    uint8_t write_byte, uint8_t *read_byte)
 {
+	struct i2c_algo_dp_aux_data *algo_data = adapter->algo_data;
 	struct intel_dp_priv *dp_priv = container_of(adapter,
 						     struct intel_dp_priv,
 						     adapter);
 	struct intel_output *intel_output = dp_priv->intel_output;
+	uint16_t address = algo_data->address;
+	uint8_t msg[5];
+	uint8_t reply[2];
+	int msg_bytes;
+	int reply_bytes;
+	int ret;
+
+	/* Set up the command byte */
+	if (mode & MODE_I2C_READ)
+		msg[0] = AUX_I2C_READ << 4;
+	else
+		msg[0] = AUX_I2C_WRITE << 4;
+
+	if (!(mode & MODE_I2C_STOP))
+		msg[0] |= AUX_I2C_MOT << 4;
+
+	msg[1] = address >> 8;
+	msg[2] = address;
+
+	switch (mode) {
+	case MODE_I2C_WRITE:
+		msg[3] = 0;
+		msg[4] = write_byte;
+		msg_bytes = 5;
+		reply_bytes = 1;
+		break;
+	case MODE_I2C_READ:
+		msg[3] = 0;
+		msg_bytes = 4;
+		reply_bytes = 2;
+		break;
+	default:
+		msg_bytes = 3;
+		reply_bytes = 1;
+		break;
+	}
 
-	return intel_dp_aux_ch(intel_output,
-			       send, send_bytes, recv, recv_bytes);
+	for (;;) {
+	  ret = intel_dp_aux_ch(intel_output,
+				msg, msg_bytes,
+				reply, reply_bytes);
+		if (ret < 0) {
+			DRM_DEBUG_KMS("aux_ch failed %d\n", ret);
+			return ret;
+		}
+		switch (reply[0] & AUX_I2C_REPLY_MASK) {
+		case AUX_I2C_REPLY_ACK:
+			if (mode == MODE_I2C_READ) {
+				*read_byte = reply[1];
+			}
+			return reply_bytes - 1;
+		case AUX_I2C_REPLY_NACK:
+			DRM_DEBUG_KMS("aux_ch nack\n");
+			return -EREMOTEIO;
+		case AUX_I2C_REPLY_DEFER:
+			DRM_DEBUG_KMS("aux_ch defer\n");
+			udelay(100);
+			break;
+		default:
+			DRM_ERROR("aux_ch invalid reply 0x%02x\n", reply[0]);
+			return -EREMOTEIO;
+		}
+	}
 }
 
 static int
@@ -400,7 +468,7 @@ intel_dp_i2c_init(struct intel_output *intel_output, const char *name)
 {
 	struct intel_dp_priv   *dp_priv = intel_output->dev_priv;
 
-	DRM_ERROR("i2c_init %s\n", name);
+	DRM_DEBUG_KMS("i2c_init %s\n", name);
 	dp_priv->algo.running = false;
 	dp_priv->algo.address = 0;
 	dp_priv->algo.aux_ch = intel_dp_i2c_aux_ch;
@@ -431,11 +499,13 @@ intel_dp_mode_fixup(struct drm_encoder *encoder, struct drm_display_mode *mode,
 		for (clock = 0; clock <= max_clock; clock++) {
 			int link_avail = intel_dp_link_clock(bws[clock]) * lane_count;
 
-			if (intel_dp_link_required(mode->clock) <= link_avail) {
+			if (intel_dp_link_required(encoder->dev, intel_output, mode->clock)
+					<= link_avail) {
 				dp_priv->link_bw = bws[clock];
 				dp_priv->lane_count = lane_count;
 				adjusted_mode->clock = intel_dp_link_clock(dp_priv->link_bw);
-				DRM_DEBUG("Display port link bw %02x lane count %d clock %d\n",
+				DRM_DEBUG_KMS("Display port link bw %02x lane "
+						"count %d clock %d\n",
 				       dp_priv->link_bw, dp_priv->lane_count,
 				       adjusted_mode->clock);
 				return true;
@@ -514,7 +584,7 @@ intel_dp_set_m_n(struct drm_crtc *crtc, struct drm_display_mode *mode,
 	intel_dp_compute_m_n(3, lane_count,
 			     mode->clock, adjusted_mode->clock, &m_n);
 
-	if (IS_IGDNG(dev)) {
+	if (HAS_PCH_SPLIT(dev)) {
 		if (intel_crtc->pipe == 0) {
 			I915_WRITE(TRANSA_DATA_M1,
 				   ((m_n.tu - 1) << PIPE_GMCH_DATA_M_TU_SIZE_SHIFT) |
@@ -606,23 +676,23 @@ intel_dp_mode_set(struct drm_encoder *encoder, struct drm_display_mode *mode,
 	}
 }
 
-static void igdng_edp_backlight_on (struct drm_device *dev)
+static void ironlake_edp_backlight_on (struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	u32 pp;
 
-	DRM_DEBUG("\n");
+	DRM_DEBUG_KMS("\n");
 	pp = I915_READ(PCH_PP_CONTROL);
 	pp |= EDP_BLC_ENABLE;
 	I915_WRITE(PCH_PP_CONTROL, pp);
 }
 
-static void igdng_edp_backlight_off (struct drm_device *dev)
+static void ironlake_edp_backlight_off (struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	u32 pp;
 
-	DRM_DEBUG("\n");
+	DRM_DEBUG_KMS("\n");
 	pp = I915_READ(PCH_PP_CONTROL);
 	pp &= ~EDP_BLC_ENABLE;
 	I915_WRITE(PCH_PP_CONTROL, pp);
@@ -641,13 +711,13 @@ intel_dp_dpms(struct drm_encoder *encoder, int mode)
 		if (dp_reg & DP_PORT_EN) {
 			intel_dp_link_down(intel_output, dp_priv->DP);
 			if (IS_eDP(intel_output))
-				igdng_edp_backlight_off(dev);
+				ironlake_edp_backlight_off(dev);
 		}
 	} else {
 		if (!(dp_reg & DP_PORT_EN)) {
 			intel_dp_link_train(intel_output, dp_priv->DP, dp_priv->link_configuration);
 			if (IS_eDP(intel_output))
-				igdng_edp_backlight_on(dev);
+				ironlake_edp_backlight_on(dev);
 		}
 	}
 	dp_priv->dpms_mode = mode;
@@ -1010,7 +1080,7 @@ intel_dp_link_down(struct intel_output *intel_output, uint32_t DP)
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct intel_dp_priv *dp_priv = intel_output->dev_priv;
 
-	DRM_DEBUG("\n");
+	DRM_DEBUG_KMS("\n");
 
 	if (IS_eDP(intel_output)) {
 		DP &= ~DP_PLL_ENABLE;
@@ -1071,7 +1141,7 @@ intel_dp_check_link_status(struct intel_output *intel_output)
 }
 
 static enum drm_connector_status
-igdng_dp_detect(struct drm_connector *connector)
+ironlake_dp_detect(struct drm_connector *connector)
 {
 	struct intel_output *intel_output = to_intel_output(connector);
 	struct intel_dp_priv *dp_priv = intel_output->dev_priv;
@@ -1106,8 +1176,8 @@ intel_dp_detect(struct drm_connector *connector)
 
 	dp_priv->has_audio = false;
 
-	if (IS_IGDNG(dev))
-		return igdng_dp_detect(connector);
+	if (HAS_PCH_SPLIT(dev))
+		return ironlake_dp_detect(connector);
 
 	temp = I915_READ(PORT_HOTPLUG_EN);
 
@@ -1254,18 +1324,17 @@ intel_dp_init(struct drm_device *dev, int output_reg)
 	else
 		intel_output->type = INTEL_OUTPUT_DISPLAYPORT;
 
-	if (output_reg == DP_B)
+	if (output_reg == DP_B || output_reg == PCH_DP_B)
 		intel_output->clone_mask = (1 << INTEL_DP_B_CLONE_BIT);
-	else if (output_reg == DP_C)
+	else if (output_reg == DP_C || output_reg == PCH_DP_C)
 		intel_output->clone_mask = (1 << INTEL_DP_C_CLONE_BIT);
-	else if (output_reg == DP_D)
+	else if (output_reg == DP_D || output_reg == PCH_DP_D)
 		intel_output->clone_mask = (1 << INTEL_DP_D_CLONE_BIT);
 
-	if (IS_eDP(intel_output)) {
-		intel_output->crtc_mask = (1 << 1);
+	if (IS_eDP(intel_output))
 		intel_output->clone_mask = (1 << INTEL_EDP_CLONE_BIT);
-	} else
-		intel_output->crtc_mask = (1 << 0) | (1 << 1);
+
+	intel_output->crtc_mask = (1 << 0) | (1 << 1);
 	connector->interlace_allowed = true;
 	connector->doublescan_allowed = 0;
 
@@ -1290,14 +1359,20 @@ intel_dp_init(struct drm_device *dev, int output_reg)
 			break;
 		case DP_B:
 		case PCH_DP_B:
+			dev_priv->hotplug_supported_mask |=
+				HDMIB_HOTPLUG_INT_STATUS;
 			name = "DPDDC-B";
 			break;
 		case DP_C:
 		case PCH_DP_C:
+			dev_priv->hotplug_supported_mask |=
+				HDMIC_HOTPLUG_INT_STATUS;
 			name = "DPDDC-C";
 			break;
 		case DP_D:
 		case PCH_DP_D:
+			dev_priv->hotplug_supported_mask |=
+				HDMID_HOTPLUG_INT_STATUS;
 			name = "DPDDC-D";
 			break;
 	}
diff --git a/drivers/gpu/drm/i915/intel_dp.h b/drivers/gpu/drm/i915/intel_dp.h
deleted file mode 100644
index 2b38054d3b6d..000000000000
--- a/drivers/gpu/drm/i915/intel_dp.h
+++ /dev/null
@@ -1,144 +0,0 @@
-/*
- * Copyright © 2008 Keith Packard
- *
- * Permission to use, copy, modify, distribute, and sell this software and its
- * documentation for any purpose is hereby granted without fee, provided that
- * the above copyright notice appear in all copies and that both that copyright
- * notice and this permission notice appear in supporting documentation, and
- * that the name of the copyright holders not be used in advertising or
- * publicity pertaining to distribution of the software without specific,
- * written prior permission.  The copyright holders make no representations
- * about the suitability of this software for any purpose.  It is provided "as
- * is" without express or implied warranty.
- *
- * THE COPYRIGHT HOLDERS DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE,
- * INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO
- * EVENT SHALL THE COPYRIGHT HOLDERS BE LIABLE FOR ANY SPECIAL, INDIRECT OR
- * CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE,
- * DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER
- * TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE
- * OF THIS SOFTWARE.
- */
-
-#ifndef _INTEL_DP_H_
-#define _INTEL_DP_H_
-
-/* From the VESA DisplayPort spec */
-
-#define AUX_NATIVE_WRITE	0x8
-#define AUX_NATIVE_READ		0x9
-#define AUX_I2C_WRITE		0x0
-#define AUX_I2C_READ		0x1
-#define AUX_I2C_STATUS		0x2
-#define AUX_I2C_MOT		0x4
-
-#define AUX_NATIVE_REPLY_ACK	(0x0 << 4)
-#define AUX_NATIVE_REPLY_NACK	(0x1 << 4)
-#define AUX_NATIVE_REPLY_DEFER	(0x2 << 4)
-#define AUX_NATIVE_REPLY_MASK	(0x3 << 4)
-
-#define AUX_I2C_REPLY_ACK	(0x0 << 6)
-#define AUX_I2C_REPLY_NACK	(0x1 << 6)
-#define AUX_I2C_REPLY_DEFER	(0x2 << 6)
-#define AUX_I2C_REPLY_MASK	(0x3 << 6)
-
-/* AUX CH addresses */
-#define	DP_LINK_BW_SET		0x100
-# define DP_LINK_BW_1_62		    0x06
-# define DP_LINK_BW_2_7			    0x0a
-
-#define DP_LANE_COUNT_SET	0x101
-# define DP_LANE_COUNT_MASK		    0x0f
-# define DP_LANE_COUNT_ENHANCED_FRAME_EN    (1 << 7)
-
-#define DP_TRAINING_PATTERN_SET	0x102
-
-# define DP_TRAINING_PATTERN_DISABLE	    0
-# define DP_TRAINING_PATTERN_1		    1
-# define DP_TRAINING_PATTERN_2		    2
-# define DP_TRAINING_PATTERN_MASK	    0x3
-
-# define DP_LINK_QUAL_PATTERN_DISABLE	    (0 << 2)
-# define DP_LINK_QUAL_PATTERN_D10_2	    (1 << 2)
-# define DP_LINK_QUAL_PATTERN_ERROR_RATE    (2 << 2)
-# define DP_LINK_QUAL_PATTERN_PRBS7	    (3 << 2)
-# define DP_LINK_QUAL_PATTERN_MASK	    (3 << 2)
-
-# define DP_RECOVERED_CLOCK_OUT_EN	    (1 << 4)
-# define DP_LINK_SCRAMBLING_DISABLE	    (1 << 5)
-
-# define DP_SYMBOL_ERROR_COUNT_BOTH	    (0 << 6)
-# define DP_SYMBOL_ERROR_COUNT_DISPARITY    (1 << 6)
-# define DP_SYMBOL_ERROR_COUNT_SYMBOL	    (2 << 6)
-# define DP_SYMBOL_ERROR_COUNT_MASK	    (3 << 6)
-
-#define DP_TRAINING_LANE0_SET		    0x103
-#define DP_TRAINING_LANE1_SET		    0x104
-#define DP_TRAINING_LANE2_SET		    0x105
-#define DP_TRAINING_LANE3_SET		    0x106
-
-# define DP_TRAIN_VOLTAGE_SWING_MASK	    0x3
-# define DP_TRAIN_VOLTAGE_SWING_SHIFT	    0
-# define DP_TRAIN_MAX_SWING_REACHED	    (1 << 2)
-# define DP_TRAIN_VOLTAGE_SWING_400	    (0 << 0)
-# define DP_TRAIN_VOLTAGE_SWING_600	    (1 << 0)
-# define DP_TRAIN_VOLTAGE_SWING_800	    (2 << 0)
-# define DP_TRAIN_VOLTAGE_SWING_1200	    (3 << 0)
-
-# define DP_TRAIN_PRE_EMPHASIS_MASK	    (3 << 3)
-# define DP_TRAIN_PRE_EMPHASIS_0	    (0 << 3)
-# define DP_TRAIN_PRE_EMPHASIS_3_5	    (1 << 3)
-# define DP_TRAIN_PRE_EMPHASIS_6	    (2 << 3)
-# define DP_TRAIN_PRE_EMPHASIS_9_5	    (3 << 3)
-
-# define DP_TRAIN_PRE_EMPHASIS_SHIFT	    3
-# define DP_TRAIN_MAX_PRE_EMPHASIS_REACHED  (1 << 5)
-
-#define DP_DOWNSPREAD_CTRL		    0x107
-# define DP_SPREAD_AMP_0_5		    (1 << 4)
-
-#define DP_MAIN_LINK_CHANNEL_CODING_SET	    0x108
-# define DP_SET_ANSI_8B10B		    (1 << 0)
-
-#define DP_LANE0_1_STATUS		    0x202
-#define DP_LANE2_3_STATUS		    0x203
-
-# define DP_LANE_CR_DONE		    (1 << 0)
-# define DP_LANE_CHANNEL_EQ_DONE	    (1 << 1)
-# define DP_LANE_SYMBOL_LOCKED		    (1 << 2)
-
-#define DP_LANE_ALIGN_STATUS_UPDATED	    0x204
-
-#define DP_INTERLANE_ALIGN_DONE		    (1 << 0)
-#define DP_DOWNSTREAM_PORT_STATUS_CHANGED   (1 << 6)
-#define DP_LINK_STATUS_UPDATED		    (1 << 7)
-
-#define DP_SINK_STATUS			    0x205
-
-#define DP_RECEIVE_PORT_0_STATUS	    (1 << 0)
-#define DP_RECEIVE_PORT_1_STATUS	    (1 << 1)
-
-#define DP_ADJUST_REQUEST_LANE0_1	    0x206
-#define DP_ADJUST_REQUEST_LANE2_3	    0x207
-
-#define DP_ADJUST_VOLTAGE_SWING_LANE0_MASK  0x03
-#define DP_ADJUST_VOLTAGE_SWING_LANE0_SHIFT 0
-#define DP_ADJUST_PRE_EMPHASIS_LANE0_MASK   0x0c
-#define DP_ADJUST_PRE_EMPHASIS_LANE0_SHIFT  2
-#define DP_ADJUST_VOLTAGE_SWING_LANE1_MASK  0x30
-#define DP_ADJUST_VOLTAGE_SWING_LANE1_SHIFT 4
-#define DP_ADJUST_PRE_EMPHASIS_LANE1_MASK   0xc0
-#define DP_ADJUST_PRE_EMPHASIS_LANE1_SHIFT  6
-
-struct i2c_algo_dp_aux_data {
-	bool running;
-	u16 address;
-	int (*aux_ch) (struct i2c_adapter *adapter,
-		       uint8_t *send, int send_bytes,
-		       uint8_t *recv, int recv_bytes);
-};
-
-int
-i2c_dp_aux_add_bus(struct i2c_adapter *adapter);
-
-#endif /* _INTEL_DP_H_ */
diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h
index 8aa4b7f30daa..3a467ca57857 100644
--- a/drivers/gpu/drm/i915/intel_drv.h
+++ b/drivers/gpu/drm/i915/intel_drv.h
@@ -110,6 +110,32 @@ struct intel_output {
 	int clone_mask;
 };
 
+struct intel_crtc;
+struct intel_overlay {
+	struct drm_device *dev;
+	struct intel_crtc *crtc;
+	struct drm_i915_gem_object *vid_bo;
+	struct drm_i915_gem_object *old_vid_bo;
+	int active;
+	int pfit_active;
+	u32 pfit_vscale_ratio; /* shifted-point number, (1<<12) == 1.0 */
+	u32 color_key;
+	u32 brightness, contrast, saturation;
+	u32 old_xscale, old_yscale;
+	/* register access */
+	u32 flip_addr;
+	struct drm_i915_gem_object *reg_bo;
+	void *virt_addr;
+	/* flip handling */
+	uint32_t last_flip_req;
+	int hw_wedged;
+#define HW_WEDGED		1
+#define NEEDS_WAIT_FOR_FLIP	2
+#define RELEASE_OLD_VID		3
+#define SWITCH_OFF_STAGE_1	4
+#define SWITCH_OFF_STAGE_2	5
+};
+
 struct intel_crtc {
 	struct drm_crtc base;
 	enum pipe pipe;
@@ -121,6 +147,8 @@ struct intel_crtc {
 	bool busy; /* is scanout buffer being updated frequently? */
 	struct timer_list idle_timer;
 	bool lowfreq_avail;
+	struct intel_overlay *overlay;
+	struct intel_unpin_work *unpin_work;
 };
 
 #define to_intel_crtc(x) container_of(x, struct intel_crtc, base)
@@ -134,6 +162,8 @@ void intel_i2c_destroy(struct i2c_adapter *adapter);
 int intel_ddc_get_modes(struct intel_output *intel_output);
 extern bool intel_ddc_probe(struct intel_output *intel_output);
 void intel_i2c_quirk_set(struct drm_device *dev, bool enable);
+void intel_i2c_reset_gmbus(struct drm_device *dev);
+
 extern void intel_crt_init(struct drm_device *dev);
 extern void intel_hdmi_init(struct drm_device *dev, int sdvox_reg);
 extern bool intel_sdvo_init(struct drm_device *dev, int output_device);
@@ -148,6 +178,7 @@ intel_dp_set_m_n(struct drm_crtc *crtc, struct drm_display_mode *mode,
 extern void intel_edp_link_config (struct intel_output *, int *, int *);
 
 
+extern int intel_panel_fitter_pipe (struct drm_device *dev);
 extern void intel_crtc_load_lut(struct drm_crtc *crtc);
 extern void intel_encoder_prepare (struct drm_encoder *encoder);
 extern void intel_encoder_commit (struct drm_encoder *encoder);
@@ -175,10 +206,27 @@ extern int intelfb_resize(struct drm_device *dev, struct drm_crtc *crtc);
 extern void intelfb_restore(void);
 extern void intel_crtc_fb_gamma_set(struct drm_crtc *crtc, u16 red, u16 green,
 				    u16 blue, int regno);
+extern void intel_crtc_fb_gamma_get(struct drm_crtc *crtc, u16 *red, u16 *green,
+				    u16 *blue, int regno);
+extern void intel_init_clock_gating(struct drm_device *dev);
+extern void ironlake_enable_drps(struct drm_device *dev);
+extern void ironlake_disable_drps(struct drm_device *dev);
 
 extern int intel_framebuffer_create(struct drm_device *dev,
 				    struct drm_mode_fb_cmd *mode_cmd,
 				    struct drm_framebuffer **fb,
 				    struct drm_gem_object *obj);
 
+extern void intel_prepare_page_flip(struct drm_device *dev, int plane);
+extern void intel_finish_page_flip(struct drm_device *dev, int pipe);
+
+extern void intel_setup_overlay(struct drm_device *dev);
+extern void intel_cleanup_overlay(struct drm_device *dev);
+extern int intel_overlay_switch_off(struct intel_overlay *overlay);
+extern int intel_overlay_recover_from_interrupt(struct intel_overlay *overlay,
+						int interruptible);
+extern int intel_overlay_put_image(struct drm_device *dev, void *data,
+				   struct drm_file *file_priv);
+extern int intel_overlay_attrs(struct drm_device *dev, void *data,
+			       struct drm_file *file_priv);
 #endif /* __INTEL_DRV_H__ */
diff --git a/drivers/gpu/drm/i915/intel_fb.c b/drivers/gpu/drm/i915/intel_fb.c
index 7ba4a232a97f..8cd791dc5b29 100644
--- a/drivers/gpu/drm/i915/intel_fb.c
+++ b/drivers/gpu/drm/i915/intel_fb.c
@@ -35,6 +35,7 @@
 #include <linux/delay.h>
 #include <linux/fb.h>
 #include <linux/init.h>
+#include <linux/vga_switcheroo.h>
 
 #include "drmP.h"
 #include "drm.h"
@@ -60,15 +61,17 @@ static struct fb_ops intelfb_ops = {
 	.fb_imageblit = cfb_imageblit,
 	.fb_pan_display = drm_fb_helper_pan_display,
 	.fb_blank = drm_fb_helper_blank,
+	.fb_setcmap = drm_fb_helper_setcmap,
 };
 
 static struct drm_fb_helper_funcs intel_fb_helper_funcs = {
 	.gamma_set = intel_crtc_fb_gamma_set,
+	.gamma_get = intel_crtc_fb_gamma_get,
 };
 
 
 /**
- * Curretly it is assumed that the old framebuffer is reused.
+ * Currently it is assumed that the old framebuffer is reused.
  *
  * LOCKING
  * caller should hold the mode config lock.
@@ -110,6 +113,7 @@ EXPORT_SYMBOL(intelfb_resize);
 static int intelfb_create(struct drm_device *dev, uint32_t fb_width,
 			  uint32_t fb_height, uint32_t surface_width,
 			  uint32_t surface_height,
+			  uint32_t surface_depth, uint32_t surface_bpp,
 			  struct drm_framebuffer **fb_p)
 {
 	struct fb_info *info;
@@ -122,12 +126,16 @@ static int intelfb_create(struct drm_device *dev, uint32_t fb_width,
 	struct device *device = &dev->pdev->dev;
 	int size, ret, mmio_bar = IS_I9XX(dev) ? 0 : 1;
 
+	/* we don't do packed 24bpp */
+	if (surface_bpp == 24)
+		surface_bpp = 32;
+
 	mode_cmd.width = surface_width;
 	mode_cmd.height = surface_height;
 
-	mode_cmd.bpp = 32;
+	mode_cmd.bpp = surface_bpp;
 	mode_cmd.pitch = ALIGN(mode_cmd.width * ((mode_cmd.bpp + 1) / 8), 64);
-	mode_cmd.depth = 24;
+	mode_cmd.depth = surface_depth;
 
 	size = mode_cmd.pitch * mode_cmd.height;
 	size = ALIGN(size, PAGE_SIZE);
@@ -141,7 +149,7 @@ static int intelfb_create(struct drm_device *dev, uint32_t fb_width,
 
 	mutex_lock(&dev->struct_mutex);
 
-	ret = i915_gem_object_pin(fbo, PAGE_SIZE);
+	ret = i915_gem_object_pin(fbo, 64*1024);
 	if (ret) {
 		DRM_ERROR("failed to pin fb: %d\n", ret);
 		goto out_unref;
@@ -205,7 +213,7 @@ static int intelfb_create(struct drm_device *dev, uint32_t fb_width,
 
 //	memset(info->screen_base, 0, size);
 
-	drm_fb_helper_fill_fix(info, fb->pitch);
+	drm_fb_helper_fill_fix(info, fb->pitch, fb->depth);
 	drm_fb_helper_fill_var(info, fb, fb_width, fb_height);
 
 	/* FIXME: we really shouldn't expose mmio space at all */
@@ -223,10 +231,12 @@ static int intelfb_create(struct drm_device *dev, uint32_t fb_width,
 	par->intel_fb = intel_fb;
 
 	/* To allow resizeing without swapping buffers */
-	DRM_DEBUG("allocated %dx%d fb: 0x%08x, bo %p\n", intel_fb->base.width,
-		  intel_fb->base.height, obj_priv->gtt_offset, fbo);
+	DRM_DEBUG_KMS("allocated %dx%d fb: 0x%08x, bo %p\n",
+			intel_fb->base.width, intel_fb->base.height,
+			obj_priv->gtt_offset, fbo);
 
 	mutex_unlock(&dev->struct_mutex);
+	vga_switcheroo_client_fb_set(dev->pdev, info);
 	return 0;
 
 out_unpin:
@@ -242,8 +252,8 @@ int intelfb_probe(struct drm_device *dev)
 {
 	int ret;
 
-	DRM_DEBUG("\n");
-	ret = drm_fb_helper_single_fb_probe(dev, intelfb_create);
+	DRM_DEBUG_KMS("\n");
+	ret = drm_fb_helper_single_fb_probe(dev, 32, intelfb_create);
 	return ret;
 }
 EXPORT_SYMBOL(intelfb_probe);
diff --git a/drivers/gpu/drm/i915/intel_hdmi.c b/drivers/gpu/drm/i915/intel_hdmi.c
index fa304e136010..a30f8bfc1985 100644
--- a/drivers/gpu/drm/i915/intel_hdmi.c
+++ b/drivers/gpu/drm/i915/intel_hdmi.c
@@ -77,14 +77,32 @@ static void intel_hdmi_dpms(struct drm_encoder *encoder, int mode)
 	struct intel_hdmi_priv *hdmi_priv = intel_output->dev_priv;
 	u32 temp;
 
-	if (mode != DRM_MODE_DPMS_ON) {
-		temp = I915_READ(hdmi_priv->sdvox_reg);
+	temp = I915_READ(hdmi_priv->sdvox_reg);
+
+	/* HW workaround, need to toggle enable bit off and on for 12bpc, but
+	 * we do this anyway which shows more stable in testing.
+	 */
+	if (HAS_PCH_SPLIT(dev)) {
 		I915_WRITE(hdmi_priv->sdvox_reg, temp & ~SDVO_ENABLE);
+		POSTING_READ(hdmi_priv->sdvox_reg);
+	}
+
+	if (mode != DRM_MODE_DPMS_ON) {
+		temp &= ~SDVO_ENABLE;
 	} else {
-		temp = I915_READ(hdmi_priv->sdvox_reg);
-		I915_WRITE(hdmi_priv->sdvox_reg, temp | SDVO_ENABLE);
+		temp |= SDVO_ENABLE;
 	}
+
+	I915_WRITE(hdmi_priv->sdvox_reg, temp);
 	POSTING_READ(hdmi_priv->sdvox_reg);
+
+	/* HW workaround, need to write this twice for issue that may result
+	 * in first write getting masked.
+	 */
+	if (HAS_PCH_SPLIT(dev)) {
+		I915_WRITE(hdmi_priv->sdvox_reg, temp);
+		POSTING_READ(hdmi_priv->sdvox_reg);
+	}
 }
 
 static void intel_hdmi_save(struct drm_connector *connector)
@@ -207,7 +225,6 @@ static const struct drm_encoder_funcs intel_hdmi_enc_funcs = {
 	.destroy = intel_hdmi_enc_destroy,
 };
 
-
 void intel_hdmi_init(struct drm_device *dev, int sdvox_reg)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
@@ -223,7 +240,7 @@ void intel_hdmi_init(struct drm_device *dev, int sdvox_reg)
 
 	connector = &intel_output->base;
 	drm_connector_init(dev, connector, &intel_hdmi_connector_funcs,
-			   DRM_MODE_CONNECTOR_DVID);
+			   DRM_MODE_CONNECTOR_HDMIA);
 	drm_connector_helper_add(connector, &intel_hdmi_connector_helper_funcs);
 
 	intel_output->type = INTEL_OUTPUT_HDMI;
@@ -236,21 +253,26 @@ void intel_hdmi_init(struct drm_device *dev, int sdvox_reg)
 	if (sdvox_reg == SDVOB) {
 		intel_output->clone_mask = (1 << INTEL_HDMIB_CLONE_BIT);
 		intel_output->ddc_bus = intel_i2c_create(dev, GPIOE, "HDMIB");
+		dev_priv->hotplug_supported_mask |= HDMIB_HOTPLUG_INT_STATUS;
 	} else if (sdvox_reg == SDVOC) {
 		intel_output->clone_mask = (1 << INTEL_HDMIC_CLONE_BIT);
 		intel_output->ddc_bus = intel_i2c_create(dev, GPIOD, "HDMIC");
+		dev_priv->hotplug_supported_mask |= HDMIC_HOTPLUG_INT_STATUS;
 	} else if (sdvox_reg == HDMIB) {
 		intel_output->clone_mask = (1 << INTEL_HDMID_CLONE_BIT);
 		intel_output->ddc_bus = intel_i2c_create(dev, PCH_GPIOE,
 								"HDMIB");
+		dev_priv->hotplug_supported_mask |= HDMIB_HOTPLUG_INT_STATUS;
 	} else if (sdvox_reg == HDMIC) {
 		intel_output->clone_mask = (1 << INTEL_HDMIE_CLONE_BIT);
 		intel_output->ddc_bus = intel_i2c_create(dev, PCH_GPIOD,
 								"HDMIC");
+		dev_priv->hotplug_supported_mask |= HDMIC_HOTPLUG_INT_STATUS;
 	} else if (sdvox_reg == HDMID) {
 		intel_output->clone_mask = (1 << INTEL_HDMIF_CLONE_BIT);
 		intel_output->ddc_bus = intel_i2c_create(dev, PCH_GPIOF,
 								"HDMID");
+		dev_priv->hotplug_supported_mask |= HDMID_HOTPLUG_INT_STATUS;
 	}
 	if (!intel_output->ddc_bus)
 		goto err_connector;
diff --git a/drivers/gpu/drm/i915/intel_i2c.c b/drivers/gpu/drm/i915/intel_i2c.c
index c7eab724c418..fcc753ca5d94 100644
--- a/drivers/gpu/drm/i915/intel_i2c.c
+++ b/drivers/gpu/drm/i915/intel_i2c.c
@@ -39,7 +39,7 @@ void intel_i2c_quirk_set(struct drm_device *dev, bool enable)
 	struct drm_i915_private *dev_priv = dev->dev_private;
 
 	/* When using bit bashing for I2C, this bit needs to be set to 1 */
-	if (!IS_IGD(dev))
+	if (!IS_PINEVIEW(dev))
 		return;
 	if (enable)
 		I915_WRITE(DSPCLK_GATE_D,
@@ -118,6 +118,23 @@ static void set_data(void *data, int state_high)
 	udelay(I2C_RISEFALL_TIME); /* wait for the line to change state */
 }
 
+/* Clears the GMBUS setup.  Our driver doesn't make use of the GMBUS I2C
+ * engine, but if the BIOS leaves it enabled, then that can break our use
+ * of the bit-banging I2C interfaces.  This is notably the case with the
+ * Mac Mini in EFI mode.
+ */
+void
+intel_i2c_reset_gmbus(struct drm_device *dev)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+
+	if (HAS_PCH_SPLIT(dev)) {
+		I915_WRITE(PCH_GMBUS0, 0);
+	} else {
+		I915_WRITE(GMBUS0, 0);
+	}
+}
+
 /**
  * intel_i2c_create - instantiate an Intel i2c bus using the specified GPIO reg
  * @dev: DRM device
@@ -168,6 +185,8 @@ struct i2c_adapter *intel_i2c_create(struct drm_device *dev, const u32 reg,
 	if(i2c_bit_add_bus(&chan->adapter))
 		goto out_free;
 
+	intel_i2c_reset_gmbus(dev);
+
 	/* JJJ:  raise SCL and SDA? */
 	intel_i2c_quirk_set(dev, true);
 	set_data(chan, 1);
diff --git a/drivers/gpu/drm/i915/intel_lvds.c b/drivers/gpu/drm/i915/intel_lvds.c
index 98ae3d73577e..14e516fdc2dd 100644
--- a/drivers/gpu/drm/i915/intel_lvds.c
+++ b/drivers/gpu/drm/i915/intel_lvds.c
@@ -56,7 +56,7 @@ static void intel_lvds_set_backlight(struct drm_device *dev, int level)
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	u32 blc_pwm_ctl, reg;
 
-	if (IS_IGDNG(dev))
+	if (HAS_PCH_SPLIT(dev))
 		reg = BLC_PWM_CPU_CTL;
 	else
 		reg = BLC_PWM_CTL;
@@ -74,7 +74,7 @@ static u32 intel_lvds_get_max_backlight(struct drm_device *dev)
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	u32 reg;
 
-	if (IS_IGDNG(dev))
+	if (HAS_PCH_SPLIT(dev))
 		reg = BLC_PWM_PCH_CTL2;
 	else
 		reg = BLC_PWM_CTL;
@@ -89,17 +89,22 @@ static u32 intel_lvds_get_max_backlight(struct drm_device *dev)
 static void intel_lvds_set_power(struct drm_device *dev, bool on)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
-	u32 pp_status, ctl_reg, status_reg;
+	u32 pp_status, ctl_reg, status_reg, lvds_reg;
 
-	if (IS_IGDNG(dev)) {
+	if (HAS_PCH_SPLIT(dev)) {
 		ctl_reg = PCH_PP_CONTROL;
 		status_reg = PCH_PP_STATUS;
+		lvds_reg = PCH_LVDS;
 	} else {
 		ctl_reg = PP_CONTROL;
 		status_reg = PP_STATUS;
+		lvds_reg = LVDS;
 	}
 
 	if (on) {
+		I915_WRITE(lvds_reg, I915_READ(lvds_reg) | LVDS_PORT_EN);
+		POSTING_READ(lvds_reg);
+
 		I915_WRITE(ctl_reg, I915_READ(ctl_reg) |
 			   POWER_TARGET_ON);
 		do {
@@ -115,6 +120,9 @@ static void intel_lvds_set_power(struct drm_device *dev, bool on)
 		do {
 			pp_status = I915_READ(status_reg);
 		} while (pp_status & PP_ON);
+
+		I915_WRITE(lvds_reg, I915_READ(lvds_reg) & ~LVDS_PORT_EN);
+		POSTING_READ(lvds_reg);
 	}
 }
 
@@ -137,7 +145,7 @@ static void intel_lvds_save(struct drm_connector *connector)
 	u32 pp_on_reg, pp_off_reg, pp_ctl_reg, pp_div_reg;
 	u32 pwm_ctl_reg;
 
-	if (IS_IGDNG(dev)) {
+	if (HAS_PCH_SPLIT(dev)) {
 		pp_on_reg = PCH_PP_ON_DELAYS;
 		pp_off_reg = PCH_PP_OFF_DELAYS;
 		pp_ctl_reg = PCH_PP_CONTROL;
@@ -174,7 +182,7 @@ static void intel_lvds_restore(struct drm_connector *connector)
 	u32 pp_on_reg, pp_off_reg, pp_ctl_reg, pp_div_reg;
 	u32 pwm_ctl_reg;
 
-	if (IS_IGDNG(dev)) {
+	if (HAS_PCH_SPLIT(dev)) {
 		pp_on_reg = PCH_PP_ON_DELAYS;
 		pp_off_reg = PCH_PP_OFF_DELAYS;
 		pp_ctl_reg = PCH_PP_CONTROL;
@@ -297,7 +305,7 @@ static bool intel_lvds_mode_fixup(struct drm_encoder *encoder,
 	}
 
 	/* full screen scale for now */
-	if (IS_IGDNG(dev))
+	if (HAS_PCH_SPLIT(dev))
 		goto out;
 
 	/* 965+ wants fuzzy fitting */
@@ -327,7 +335,7 @@ static bool intel_lvds_mode_fixup(struct drm_encoder *encoder,
 	 * to register description and PRM.
 	 * Change the value here to see the borders for debugging
 	 */
-	if (!IS_IGDNG(dev)) {
+	if (!HAS_PCH_SPLIT(dev)) {
 		I915_WRITE(BCLRPAT_A, 0);
 		I915_WRITE(BCLRPAT_B, 0);
 	}
@@ -380,7 +388,7 @@ static bool intel_lvds_mode_fixup(struct drm_encoder *encoder,
 				adjusted_mode->crtc_vblank_start + vsync_pos;
 		/* keep the vsync width constant */
 		adjusted_mode->crtc_vsync_end =
-				adjusted_mode->crtc_vblank_start + vsync_width;
+				adjusted_mode->crtc_vsync_start + vsync_width;
 		border = 1;
 		break;
 	case DRM_MODE_SCALE_ASPECT:
@@ -526,6 +534,14 @@ out:
 	lvds_priv->pfit_control = pfit_control;
 	lvds_priv->pfit_pgm_ratios = pfit_pgm_ratios;
 	/*
+	 * When there exists the border, it means that the LVDS_BORDR
+	 * should be enabled.
+	 */
+	if (border)
+		dev_priv->lvds_border_bits |= LVDS_BORDER_ENABLE;
+	else
+		dev_priv->lvds_border_bits &= ~(LVDS_BORDER_ENABLE);
+	/*
 	 * XXX: It would be nice to support lower refresh rates on the
 	 * panels to reduce power consumption, and perhaps match the
 	 * user's requested refresh rate.
@@ -540,7 +556,7 @@ static void intel_lvds_prepare(struct drm_encoder *encoder)
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	u32 reg;
 
-	if (IS_IGDNG(dev))
+	if (HAS_PCH_SPLIT(dev))
 		reg = BLC_PWM_CPU_CTL;
 	else
 		reg = BLC_PWM_CTL;
@@ -579,7 +595,7 @@ static void intel_lvds_mode_set(struct drm_encoder *encoder,
 	 * settings.
 	 */
 
-	if (IS_IGDNG(dev))
+	if (HAS_PCH_SPLIT(dev))
 		return;
 
 	/*
@@ -594,12 +610,47 @@ static void intel_lvds_mode_set(struct drm_encoder *encoder,
 /* Some lid devices report incorrect lid status, assume they're connected */
 static const struct dmi_system_id bad_lid_status[] = {
 	{
+		.ident = "Compaq nx9020",
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "Hewlett-Packard"),
+			DMI_MATCH(DMI_BOARD_NAME, "3084"),
+		},
+	},
+	{
+		.ident = "Samsung SX20S",
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "Samsung Electronics"),
+			DMI_MATCH(DMI_BOARD_NAME, "SX20S"),
+		},
+	},
+	{
 		.ident = "Aspire One",
 		.matches = {
 			DMI_MATCH(DMI_SYS_VENDOR, "Acer"),
 			DMI_MATCH(DMI_PRODUCT_NAME, "Aspire one"),
 		},
 	},
+	{
+		.ident = "Aspire 1810T",
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "Acer"),
+			DMI_MATCH(DMI_PRODUCT_NAME, "Aspire 1810T"),
+		},
+	},
+	{
+		.ident = "PC-81005",
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "MALATA"),
+			DMI_MATCH(DMI_PRODUCT_NAME, "PC-81005"),
+		},
+	},
+	{
+		.ident = "Clevo M5x0N",
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "CLEVO Co."),
+			DMI_MATCH(DMI_BOARD_NAME, "M5x0N"),
+		},
+	},
 	{ }
 };
 
@@ -612,9 +663,16 @@ static const struct dmi_system_id bad_lid_status[] = {
  */
 static enum drm_connector_status intel_lvds_detect(struct drm_connector *connector)
 {
+	struct drm_device *dev = connector->dev;
 	enum drm_connector_status status = connector_status_connected;
 
-	if (!acpi_lid_open() && !dmi_check_system(bad_lid_status))
+	/* ACPI lid methods were generally unreliable in this generation, so
+	 * don't even bother.
+	 */
+	if (IS_GEN2(dev))
+		return connector_status_connected;
+
+	if (!dmi_check_system(bad_lid_status) && !acpi_lid_open())
 		status = connector_status_disconnected;
 
 	return status;
@@ -656,20 +714,42 @@ static int intel_lvds_get_modes(struct drm_connector *connector)
 	return 0;
 }
 
+/*
+ * Lid events. Note the use of 'modeset_on_lid':
+ *  - we set it on lid close, and reset it on open
+ *  - we use it as a "only once" bit (ie we ignore
+ *    duplicate events where it was already properly
+ *    set/reset)
+ *  - the suspend/resume paths will also set it to
+ *    zero, since they restore the mode ("lid open").
+ */
 static int intel_lid_notify(struct notifier_block *nb, unsigned long val,
 			    void *unused)
 {
 	struct drm_i915_private *dev_priv =
 		container_of(nb, struct drm_i915_private, lid_notifier);
 	struct drm_device *dev = dev_priv->dev;
+	struct drm_connector *connector = dev_priv->int_lvds_connector;
 
-	if (acpi_lid_open() && !dev_priv->suspended) {
-		mutex_lock(&dev->mode_config.mutex);
-		drm_helper_resume_force_mode(dev);
-		mutex_unlock(&dev->mode_config.mutex);
+	/*
+	 * check and update the status of LVDS connector after receiving
+	 * the LID nofication event.
+	 */
+	if (connector)
+		connector->status = connector->funcs->detect(connector);
+	if (!acpi_lid_open()) {
+		dev_priv->modeset_on_lid = 1;
+		return NOTIFY_OK;
 	}
 
-	drm_sysfs_hotplug_event(dev_priv->dev);
+	if (!dev_priv->modeset_on_lid)
+		return NOTIFY_OK;
+
+	dev_priv->modeset_on_lid = 0;
+
+	mutex_lock(&dev->mode_config.mutex);
+	drm_helper_resume_force_mode(dev);
+	mutex_unlock(&dev->mode_config.mutex);
 
 	return NOTIFY_OK;
 }
@@ -831,64 +911,101 @@ static const struct dmi_system_id intel_no_lvds[] = {
 	{ }	/* terminating entry */
 };
 
-#ifdef CONFIG_ACPI
-/*
- * check_lid_device -- check whether @handle is an ACPI LID device.
- * @handle: ACPI device handle
- * @level : depth in the ACPI namespace tree
- * @context: the number of LID device when we find the device
- * @rv: a return value to fill if desired (Not use)
+/**
+ * intel_find_lvds_downclock - find the reduced downclock for LVDS in EDID
+ * @dev: drm device
+ * @connector: LVDS connector
+ *
+ * Find the reduced downclock for LVDS in EDID.
  */
-static acpi_status
-check_lid_device(acpi_handle handle, u32 level, void *context,
-			void **return_value)
+static void intel_find_lvds_downclock(struct drm_device *dev,
+				struct drm_connector *connector)
 {
-	struct acpi_device *acpi_dev;
-	int *lid_present = context;
-
-	acpi_dev = NULL;
-	/* Get the acpi device for device handle */
-	if (acpi_bus_get_device(handle, &acpi_dev) || !acpi_dev) {
-		/* If there is no ACPI device for handle, return */
-		return AE_OK;
-	}
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct drm_display_mode *scan, *panel_fixed_mode;
+	int temp_downclock;
 
-	if (!strncmp(acpi_device_hid(acpi_dev), "PNP0C0D", 7))
-		*lid_present = 1;
+	panel_fixed_mode = dev_priv->panel_fixed_mode;
+	temp_downclock = panel_fixed_mode->clock;
 
-	return AE_OK;
+	mutex_lock(&dev->mode_config.mutex);
+	list_for_each_entry(scan, &connector->probed_modes, head) {
+		/*
+		 * If one mode has the same resolution with the fixed_panel
+		 * mode while they have the different refresh rate, it means
+		 * that the reduced downclock is found for the LVDS. In such
+		 * case we can set the different FPx0/1 to dynamically select
+		 * between low and high frequency.
+		 */
+		if (scan->hdisplay == panel_fixed_mode->hdisplay &&
+			scan->hsync_start == panel_fixed_mode->hsync_start &&
+			scan->hsync_end == panel_fixed_mode->hsync_end &&
+			scan->htotal == panel_fixed_mode->htotal &&
+			scan->vdisplay == panel_fixed_mode->vdisplay &&
+			scan->vsync_start == panel_fixed_mode->vsync_start &&
+			scan->vsync_end == panel_fixed_mode->vsync_end &&
+			scan->vtotal == panel_fixed_mode->vtotal) {
+			if (scan->clock < temp_downclock) {
+				/*
+				 * The downclock is already found. But we
+				 * expect to find the lower downclock.
+				 */
+				temp_downclock = scan->clock;
+			}
+		}
+	}
+	mutex_unlock(&dev->mode_config.mutex);
+	if (temp_downclock < panel_fixed_mode->clock &&
+	    i915_lvds_downclock) {
+		/* We found the downclock for LVDS. */
+		dev_priv->lvds_downclock_avail = 1;
+		dev_priv->lvds_downclock = temp_downclock;
+		DRM_DEBUG_KMS("LVDS downclock is found in EDID. "
+				"Normal clock %dKhz, downclock %dKhz\n",
+				panel_fixed_mode->clock, temp_downclock);
+	}
+	return;
 }
 
-/**
- * check whether there exists the ACPI LID device by enumerating the ACPI
- * device tree.
+/*
+ * Enumerate the child dev array parsed from VBT to check whether
+ * the LVDS is present.
+ * If it is present, return 1.
+ * If it is not present, return false.
+ * If no child dev is parsed from VBT, it assumes that the LVDS is present.
+ * Note: The addin_offset should also be checked for LVDS panel.
+ * Only when it is non-zero, it is assumed that it is present.
  */
-static int intel_lid_present(void)
+static int lvds_is_present_in_vbt(struct drm_device *dev)
 {
-	int lid_present = 0;
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct child_device_config *p_child;
+	int i, ret;
 
-	if (acpi_disabled) {
-		/* If ACPI is disabled, there is no ACPI device tree to
-		 * check, so assume the LID device would have been present.
-		 */
+	if (!dev_priv->child_dev_num)
 		return 1;
-	}
 
-	acpi_walk_namespace(ACPI_TYPE_DEVICE, ACPI_ROOT_OBJECT,
-				ACPI_UINT32_MAX,
-				check_lid_device, &lid_present, NULL);
+	ret = 0;
+	for (i = 0; i < dev_priv->child_dev_num; i++) {
+		p_child = dev_priv->child_dev + i;
+		/*
+		 * If the device type is not LFP, continue.
+		 * If the device type is 0x22, it is also regarded as LFP.
+		 */
+		if (p_child->device_type != DEVICE_TYPE_INT_LFP &&
+			p_child->device_type != DEVICE_TYPE_LFP)
+			continue;
 
-	return lid_present;
-}
-#else
-static int intel_lid_present(void)
-{
-	/* In the absence of ACPI built in, assume that the LID device would
-	 * have been present.
-	 */
-	return 1;
+		/* The addin_offset should be checked. Only when it is
+		 * non-zero, it is regarded as present.
+		 */
+		if (p_child->addin_offset) {
+			ret = 1;
+			break;
+		}
+	}
+	return ret;
 }
-#endif
 
 /**
  * intel_lvds_init - setup LVDS connectors on this device
@@ -913,21 +1030,16 @@ void intel_lvds_init(struct drm_device *dev)
 	if (dmi_check_system(intel_no_lvds))
 		return;
 
-	/* Assume that any device without an ACPI LID device also doesn't
-	 * have an integrated LVDS.  We would be better off parsing the BIOS
-	 * to get a reliable indicator, but that code isn't written yet.
-	 *
-	 * In the case of all-in-one desktops using LVDS that we've seen,
-	 * they're using SDVO LVDS.
-	 */
-	if (!intel_lid_present())
+	if (!lvds_is_present_in_vbt(dev)) {
+		DRM_DEBUG_KMS("LVDS is not present in VBT\n");
 		return;
+	}
 
-	if (IS_IGDNG(dev)) {
+	if (HAS_PCH_SPLIT(dev)) {
 		if ((I915_READ(PCH_LVDS) & LVDS_DETECTED) == 0)
 			return;
 		if (dev_priv->edp_support) {
-			DRM_DEBUG("disable LVDS for eDP support\n");
+			DRM_DEBUG_KMS("disable LVDS for eDP support\n");
 			return;
 		}
 		gpio = PCH_GPIOC;
@@ -1000,6 +1112,7 @@ void intel_lvds_init(struct drm_device *dev)
 			dev_priv->panel_fixed_mode =
 				drm_mode_duplicate(dev, scan);
 			mutex_unlock(&dev->mode_config.mutex);
+			intel_find_lvds_downclock(dev, connector);
 			goto out;
 		}
 		mutex_unlock(&dev->mode_config.mutex);
@@ -1024,8 +1137,8 @@ void intel_lvds_init(struct drm_device *dev)
 	 * correct mode.
 	 */
 
-	/* IGDNG: FIXME if still fail, not try pipe mode now */
-	if (IS_IGDNG(dev))
+	/* Ironlake: FIXME if still fail, not try pipe mode now */
+	if (HAS_PCH_SPLIT(dev))
 		goto failed;
 
 	lvds = I915_READ(LVDS);
@@ -1046,7 +1159,7 @@ void intel_lvds_init(struct drm_device *dev)
 		goto failed;
 
 out:
-	if (IS_IGDNG(dev)) {
+	if (HAS_PCH_SPLIT(dev)) {
 		u32 pwm;
 		/* make sure PWM is enabled */
 		pwm = I915_READ(BLC_PWM_CPU_CTL2);
@@ -1059,9 +1172,11 @@ out:
 	}
 	dev_priv->lid_notifier.notifier_call = intel_lid_notify;
 	if (acpi_lid_notifier_register(&dev_priv->lid_notifier)) {
-		DRM_DEBUG("lid notifier registration failed\n");
+		DRM_DEBUG_KMS("lid notifier registration failed\n");
 		dev_priv->lid_notifier.notifier_call = NULL;
 	}
+	/* keep the LVDS connector */
+	dev_priv->int_lvds_connector = connector;
 	drm_sysfs_connector_add(connector);
 	return;
 
@@ -1070,5 +1185,6 @@ failed:
 	if (intel_output->ddc_bus)
 		intel_i2c_destroy(intel_output->ddc_bus);
 	drm_connector_cleanup(connector);
+	drm_encoder_cleanup(encoder);
 	kfree(intel_output);
 }
diff --git a/drivers/gpu/drm/i915/intel_overlay.c b/drivers/gpu/drm/i915/intel_overlay.c
new file mode 100644
index 000000000000..d355d1d527e7
--- /dev/null
+++ b/drivers/gpu/drm/i915/intel_overlay.c
@@ -0,0 +1,1401 @@
+/*
+ * Copyright © 2009
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ * Authors:
+ *    Daniel Vetter <daniel@ffwll.ch>
+ *
+ * Derived from Xorg ddx, xf86-video-intel, src/i830_video.c
+ */
+#include "drmP.h"
+#include "drm.h"
+#include "i915_drm.h"
+#include "i915_drv.h"
+#include "i915_reg.h"
+#include "intel_drv.h"
+
+/* Limits for overlay size. According to intel doc, the real limits are:
+ * Y width: 4095, UV width (planar): 2047, Y height: 2047,
+ * UV width (planar): * 1023. But the xorg thinks 2048 for height and width. Use
+ * the mininum of both.  */
+#define IMAGE_MAX_WIDTH		2048
+#define IMAGE_MAX_HEIGHT	2046 /* 2 * 1023 */
+/* on 830 and 845 these large limits result in the card hanging */
+#define IMAGE_MAX_WIDTH_LEGACY	1024
+#define IMAGE_MAX_HEIGHT_LEGACY	1088
+
+/* overlay register definitions */
+/* OCMD register */
+#define OCMD_TILED_SURFACE	(0x1<<19)
+#define OCMD_MIRROR_MASK	(0x3<<17)
+#define OCMD_MIRROR_MODE	(0x3<<17)
+#define OCMD_MIRROR_HORIZONTAL	(0x1<<17)
+#define OCMD_MIRROR_VERTICAL	(0x2<<17)
+#define OCMD_MIRROR_BOTH	(0x3<<17)
+#define OCMD_BYTEORDER_MASK	(0x3<<14) /* zero for YUYV or FOURCC YUY2 */
+#define OCMD_UV_SWAP		(0x1<<14) /* YVYU */
+#define OCMD_Y_SWAP		(0x2<<14) /* UYVY or FOURCC UYVY */
+#define OCMD_Y_AND_UV_SWAP	(0x3<<14) /* VYUY */
+#define OCMD_SOURCE_FORMAT_MASK (0xf<<10)
+#define OCMD_RGB_888		(0x1<<10) /* not in i965 Intel docs */
+#define OCMD_RGB_555		(0x2<<10) /* not in i965 Intel docs */
+#define OCMD_RGB_565		(0x3<<10) /* not in i965 Intel docs */
+#define OCMD_YUV_422_PACKED	(0x8<<10)
+#define OCMD_YUV_411_PACKED	(0x9<<10) /* not in i965 Intel docs */
+#define OCMD_YUV_420_PLANAR	(0xc<<10)
+#define OCMD_YUV_422_PLANAR	(0xd<<10)
+#define OCMD_YUV_410_PLANAR	(0xe<<10) /* also 411 */
+#define OCMD_TVSYNCFLIP_PARITY	(0x1<<9)
+#define OCMD_TVSYNCFLIP_ENABLE	(0x1<<7)
+#define OCMD_BUF_TYPE_MASK	(Ox1<<5)
+#define OCMD_BUF_TYPE_FRAME	(0x0<<5)
+#define OCMD_BUF_TYPE_FIELD	(0x1<<5)
+#define OCMD_TEST_MODE		(0x1<<4)
+#define OCMD_BUFFER_SELECT	(0x3<<2)
+#define OCMD_BUFFER0		(0x0<<2)
+#define OCMD_BUFFER1		(0x1<<2)
+#define OCMD_FIELD_SELECT	(0x1<<2)
+#define OCMD_FIELD0		(0x0<<1)
+#define OCMD_FIELD1		(0x1<<1)
+#define OCMD_ENABLE		(0x1<<0)
+
+/* OCONFIG register */
+#define OCONF_PIPE_MASK		(0x1<<18)
+#define OCONF_PIPE_A		(0x0<<18)
+#define OCONF_PIPE_B		(0x1<<18)
+#define OCONF_GAMMA2_ENABLE	(0x1<<16)
+#define OCONF_CSC_MODE_BT601	(0x0<<5)
+#define OCONF_CSC_MODE_BT709	(0x1<<5)
+#define OCONF_CSC_BYPASS	(0x1<<4)
+#define OCONF_CC_OUT_8BIT	(0x1<<3)
+#define OCONF_TEST_MODE		(0x1<<2)
+#define OCONF_THREE_LINE_BUFFER	(0x1<<0)
+#define OCONF_TWO_LINE_BUFFER	(0x0<<0)
+
+/* DCLRKM (dst-key) register */
+#define DST_KEY_ENABLE		(0x1<<31)
+#define CLK_RGB24_MASK		0x0
+#define CLK_RGB16_MASK		0x070307
+#define CLK_RGB15_MASK		0x070707
+#define CLK_RGB8I_MASK		0xffffff
+
+#define RGB16_TO_COLORKEY(c) \
+	(((c & 0xF800) << 8) | ((c & 0x07E0) << 5) | ((c & 0x001F) << 3))
+#define RGB15_TO_COLORKEY(c) \
+	(((c & 0x7c00) << 9) | ((c & 0x03E0) << 6) | ((c & 0x001F) << 3))
+
+/* overlay flip addr flag */
+#define OFC_UPDATE		0x1
+
+/* polyphase filter coefficients */
+#define N_HORIZ_Y_TAPS          5
+#define N_VERT_Y_TAPS           3
+#define N_HORIZ_UV_TAPS         3
+#define N_VERT_UV_TAPS          3
+#define N_PHASES                17
+#define MAX_TAPS                5
+
+/* memory bufferd overlay registers */
+struct overlay_registers {
+    u32 OBUF_0Y;
+    u32 OBUF_1Y;
+    u32 OBUF_0U;
+    u32 OBUF_0V;
+    u32 OBUF_1U;
+    u32 OBUF_1V;
+    u32 OSTRIDE;
+    u32 YRGB_VPH;
+    u32 UV_VPH;
+    u32 HORZ_PH;
+    u32 INIT_PHS;
+    u32 DWINPOS;
+    u32 DWINSZ;
+    u32 SWIDTH;
+    u32 SWIDTHSW;
+    u32 SHEIGHT;
+    u32 YRGBSCALE;
+    u32 UVSCALE;
+    u32 OCLRC0;
+    u32 OCLRC1;
+    u32 DCLRKV;
+    u32 DCLRKM;
+    u32 SCLRKVH;
+    u32 SCLRKVL;
+    u32 SCLRKEN;
+    u32 OCONFIG;
+    u32 OCMD;
+    u32 RESERVED1; /* 0x6C */
+    u32 OSTART_0Y;
+    u32 OSTART_1Y;
+    u32 OSTART_0U;
+    u32 OSTART_0V;
+    u32 OSTART_1U;
+    u32 OSTART_1V;
+    u32 OTILEOFF_0Y;
+    u32 OTILEOFF_1Y;
+    u32 OTILEOFF_0U;
+    u32 OTILEOFF_0V;
+    u32 OTILEOFF_1U;
+    u32 OTILEOFF_1V;
+    u32 FASTHSCALE; /* 0xA0 */
+    u32 UVSCALEV; /* 0xA4 */
+    u32 RESERVEDC[(0x200 - 0xA8) / 4]; /* 0xA8 - 0x1FC */
+    u16 Y_VCOEFS[N_VERT_Y_TAPS * N_PHASES]; /* 0x200 */
+    u16 RESERVEDD[0x100 / 2 - N_VERT_Y_TAPS * N_PHASES];
+    u16 Y_HCOEFS[N_HORIZ_Y_TAPS * N_PHASES]; /* 0x300 */
+    u16 RESERVEDE[0x200 / 2 - N_HORIZ_Y_TAPS * N_PHASES];
+    u16 UV_VCOEFS[N_VERT_UV_TAPS * N_PHASES]; /* 0x500 */
+    u16 RESERVEDF[0x100 / 2 - N_VERT_UV_TAPS * N_PHASES];
+    u16 UV_HCOEFS[N_HORIZ_UV_TAPS * N_PHASES]; /* 0x600 */
+    u16 RESERVEDG[0x100 / 2 - N_HORIZ_UV_TAPS * N_PHASES];
+};
+
+/* overlay flip addr flag */
+#define OFC_UPDATE		0x1
+
+#define OVERLAY_NONPHYSICAL(dev) (IS_G33(dev) || IS_I965G(dev))
+#define OVERLAY_EXISTS(dev) (!IS_G4X(dev) && !IS_IRONLAKE(dev) && !IS_GEN6(dev))
+
+
+static struct overlay_registers *intel_overlay_map_regs_atomic(struct intel_overlay *overlay)
+{
+        drm_i915_private_t *dev_priv = overlay->dev->dev_private;
+	struct overlay_registers *regs;
+
+	/* no recursive mappings */
+	BUG_ON(overlay->virt_addr);
+
+	if (OVERLAY_NONPHYSICAL(overlay->dev)) {
+		regs = io_mapping_map_atomic_wc(dev_priv->mm.gtt_mapping,
+				overlay->reg_bo->gtt_offset);
+
+		if (!regs) {
+			DRM_ERROR("failed to map overlay regs in GTT\n");
+			return NULL;
+		}
+	} else
+		regs = overlay->reg_bo->phys_obj->handle->vaddr;
+
+	return overlay->virt_addr = regs;
+}
+
+static void intel_overlay_unmap_regs_atomic(struct intel_overlay *overlay)
+{
+	if (OVERLAY_NONPHYSICAL(overlay->dev))
+		io_mapping_unmap_atomic(overlay->virt_addr);
+
+	overlay->virt_addr = NULL;
+
+	return;
+}
+
+/* overlay needs to be disable in OCMD reg */
+static int intel_overlay_on(struct intel_overlay *overlay)
+{
+	struct drm_device *dev = overlay->dev;
+        drm_i915_private_t *dev_priv = dev->dev_private;
+	int ret;
+	RING_LOCALS;
+
+	BUG_ON(overlay->active);
+
+	overlay->active = 1;
+	overlay->hw_wedged = NEEDS_WAIT_FOR_FLIP;
+
+	BEGIN_LP_RING(4);
+	OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_ON);
+	OUT_RING(overlay->flip_addr | OFC_UPDATE);
+	OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
+	OUT_RING(MI_NOOP);
+	ADVANCE_LP_RING();
+
+	overlay->last_flip_req = i915_add_request(dev, NULL, 0);
+	if (overlay->last_flip_req == 0)
+		return -ENOMEM;
+
+	ret = i915_do_wait_request(dev, overlay->last_flip_req, 1);
+	if (ret != 0)
+		return ret;
+
+	overlay->hw_wedged = 0;
+	overlay->last_flip_req = 0;
+	return 0;
+}
+
+/* overlay needs to be enabled in OCMD reg */
+static void intel_overlay_continue(struct intel_overlay *overlay,
+			    bool load_polyphase_filter)
+{
+	struct drm_device *dev = overlay->dev;
+        drm_i915_private_t *dev_priv = dev->dev_private;
+	u32 flip_addr = overlay->flip_addr;
+	u32 tmp;
+	RING_LOCALS;
+
+	BUG_ON(!overlay->active);
+
+	if (load_polyphase_filter)
+		flip_addr |= OFC_UPDATE;
+
+	/* check for underruns */
+	tmp = I915_READ(DOVSTA);
+	if (tmp & (1 << 17))
+		DRM_DEBUG("overlay underrun, DOVSTA: %x\n", tmp);
+
+	BEGIN_LP_RING(2);
+	OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_CONTINUE);
+	OUT_RING(flip_addr);
+        ADVANCE_LP_RING();
+
+	overlay->last_flip_req = i915_add_request(dev, NULL, 0);
+}
+
+static int intel_overlay_wait_flip(struct intel_overlay *overlay)
+{
+	struct drm_device *dev = overlay->dev;
+        drm_i915_private_t *dev_priv = dev->dev_private;
+	int ret;
+	u32 tmp;
+	RING_LOCALS;
+
+	if (overlay->last_flip_req != 0) {
+		ret = i915_do_wait_request(dev, overlay->last_flip_req, 1);
+		if (ret == 0) {
+			overlay->last_flip_req = 0;
+
+			tmp = I915_READ(ISR);
+
+			if (!(tmp & I915_OVERLAY_PLANE_FLIP_PENDING_INTERRUPT))
+				return 0;
+		}
+	}
+
+	/* synchronous slowpath */
+	overlay->hw_wedged = RELEASE_OLD_VID;
+
+	BEGIN_LP_RING(2);
+        OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
+        OUT_RING(MI_NOOP);
+        ADVANCE_LP_RING();
+
+	overlay->last_flip_req = i915_add_request(dev, NULL, 0);
+	if (overlay->last_flip_req == 0)
+		return -ENOMEM;
+
+	ret = i915_do_wait_request(dev, overlay->last_flip_req, 1);
+	if (ret != 0)
+		return ret;
+
+	overlay->hw_wedged = 0;
+	overlay->last_flip_req = 0;
+	return 0;
+}
+
+/* overlay needs to be disabled in OCMD reg */
+static int intel_overlay_off(struct intel_overlay *overlay)
+{
+	u32 flip_addr = overlay->flip_addr;
+	struct drm_device *dev = overlay->dev;
+        drm_i915_private_t *dev_priv = dev->dev_private;
+	int ret;
+	RING_LOCALS;
+
+	BUG_ON(!overlay->active);
+
+	/* According to intel docs the overlay hw may hang (when switching
+	 * off) without loading the filter coeffs. It is however unclear whether
+	 * this applies to the disabling of the overlay or to the switching off
+	 * of the hw. Do it in both cases */
+	flip_addr |= OFC_UPDATE;
+
+	/* wait for overlay to go idle */
+	overlay->hw_wedged = SWITCH_OFF_STAGE_1;
+
+	BEGIN_LP_RING(4);
+	OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_CONTINUE);
+	OUT_RING(flip_addr);
+        OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
+        OUT_RING(MI_NOOP);
+        ADVANCE_LP_RING();
+
+	overlay->last_flip_req = i915_add_request(dev, NULL, 0);
+	if (overlay->last_flip_req == 0)
+		return -ENOMEM;
+
+	ret = i915_do_wait_request(dev, overlay->last_flip_req, 1);
+	if (ret != 0)
+		return ret;
+
+	/* turn overlay off */
+	overlay->hw_wedged = SWITCH_OFF_STAGE_2;
+
+	BEGIN_LP_RING(4);
+        OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_OFF);
+	OUT_RING(flip_addr);
+        OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
+        OUT_RING(MI_NOOP);
+	ADVANCE_LP_RING();
+
+	overlay->last_flip_req = i915_add_request(dev, NULL, 0);
+	if (overlay->last_flip_req == 0)
+		return -ENOMEM;
+
+	ret = i915_do_wait_request(dev, overlay->last_flip_req, 1);
+	if (ret != 0)
+		return ret;
+
+	overlay->hw_wedged = 0;
+	overlay->last_flip_req = 0;
+	return ret;
+}
+
+static void intel_overlay_off_tail(struct intel_overlay *overlay)
+{
+	struct drm_gem_object *obj;
+
+	/* never have the overlay hw on without showing a frame */
+	BUG_ON(!overlay->vid_bo);
+	obj = overlay->vid_bo->obj;
+
+	i915_gem_object_unpin(obj);
+	drm_gem_object_unreference(obj);
+	overlay->vid_bo = NULL;
+
+	overlay->crtc->overlay = NULL;
+	overlay->crtc = NULL;
+	overlay->active = 0;
+}
+
+/* recover from an interruption due to a signal
+ * We have to be careful not to repeat work forever an make forward progess. */
+int intel_overlay_recover_from_interrupt(struct intel_overlay *overlay,
+					 int interruptible)
+{
+	struct drm_device *dev = overlay->dev;
+        drm_i915_private_t *dev_priv = dev->dev_private;
+	struct drm_gem_object *obj;
+	u32 flip_addr;
+	int ret;
+	RING_LOCALS;
+
+	if (overlay->hw_wedged == HW_WEDGED)
+		return -EIO;
+
+	if (overlay->last_flip_req == 0) {
+		overlay->last_flip_req = i915_add_request(dev, NULL, 0);
+		if (overlay->last_flip_req == 0)
+			return -ENOMEM;
+	}
+
+	ret = i915_do_wait_request(dev, overlay->last_flip_req, interruptible);
+	if (ret != 0)
+		return ret;
+
+	switch (overlay->hw_wedged) {
+		case RELEASE_OLD_VID:
+			obj = overlay->old_vid_bo->obj;
+			i915_gem_object_unpin(obj);
+			drm_gem_object_unreference(obj);
+			overlay->old_vid_bo = NULL;
+			break;
+		case SWITCH_OFF_STAGE_1:
+			flip_addr = overlay->flip_addr;
+			flip_addr |= OFC_UPDATE;
+
+			overlay->hw_wedged = SWITCH_OFF_STAGE_2;
+
+			BEGIN_LP_RING(4);
+			OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_OFF);
+			OUT_RING(flip_addr);
+			OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
+			OUT_RING(MI_NOOP);
+			ADVANCE_LP_RING();
+
+			overlay->last_flip_req = i915_add_request(dev, NULL, 0);
+			if (overlay->last_flip_req == 0)
+				return -ENOMEM;
+
+			ret = i915_do_wait_request(dev, overlay->last_flip_req,
+					interruptible);
+			if (ret != 0)
+				return ret;
+
+		case SWITCH_OFF_STAGE_2:
+			intel_overlay_off_tail(overlay);
+			break;
+		default:
+			BUG_ON(overlay->hw_wedged != NEEDS_WAIT_FOR_FLIP);
+	}
+
+	overlay->hw_wedged = 0;
+	overlay->last_flip_req = 0;
+	return 0;
+}
+
+/* Wait for pending overlay flip and release old frame.
+ * Needs to be called before the overlay register are changed
+ * via intel_overlay_(un)map_regs_atomic */
+static int intel_overlay_release_old_vid(struct intel_overlay *overlay)
+{
+	int ret;
+	struct drm_gem_object *obj;
+
+	/* only wait if there is actually an old frame to release to
+	 * guarantee forward progress */
+	if (!overlay->old_vid_bo)
+		return 0;
+
+	ret = intel_overlay_wait_flip(overlay);
+	if (ret != 0)
+		return ret;
+
+	obj = overlay->old_vid_bo->obj;
+	i915_gem_object_unpin(obj);
+	drm_gem_object_unreference(obj);
+	overlay->old_vid_bo = NULL;
+
+	return 0;
+}
+
+struct put_image_params {
+	int format;
+	short dst_x;
+	short dst_y;
+	short dst_w;
+	short dst_h;
+	short src_w;
+	short src_scan_h;
+	short src_scan_w;
+	short src_h;
+	short stride_Y;
+	short stride_UV;
+	int offset_Y;
+	int offset_U;
+	int offset_V;
+};
+
+static int packed_depth_bytes(u32 format)
+{
+	switch (format & I915_OVERLAY_DEPTH_MASK) {
+		case I915_OVERLAY_YUV422:
+			return 4;
+		case I915_OVERLAY_YUV411:
+			/* return 6; not implemented */
+		default:
+			return -EINVAL;
+	}
+}
+
+static int packed_width_bytes(u32 format, short width)
+{
+	switch (format & I915_OVERLAY_DEPTH_MASK) {
+		case I915_OVERLAY_YUV422:
+			return width << 1;
+		default:
+			return -EINVAL;
+	}
+}
+
+static int uv_hsubsampling(u32 format)
+{
+	switch (format & I915_OVERLAY_DEPTH_MASK) {
+		case I915_OVERLAY_YUV422:
+		case I915_OVERLAY_YUV420:
+			return 2;
+		case I915_OVERLAY_YUV411:
+		case I915_OVERLAY_YUV410:
+			return 4;
+		default:
+			return -EINVAL;
+	}
+}
+
+static int uv_vsubsampling(u32 format)
+{
+	switch (format & I915_OVERLAY_DEPTH_MASK) {
+		case I915_OVERLAY_YUV420:
+		case I915_OVERLAY_YUV410:
+			return 2;
+		case I915_OVERLAY_YUV422:
+		case I915_OVERLAY_YUV411:
+			return 1;
+		default:
+			return -EINVAL;
+	}
+}
+
+static u32 calc_swidthsw(struct drm_device *dev, u32 offset, u32 width)
+{
+	u32 mask, shift, ret;
+	if (IS_I9XX(dev)) {
+		mask = 0x3f;
+		shift = 6;
+	} else {
+		mask = 0x1f;
+		shift = 5;
+	}
+	ret = ((offset + width + mask) >> shift) - (offset >> shift);
+	if (IS_I9XX(dev))
+		ret <<= 1;
+	ret -=1;
+	return ret << 2;
+}
+
+static const u16 y_static_hcoeffs[N_HORIZ_Y_TAPS * N_PHASES] = {
+	0x3000, 0xb4a0, 0x1930, 0x1920, 0xb4a0,
+	0x3000, 0xb500, 0x19d0, 0x1880, 0xb440,
+	0x3000, 0xb540, 0x1a88, 0x2f80, 0xb3e0,
+	0x3000, 0xb580, 0x1b30, 0x2e20, 0xb380,
+	0x3000, 0xb5c0, 0x1bd8, 0x2cc0, 0xb320,
+	0x3020, 0xb5e0, 0x1c60, 0x2b80, 0xb2c0,
+	0x3020, 0xb5e0, 0x1cf8, 0x2a20, 0xb260,
+	0x3020, 0xb5e0, 0x1d80, 0x28e0, 0xb200,
+	0x3020, 0xb5c0, 0x1e08, 0x3f40, 0xb1c0,
+	0x3020, 0xb580, 0x1e78, 0x3ce0, 0xb160,
+	0x3040, 0xb520, 0x1ed8, 0x3aa0, 0xb120,
+	0x3040, 0xb4a0, 0x1f30, 0x3880, 0xb0e0,
+	0x3040, 0xb400, 0x1f78, 0x3680, 0xb0a0,
+	0x3020, 0xb340, 0x1fb8, 0x34a0, 0xb060,
+	0x3020, 0xb240, 0x1fe0, 0x32e0, 0xb040,
+	0x3020, 0xb140, 0x1ff8, 0x3160, 0xb020,
+	0xb000, 0x3000, 0x0800, 0x3000, 0xb000};
+static const u16 uv_static_hcoeffs[N_HORIZ_UV_TAPS * N_PHASES] = {
+	0x3000, 0x1800, 0x1800, 0xb000, 0x18d0, 0x2e60,
+	0xb000, 0x1990, 0x2ce0, 0xb020, 0x1a68, 0x2b40,
+	0xb040, 0x1b20, 0x29e0, 0xb060, 0x1bd8, 0x2880,
+	0xb080, 0x1c88, 0x3e60, 0xb0a0, 0x1d28, 0x3c00,
+	0xb0c0, 0x1db8, 0x39e0, 0xb0e0, 0x1e40, 0x37e0,
+	0xb100, 0x1eb8, 0x3620, 0xb100, 0x1f18, 0x34a0,
+	0xb100, 0x1f68, 0x3360, 0xb0e0, 0x1fa8, 0x3240,
+	0xb0c0, 0x1fe0, 0x3140, 0xb060, 0x1ff0, 0x30a0,
+	0x3000, 0x0800, 0x3000};
+
+static void update_polyphase_filter(struct overlay_registers *regs)
+{
+	memcpy(regs->Y_HCOEFS, y_static_hcoeffs, sizeof(y_static_hcoeffs));
+	memcpy(regs->UV_HCOEFS, uv_static_hcoeffs, sizeof(uv_static_hcoeffs));
+}
+
+static bool update_scaling_factors(struct intel_overlay *overlay,
+				   struct overlay_registers *regs,
+				   struct put_image_params *params)
+{
+	/* fixed point with a 12 bit shift */
+	u32 xscale, yscale, xscale_UV, yscale_UV;
+#define FP_SHIFT 12
+#define FRACT_MASK 0xfff
+	bool scale_changed = false;
+	int uv_hscale = uv_hsubsampling(params->format);
+	int uv_vscale = uv_vsubsampling(params->format);
+
+	if (params->dst_w > 1)
+		xscale = ((params->src_scan_w - 1) << FP_SHIFT)
+			/(params->dst_w);
+	else
+		xscale = 1 << FP_SHIFT;
+
+	if (params->dst_h > 1)
+		yscale = ((params->src_scan_h - 1) << FP_SHIFT)
+			/(params->dst_h);
+	else
+		yscale = 1 << FP_SHIFT;
+
+	/*if (params->format & I915_OVERLAY_YUV_PLANAR) {*/
+		xscale_UV = xscale/uv_hscale;
+		yscale_UV = yscale/uv_vscale;
+		/* make the Y scale to UV scale ratio an exact multiply */
+		xscale = xscale_UV * uv_hscale;
+		yscale = yscale_UV * uv_vscale;
+	/*} else {
+		xscale_UV = 0;
+		yscale_UV = 0;
+	}*/
+
+	if (xscale != overlay->old_xscale || yscale != overlay->old_yscale)
+		scale_changed = true;
+	overlay->old_xscale = xscale;
+	overlay->old_yscale = yscale;
+
+	regs->YRGBSCALE = ((yscale & FRACT_MASK) << 20)
+		| ((xscale >> FP_SHIFT) << 16)
+		| ((xscale & FRACT_MASK) << 3);
+	regs->UVSCALE = ((yscale_UV & FRACT_MASK) << 20)
+		| ((xscale_UV >> FP_SHIFT) << 16)
+		| ((xscale_UV & FRACT_MASK) << 3);
+	regs->UVSCALEV = ((yscale >> FP_SHIFT) << 16)
+		| ((yscale_UV >> FP_SHIFT) << 0);
+
+	if (scale_changed)
+		update_polyphase_filter(regs);
+
+	return scale_changed;
+}
+
+static void update_colorkey(struct intel_overlay *overlay,
+			    struct overlay_registers *regs)
+{
+	u32 key = overlay->color_key;
+	switch (overlay->crtc->base.fb->bits_per_pixel) {
+		case 8:
+			regs->DCLRKV = 0;
+			regs->DCLRKM = CLK_RGB8I_MASK | DST_KEY_ENABLE;
+		case 16:
+			if (overlay->crtc->base.fb->depth == 15) {
+				regs->DCLRKV = RGB15_TO_COLORKEY(key);
+				regs->DCLRKM = CLK_RGB15_MASK | DST_KEY_ENABLE;
+			} else {
+				regs->DCLRKV = RGB16_TO_COLORKEY(key);
+				regs->DCLRKM = CLK_RGB16_MASK | DST_KEY_ENABLE;
+			}
+		case 24:
+		case 32:
+			regs->DCLRKV = key;
+			regs->DCLRKM = CLK_RGB24_MASK | DST_KEY_ENABLE;
+	}
+}
+
+static u32 overlay_cmd_reg(struct put_image_params *params)
+{
+	u32 cmd = OCMD_ENABLE | OCMD_BUF_TYPE_FRAME | OCMD_BUFFER0;
+
+	if (params->format & I915_OVERLAY_YUV_PLANAR) {
+		switch (params->format & I915_OVERLAY_DEPTH_MASK) {
+			case I915_OVERLAY_YUV422:
+				cmd |= OCMD_YUV_422_PLANAR;
+				break;
+			case I915_OVERLAY_YUV420:
+				cmd |= OCMD_YUV_420_PLANAR;
+				break;
+			case I915_OVERLAY_YUV411:
+			case I915_OVERLAY_YUV410:
+				cmd |= OCMD_YUV_410_PLANAR;
+				break;
+		}
+	} else { /* YUV packed */
+		switch (params->format & I915_OVERLAY_DEPTH_MASK) {
+			case I915_OVERLAY_YUV422:
+				cmd |= OCMD_YUV_422_PACKED;
+				break;
+			case I915_OVERLAY_YUV411:
+				cmd |= OCMD_YUV_411_PACKED;
+				break;
+		}
+
+		switch (params->format & I915_OVERLAY_SWAP_MASK) {
+			case I915_OVERLAY_NO_SWAP:
+				break;
+			case I915_OVERLAY_UV_SWAP:
+				cmd |= OCMD_UV_SWAP;
+				break;
+			case I915_OVERLAY_Y_SWAP:
+				cmd |= OCMD_Y_SWAP;
+				break;
+			case I915_OVERLAY_Y_AND_UV_SWAP:
+				cmd |= OCMD_Y_AND_UV_SWAP;
+				break;
+		}
+	}
+
+	return cmd;
+}
+
+int intel_overlay_do_put_image(struct intel_overlay *overlay,
+			       struct drm_gem_object *new_bo,
+			       struct put_image_params *params)
+{
+	int ret, tmp_width;
+	struct overlay_registers *regs;
+	bool scale_changed = false;
+	struct drm_i915_gem_object *bo_priv = new_bo->driver_private;
+	struct drm_device *dev = overlay->dev;
+
+	BUG_ON(!mutex_is_locked(&dev->struct_mutex));
+	BUG_ON(!mutex_is_locked(&dev->mode_config.mutex));
+	BUG_ON(!overlay);
+
+	ret = intel_overlay_release_old_vid(overlay);
+	if (ret != 0)
+		return ret;
+
+	ret = i915_gem_object_pin(new_bo, PAGE_SIZE);
+	if (ret != 0)
+		return ret;
+
+	ret = i915_gem_object_set_to_gtt_domain(new_bo, 0);
+	if (ret != 0)
+		goto out_unpin;
+
+	if (!overlay->active) {
+		regs = intel_overlay_map_regs_atomic(overlay);
+		if (!regs) {
+			ret = -ENOMEM;
+			goto out_unpin;
+		}
+		regs->OCONFIG = OCONF_CC_OUT_8BIT;
+		if (IS_I965GM(overlay->dev))
+			regs->OCONFIG |= OCONF_CSC_MODE_BT709;
+		regs->OCONFIG |= overlay->crtc->pipe == 0 ?
+			OCONF_PIPE_A : OCONF_PIPE_B;
+		intel_overlay_unmap_regs_atomic(overlay);
+
+		ret = intel_overlay_on(overlay);
+		if (ret != 0)
+			goto out_unpin;
+	}
+
+	regs = intel_overlay_map_regs_atomic(overlay);
+	if (!regs) {
+		ret = -ENOMEM;
+		goto out_unpin;
+	}
+
+	regs->DWINPOS = (params->dst_y << 16) | params->dst_x;
+	regs->DWINSZ = (params->dst_h << 16) | params->dst_w;
+
+	if (params->format & I915_OVERLAY_YUV_PACKED)
+		tmp_width = packed_width_bytes(params->format, params->src_w);
+	else
+		tmp_width = params->src_w;
+
+	regs->SWIDTH = params->src_w;
+	regs->SWIDTHSW = calc_swidthsw(overlay->dev,
+			params->offset_Y, tmp_width);
+	regs->SHEIGHT = params->src_h;
+	regs->OBUF_0Y = bo_priv->gtt_offset + params-> offset_Y;
+	regs->OSTRIDE = params->stride_Y;
+
+	if (params->format & I915_OVERLAY_YUV_PLANAR) {
+		int uv_hscale = uv_hsubsampling(params->format);
+		int uv_vscale = uv_vsubsampling(params->format);
+		u32 tmp_U, tmp_V;
+		regs->SWIDTH |= (params->src_w/uv_hscale) << 16;
+		tmp_U = calc_swidthsw(overlay->dev, params->offset_U,
+				params->src_w/uv_hscale);
+		tmp_V = calc_swidthsw(overlay->dev, params->offset_V,
+				params->src_w/uv_hscale);
+		regs->SWIDTHSW |= max_t(u32, tmp_U, tmp_V) << 16;
+		regs->SHEIGHT |= (params->src_h/uv_vscale) << 16;
+		regs->OBUF_0U = bo_priv->gtt_offset + params->offset_U;
+		regs->OBUF_0V = bo_priv->gtt_offset + params->offset_V;
+		regs->OSTRIDE |= params->stride_UV << 16;
+	}
+
+	scale_changed = update_scaling_factors(overlay, regs, params);
+
+	update_colorkey(overlay, regs);
+
+	regs->OCMD = overlay_cmd_reg(params);
+
+	intel_overlay_unmap_regs_atomic(overlay);
+
+	intel_overlay_continue(overlay, scale_changed);
+
+	overlay->old_vid_bo = overlay->vid_bo;
+	overlay->vid_bo = new_bo->driver_private;
+
+	return 0;
+
+out_unpin:
+	i915_gem_object_unpin(new_bo);
+	return ret;
+}
+
+int intel_overlay_switch_off(struct intel_overlay *overlay)
+{
+	int ret;
+	struct overlay_registers *regs;
+	struct drm_device *dev = overlay->dev;
+
+	BUG_ON(!mutex_is_locked(&dev->struct_mutex));
+	BUG_ON(!mutex_is_locked(&dev->mode_config.mutex));
+
+	if (overlay->hw_wedged) {
+		ret = intel_overlay_recover_from_interrupt(overlay, 1);
+		if (ret != 0)
+			return ret;
+	}
+
+	if (!overlay->active)
+		return 0;
+
+	ret = intel_overlay_release_old_vid(overlay);
+	if (ret != 0)
+		return ret;
+
+	regs = intel_overlay_map_regs_atomic(overlay);
+	regs->OCMD = 0;
+	intel_overlay_unmap_regs_atomic(overlay);
+
+	ret = intel_overlay_off(overlay);
+	if (ret != 0)
+		return ret;
+
+	intel_overlay_off_tail(overlay);
+
+	return 0;
+}
+
+static int check_overlay_possible_on_crtc(struct intel_overlay *overlay,
+					  struct intel_crtc *crtc)
+{
+        drm_i915_private_t *dev_priv = overlay->dev->dev_private;
+	u32 pipeconf;
+	int pipeconf_reg = (crtc->pipe == 0) ? PIPEACONF : PIPEBCONF;
+
+	if (!crtc->base.enabled || crtc->dpms_mode != DRM_MODE_DPMS_ON)
+		return -EINVAL;
+
+	pipeconf = I915_READ(pipeconf_reg);
+
+	/* can't use the overlay with double wide pipe */
+	if (!IS_I965G(overlay->dev) && pipeconf & PIPEACONF_DOUBLE_WIDE)
+		return -EINVAL;
+
+	return 0;
+}
+
+static void update_pfit_vscale_ratio(struct intel_overlay *overlay)
+{
+	struct drm_device *dev = overlay->dev;
+        drm_i915_private_t *dev_priv = dev->dev_private;
+	u32 ratio;
+	u32 pfit_control = I915_READ(PFIT_CONTROL);
+
+	/* XXX: This is not the same logic as in the xorg driver, but more in
+	 * line with the intel documentation for the i965 */
+	if (!IS_I965G(dev) && (pfit_control & VERT_AUTO_SCALE)) {
+		ratio = I915_READ(PFIT_AUTO_RATIOS) >> PFIT_VERT_SCALE_SHIFT;
+	} else { /* on i965 use the PGM reg to read out the autoscaler values */
+		ratio = I915_READ(PFIT_PGM_RATIOS);
+		if (IS_I965G(dev))
+			ratio >>= PFIT_VERT_SCALE_SHIFT_965;
+		else
+			ratio >>= PFIT_VERT_SCALE_SHIFT;
+	}
+
+	overlay->pfit_vscale_ratio = ratio;
+}
+
+static int check_overlay_dst(struct intel_overlay *overlay,
+			     struct drm_intel_overlay_put_image *rec)
+{
+	struct drm_display_mode *mode = &overlay->crtc->base.mode;
+
+	if ((rec->dst_x < mode->crtc_hdisplay)
+	    && (rec->dst_x + rec->dst_width
+		    <= mode->crtc_hdisplay)
+	    && (rec->dst_y < mode->crtc_vdisplay)
+	    && (rec->dst_y + rec->dst_height
+		    <= mode->crtc_vdisplay))
+		return 0;
+	else
+		return -EINVAL;
+}
+
+static int check_overlay_scaling(struct put_image_params *rec)
+{
+	u32 tmp;
+
+	/* downscaling limit is 8.0 */
+	tmp = ((rec->src_scan_h << 16) / rec->dst_h) >> 16;
+	if (tmp > 7)
+		return -EINVAL;
+	tmp = ((rec->src_scan_w << 16) / rec->dst_w) >> 16;
+	if (tmp > 7)
+		return -EINVAL;
+
+	return 0;
+}
+
+static int check_overlay_src(struct drm_device *dev,
+			     struct drm_intel_overlay_put_image *rec,
+			     struct drm_gem_object *new_bo)
+{
+	u32 stride_mask;
+	int depth;
+	int uv_hscale = uv_hsubsampling(rec->flags);
+	int uv_vscale = uv_vsubsampling(rec->flags);
+	size_t tmp;
+
+	/* check src dimensions */
+	if (IS_845G(dev) || IS_I830(dev)) {
+		if (rec->src_height > IMAGE_MAX_HEIGHT_LEGACY
+		    || rec->src_width > IMAGE_MAX_WIDTH_LEGACY)
+			return -EINVAL;
+	} else {
+		if (rec->src_height > IMAGE_MAX_HEIGHT
+		    || rec->src_width > IMAGE_MAX_WIDTH)
+			return -EINVAL;
+	}
+	/* better safe than sorry, use 4 as the maximal subsampling ratio */
+	if (rec->src_height < N_VERT_Y_TAPS*4
+	    || rec->src_width < N_HORIZ_Y_TAPS*4)
+		return -EINVAL;
+
+	/* check alingment constrains */
+	switch (rec->flags & I915_OVERLAY_TYPE_MASK) {
+		case I915_OVERLAY_RGB:
+			/* not implemented */
+			return -EINVAL;
+		case I915_OVERLAY_YUV_PACKED:
+			depth = packed_depth_bytes(rec->flags);
+			if (uv_vscale != 1)
+				return -EINVAL;
+			if (depth < 0)
+				return depth;
+			/* ignore UV planes */
+			rec->stride_UV = 0;
+			rec->offset_U = 0;
+			rec->offset_V = 0;
+			/* check pixel alignment */
+			if (rec->offset_Y % depth)
+				return -EINVAL;
+			break;
+		case I915_OVERLAY_YUV_PLANAR:
+			if (uv_vscale < 0 || uv_hscale < 0)
+				return -EINVAL;
+			/* no offset restrictions for planar formats */
+			break;
+		default:
+			return -EINVAL;
+	}
+
+	if (rec->src_width % uv_hscale)
+		return -EINVAL;
+
+	/* stride checking */
+	stride_mask = 63;
+
+	if (rec->stride_Y & stride_mask || rec->stride_UV & stride_mask)
+		return -EINVAL;
+	if (IS_I965G(dev) && rec->stride_Y < 512)
+		return -EINVAL;
+
+	tmp = (rec->flags & I915_OVERLAY_TYPE_MASK) == I915_OVERLAY_YUV_PLANAR ?
+		4 : 8;
+	if (rec->stride_Y > tmp*1024 || rec->stride_UV > 2*1024)
+		return -EINVAL;
+
+	/* check buffer dimensions */
+	switch (rec->flags & I915_OVERLAY_TYPE_MASK) {
+		case I915_OVERLAY_RGB:
+		case I915_OVERLAY_YUV_PACKED:
+			/* always 4 Y values per depth pixels */
+			if (packed_width_bytes(rec->flags, rec->src_width)
+					> rec->stride_Y)
+				return -EINVAL;
+
+			tmp = rec->stride_Y*rec->src_height;
+			if (rec->offset_Y + tmp > new_bo->size)
+				return -EINVAL;
+			break;
+		case I915_OVERLAY_YUV_PLANAR:
+			if (rec->src_width > rec->stride_Y)
+				return -EINVAL;
+			if (rec->src_width/uv_hscale > rec->stride_UV)
+				return -EINVAL;
+
+			tmp = rec->stride_Y*rec->src_height;
+			if (rec->offset_Y + tmp > new_bo->size)
+				return -EINVAL;
+			tmp = rec->stride_UV*rec->src_height;
+			tmp /= uv_vscale;
+			if (rec->offset_U + tmp > new_bo->size
+			    || rec->offset_V + tmp > new_bo->size)
+				return -EINVAL;
+			break;
+	}
+
+	return 0;
+}
+
+int intel_overlay_put_image(struct drm_device *dev, void *data,
+                            struct drm_file *file_priv)
+{
+	struct drm_intel_overlay_put_image *put_image_rec = data;
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	struct intel_overlay *overlay;
+	struct drm_mode_object *drmmode_obj;
+	struct intel_crtc *crtc;
+	struct drm_gem_object *new_bo;
+	struct put_image_params *params;
+	int ret;
+
+	if (!dev_priv) {
+		DRM_ERROR("called with no initialization\n");
+		return -EINVAL;
+	}
+
+	overlay = dev_priv->overlay;
+	if (!overlay) {
+		DRM_DEBUG("userspace bug: no overlay\n");
+		return -ENODEV;
+	}
+
+	if (!(put_image_rec->flags & I915_OVERLAY_ENABLE)) {
+		mutex_lock(&dev->mode_config.mutex);
+		mutex_lock(&dev->struct_mutex);
+
+		ret = intel_overlay_switch_off(overlay);
+
+		mutex_unlock(&dev->struct_mutex);
+		mutex_unlock(&dev->mode_config.mutex);
+
+		return ret;
+	}
+
+	params = kmalloc(sizeof(struct put_image_params), GFP_KERNEL);
+	if (!params)
+		return -ENOMEM;
+
+	drmmode_obj = drm_mode_object_find(dev, put_image_rec->crtc_id,
+                        DRM_MODE_OBJECT_CRTC);
+	if (!drmmode_obj)
+		return -ENOENT;
+	crtc = to_intel_crtc(obj_to_crtc(drmmode_obj));
+
+	new_bo = drm_gem_object_lookup(dev, file_priv,
+			put_image_rec->bo_handle);
+	if (!new_bo)
+		return -ENOENT;
+
+	mutex_lock(&dev->mode_config.mutex);
+	mutex_lock(&dev->struct_mutex);
+
+	if (overlay->hw_wedged) {
+		ret = intel_overlay_recover_from_interrupt(overlay, 1);
+		if (ret != 0)
+			goto out_unlock;
+	}
+
+	if (overlay->crtc != crtc) {
+		struct drm_display_mode *mode = &crtc->base.mode;
+		ret = intel_overlay_switch_off(overlay);
+		if (ret != 0)
+			goto out_unlock;
+
+		ret = check_overlay_possible_on_crtc(overlay, crtc);
+		if (ret != 0)
+			goto out_unlock;
+
+		overlay->crtc = crtc;
+		crtc->overlay = overlay;
+
+		if (intel_panel_fitter_pipe(dev) == crtc->pipe
+		    /* and line to wide, i.e. one-line-mode */
+		    && mode->hdisplay > 1024) {
+			overlay->pfit_active = 1;
+			update_pfit_vscale_ratio(overlay);
+		} else
+			overlay->pfit_active = 0;
+	}
+
+	ret = check_overlay_dst(overlay, put_image_rec);
+	if (ret != 0)
+		goto out_unlock;
+
+	if (overlay->pfit_active) {
+		params->dst_y = ((((u32)put_image_rec->dst_y) << 12) /
+			overlay->pfit_vscale_ratio);
+		/* shifting right rounds downwards, so add 1 */
+		params->dst_h = ((((u32)put_image_rec->dst_height) << 12) /
+			overlay->pfit_vscale_ratio) + 1;
+	} else {
+		params->dst_y = put_image_rec->dst_y;
+		params->dst_h = put_image_rec->dst_height;
+	}
+	params->dst_x = put_image_rec->dst_x;
+	params->dst_w = put_image_rec->dst_width;
+
+	params->src_w = put_image_rec->src_width;
+	params->src_h = put_image_rec->src_height;
+	params->src_scan_w = put_image_rec->src_scan_width;
+	params->src_scan_h = put_image_rec->src_scan_height;
+	if (params->src_scan_h > params->src_h
+	    || params->src_scan_w > params->src_w) {
+		ret = -EINVAL;
+		goto out_unlock;
+	}
+
+	ret = check_overlay_src(dev, put_image_rec, new_bo);
+	if (ret != 0)
+		goto out_unlock;
+	params->format = put_image_rec->flags & ~I915_OVERLAY_FLAGS_MASK;
+	params->stride_Y = put_image_rec->stride_Y;
+	params->stride_UV = put_image_rec->stride_UV;
+	params->offset_Y = put_image_rec->offset_Y;
+	params->offset_U = put_image_rec->offset_U;
+	params->offset_V = put_image_rec->offset_V;
+
+	/* Check scaling after src size to prevent a divide-by-zero. */
+	ret = check_overlay_scaling(params);
+	if (ret != 0)
+		goto out_unlock;
+
+	ret = intel_overlay_do_put_image(overlay, new_bo, params);
+	if (ret != 0)
+		goto out_unlock;
+
+	mutex_unlock(&dev->struct_mutex);
+	mutex_unlock(&dev->mode_config.mutex);
+
+	kfree(params);
+
+	return 0;
+
+out_unlock:
+	mutex_unlock(&dev->struct_mutex);
+	mutex_unlock(&dev->mode_config.mutex);
+	drm_gem_object_unreference_unlocked(new_bo);
+	kfree(params);
+
+	return ret;
+}
+
+static void update_reg_attrs(struct intel_overlay *overlay,
+			     struct overlay_registers *regs)
+{
+	regs->OCLRC0 = (overlay->contrast << 18) | (overlay->brightness & 0xff);
+	regs->OCLRC1 = overlay->saturation;
+}
+
+static bool check_gamma_bounds(u32 gamma1, u32 gamma2)
+{
+	int i;
+
+	if (gamma1 & 0xff000000 || gamma2 & 0xff000000)
+		return false;
+
+	for (i = 0; i < 3; i++) {
+		if (((gamma1 >> i * 8) & 0xff) >= ((gamma2 >> i*8) & 0xff))
+			return false;
+	}
+
+	return true;
+}
+
+static bool check_gamma5_errata(u32 gamma5)
+{
+	int i;
+
+	for (i = 0; i < 3; i++) {
+		if (((gamma5 >> i*8) & 0xff) == 0x80)
+			return false;
+	}
+
+	return true;
+}
+
+static int check_gamma(struct drm_intel_overlay_attrs *attrs)
+{
+	if (!check_gamma_bounds(0, attrs->gamma0)
+	    || !check_gamma_bounds(attrs->gamma0, attrs->gamma1)
+	    || !check_gamma_bounds(attrs->gamma1, attrs->gamma2)
+	    || !check_gamma_bounds(attrs->gamma2, attrs->gamma3)
+	    || !check_gamma_bounds(attrs->gamma3, attrs->gamma4)
+	    || !check_gamma_bounds(attrs->gamma4, attrs->gamma5)
+	    || !check_gamma_bounds(attrs->gamma5, 0x00ffffff))
+		return -EINVAL;
+	if (!check_gamma5_errata(attrs->gamma5))
+		return -EINVAL;
+	return 0;
+}
+
+int intel_overlay_attrs(struct drm_device *dev, void *data,
+                        struct drm_file *file_priv)
+{
+	struct drm_intel_overlay_attrs *attrs = data;
+        drm_i915_private_t *dev_priv = dev->dev_private;
+	struct intel_overlay *overlay;
+	struct overlay_registers *regs;
+	int ret;
+
+	if (!dev_priv) {
+		DRM_ERROR("called with no initialization\n");
+		return -EINVAL;
+	}
+
+	overlay = dev_priv->overlay;
+	if (!overlay) {
+		DRM_DEBUG("userspace bug: no overlay\n");
+		return -ENODEV;
+	}
+
+	mutex_lock(&dev->mode_config.mutex);
+	mutex_lock(&dev->struct_mutex);
+
+	if (!(attrs->flags & I915_OVERLAY_UPDATE_ATTRS)) {
+		attrs->color_key = overlay->color_key;
+		attrs->brightness = overlay->brightness;
+		attrs->contrast = overlay->contrast;
+		attrs->saturation = overlay->saturation;
+
+		if (IS_I9XX(dev)) {
+			attrs->gamma0 = I915_READ(OGAMC0);
+			attrs->gamma1 = I915_READ(OGAMC1);
+			attrs->gamma2 = I915_READ(OGAMC2);
+			attrs->gamma3 = I915_READ(OGAMC3);
+			attrs->gamma4 = I915_READ(OGAMC4);
+			attrs->gamma5 = I915_READ(OGAMC5);
+		}
+		ret = 0;
+	} else {
+		overlay->color_key = attrs->color_key;
+		if (attrs->brightness >= -128 && attrs->brightness <= 127) {
+			overlay->brightness = attrs->brightness;
+		} else {
+			ret = -EINVAL;
+			goto out_unlock;
+		}
+		if (attrs->contrast <= 255) {
+			overlay->contrast = attrs->contrast;
+		} else {
+			ret = -EINVAL;
+			goto out_unlock;
+		}
+		if (attrs->saturation <= 1023) {
+			overlay->saturation = attrs->saturation;
+		} else {
+			ret = -EINVAL;
+			goto out_unlock;
+		}
+
+		regs = intel_overlay_map_regs_atomic(overlay);
+		if (!regs) {
+			ret = -ENOMEM;
+			goto out_unlock;
+		}
+
+		update_reg_attrs(overlay, regs);
+
+		intel_overlay_unmap_regs_atomic(overlay);
+
+		if (attrs->flags & I915_OVERLAY_UPDATE_GAMMA) {
+			if (!IS_I9XX(dev)) {
+				ret = -EINVAL;
+				goto out_unlock;
+			}
+
+			if (overlay->active) {
+				ret = -EBUSY;
+				goto out_unlock;
+			}
+
+			ret = check_gamma(attrs);
+			if (ret != 0)
+				goto out_unlock;
+
+			I915_WRITE(OGAMC0, attrs->gamma0);
+			I915_WRITE(OGAMC1, attrs->gamma1);
+			I915_WRITE(OGAMC2, attrs->gamma2);
+			I915_WRITE(OGAMC3, attrs->gamma3);
+			I915_WRITE(OGAMC4, attrs->gamma4);
+			I915_WRITE(OGAMC5, attrs->gamma5);
+		}
+		ret = 0;
+	}
+
+out_unlock:
+	mutex_unlock(&dev->struct_mutex);
+	mutex_unlock(&dev->mode_config.mutex);
+
+	return ret;
+}
+
+void intel_setup_overlay(struct drm_device *dev)
+{
+        drm_i915_private_t *dev_priv = dev->dev_private;
+	struct intel_overlay *overlay;
+	struct drm_gem_object *reg_bo;
+	struct overlay_registers *regs;
+	int ret;
+
+	if (!OVERLAY_EXISTS(dev))
+		return;
+
+	overlay = kzalloc(sizeof(struct intel_overlay), GFP_KERNEL);
+	if (!overlay)
+		return;
+	overlay->dev = dev;
+
+	reg_bo = drm_gem_object_alloc(dev, PAGE_SIZE);
+	if (!reg_bo)
+		goto out_free;
+	overlay->reg_bo = reg_bo->driver_private;
+
+	if (OVERLAY_NONPHYSICAL(dev)) {
+		ret = i915_gem_object_pin(reg_bo, PAGE_SIZE);
+		if (ret) {
+                        DRM_ERROR("failed to pin overlay register bo\n");
+                        goto out_free_bo;
+                }
+		overlay->flip_addr = overlay->reg_bo->gtt_offset;
+	} else {
+		ret = i915_gem_attach_phys_object(dev, reg_bo,
+				I915_GEM_PHYS_OVERLAY_REGS);
+                if (ret) {
+                        DRM_ERROR("failed to attach phys overlay regs\n");
+                        goto out_free_bo;
+                }
+		overlay->flip_addr = overlay->reg_bo->phys_obj->handle->busaddr;
+	}
+
+	/* init all values */
+	overlay->color_key = 0x0101fe;
+	overlay->brightness = -19;
+	overlay->contrast = 75;
+	overlay->saturation = 146;
+
+	regs = intel_overlay_map_regs_atomic(overlay);
+	if (!regs)
+		goto out_free_bo;
+
+	memset(regs, 0, sizeof(struct overlay_registers));
+	update_polyphase_filter(regs);
+
+	update_reg_attrs(overlay, regs);
+
+	intel_overlay_unmap_regs_atomic(overlay);
+
+	dev_priv->overlay = overlay;
+	DRM_INFO("initialized overlay support\n");
+	return;
+
+out_free_bo:
+	drm_gem_object_unreference(reg_bo);
+out_free:
+	kfree(overlay);
+	return;
+}
+
+void intel_cleanup_overlay(struct drm_device *dev)
+{
+        drm_i915_private_t *dev_priv = dev->dev_private;
+
+	if (dev_priv->overlay) {
+		/* The bo's should be free'd by the generic code already.
+		 * Furthermore modesetting teardown happens beforehand so the
+		 * hardware should be off already */
+		BUG_ON(dev_priv->overlay->active);
+
+		kfree(dev_priv->overlay);
+	}
+}
diff --git a/drivers/gpu/drm/i915/intel_sdvo.c b/drivers/gpu/drm/i915/intel_sdvo.c
index 083bec2e50f9..48daee5c9c63 100644
--- a/drivers/gpu/drm/i915/intel_sdvo.c
+++ b/drivers/gpu/drm/i915/intel_sdvo.c
@@ -35,8 +35,7 @@
 #include "i915_drm.h"
 #include "i915_drv.h"
 #include "intel_sdvo_regs.h"
-
-#undef SDVO_DEBUG
+#include <linux/dmi.h>
 
 static char *tv_format_names[] = {
 	"NTSC_M"   , "NTSC_J"  , "NTSC_443",
@@ -356,7 +355,6 @@ static const struct _sdvo_cmd_name {
 #define SDVO_NAME(dev_priv) ((dev_priv)->output_device == SDVOB ? "SDVOB" : "SDVOC")
 #define SDVO_PRIV(output)   ((struct intel_sdvo_priv *) (output)->dev_priv)
 
-#ifdef SDVO_DEBUG
 static void intel_sdvo_debug_write(struct intel_output *intel_output, u8 cmd,
 				   void *args, int args_len)
 {
@@ -379,9 +377,6 @@ static void intel_sdvo_debug_write(struct intel_output *intel_output, u8 cmd,
 		DRM_LOG_KMS("(%02X)", cmd);
 	DRM_LOG_KMS("\n");
 }
-#else
-#define intel_sdvo_debug_write(o, c, a, l)
-#endif
 
 static void intel_sdvo_write_cmd(struct intel_output *intel_output, u8 cmd,
 				 void *args, int args_len)
@@ -398,7 +393,6 @@ static void intel_sdvo_write_cmd(struct intel_output *intel_output, u8 cmd,
 	intel_sdvo_write_byte(intel_output, SDVO_I2C_OPCODE, cmd);
 }
 
-#ifdef SDVO_DEBUG
 static const char *cmd_status_names[] = {
 	"Power on",
 	"Success",
@@ -427,9 +421,6 @@ static void intel_sdvo_debug_response(struct intel_output *intel_output,
 		DRM_LOG_KMS("(??? %d)", status);
 	DRM_LOG_KMS("\n");
 }
-#else
-#define intel_sdvo_debug_response(o, r, l, s)
-#endif
 
 static u8 intel_sdvo_read_response(struct intel_output *intel_output,
 				   void *response, int response_len)
@@ -472,14 +463,63 @@ static int intel_sdvo_get_pixel_multiplier(struct drm_display_mode *mode)
 }
 
 /**
- * Don't check status code from this as it switches the bus back to the
- * SDVO chips which defeats the purpose of doing a bus switch in the first
- * place.
+ * Try to read the response after issuie the DDC switch command. But it
+ * is noted that we must do the action of reading response and issuing DDC
+ * switch command in one I2C transaction. Otherwise when we try to start
+ * another I2C transaction after issuing the DDC bus switch, it will be
+ * switched to the internal SDVO register.
  */
 static void intel_sdvo_set_control_bus_switch(struct intel_output *intel_output,
 					      u8 target)
 {
-	intel_sdvo_write_cmd(intel_output, SDVO_CMD_SET_CONTROL_BUS_SWITCH, &target, 1);
+	struct intel_sdvo_priv *sdvo_priv = intel_output->dev_priv;
+	u8 out_buf[2], cmd_buf[2], ret_value[2], ret;
+	struct i2c_msg msgs[] = {
+		{
+			.addr = sdvo_priv->slave_addr >> 1,
+			.flags = 0,
+			.len = 2,
+			.buf = out_buf,
+		},
+		/* the following two are to read the response */
+		{
+			.addr = sdvo_priv->slave_addr >> 1,
+			.flags = 0,
+			.len = 1,
+			.buf = cmd_buf,
+		},
+		{
+			.addr = sdvo_priv->slave_addr >> 1,
+			.flags = I2C_M_RD,
+			.len = 1,
+			.buf = ret_value,
+		},
+	};
+
+	intel_sdvo_debug_write(intel_output, SDVO_CMD_SET_CONTROL_BUS_SWITCH,
+					&target, 1);
+	/* write the DDC switch command argument */
+	intel_sdvo_write_byte(intel_output, SDVO_I2C_ARG_0, target);
+
+	out_buf[0] = SDVO_I2C_OPCODE;
+	out_buf[1] = SDVO_CMD_SET_CONTROL_BUS_SWITCH;
+	cmd_buf[0] = SDVO_I2C_CMD_STATUS;
+	cmd_buf[1] = 0;
+	ret_value[0] = 0;
+	ret_value[1] = 0;
+
+	ret = i2c_transfer(intel_output->i2c_bus, msgs, 3);
+	if (ret != 3) {
+		/* failure in I2C transfer */
+		DRM_DEBUG_KMS("I2c transfer returned %d\n", ret);
+		return;
+	}
+	if (ret_value[0] != SDVO_CMD_STATUS_SUCCESS) {
+		DRM_DEBUG_KMS("DDC switch command returns response %d\n",
+					ret_value[0]);
+		return;
+	}
+	return;
 }
 
 static bool intel_sdvo_set_target_input(struct intel_output *intel_output, bool target_0, bool target_1)
@@ -1589,6 +1629,32 @@ intel_sdvo_hdmi_sink_detect(struct drm_connector *connector, u16 response)
 	edid = drm_get_edid(&intel_output->base,
 			    intel_output->ddc_bus);
 
+	/* This is only applied to SDVO cards with multiple outputs */
+	if (edid == NULL && intel_sdvo_multifunc_encoder(intel_output)) {
+		uint8_t saved_ddc, temp_ddc;
+		saved_ddc = sdvo_priv->ddc_bus;
+		temp_ddc = sdvo_priv->ddc_bus >> 1;
+		/*
+		 * Don't use the 1 as the argument of DDC bus switch to get
+		 * the EDID. It is used for SDVO SPD ROM.
+		 */
+		while(temp_ddc > 1) {
+			sdvo_priv->ddc_bus = temp_ddc;
+			edid = drm_get_edid(&intel_output->base,
+				intel_output->ddc_bus);
+			if (edid) {
+				/*
+				 * When we can get the EDID, maybe it is the
+				 * correct DDC bus. Update it.
+				 */
+				sdvo_priv->ddc_bus = temp_ddc;
+				break;
+			}
+			temp_ddc >>= 1;
+		}
+		if (edid == NULL)
+			sdvo_priv->ddc_bus = saved_ddc;
+	}
 	/* when there is no edid and no monitor is connected with VGA
 	 * port, try to use the CRT ddc to read the EDID for DVI-connector
 	 */
@@ -1627,6 +1693,10 @@ static enum drm_connector_status intel_sdvo_detect(struct drm_connector *connect
 
 	intel_sdvo_write_cmd(intel_output,
 			     SDVO_CMD_GET_ATTACHED_DISPLAYS, NULL, 0);
+	if (sdvo_priv->is_tv) {
+		/* add 30ms delay when the output type is SDVO-TV */
+		mdelay(30);
+	}
 	status = intel_sdvo_read_response(intel_output, &response, 2);
 
 	DRM_DEBUG_KMS("SDVO response %d %d\n", response & 0xff, response >> 8);
@@ -2214,6 +2284,25 @@ intel_sdvo_get_slave_addr(struct drm_device *dev, int output_device)
 		return 0x72;
 }
 
+static int intel_sdvo_bad_tv_callback(const struct dmi_system_id *id)
+{
+	DRM_DEBUG_KMS("Ignoring bad SDVO TV connector for %s\n", id->ident);
+	return 1;
+}
+
+static struct dmi_system_id intel_sdvo_bad_tv[] = {
+	{
+		.callback = intel_sdvo_bad_tv_callback,
+		.ident = "IntelG45/ICH10R/DME1737",
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "IBM CORPORATION"),
+			DMI_MATCH(DMI_PRODUCT_NAME, "4800784"),
+		},
+	},
+
+	{ }	/* terminating entry */
+};
+
 static bool
 intel_sdvo_output_setup(struct intel_output *intel_output, uint16_t flags)
 {
@@ -2254,7 +2343,8 @@ intel_sdvo_output_setup(struct intel_output *intel_output, uint16_t flags)
 					(1 << INTEL_SDVO_NON_TV_CLONE_BIT) |
 					(1 << INTEL_ANALOG_CLONE_BIT);
 		}
-	} else if (flags & SDVO_OUTPUT_SVID0) {
+	} else if ((flags & SDVO_OUTPUT_SVID0) &&
+		   !dmi_check_system(intel_sdvo_bad_tv)) {
 
 		sdvo_priv->controlled_output = SDVO_OUTPUT_SVID0;
 		encoder->encoder_type = DRM_MODE_ENCODER_TVDAC;
@@ -2276,6 +2366,14 @@ intel_sdvo_output_setup(struct intel_output *intel_output, uint16_t flags)
 		connector->connector_type = DRM_MODE_CONNECTOR_VGA;
 		intel_output->clone_mask = (1 << INTEL_SDVO_NON_TV_CLONE_BIT) |
 					(1 << INTEL_ANALOG_CLONE_BIT);
+	} else if (flags & SDVO_OUTPUT_CVBS0) {
+
+		sdvo_priv->controlled_output = SDVO_OUTPUT_CVBS0;
+		encoder->encoder_type = DRM_MODE_ENCODER_TVDAC;
+		connector->connector_type = DRM_MODE_CONNECTOR_SVIDEO;
+		sdvo_priv->is_tv = true;
+		intel_output->needs_tv_clock = true;
+		intel_output->clone_mask = 1 << INTEL_SDVO_TV_CLONE_BIT;
 	} else if (flags & SDVO_OUTPUT_LVDS0) {
 
 		sdvo_priv->controlled_output = SDVO_OUTPUT_LVDS0;
@@ -2668,6 +2766,7 @@ static void intel_sdvo_create_enhance_property(struct drm_connector *connector)
 
 bool intel_sdvo_init(struct drm_device *dev, int output_device)
 {
+	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct drm_connector *connector;
 	struct intel_output *intel_output;
 	struct intel_sdvo_priv *sdvo_priv;
@@ -2714,10 +2813,12 @@ bool intel_sdvo_init(struct drm_device *dev, int output_device)
 		intel_output->ddc_bus = intel_i2c_create(dev, GPIOE, "SDVOB DDC BUS");
 		sdvo_priv->analog_ddc_bus = intel_i2c_create(dev, GPIOA,
 						"SDVOB/VGA DDC BUS");
+		dev_priv->hotplug_supported_mask |= SDVOB_HOTPLUG_INT_STATUS;
 	} else {
 		intel_output->ddc_bus = intel_i2c_create(dev, GPIOE, "SDVOC DDC BUS");
 		sdvo_priv->analog_ddc_bus = intel_i2c_create(dev, GPIOA,
 						"SDVOC/VGA DDC BUS");
+		dev_priv->hotplug_supported_mask |= SDVOC_HOTPLUG_INT_STATUS;
 	}
 
 	if (intel_output->ddc_bus == NULL)
@@ -2726,7 +2827,7 @@ bool intel_sdvo_init(struct drm_device *dev, int output_device)
 	/* Wrap with our custom algo which switches to DDC mode */
 	intel_output->ddc_bus->algo = &intel_sdvo_i2c_bit_algo;
 
-	/* In defaut case sdvo lvds is false */
+	/* In default case sdvo lvds is false */
 	intel_sdvo_get_capabilities(intel_output, &sdvo_priv->caps);
 
 	if (intel_sdvo_output_setup(intel_output,
diff --git a/drivers/gpu/drm/i915/intel_tv.c b/drivers/gpu/drm/i915/intel_tv.c
index c64eab493fb0..552ec110b741 100644
--- a/drivers/gpu/drm/i915/intel_tv.c
+++ b/drivers/gpu/drm/i915/intel_tv.c
@@ -1082,7 +1082,8 @@ intel_tv_mode_valid(struct drm_connector *connector, struct drm_display_mode *mo
 	const struct tv_mode *tv_mode = intel_tv_mode_find(intel_output);
 
 	/* Ensure TV refresh is close to desired refresh */
-	if (tv_mode && abs(tv_mode->refresh - drm_mode_vrefresh(mode)) < 10)
+	if (tv_mode && abs(tv_mode->refresh - drm_mode_vrefresh(mode) * 1000)
+				< 1000)
 		return MODE_OK;
 	return MODE_CLOCK_RANGE;
 }
@@ -1212,20 +1213,17 @@ intel_tv_mode_set(struct drm_encoder *encoder, struct drm_display_mode *mode,
 		tv_ctl |= TV_TRILEVEL_SYNC;
 	if (tv_mode->pal_burst)
 		tv_ctl |= TV_PAL_BURST;
+
 	scctl1 = 0;
-	/* dda1 implies valid video levels */
-	if (tv_mode->dda1_inc) {
+	if (tv_mode->dda1_inc)
 		scctl1 |= TV_SC_DDA1_EN;
-	}
-
 	if (tv_mode->dda2_inc)
 		scctl1 |= TV_SC_DDA2_EN;
-
 	if (tv_mode->dda3_inc)
 		scctl1 |= TV_SC_DDA3_EN;
-
 	scctl1 |= tv_mode->sc_reset;
-	scctl1 |= video_levels->burst << TV_BURST_LEVEL_SHIFT;
+	if (video_levels)
+		scctl1 |= video_levels->burst << TV_BURST_LEVEL_SHIFT;
 	scctl1 |= tv_mode->dda1_inc << TV_SCDDA1_INC_SHIFT;
 
 	scctl2 = tv_mode->dda2_size << TV_SCDDA2_SIZE_SHIFT |
@@ -1415,16 +1413,16 @@ intel_tv_detect_type (struct drm_crtc *crtc, struct intel_output *intel_output)
 	 *  0 0 0 Component
 	 */
 	if ((tv_dac & TVDAC_SENSE_MASK) == (TVDAC_B_SENSE | TVDAC_C_SENSE)) {
-		DRM_DEBUG("Detected Composite TV connection\n");
+		DRM_DEBUG_KMS("Detected Composite TV connection\n");
 		type = DRM_MODE_CONNECTOR_Composite;
 	} else if ((tv_dac & (TVDAC_A_SENSE|TVDAC_B_SENSE)) == TVDAC_A_SENSE) {
-		DRM_DEBUG("Detected S-Video TV connection\n");
+		DRM_DEBUG_KMS("Detected S-Video TV connection\n");
 		type = DRM_MODE_CONNECTOR_SVIDEO;
 	} else if ((tv_dac & TVDAC_SENSE_MASK) == 0) {
-		DRM_DEBUG("Detected Component TV connection\n");
+		DRM_DEBUG_KMS("Detected Component TV connection\n");
 		type = DRM_MODE_CONNECTOR_Component;
 	} else {
-		DRM_DEBUG("No TV connection detected\n");
+		DRM_DEBUG_KMS("No TV connection detected\n");
 		type = -1;
 	}
 
@@ -1701,6 +1699,41 @@ static const struct drm_encoder_funcs intel_tv_enc_funcs = {
 	.destroy = intel_tv_enc_destroy,
 };
 
+/*
+ * Enumerate the child dev array parsed from VBT to check whether
+ * the integrated TV is present.
+ * If it is present, return 1.
+ * If it is not present, return false.
+ * If no child dev is parsed from VBT, it assumes that the TV is present.
+ */
+static int tv_is_present_in_vbt(struct drm_device *dev)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct child_device_config *p_child;
+	int i, ret;
+
+	if (!dev_priv->child_dev_num)
+		return 1;
+
+	ret = 0;
+	for (i = 0; i < dev_priv->child_dev_num; i++) {
+		p_child = dev_priv->child_dev + i;
+		/*
+		 * If the device type is not TV, continue.
+		 */
+		if (p_child->device_type != DEVICE_TYPE_INT_TV &&
+			p_child->device_type != DEVICE_TYPE_TV)
+			continue;
+		/* Only when the addin_offset is non-zero, it is regarded
+		 * as present.
+		 */
+		if (p_child->addin_offset) {
+			ret = 1;
+			break;
+		}
+	}
+	return ret;
+}
 
 void
 intel_tv_init(struct drm_device *dev)
@@ -1716,6 +1749,10 @@ intel_tv_init(struct drm_device *dev)
 	if ((I915_READ(TV_CTL) & TV_FUSE_STATE_MASK) == TV_FUSE_STATE_DISABLED)
 		return;
 
+	if (!tv_is_present_in_vbt(dev)) {
+		DRM_DEBUG_KMS("Integrated TV is not present.\n");
+		return;
+	}
 	/* Even if we have an encoder we may not have a connector */
 	if (!dev_priv->int_tv_support)
 		return;
diff --git a/drivers/gpu/drm/mga/mga_drv.c b/drivers/gpu/drm/mga/mga_drv.c
index 97ee566ef749..ddfe16197b59 100644
--- a/drivers/gpu/drm/mga/mga_drv.c
+++ b/drivers/gpu/drm/mga/mga_drv.c
@@ -68,7 +68,7 @@ static struct drm_driver driver = {
 		.owner = THIS_MODULE,
 		.open = drm_open,
 		.release = drm_release,
-		.ioctl = drm_ioctl,
+		.unlocked_ioctl = drm_ioctl,
 		.mmap = drm_mmap,
 		.poll = drm_poll,
 		.fasync = drm_fasync,
diff --git a/drivers/gpu/drm/mga/mga_ioc32.c b/drivers/gpu/drm/mga/mga_ioc32.c
index 30d00478ddee..c1f877b7bac1 100644
--- a/drivers/gpu/drm/mga/mga_ioc32.c
+++ b/drivers/gpu/drm/mga/mga_ioc32.c
@@ -100,8 +100,7 @@ static int compat_mga_init(struct file *file, unsigned int cmd,
 	if (err)
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_MGA_INIT, (unsigned long)init);
+	return drm_ioctl(file, DRM_IOCTL_MGA_INIT, (unsigned long)init);
 }
 
 typedef struct drm_mga_getparam32 {
@@ -125,8 +124,7 @@ static int compat_mga_getparam(struct file *file, unsigned int cmd,
 			  &getparam->value))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_MGA_GETPARAM, (unsigned long)getparam);
+	return drm_ioctl(file, DRM_IOCTL_MGA_GETPARAM, (unsigned long)getparam);
 }
 
 typedef struct drm_mga_drm_bootstrap32 {
@@ -166,8 +164,7 @@ static int compat_mga_dma_bootstrap(struct file *file, unsigned int cmd,
 	    || __put_user(dma_bootstrap32.agp_size, &dma_bootstrap->agp_size))
 		return -EFAULT;
 
-	err = drm_ioctl(file->f_path.dentry->d_inode, file,
-			DRM_IOCTL_MGA_DMA_BOOTSTRAP,
+	err = drm_ioctl(file, DRM_IOCTL_MGA_DMA_BOOTSTRAP,
 			(unsigned long)dma_bootstrap);
 	if (err)
 		return err;
@@ -220,12 +217,10 @@ long mga_compat_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
 	if (nr < DRM_COMMAND_BASE + DRM_ARRAY_SIZE(mga_compat_ioctls))
 		fn = mga_compat_ioctls[nr - DRM_COMMAND_BASE];
 
-	lock_kernel();		/* XXX for now */
 	if (fn != NULL)
 		ret = (*fn) (filp, cmd, arg);
 	else
-		ret = drm_ioctl(filp->f_path.dentry->d_inode, filp, cmd, arg);
-	unlock_kernel();
+		ret = drm_ioctl(filp, cmd, arg);
 
 	return ret;
 }
diff --git a/drivers/gpu/drm/nouveau/Kconfig b/drivers/gpu/drm/nouveau/Kconfig
new file mode 100644
index 000000000000..1175429da102
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/Kconfig
@@ -0,0 +1,44 @@
+config DRM_NOUVEAU
+	tristate "Nouveau (nVidia) cards"
+	depends on DRM
+        select FW_LOADER
+	select DRM_KMS_HELPER
+	select DRM_TTM
+	select FB_CFB_FILLRECT
+	select FB_CFB_COPYAREA
+	select FB_CFB_IMAGEBLIT
+	select FB
+	select FRAMEBUFFER_CONSOLE if !EMBEDDED
+	select FB_BACKLIGHT if DRM_NOUVEAU_BACKLIGHT
+	help
+	  Choose this option for open-source nVidia support.
+
+config DRM_NOUVEAU_BACKLIGHT
+	bool "Support for backlight control"
+	depends on DRM_NOUVEAU
+	default y
+	help
+	  Say Y here if you want to control the backlight of your display
+	  (e.g. a laptop panel).
+
+config DRM_NOUVEAU_DEBUG
+	bool "Build in Nouveau's debugfs support"
+	depends on DRM_NOUVEAU && DEBUG_FS
+	default y
+	help
+	  Say Y here if you want Nouveau to output debugging information
+	  via debugfs.
+
+menu "I2C encoder or helper chips"
+     depends on DRM && DRM_KMS_HELPER && I2C
+
+config DRM_I2C_CH7006
+	tristate "Chrontel ch7006 TV encoder"
+	default m if DRM_NOUVEAU
+	help
+	  Support for Chrontel ch7006 and similar TV encoders, found
+	  on some nVidia video cards.
+
+	  This driver is currently only useful if you're also using
+	  the nouveau driver.
+endmenu
diff --git a/drivers/gpu/drm/nouveau/Makefile b/drivers/gpu/drm/nouveau/Makefile
new file mode 100644
index 000000000000..32db806f3b5a
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/Makefile
@@ -0,0 +1,32 @@
+#
+# Makefile for the drm device driver.  This driver provides support for the
+# Direct Rendering Infrastructure (DRI) in XFree86 4.1.0 and higher.
+
+ccflags-y := -Iinclude/drm
+nouveau-y := nouveau_drv.o nouveau_state.o nouveau_channel.o nouveau_mem.o \
+             nouveau_object.o nouveau_irq.o nouveau_notifier.o \
+             nouveau_sgdma.o nouveau_dma.o \
+             nouveau_bo.o nouveau_fence.o nouveau_gem.o nouveau_ttm.o \
+             nouveau_hw.o nouveau_calc.o nouveau_bios.o nouveau_i2c.o \
+             nouveau_display.o nouveau_connector.o nouveau_fbcon.o \
+             nouveau_dp.o nouveau_grctx.o \
+             nv04_timer.o \
+             nv04_mc.o nv40_mc.o nv50_mc.o \
+             nv04_fb.o nv10_fb.o nv40_fb.o \
+             nv04_fifo.o nv10_fifo.o nv40_fifo.o nv50_fifo.o \
+             nv04_graph.o nv10_graph.o nv20_graph.o \
+             nv40_graph.o nv50_graph.o \
+             nv40_grctx.o nv50_grctx.o \
+             nv04_instmem.o nv50_instmem.o \
+             nv50_crtc.o nv50_dac.o nv50_sor.o \
+             nv50_cursor.o nv50_display.o nv50_fbcon.o \
+             nv04_dac.o nv04_dfp.o nv04_tv.o nv17_tv.o nv17_tv_modes.o \
+             nv04_crtc.o nv04_display.o nv04_cursor.o nv04_fbcon.o \
+             nv17_gpio.o
+
+nouveau-$(CONFIG_DRM_NOUVEAU_DEBUG) += nouveau_debugfs.o
+nouveau-$(CONFIG_COMPAT) += nouveau_ioc32.o
+nouveau-$(CONFIG_DRM_NOUVEAU_BACKLIGHT) += nouveau_backlight.o
+nouveau-$(CONFIG_ACPI) += nouveau_acpi.o
+
+obj-$(CONFIG_DRM_NOUVEAU)+= nouveau.o
diff --git a/drivers/gpu/drm/nouveau/nouveau_acpi.c b/drivers/gpu/drm/nouveau/nouveau_acpi.c
new file mode 100644
index 000000000000..0e0730a53137
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_acpi.c
@@ -0,0 +1,205 @@
+#include <linux/pci.h>
+#include <linux/acpi.h>
+#include <acpi/acpi_drivers.h>
+#include <acpi/acpi_bus.h>
+
+#include "drmP.h"
+#include "drm.h"
+#include "drm_sarea.h"
+#include "drm_crtc_helper.h"
+#include "nouveau_drv.h"
+#include "nouveau_drm.h"
+#include "nv50_display.h"
+
+#include <linux/vga_switcheroo.h>
+
+#define NOUVEAU_DSM_SUPPORTED 0x00
+#define NOUVEAU_DSM_SUPPORTED_FUNCTIONS 0x00
+
+#define NOUVEAU_DSM_ACTIVE 0x01
+#define NOUVEAU_DSM_ACTIVE_QUERY 0x00
+
+#define NOUVEAU_DSM_LED 0x02
+#define NOUVEAU_DSM_LED_STATE 0x00
+#define NOUVEAU_DSM_LED_OFF 0x10
+#define NOUVEAU_DSM_LED_STAMINA 0x11
+#define NOUVEAU_DSM_LED_SPEED 0x12
+
+#define NOUVEAU_DSM_POWER 0x03
+#define NOUVEAU_DSM_POWER_STATE 0x00
+#define NOUVEAU_DSM_POWER_SPEED 0x01
+#define NOUVEAU_DSM_POWER_STAMINA 0x02
+
+static struct nouveau_dsm_priv {
+	bool dsm_detected;
+	acpi_handle dhandle;
+	acpi_handle dsm_handle;
+} nouveau_dsm_priv;
+
+static const char nouveau_dsm_muid[] = {
+	0xA0, 0xA0, 0x95, 0x9D, 0x60, 0x00, 0x48, 0x4D,
+	0xB3, 0x4D, 0x7E, 0x5F, 0xEA, 0x12, 0x9F, 0xD4,
+};
+
+static int nouveau_dsm(acpi_handle handle, int func, int arg, int *result)
+{
+	struct acpi_buffer output = { ACPI_ALLOCATE_BUFFER, NULL };
+	struct acpi_object_list input;
+	union acpi_object params[4];
+	union acpi_object *obj;
+	int err;
+
+	input.count = 4;
+	input.pointer = params;
+	params[0].type = ACPI_TYPE_BUFFER;
+	params[0].buffer.length = sizeof(nouveau_dsm_muid);
+	params[0].buffer.pointer = (char *)nouveau_dsm_muid;
+	params[1].type = ACPI_TYPE_INTEGER;
+	params[1].integer.value = 0x00000102;
+	params[2].type = ACPI_TYPE_INTEGER;
+	params[2].integer.value = func;
+	params[3].type = ACPI_TYPE_INTEGER;
+	params[3].integer.value = arg;
+
+	err = acpi_evaluate_object(handle, "_DSM", &input, &output);
+	if (err) {
+		printk(KERN_INFO "failed to evaluate _DSM: %d\n", err);
+		return err;
+	}
+
+	obj = (union acpi_object *)output.pointer;
+
+	if (obj->type == ACPI_TYPE_INTEGER)
+		if (obj->integer.value == 0x80000002)
+			return -ENODEV;
+
+	if (obj->type == ACPI_TYPE_BUFFER) {
+		if (obj->buffer.length == 4 && result) {
+			*result = 0;
+			*result |= obj->buffer.pointer[0];
+			*result |= (obj->buffer.pointer[1] << 8);
+			*result |= (obj->buffer.pointer[2] << 16);
+			*result |= (obj->buffer.pointer[3] << 24);
+		}
+	}
+
+	kfree(output.pointer);
+	return 0;
+}
+
+static int nouveau_dsm_switch_mux(acpi_handle handle, int mux_id)
+{
+	return nouveau_dsm(handle, NOUVEAU_DSM_LED, mux_id, NULL);
+}
+
+static int nouveau_dsm_set_discrete_state(acpi_handle handle, enum vga_switcheroo_state state)
+{
+	int arg;
+	if (state == VGA_SWITCHEROO_ON)
+		arg = NOUVEAU_DSM_POWER_SPEED;
+	else
+		arg = NOUVEAU_DSM_POWER_STAMINA;
+	nouveau_dsm(handle, NOUVEAU_DSM_POWER, arg, NULL);
+	return 0;
+}
+
+static int nouveau_dsm_switchto(enum vga_switcheroo_client_id id)
+{
+	if (id == VGA_SWITCHEROO_IGD)
+		return nouveau_dsm_switch_mux(nouveau_dsm_priv.dsm_handle, NOUVEAU_DSM_LED_STAMINA);
+	else
+		return nouveau_dsm_switch_mux(nouveau_dsm_priv.dsm_handle, NOUVEAU_DSM_LED_SPEED);
+}
+
+static int nouveau_dsm_power_state(enum vga_switcheroo_client_id id,
+				   enum vga_switcheroo_state state)
+{
+	if (id == VGA_SWITCHEROO_IGD)
+		return 0;
+
+	return nouveau_dsm_set_discrete_state(nouveau_dsm_priv.dsm_handle, state);
+}
+
+static int nouveau_dsm_init(void)
+{
+	return 0;
+}
+
+static int nouveau_dsm_get_client_id(struct pci_dev *pdev)
+{
+	if (nouveau_dsm_priv.dhandle == DEVICE_ACPI_HANDLE(&pdev->dev))
+		return VGA_SWITCHEROO_IGD;
+	else
+		return VGA_SWITCHEROO_DIS;
+}
+
+static struct vga_switcheroo_handler nouveau_dsm_handler = {
+	.switchto = nouveau_dsm_switchto,
+	.power_state = nouveau_dsm_power_state,
+	.init = nouveau_dsm_init,
+	.get_client_id = nouveau_dsm_get_client_id,
+};
+
+static bool nouveau_dsm_pci_probe(struct pci_dev *pdev)
+{
+	acpi_handle dhandle, nvidia_handle;
+	acpi_status status;
+	int ret;
+	uint32_t result;
+
+	dhandle = DEVICE_ACPI_HANDLE(&pdev->dev);
+	if (!dhandle)
+		return false;
+	status = acpi_get_handle(dhandle, "_DSM", &nvidia_handle);
+	if (ACPI_FAILURE(status)) {
+		return false;
+	}
+
+	ret= nouveau_dsm(nvidia_handle, NOUVEAU_DSM_SUPPORTED,
+			 NOUVEAU_DSM_SUPPORTED_FUNCTIONS, &result);
+	if (ret < 0)
+		return false;
+
+	nouveau_dsm_priv.dhandle = dhandle;
+	nouveau_dsm_priv.dsm_handle = nvidia_handle;
+	return true;
+}
+
+static bool nouveau_dsm_detect(void)
+{
+	char acpi_method_name[255] = { 0 };
+	struct acpi_buffer buffer = {sizeof(acpi_method_name), acpi_method_name};
+	struct pci_dev *pdev = NULL;
+	int has_dsm = 0;
+	int vga_count = 0;
+	while ((pdev = pci_get_class(PCI_CLASS_DISPLAY_VGA << 8, pdev)) != NULL) {
+		vga_count++;
+
+		has_dsm |= (nouveau_dsm_pci_probe(pdev) == true);
+	}
+
+	if (vga_count == 2 && has_dsm) {
+		acpi_get_name(nouveau_dsm_priv.dsm_handle, ACPI_FULL_PATHNAME, &buffer);
+		printk(KERN_INFO "VGA switcheroo: detected DSM switching method %s handle\n",
+		       acpi_method_name);
+		nouveau_dsm_priv.dsm_detected = true;
+		return true;
+	}
+	return false;
+}
+
+void nouveau_register_dsm_handler(void)
+{
+	bool r;
+
+	r = nouveau_dsm_detect();
+	if (!r)
+		return;
+
+	vga_switcheroo_register_handler(&nouveau_dsm_handler);
+}
+
+void nouveau_unregister_dsm_handler(void)
+{
+	vga_switcheroo_unregister_handler();
+}
diff --git a/drivers/gpu/drm/nouveau/nouveau_backlight.c b/drivers/gpu/drm/nouveau/nouveau_backlight.c
new file mode 100644
index 000000000000..406228f4a2a0
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_backlight.c
@@ -0,0 +1,159 @@
+/*
+ * Copyright (C) 2009 Red Hat <mjg@redhat.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+/*
+ * Authors:
+ *  Matthew Garrett <mjg@redhat.com>
+ *
+ * Register locations derived from NVClock by Roderick Colenbrander
+ */
+
+#include <linux/backlight.h>
+
+#include "drmP.h"
+#include "nouveau_drv.h"
+#include "nouveau_drm.h"
+#include "nouveau_reg.h"
+
+static int nv40_get_intensity(struct backlight_device *bd)
+{
+	struct drm_device *dev = bl_get_data(bd);
+	int val = (nv_rd32(dev, NV40_PMC_BACKLIGHT) & NV40_PMC_BACKLIGHT_MASK)
+									>> 16;
+
+	return val;
+}
+
+static int nv40_set_intensity(struct backlight_device *bd)
+{
+	struct drm_device *dev = bl_get_data(bd);
+	int val = bd->props.brightness;
+	int reg = nv_rd32(dev, NV40_PMC_BACKLIGHT);
+
+	nv_wr32(dev, NV40_PMC_BACKLIGHT,
+		 (val << 16) | (reg & ~NV40_PMC_BACKLIGHT_MASK));
+
+	return 0;
+}
+
+static struct backlight_ops nv40_bl_ops = {
+	.options = BL_CORE_SUSPENDRESUME,
+	.get_brightness = nv40_get_intensity,
+	.update_status = nv40_set_intensity,
+};
+
+static int nv50_get_intensity(struct backlight_device *bd)
+{
+	struct drm_device *dev = bl_get_data(bd);
+
+	return nv_rd32(dev, NV50_PDISPLAY_SOR_BACKLIGHT);
+}
+
+static int nv50_set_intensity(struct backlight_device *bd)
+{
+	struct drm_device *dev = bl_get_data(bd);
+	int val = bd->props.brightness;
+
+	nv_wr32(dev, NV50_PDISPLAY_SOR_BACKLIGHT,
+		val | NV50_PDISPLAY_SOR_BACKLIGHT_ENABLE);
+	return 0;
+}
+
+static struct backlight_ops nv50_bl_ops = {
+	.options = BL_CORE_SUSPENDRESUME,
+	.get_brightness = nv50_get_intensity,
+	.update_status = nv50_set_intensity,
+};
+
+static int nouveau_nv40_backlight_init(struct drm_device *dev)
+{
+	struct backlight_properties props;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct backlight_device *bd;
+
+	if (!(nv_rd32(dev, NV40_PMC_BACKLIGHT) & NV40_PMC_BACKLIGHT_MASK))
+		return 0;
+
+	memset(&props, 0, sizeof(struct backlight_properties));
+	props.max_brightness = 31;
+	bd = backlight_device_register("nv_backlight", &dev->pdev->dev, dev,
+				       &nv40_bl_ops, &props);
+	if (IS_ERR(bd))
+		return PTR_ERR(bd);
+
+	dev_priv->backlight = bd;
+	bd->props.brightness = nv40_get_intensity(bd);
+	backlight_update_status(bd);
+
+	return 0;
+}
+
+static int nouveau_nv50_backlight_init(struct drm_device *dev)
+{
+	struct backlight_properties props;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct backlight_device *bd;
+
+	if (!nv_rd32(dev, NV50_PDISPLAY_SOR_BACKLIGHT))
+		return 0;
+
+	memset(&props, 0, sizeof(struct backlight_properties));
+	props.max_brightness = 1025;
+	bd = backlight_device_register("nv_backlight", &dev->pdev->dev, dev,
+				       &nv50_bl_ops, &props);
+	if (IS_ERR(bd))
+		return PTR_ERR(bd);
+
+	dev_priv->backlight = bd;
+	bd->props.brightness = nv50_get_intensity(bd);
+	backlight_update_status(bd);
+	return 0;
+}
+
+int nouveau_backlight_init(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	switch (dev_priv->card_type) {
+	case NV_40:
+		return nouveau_nv40_backlight_init(dev);
+	case NV_50:
+		return nouveau_nv50_backlight_init(dev);
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+void nouveau_backlight_exit(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	if (dev_priv->backlight) {
+		backlight_device_unregister(dev_priv->backlight);
+		dev_priv->backlight = NULL;
+	}
+}
diff --git a/drivers/gpu/drm/nouveau/nouveau_bios.c b/drivers/gpu/drm/nouveau/nouveau_bios.c
new file mode 100644
index 000000000000..75bceee76044
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_bios.c
@@ -0,0 +1,6112 @@
+/*
+ * Copyright 2005-2006 Erik Waling
+ * Copyright 2006 Stephane Marchesin
+ * Copyright 2007-2009 Stuart Bennett
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF
+ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include "drmP.h"
+#define NV_DEBUG_NOTRACE
+#include "nouveau_drv.h"
+#include "nouveau_hw.h"
+
+/* these defines are made up */
+#define NV_CIO_CRE_44_HEADA 0x0
+#define NV_CIO_CRE_44_HEADB 0x3
+#define FEATURE_MOBILE 0x10	/* also FEATURE_QUADRO for BMP */
+#define LEGACY_I2C_CRT 0x80
+#define LEGACY_I2C_PANEL 0x81
+#define LEGACY_I2C_TV 0x82
+
+#define EDID1_LEN 128
+
+#define BIOSLOG(sip, fmt, arg...) NV_DEBUG(sip->dev, fmt, ##arg)
+#define LOG_OLD_VALUE(x)
+
+#define ROM16(x) le16_to_cpu(*(uint16_t *)&(x))
+#define ROM32(x) le32_to_cpu(*(uint32_t *)&(x))
+
+struct init_exec {
+	bool execute;
+	bool repeat;
+};
+
+static bool nv_cksum(const uint8_t *data, unsigned int length)
+{
+	/*
+	 * There's a few checksums in the BIOS, so here's a generic checking
+	 * function.
+	 */
+	int i;
+	uint8_t sum = 0;
+
+	for (i = 0; i < length; i++)
+		sum += data[i];
+
+	if (sum)
+		return true;
+
+	return false;
+}
+
+static int
+score_vbios(struct drm_device *dev, const uint8_t *data, const bool writeable)
+{
+	if (!(data[0] == 0x55 && data[1] == 0xAA)) {
+		NV_TRACEWARN(dev, "... BIOS signature not found\n");
+		return 0;
+	}
+
+	if (nv_cksum(data, data[2] * 512)) {
+		NV_TRACEWARN(dev, "... BIOS checksum invalid\n");
+		/* if a ro image is somewhat bad, it's probably all rubbish */
+		return writeable ? 2 : 1;
+	} else
+		NV_TRACE(dev, "... appears to be valid\n");
+
+	return 3;
+}
+
+static void load_vbios_prom(struct drm_device *dev, uint8_t *data)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t pci_nv_20, save_pci_nv_20;
+	int pcir_ptr;
+	int i;
+
+	if (dev_priv->card_type >= NV_50)
+		pci_nv_20 = 0x88050;
+	else
+		pci_nv_20 = NV_PBUS_PCI_NV_20;
+
+	/* enable ROM access */
+	save_pci_nv_20 = nvReadMC(dev, pci_nv_20);
+	nvWriteMC(dev, pci_nv_20,
+		  save_pci_nv_20 & ~NV_PBUS_PCI_NV_20_ROM_SHADOW_ENABLED);
+
+	/* bail if no rom signature */
+	if (nv_rd08(dev, NV_PROM_OFFSET) != 0x55 ||
+	    nv_rd08(dev, NV_PROM_OFFSET + 1) != 0xaa)
+		goto out;
+
+	/* additional check (see note below) - read PCI record header */
+	pcir_ptr = nv_rd08(dev, NV_PROM_OFFSET + 0x18) |
+		   nv_rd08(dev, NV_PROM_OFFSET + 0x19) << 8;
+	if (nv_rd08(dev, NV_PROM_OFFSET + pcir_ptr) != 'P' ||
+	    nv_rd08(dev, NV_PROM_OFFSET + pcir_ptr + 1) != 'C' ||
+	    nv_rd08(dev, NV_PROM_OFFSET + pcir_ptr + 2) != 'I' ||
+	    nv_rd08(dev, NV_PROM_OFFSET + pcir_ptr + 3) != 'R')
+		goto out;
+
+	/* on some 6600GT/6800LE prom reads are messed up.  nvclock alleges a
+	 * a good read may be obtained by waiting or re-reading (cargocult: 5x)
+	 * each byte.  we'll hope pramin has something usable instead
+	 */
+	for (i = 0; i < NV_PROM_SIZE; i++)
+		data[i] = nv_rd08(dev, NV_PROM_OFFSET + i);
+
+out:
+	/* disable ROM access */
+	nvWriteMC(dev, pci_nv_20,
+		  save_pci_nv_20 | NV_PBUS_PCI_NV_20_ROM_SHADOW_ENABLED);
+}
+
+static void load_vbios_pramin(struct drm_device *dev, uint8_t *data)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t old_bar0_pramin = 0;
+	int i;
+
+	if (dev_priv->card_type >= NV_50) {
+		uint32_t vbios_vram = (nv_rd32(dev, 0x619f04) & ~0xff) << 8;
+
+		if (!vbios_vram)
+			vbios_vram = (nv_rd32(dev, 0x1700) << 16) + 0xf0000;
+
+		old_bar0_pramin = nv_rd32(dev, 0x1700);
+		nv_wr32(dev, 0x1700, vbios_vram >> 16);
+	}
+
+	/* bail if no rom signature */
+	if (nv_rd08(dev, NV_PRAMIN_OFFSET) != 0x55 ||
+	    nv_rd08(dev, NV_PRAMIN_OFFSET + 1) != 0xaa)
+		goto out;
+
+	for (i = 0; i < NV_PROM_SIZE; i++)
+		data[i] = nv_rd08(dev, NV_PRAMIN_OFFSET + i);
+
+out:
+	if (dev_priv->card_type >= NV_50)
+		nv_wr32(dev, 0x1700, old_bar0_pramin);
+}
+
+static void load_vbios_pci(struct drm_device *dev, uint8_t *data)
+{
+	void __iomem *rom = NULL;
+	size_t rom_len;
+	int ret;
+
+	ret = pci_enable_rom(dev->pdev);
+	if (ret)
+		return;
+
+	rom = pci_map_rom(dev->pdev, &rom_len);
+	if (!rom)
+		goto out;
+	memcpy_fromio(data, rom, rom_len);
+	pci_unmap_rom(dev->pdev, rom);
+
+out:
+	pci_disable_rom(dev->pdev);
+}
+
+struct methods {
+	const char desc[8];
+	void (*loadbios)(struct drm_device *, uint8_t *);
+	const bool rw;
+};
+
+static struct methods nv04_methods[] = {
+	{ "PROM", load_vbios_prom, false },
+	{ "PRAMIN", load_vbios_pramin, true },
+	{ "PCIROM", load_vbios_pci, true },
+};
+
+static struct methods nv50_methods[] = {
+	{ "PRAMIN", load_vbios_pramin, true },
+	{ "PROM", load_vbios_prom, false },
+	{ "PCIROM", load_vbios_pci, true },
+};
+
+#define METHODCNT 3
+
+static bool NVShadowVBIOS(struct drm_device *dev, uint8_t *data)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct methods *methods;
+	int i;
+	int testscore = 3;
+	int scores[METHODCNT];
+
+	if (nouveau_vbios) {
+		methods = nv04_methods;
+		for (i = 0; i < METHODCNT; i++)
+			if (!strcasecmp(nouveau_vbios, methods[i].desc))
+				break;
+
+		if (i < METHODCNT) {
+			NV_INFO(dev, "Attempting to use BIOS image from %s\n",
+				methods[i].desc);
+
+			methods[i].loadbios(dev, data);
+			if (score_vbios(dev, data, methods[i].rw))
+				return true;
+		}
+
+		NV_ERROR(dev, "VBIOS source \'%s\' invalid\n", nouveau_vbios);
+	}
+
+	if (dev_priv->card_type < NV_50)
+		methods = nv04_methods;
+	else
+		methods = nv50_methods;
+
+	for (i = 0; i < METHODCNT; i++) {
+		NV_TRACE(dev, "Attempting to load BIOS image from %s\n",
+			 methods[i].desc);
+		data[0] = data[1] = 0;	/* avoid reuse of previous image */
+		methods[i].loadbios(dev, data);
+		scores[i] = score_vbios(dev, data, methods[i].rw);
+		if (scores[i] == testscore)
+			return true;
+	}
+
+	while (--testscore > 0) {
+		for (i = 0; i < METHODCNT; i++) {
+			if (scores[i] == testscore) {
+				NV_TRACE(dev, "Using BIOS image from %s\n",
+					 methods[i].desc);
+				methods[i].loadbios(dev, data);
+				return true;
+			}
+		}
+	}
+
+	NV_ERROR(dev, "No valid BIOS image found\n");
+	return false;
+}
+
+struct init_tbl_entry {
+	char *name;
+	uint8_t id;
+	int (*handler)(struct nvbios *, uint16_t, struct init_exec *);
+};
+
+struct bit_entry {
+	uint8_t id[2];
+	uint16_t length;
+	uint16_t offset;
+};
+
+static int parse_init_table(struct nvbios *, unsigned int, struct init_exec *);
+
+#define MACRO_INDEX_SIZE	2
+#define MACRO_SIZE		8
+#define CONDITION_SIZE		12
+#define IO_FLAG_CONDITION_SIZE	9
+#define IO_CONDITION_SIZE	5
+#define MEM_INIT_SIZE		66
+
+static void still_alive(void)
+{
+#if 0
+	sync();
+	msleep(2);
+#endif
+}
+
+static uint32_t
+munge_reg(struct nvbios *bios, uint32_t reg)
+{
+	struct drm_nouveau_private *dev_priv = bios->dev->dev_private;
+	struct dcb_entry *dcbent = bios->display.output;
+
+	if (dev_priv->card_type < NV_50)
+		return reg;
+
+	if (reg & 0x40000000) {
+		BUG_ON(!dcbent);
+
+		reg += (ffs(dcbent->or) - 1) * 0x800;
+		if ((reg & 0x20000000) && !(dcbent->sorconf.link & 1))
+			reg += 0x00000080;
+	}
+
+	reg &= ~0x60000000;
+	return reg;
+}
+
+static int
+valid_reg(struct nvbios *bios, uint32_t reg)
+{
+	struct drm_nouveau_private *dev_priv = bios->dev->dev_private;
+	struct drm_device *dev = bios->dev;
+
+	/* C51 has misaligned regs on purpose. Marvellous */
+	if (reg & 0x2 ||
+	    (reg & 0x1 && dev_priv->vbios.chip_version != 0x51))
+		NV_ERROR(dev, "======= misaligned reg 0x%08X =======\n", reg);
+
+	/* warn on C51 regs that haven't been verified accessible in tracing */
+	if (reg & 0x1 && dev_priv->vbios.chip_version == 0x51 &&
+	    reg != 0x130d && reg != 0x1311 && reg != 0x60081d)
+		NV_WARN(dev, "=== C51 misaligned reg 0x%08X not verified ===\n",
+			reg);
+
+	if (reg >= (8*1024*1024)) {
+		NV_ERROR(dev, "=== reg 0x%08x out of mapped bounds ===\n", reg);
+		return 0;
+	}
+
+	return 1;
+}
+
+static bool
+valid_idx_port(struct nvbios *bios, uint16_t port)
+{
+	struct drm_nouveau_private *dev_priv = bios->dev->dev_private;
+	struct drm_device *dev = bios->dev;
+
+	/*
+	 * If adding more ports here, the read/write functions below will need
+	 * updating so that the correct mmio range (PRMCIO, PRMDIO, PRMVIO) is
+	 * used for the port in question
+	 */
+	if (dev_priv->card_type < NV_50) {
+		if (port == NV_CIO_CRX__COLOR)
+			return true;
+		if (port == NV_VIO_SRX)
+			return true;
+	} else {
+		if (port == NV_CIO_CRX__COLOR)
+			return true;
+	}
+
+	NV_ERROR(dev, "========== unknown indexed io port 0x%04X ==========\n",
+		 port);
+
+	return false;
+}
+
+static bool
+valid_port(struct nvbios *bios, uint16_t port)
+{
+	struct drm_device *dev = bios->dev;
+
+	/*
+	 * If adding more ports here, the read/write functions below will need
+	 * updating so that the correct mmio range (PRMCIO, PRMDIO, PRMVIO) is
+	 * used for the port in question
+	 */
+	if (port == NV_VIO_VSE2)
+		return true;
+
+	NV_ERROR(dev, "========== unknown io port 0x%04X ==========\n", port);
+
+	return false;
+}
+
+static uint32_t
+bios_rd32(struct nvbios *bios, uint32_t reg)
+{
+	uint32_t data;
+
+	reg = munge_reg(bios, reg);
+	if (!valid_reg(bios, reg))
+		return 0;
+
+	/*
+	 * C51 sometimes uses regs with bit0 set in the address. For these
+	 * cases there should exist a translation in a BIOS table to an IO
+	 * port address which the BIOS uses for accessing the reg
+	 *
+	 * These only seem to appear for the power control regs to a flat panel,
+	 * and the GPIO regs at 0x60081*.  In C51 mmio traces the normal regs
+	 * for 0x1308 and 0x1310 are used - hence the mask below.  An S3
+	 * suspend-resume mmio trace from a C51 will be required to see if this
+	 * is true for the power microcode in 0x14.., or whether the direct IO
+	 * port access method is needed
+	 */
+	if (reg & 0x1)
+		reg &= ~0x1;
+
+	data = nv_rd32(bios->dev, reg);
+
+	BIOSLOG(bios, "	Read:  Reg: 0x%08X, Data: 0x%08X\n", reg, data);
+
+	return data;
+}
+
+static void
+bios_wr32(struct nvbios *bios, uint32_t reg, uint32_t data)
+{
+	struct drm_nouveau_private *dev_priv = bios->dev->dev_private;
+
+	reg = munge_reg(bios, reg);
+	if (!valid_reg(bios, reg))
+		return;
+
+	/* see note in bios_rd32 */
+	if (reg & 0x1)
+		reg &= 0xfffffffe;
+
+	LOG_OLD_VALUE(bios_rd32(bios, reg));
+	BIOSLOG(bios, "	Write: Reg: 0x%08X, Data: 0x%08X\n", reg, data);
+
+	if (dev_priv->vbios.execute) {
+		still_alive();
+		nv_wr32(bios->dev, reg, data);
+	}
+}
+
+static uint8_t
+bios_idxprt_rd(struct nvbios *bios, uint16_t port, uint8_t index)
+{
+	struct drm_nouveau_private *dev_priv = bios->dev->dev_private;
+	struct drm_device *dev = bios->dev;
+	uint8_t data;
+
+	if (!valid_idx_port(bios, port))
+		return 0;
+
+	if (dev_priv->card_type < NV_50) {
+		if (port == NV_VIO_SRX)
+			data = NVReadVgaSeq(dev, bios->state.crtchead, index);
+		else	/* assume NV_CIO_CRX__COLOR */
+			data = NVReadVgaCrtc(dev, bios->state.crtchead, index);
+	} else {
+		uint32_t data32;
+
+		data32 = bios_rd32(bios, NV50_PDISPLAY_VGACRTC(index & ~3));
+		data = (data32 >> ((index & 3) << 3)) & 0xff;
+	}
+
+	BIOSLOG(bios, "	Indexed IO read:  Port: 0x%04X, Index: 0x%02X, "
+		      "Head: 0x%02X, Data: 0x%02X\n",
+		port, index, bios->state.crtchead, data);
+	return data;
+}
+
+static void
+bios_idxprt_wr(struct nvbios *bios, uint16_t port, uint8_t index, uint8_t data)
+{
+	struct drm_nouveau_private *dev_priv = bios->dev->dev_private;
+	struct drm_device *dev = bios->dev;
+
+	if (!valid_idx_port(bios, port))
+		return;
+
+	/*
+	 * The current head is maintained in the nvbios member  state.crtchead.
+	 * We trap changes to CR44 and update the head variable and hence the
+	 * register set written.
+	 * As CR44 only exists on CRTC0, we update crtchead to head0 in advance
+	 * of the write, and to head1 after the write
+	 */
+	if (port == NV_CIO_CRX__COLOR && index == NV_CIO_CRE_44 &&
+	    data != NV_CIO_CRE_44_HEADB)
+		bios->state.crtchead = 0;
+
+	LOG_OLD_VALUE(bios_idxprt_rd(bios, port, index));
+	BIOSLOG(bios, "	Indexed IO write: Port: 0x%04X, Index: 0x%02X, "
+		      "Head: 0x%02X, Data: 0x%02X\n",
+		port, index, bios->state.crtchead, data);
+
+	if (bios->execute && dev_priv->card_type < NV_50) {
+		still_alive();
+		if (port == NV_VIO_SRX)
+			NVWriteVgaSeq(dev, bios->state.crtchead, index, data);
+		else	/* assume NV_CIO_CRX__COLOR */
+			NVWriteVgaCrtc(dev, bios->state.crtchead, index, data);
+	} else
+	if (bios->execute) {
+		uint32_t data32, shift = (index & 3) << 3;
+
+		still_alive();
+
+		data32  = bios_rd32(bios, NV50_PDISPLAY_VGACRTC(index & ~3));
+		data32 &= ~(0xff << shift);
+		data32 |= (data << shift);
+		bios_wr32(bios, NV50_PDISPLAY_VGACRTC(index & ~3), data32);
+	}
+
+	if (port == NV_CIO_CRX__COLOR &&
+	    index == NV_CIO_CRE_44 && data == NV_CIO_CRE_44_HEADB)
+		bios->state.crtchead = 1;
+}
+
+static uint8_t
+bios_port_rd(struct nvbios *bios, uint16_t port)
+{
+	uint8_t data, head = bios->state.crtchead;
+
+	if (!valid_port(bios, port))
+		return 0;
+
+	data = NVReadPRMVIO(bios->dev, head, NV_PRMVIO0_OFFSET + port);
+
+	BIOSLOG(bios, "	IO read:  Port: 0x%04X, Head: 0x%02X, Data: 0x%02X\n",
+		port, head, data);
+
+	return data;
+}
+
+static void
+bios_port_wr(struct nvbios *bios, uint16_t port, uint8_t data)
+{
+	int head = bios->state.crtchead;
+
+	if (!valid_port(bios, port))
+		return;
+
+	LOG_OLD_VALUE(bios_port_rd(bios, port));
+	BIOSLOG(bios, "	IO write: Port: 0x%04X, Head: 0x%02X, Data: 0x%02X\n",
+		port, head, data);
+
+	if (!bios->execute)
+		return;
+
+	still_alive();
+	NVWritePRMVIO(bios->dev, head, NV_PRMVIO0_OFFSET + port, data);
+}
+
+static bool
+io_flag_condition_met(struct nvbios *bios, uint16_t offset, uint8_t cond)
+{
+	/*
+	 * The IO flag condition entry has 2 bytes for the CRTC port; 1 byte
+	 * for the CRTC index; 1 byte for the mask to apply to the value
+	 * retrieved from the CRTC; 1 byte for the shift right to apply to the
+	 * masked CRTC value; 2 bytes for the offset to the flag array, to
+	 * which the shifted value is added; 1 byte for the mask applied to the
+	 * value read from the flag array; and 1 byte for the value to compare
+	 * against the masked byte from the flag table.
+	 */
+
+	uint16_t condptr = bios->io_flag_condition_tbl_ptr + cond * IO_FLAG_CONDITION_SIZE;
+	uint16_t crtcport = ROM16(bios->data[condptr]);
+	uint8_t crtcindex = bios->data[condptr + 2];
+	uint8_t mask = bios->data[condptr + 3];
+	uint8_t shift = bios->data[condptr + 4];
+	uint16_t flagarray = ROM16(bios->data[condptr + 5]);
+	uint8_t flagarraymask = bios->data[condptr + 7];
+	uint8_t cmpval = bios->data[condptr + 8];
+	uint8_t data;
+
+	BIOSLOG(bios, "0x%04X: Port: 0x%04X, Index: 0x%02X, Mask: 0x%02X, "
+		      "Shift: 0x%02X, FlagArray: 0x%04X, FAMask: 0x%02X, "
+		      "Cmpval: 0x%02X\n",
+		offset, crtcport, crtcindex, mask, shift, flagarray, flagarraymask, cmpval);
+
+	data = bios_idxprt_rd(bios, crtcport, crtcindex);
+
+	data = bios->data[flagarray + ((data & mask) >> shift)];
+	data &= flagarraymask;
+
+	BIOSLOG(bios, "0x%04X: Checking if 0x%02X equals 0x%02X\n",
+		offset, data, cmpval);
+
+	return (data == cmpval);
+}
+
+static bool
+bios_condition_met(struct nvbios *bios, uint16_t offset, uint8_t cond)
+{
+	/*
+	 * The condition table entry has 4 bytes for the address of the
+	 * register to check, 4 bytes for a mask to apply to the register and
+	 * 4 for a test comparison value
+	 */
+
+	uint16_t condptr = bios->condition_tbl_ptr + cond * CONDITION_SIZE;
+	uint32_t reg = ROM32(bios->data[condptr]);
+	uint32_t mask = ROM32(bios->data[condptr + 4]);
+	uint32_t cmpval = ROM32(bios->data[condptr + 8]);
+	uint32_t data;
+
+	BIOSLOG(bios, "0x%04X: Cond: 0x%02X, Reg: 0x%08X, Mask: 0x%08X\n",
+		offset, cond, reg, mask);
+
+	data = bios_rd32(bios, reg) & mask;
+
+	BIOSLOG(bios, "0x%04X: Checking if 0x%08X equals 0x%08X\n",
+		offset, data, cmpval);
+
+	return (data == cmpval);
+}
+
+static bool
+io_condition_met(struct nvbios *bios, uint16_t offset, uint8_t cond)
+{
+	/*
+	 * The IO condition entry has 2 bytes for the IO port address; 1 byte
+	 * for the index to write to io_port; 1 byte for the mask to apply to
+	 * the byte read from io_port+1; and 1 byte for the value to compare
+	 * against the masked byte.
+	 */
+
+	uint16_t condptr = bios->io_condition_tbl_ptr + cond * IO_CONDITION_SIZE;
+	uint16_t io_port = ROM16(bios->data[condptr]);
+	uint8_t port_index = bios->data[condptr + 2];
+	uint8_t mask = bios->data[condptr + 3];
+	uint8_t cmpval = bios->data[condptr + 4];
+
+	uint8_t data = bios_idxprt_rd(bios, io_port, port_index) & mask;
+
+	BIOSLOG(bios, "0x%04X: Checking if 0x%02X equals 0x%02X\n",
+		offset, data, cmpval);
+
+	return (data == cmpval);
+}
+
+static int
+nv50_pll_set(struct drm_device *dev, uint32_t reg, uint32_t clk)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t reg0 = nv_rd32(dev, reg + 0);
+	uint32_t reg1 = nv_rd32(dev, reg + 4);
+	struct nouveau_pll_vals pll;
+	struct pll_lims pll_limits;
+	int ret;
+
+	ret = get_pll_limits(dev, reg, &pll_limits);
+	if (ret)
+		return ret;
+
+	clk = nouveau_calc_pll_mnp(dev, &pll_limits, clk, &pll);
+	if (!clk)
+		return -ERANGE;
+
+	reg0 = (reg0 & 0xfff8ffff) | (pll.log2P << 16);
+	reg1 = (reg1 & 0xffff0000) | (pll.N1 << 8) | pll.M1;
+
+	if (dev_priv->vbios.execute) {
+		still_alive();
+		nv_wr32(dev, reg + 4, reg1);
+		nv_wr32(dev, reg + 0, reg0);
+	}
+
+	return 0;
+}
+
+static int
+setPLL(struct nvbios *bios, uint32_t reg, uint32_t clk)
+{
+	struct drm_device *dev = bios->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	/* clk in kHz */
+	struct pll_lims pll_lim;
+	struct nouveau_pll_vals pllvals;
+	int ret;
+
+	if (dev_priv->card_type >= NV_50)
+		return nv50_pll_set(dev, reg, clk);
+
+	/* high regs (such as in the mac g5 table) are not -= 4 */
+	ret = get_pll_limits(dev, reg > 0x405c ? reg : reg - 4, &pll_lim);
+	if (ret)
+		return ret;
+
+	clk = nouveau_calc_pll_mnp(dev, &pll_lim, clk, &pllvals);
+	if (!clk)
+		return -ERANGE;
+
+	if (bios->execute) {
+		still_alive();
+		nouveau_hw_setpll(dev, reg, &pllvals);
+	}
+
+	return 0;
+}
+
+static int dcb_entry_idx_from_crtchead(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nvbios *bios = &dev_priv->vbios;
+
+	/*
+	 * For the results of this function to be correct, CR44 must have been
+	 * set (using bios_idxprt_wr to set crtchead), CR58 set for CR57 = 0,
+	 * and the DCB table parsed, before the script calling the function is
+	 * run.  run_digital_op_script is example of how to do such setup
+	 */
+
+	uint8_t dcb_entry = NVReadVgaCrtc5758(dev, bios->state.crtchead, 0);
+
+	if (dcb_entry > bios->dcb.entries) {
+		NV_ERROR(dev, "CR58 doesn't have a valid DCB entry currently "
+				"(%02X)\n", dcb_entry);
+		dcb_entry = 0x7f;	/* unused / invalid marker */
+	}
+
+	return dcb_entry;
+}
+
+static struct nouveau_i2c_chan *
+init_i2c_device_find(struct drm_device *dev, int i2c_index)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct dcb_table *dcb = &dev_priv->vbios.dcb;
+
+	if (i2c_index == 0xff) {
+		/* note: dcb_entry_idx_from_crtchead needs pre-script set-up */
+		int idx = dcb_entry_idx_from_crtchead(dev), shift = 0;
+		int default_indices = dcb->i2c_default_indices;
+
+		if (idx != 0x7f && dcb->entry[idx].i2c_upper_default)
+			shift = 4;
+
+		i2c_index = (default_indices >> shift) & 0xf;
+	}
+	if (i2c_index == 0x80)	/* g80+ */
+		i2c_index = dcb->i2c_default_indices & 0xf;
+
+	return nouveau_i2c_find(dev, i2c_index);
+}
+
+static uint32_t
+get_tmds_index_reg(struct drm_device *dev, uint8_t mlv)
+{
+	/*
+	 * For mlv < 0x80, it is an index into a table of TMDS base addresses.
+	 * For mlv == 0x80 use the "or" value of the dcb_entry indexed by
+	 * CR58 for CR57 = 0 to index a table of offsets to the basic
+	 * 0x6808b0 address.
+	 * For mlv == 0x81 use the "or" value of the dcb_entry indexed by
+	 * CR58 for CR57 = 0 to index a table of offsets to the basic
+	 * 0x6808b0 address, and then flip the offset by 8.
+	 */
+
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nvbios *bios = &dev_priv->vbios;
+	const int pramdac_offset[13] = {
+		0, 0, 0x8, 0, 0x2000, 0, 0, 0, 0x2008, 0, 0, 0, 0x2000 };
+	const uint32_t pramdac_table[4] = {
+		0x6808b0, 0x6808b8, 0x6828b0, 0x6828b8 };
+
+	if (mlv >= 0x80) {
+		int dcb_entry, dacoffset;
+
+		/* note: dcb_entry_idx_from_crtchead needs pre-script set-up */
+		dcb_entry = dcb_entry_idx_from_crtchead(dev);
+		if (dcb_entry == 0x7f)
+			return 0;
+		dacoffset = pramdac_offset[bios->dcb.entry[dcb_entry].or];
+		if (mlv == 0x81)
+			dacoffset ^= 8;
+		return 0x6808b0 + dacoffset;
+	} else {
+		if (mlv >= ARRAY_SIZE(pramdac_table)) {
+			NV_ERROR(dev, "Magic Lookup Value too big (%02X)\n",
+									mlv);
+			return 0;
+		}
+		return pramdac_table[mlv];
+	}
+}
+
+static int
+init_io_restrict_prog(struct nvbios *bios, uint16_t offset,
+		      struct init_exec *iexec)
+{
+	/*
+	 * INIT_IO_RESTRICT_PROG   opcode: 0x32 ('2')
+	 *
+	 * offset      (8  bit): opcode
+	 * offset + 1  (16 bit): CRTC port
+	 * offset + 3  (8  bit): CRTC index
+	 * offset + 4  (8  bit): mask
+	 * offset + 5  (8  bit): shift
+	 * offset + 6  (8  bit): count
+	 * offset + 7  (32 bit): register
+	 * offset + 11 (32 bit): configuration 1
+	 * ...
+	 *
+	 * Starting at offset + 11 there are "count" 32 bit values.
+	 * To find out which value to use read index "CRTC index" on "CRTC
+	 * port", AND this value with "mask" and then bit shift right "shift"
+	 * bits.  Read the appropriate value using this index and write to
+	 * "register"
+	 */
+
+	uint16_t crtcport = ROM16(bios->data[offset + 1]);
+	uint8_t crtcindex = bios->data[offset + 3];
+	uint8_t mask = bios->data[offset + 4];
+	uint8_t shift = bios->data[offset + 5];
+	uint8_t count = bios->data[offset + 6];
+	uint32_t reg = ROM32(bios->data[offset + 7]);
+	uint8_t config;
+	uint32_t configval;
+	int len = 11 + count * 4;
+
+	if (!iexec->execute)
+		return len;
+
+	BIOSLOG(bios, "0x%04X: Port: 0x%04X, Index: 0x%02X, Mask: 0x%02X, "
+		      "Shift: 0x%02X, Count: 0x%02X, Reg: 0x%08X\n",
+		offset, crtcport, crtcindex, mask, shift, count, reg);
+
+	config = (bios_idxprt_rd(bios, crtcport, crtcindex) & mask) >> shift;
+	if (config > count) {
+		NV_ERROR(bios->dev,
+			 "0x%04X: Config 0x%02X exceeds maximal bound 0x%02X\n",
+			 offset, config, count);
+		return 0;
+	}
+
+	configval = ROM32(bios->data[offset + 11 + config * 4]);
+
+	BIOSLOG(bios, "0x%04X: Writing config %02X\n", offset, config);
+
+	bios_wr32(bios, reg, configval);
+
+	return len;
+}
+
+static int
+init_repeat(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_REPEAT   opcode: 0x33 ('3')
+	 *
+	 * offset      (8 bit): opcode
+	 * offset + 1  (8 bit): count
+	 *
+	 * Execute script following this opcode up to INIT_REPEAT_END
+	 * "count" times
+	 */
+
+	uint8_t count = bios->data[offset + 1];
+	uint8_t i;
+
+	/* no iexec->execute check by design */
+
+	BIOSLOG(bios, "0x%04X: Repeating following segment %d times\n",
+		offset, count);
+
+	iexec->repeat = true;
+
+	/*
+	 * count - 1, as the script block will execute once when we leave this
+	 * opcode -- this is compatible with bios behaviour as:
+	 * a) the block is always executed at least once, even if count == 0
+	 * b) the bios interpreter skips to the op following INIT_END_REPEAT,
+	 * while we don't
+	 */
+	for (i = 0; i < count - 1; i++)
+		parse_init_table(bios, offset + 2, iexec);
+
+	iexec->repeat = false;
+
+	return 2;
+}
+
+static int
+init_io_restrict_pll(struct nvbios *bios, uint16_t offset,
+		     struct init_exec *iexec)
+{
+	/*
+	 * INIT_IO_RESTRICT_PLL   opcode: 0x34 ('4')
+	 *
+	 * offset      (8  bit): opcode
+	 * offset + 1  (16 bit): CRTC port
+	 * offset + 3  (8  bit): CRTC index
+	 * offset + 4  (8  bit): mask
+	 * offset + 5  (8  bit): shift
+	 * offset + 6  (8  bit): IO flag condition index
+	 * offset + 7  (8  bit): count
+	 * offset + 8  (32 bit): register
+	 * offset + 12 (16 bit): frequency 1
+	 * ...
+	 *
+	 * Starting at offset + 12 there are "count" 16 bit frequencies (10kHz).
+	 * Set PLL register "register" to coefficients for frequency n,
+	 * selected by reading index "CRTC index" of "CRTC port" ANDed with
+	 * "mask" and shifted right by "shift".
+	 *
+	 * If "IO flag condition index" > 0, and condition met, double
+	 * frequency before setting it.
+	 */
+
+	uint16_t crtcport = ROM16(bios->data[offset + 1]);
+	uint8_t crtcindex = bios->data[offset + 3];
+	uint8_t mask = bios->data[offset + 4];
+	uint8_t shift = bios->data[offset + 5];
+	int8_t io_flag_condition_idx = bios->data[offset + 6];
+	uint8_t count = bios->data[offset + 7];
+	uint32_t reg = ROM32(bios->data[offset + 8]);
+	uint8_t config;
+	uint16_t freq;
+	int len = 12 + count * 2;
+
+	if (!iexec->execute)
+		return len;
+
+	BIOSLOG(bios, "0x%04X: Port: 0x%04X, Index: 0x%02X, Mask: 0x%02X, "
+		      "Shift: 0x%02X, IO Flag Condition: 0x%02X, "
+		      "Count: 0x%02X, Reg: 0x%08X\n",
+		offset, crtcport, crtcindex, mask, shift,
+		io_flag_condition_idx, count, reg);
+
+	config = (bios_idxprt_rd(bios, crtcport, crtcindex) & mask) >> shift;
+	if (config > count) {
+		NV_ERROR(bios->dev,
+			 "0x%04X: Config 0x%02X exceeds maximal bound 0x%02X\n",
+			 offset, config, count);
+		return 0;
+	}
+
+	freq = ROM16(bios->data[offset + 12 + config * 2]);
+
+	if (io_flag_condition_idx > 0) {
+		if (io_flag_condition_met(bios, offset, io_flag_condition_idx)) {
+			BIOSLOG(bios, "0x%04X: Condition fulfilled -- "
+				      "frequency doubled\n", offset);
+			freq *= 2;
+		} else
+			BIOSLOG(bios, "0x%04X: Condition not fulfilled -- "
+				      "frequency unchanged\n", offset);
+	}
+
+	BIOSLOG(bios, "0x%04X: Reg: 0x%08X, Config: 0x%02X, Freq: %d0kHz\n",
+		offset, reg, config, freq);
+
+	setPLL(bios, reg, freq * 10);
+
+	return len;
+}
+
+static int
+init_end_repeat(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_END_REPEAT   opcode: 0x36 ('6')
+	 *
+	 * offset      (8 bit): opcode
+	 *
+	 * Marks the end of the block for INIT_REPEAT to repeat
+	 */
+
+	/* no iexec->execute check by design */
+
+	/*
+	 * iexec->repeat flag necessary to go past INIT_END_REPEAT opcode when
+	 * we're not in repeat mode
+	 */
+	if (iexec->repeat)
+		return 0;
+
+	return 1;
+}
+
+static int
+init_copy(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_COPY   opcode: 0x37 ('7')
+	 *
+	 * offset      (8  bit): opcode
+	 * offset + 1  (32 bit): register
+	 * offset + 5  (8  bit): shift
+	 * offset + 6  (8  bit): srcmask
+	 * offset + 7  (16 bit): CRTC port
+	 * offset + 9  (8 bit): CRTC index
+	 * offset + 10  (8 bit): mask
+	 *
+	 * Read index "CRTC index" on "CRTC port", AND with "mask", OR with
+	 * (REGVAL("register") >> "shift" & "srcmask") and write-back to CRTC
+	 * port
+	 */
+
+	uint32_t reg = ROM32(bios->data[offset + 1]);
+	uint8_t shift = bios->data[offset + 5];
+	uint8_t srcmask = bios->data[offset + 6];
+	uint16_t crtcport = ROM16(bios->data[offset + 7]);
+	uint8_t crtcindex = bios->data[offset + 9];
+	uint8_t mask = bios->data[offset + 10];
+	uint32_t data;
+	uint8_t crtcdata;
+
+	if (!iexec->execute)
+		return 11;
+
+	BIOSLOG(bios, "0x%04X: Reg: 0x%08X, Shift: 0x%02X, SrcMask: 0x%02X, "
+		      "Port: 0x%04X, Index: 0x%02X, Mask: 0x%02X\n",
+		offset, reg, shift, srcmask, crtcport, crtcindex, mask);
+
+	data = bios_rd32(bios, reg);
+
+	if (shift < 0x80)
+		data >>= shift;
+	else
+		data <<= (0x100 - shift);
+
+	data &= srcmask;
+
+	crtcdata  = bios_idxprt_rd(bios, crtcport, crtcindex) & mask;
+	crtcdata |= (uint8_t)data;
+	bios_idxprt_wr(bios, crtcport, crtcindex, crtcdata);
+
+	return 11;
+}
+
+static int
+init_not(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_NOT   opcode: 0x38 ('8')
+	 *
+	 * offset      (8  bit): opcode
+	 *
+	 * Invert the current execute / no-execute condition (i.e. "else")
+	 */
+	if (iexec->execute)
+		BIOSLOG(bios, "0x%04X: ------ Skipping following commands  ------\n", offset);
+	else
+		BIOSLOG(bios, "0x%04X: ------ Executing following commands ------\n", offset);
+
+	iexec->execute = !iexec->execute;
+	return 1;
+}
+
+static int
+init_io_flag_condition(struct nvbios *bios, uint16_t offset,
+		       struct init_exec *iexec)
+{
+	/*
+	 * INIT_IO_FLAG_CONDITION   opcode: 0x39 ('9')
+	 *
+	 * offset      (8 bit): opcode
+	 * offset + 1  (8 bit): condition number
+	 *
+	 * Check condition "condition number" in the IO flag condition table.
+	 * If condition not met skip subsequent opcodes until condition is
+	 * inverted (INIT_NOT), or we hit INIT_RESUME
+	 */
+
+	uint8_t cond = bios->data[offset + 1];
+
+	if (!iexec->execute)
+		return 2;
+
+	if (io_flag_condition_met(bios, offset, cond))
+		BIOSLOG(bios, "0x%04X: Condition fulfilled -- continuing to execute\n", offset);
+	else {
+		BIOSLOG(bios, "0x%04X: Condition not fulfilled -- skipping following commands\n", offset);
+		iexec->execute = false;
+	}
+
+	return 2;
+}
+
+static int
+init_idx_addr_latched(struct nvbios *bios, uint16_t offset,
+		      struct init_exec *iexec)
+{
+	/*
+	 * INIT_INDEX_ADDRESS_LATCHED   opcode: 0x49 ('I')
+	 *
+	 * offset      (8  bit): opcode
+	 * offset + 1  (32 bit): control register
+	 * offset + 5  (32 bit): data register
+	 * offset + 9  (32 bit): mask
+	 * offset + 13 (32 bit): data
+	 * offset + 17 (8  bit): count
+	 * offset + 18 (8  bit): address 1
+	 * offset + 19 (8  bit): data 1
+	 * ...
+	 *
+	 * For each of "count" address and data pairs, write "data n" to
+	 * "data register", read the current value of "control register",
+	 * and write it back once ANDed with "mask", ORed with "data",
+	 * and ORed with "address n"
+	 */
+
+	uint32_t controlreg = ROM32(bios->data[offset + 1]);
+	uint32_t datareg = ROM32(bios->data[offset + 5]);
+	uint32_t mask = ROM32(bios->data[offset + 9]);
+	uint32_t data = ROM32(bios->data[offset + 13]);
+	uint8_t count = bios->data[offset + 17];
+	int len = 18 + count * 2;
+	uint32_t value;
+	int i;
+
+	if (!iexec->execute)
+		return len;
+
+	BIOSLOG(bios, "0x%04X: ControlReg: 0x%08X, DataReg: 0x%08X, "
+		      "Mask: 0x%08X, Data: 0x%08X, Count: 0x%02X\n",
+		offset, controlreg, datareg, mask, data, count);
+
+	for (i = 0; i < count; i++) {
+		uint8_t instaddress = bios->data[offset + 18 + i * 2];
+		uint8_t instdata = bios->data[offset + 19 + i * 2];
+
+		BIOSLOG(bios, "0x%04X: Address: 0x%02X, Data: 0x%02X\n",
+			offset, instaddress, instdata);
+
+		bios_wr32(bios, datareg, instdata);
+		value  = bios_rd32(bios, controlreg) & mask;
+		value |= data;
+		value |= instaddress;
+		bios_wr32(bios, controlreg, value);
+	}
+
+	return len;
+}
+
+static int
+init_io_restrict_pll2(struct nvbios *bios, uint16_t offset,
+		      struct init_exec *iexec)
+{
+	/*
+	 * INIT_IO_RESTRICT_PLL2   opcode: 0x4A ('J')
+	 *
+	 * offset      (8  bit): opcode
+	 * offset + 1  (16 bit): CRTC port
+	 * offset + 3  (8  bit): CRTC index
+	 * offset + 4  (8  bit): mask
+	 * offset + 5  (8  bit): shift
+	 * offset + 6  (8  bit): count
+	 * offset + 7  (32 bit): register
+	 * offset + 11 (32 bit): frequency 1
+	 * ...
+	 *
+	 * Starting at offset + 11 there are "count" 32 bit frequencies (kHz).
+	 * Set PLL register "register" to coefficients for frequency n,
+	 * selected by reading index "CRTC index" of "CRTC port" ANDed with
+	 * "mask" and shifted right by "shift".
+	 */
+
+	uint16_t crtcport = ROM16(bios->data[offset + 1]);
+	uint8_t crtcindex = bios->data[offset + 3];
+	uint8_t mask = bios->data[offset + 4];
+	uint8_t shift = bios->data[offset + 5];
+	uint8_t count = bios->data[offset + 6];
+	uint32_t reg = ROM32(bios->data[offset + 7]);
+	int len = 11 + count * 4;
+	uint8_t config;
+	uint32_t freq;
+
+	if (!iexec->execute)
+		return len;
+
+	BIOSLOG(bios, "0x%04X: Port: 0x%04X, Index: 0x%02X, Mask: 0x%02X, "
+		      "Shift: 0x%02X, Count: 0x%02X, Reg: 0x%08X\n",
+		offset, crtcport, crtcindex, mask, shift, count, reg);
+
+	if (!reg)
+		return len;
+
+	config = (bios_idxprt_rd(bios, crtcport, crtcindex) & mask) >> shift;
+	if (config > count) {
+		NV_ERROR(bios->dev,
+			 "0x%04X: Config 0x%02X exceeds maximal bound 0x%02X\n",
+			 offset, config, count);
+		return 0;
+	}
+
+	freq = ROM32(bios->data[offset + 11 + config * 4]);
+
+	BIOSLOG(bios, "0x%04X: Reg: 0x%08X, Config: 0x%02X, Freq: %dkHz\n",
+		offset, reg, config, freq);
+
+	setPLL(bios, reg, freq);
+
+	return len;
+}
+
+static int
+init_pll2(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_PLL2   opcode: 0x4B ('K')
+	 *
+	 * offset      (8  bit): opcode
+	 * offset + 1  (32 bit): register
+	 * offset + 5  (32 bit): freq
+	 *
+	 * Set PLL register "register" to coefficients for frequency "freq"
+	 */
+
+	uint32_t reg = ROM32(bios->data[offset + 1]);
+	uint32_t freq = ROM32(bios->data[offset + 5]);
+
+	if (!iexec->execute)
+		return 9;
+
+	BIOSLOG(bios, "0x%04X: Reg: 0x%04X, Freq: %dkHz\n",
+		offset, reg, freq);
+
+	setPLL(bios, reg, freq);
+	return 9;
+}
+
+static int
+init_i2c_byte(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_I2C_BYTE   opcode: 0x4C ('L')
+	 *
+	 * offset      (8 bit): opcode
+	 * offset + 1  (8 bit): DCB I2C table entry index
+	 * offset + 2  (8 bit): I2C slave address
+	 * offset + 3  (8 bit): count
+	 * offset + 4  (8 bit): I2C register 1
+	 * offset + 5  (8 bit): mask 1
+	 * offset + 6  (8 bit): data 1
+	 * ...
+	 *
+	 * For each of "count" registers given by "I2C register n" on the device
+	 * addressed by "I2C slave address" on the I2C bus given by
+	 * "DCB I2C table entry index", read the register, AND the result with
+	 * "mask n" and OR it with "data n" before writing it back to the device
+	 */
+
+	uint8_t i2c_index = bios->data[offset + 1];
+	uint8_t i2c_address = bios->data[offset + 2];
+	uint8_t count = bios->data[offset + 3];
+	int len = 4 + count * 3;
+	struct nouveau_i2c_chan *chan;
+	struct i2c_msg msg;
+	int i;
+
+	if (!iexec->execute)
+		return len;
+
+	BIOSLOG(bios, "0x%04X: DCBI2CIndex: 0x%02X, I2CAddress: 0x%02X, "
+		      "Count: 0x%02X\n",
+		offset, i2c_index, i2c_address, count);
+
+	chan = init_i2c_device_find(bios->dev, i2c_index);
+	if (!chan)
+		return 0;
+
+	for (i = 0; i < count; i++) {
+		uint8_t i2c_reg = bios->data[offset + 4 + i * 3];
+		uint8_t mask = bios->data[offset + 5 + i * 3];
+		uint8_t data = bios->data[offset + 6 + i * 3];
+		uint8_t value;
+
+		msg.addr = i2c_address;
+		msg.flags = I2C_M_RD;
+		msg.len = 1;
+		msg.buf = &value;
+		if (i2c_transfer(&chan->adapter, &msg, 1) != 1)
+			return 0;
+
+		BIOSLOG(bios, "0x%04X: I2CReg: 0x%02X, Value: 0x%02X, "
+			      "Mask: 0x%02X, Data: 0x%02X\n",
+			offset, i2c_reg, value, mask, data);
+
+		value = (value & mask) | data;
+
+		if (bios->execute) {
+			msg.addr = i2c_address;
+			msg.flags = 0;
+			msg.len = 1;
+			msg.buf = &value;
+			if (i2c_transfer(&chan->adapter, &msg, 1) != 1)
+				return 0;
+		}
+	}
+
+	return len;
+}
+
+static int
+init_zm_i2c_byte(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_ZM_I2C_BYTE   opcode: 0x4D ('M')
+	 *
+	 * offset      (8 bit): opcode
+	 * offset + 1  (8 bit): DCB I2C table entry index
+	 * offset + 2  (8 bit): I2C slave address
+	 * offset + 3  (8 bit): count
+	 * offset + 4  (8 bit): I2C register 1
+	 * offset + 5  (8 bit): data 1
+	 * ...
+	 *
+	 * For each of "count" registers given by "I2C register n" on the device
+	 * addressed by "I2C slave address" on the I2C bus given by
+	 * "DCB I2C table entry index", set the register to "data n"
+	 */
+
+	uint8_t i2c_index = bios->data[offset + 1];
+	uint8_t i2c_address = bios->data[offset + 2];
+	uint8_t count = bios->data[offset + 3];
+	int len = 4 + count * 2;
+	struct nouveau_i2c_chan *chan;
+	struct i2c_msg msg;
+	int i;
+
+	if (!iexec->execute)
+		return len;
+
+	BIOSLOG(bios, "0x%04X: DCBI2CIndex: 0x%02X, I2CAddress: 0x%02X, "
+		      "Count: 0x%02X\n",
+		offset, i2c_index, i2c_address, count);
+
+	chan = init_i2c_device_find(bios->dev, i2c_index);
+	if (!chan)
+		return 0;
+
+	for (i = 0; i < count; i++) {
+		uint8_t i2c_reg = bios->data[offset + 4 + i * 2];
+		uint8_t data = bios->data[offset + 5 + i * 2];
+
+		BIOSLOG(bios, "0x%04X: I2CReg: 0x%02X, Data: 0x%02X\n",
+			offset, i2c_reg, data);
+
+		if (bios->execute) {
+			msg.addr = i2c_address;
+			msg.flags = 0;
+			msg.len = 1;
+			msg.buf = &data;
+			if (i2c_transfer(&chan->adapter, &msg, 1) != 1)
+				return 0;
+		}
+	}
+
+	return len;
+}
+
+static int
+init_zm_i2c(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_ZM_I2C   opcode: 0x4E ('N')
+	 *
+	 * offset      (8 bit): opcode
+	 * offset + 1  (8 bit): DCB I2C table entry index
+	 * offset + 2  (8 bit): I2C slave address
+	 * offset + 3  (8 bit): count
+	 * offset + 4  (8 bit): data 1
+	 * ...
+	 *
+	 * Send "count" bytes ("data n") to the device addressed by "I2C slave
+	 * address" on the I2C bus given by "DCB I2C table entry index"
+	 */
+
+	uint8_t i2c_index = bios->data[offset + 1];
+	uint8_t i2c_address = bios->data[offset + 2];
+	uint8_t count = bios->data[offset + 3];
+	int len = 4 + count;
+	struct nouveau_i2c_chan *chan;
+	struct i2c_msg msg;
+	uint8_t data[256];
+	int i;
+
+	if (!iexec->execute)
+		return len;
+
+	BIOSLOG(bios, "0x%04X: DCBI2CIndex: 0x%02X, I2CAddress: 0x%02X, "
+		      "Count: 0x%02X\n",
+		offset, i2c_index, i2c_address, count);
+
+	chan = init_i2c_device_find(bios->dev, i2c_index);
+	if (!chan)
+		return 0;
+
+	for (i = 0; i < count; i++) {
+		data[i] = bios->data[offset + 4 + i];
+
+		BIOSLOG(bios, "0x%04X: Data: 0x%02X\n", offset, data[i]);
+	}
+
+	if (bios->execute) {
+		msg.addr = i2c_address;
+		msg.flags = 0;
+		msg.len = count;
+		msg.buf = data;
+		if (i2c_transfer(&chan->adapter, &msg, 1) != 1)
+			return 0;
+	}
+
+	return len;
+}
+
+static int
+init_tmds(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_TMDS   opcode: 0x4F ('O')	(non-canon name)
+	 *
+	 * offset      (8 bit): opcode
+	 * offset + 1  (8 bit): magic lookup value
+	 * offset + 2  (8 bit): TMDS address
+	 * offset + 3  (8 bit): mask
+	 * offset + 4  (8 bit): data
+	 *
+	 * Read the data reg for TMDS address "TMDS address", AND it with mask
+	 * and OR it with data, then write it back
+	 * "magic lookup value" determines which TMDS base address register is
+	 * used -- see get_tmds_index_reg()
+	 */
+
+	uint8_t mlv = bios->data[offset + 1];
+	uint32_t tmdsaddr = bios->data[offset + 2];
+	uint8_t mask = bios->data[offset + 3];
+	uint8_t data = bios->data[offset + 4];
+	uint32_t reg, value;
+
+	if (!iexec->execute)
+		return 5;
+
+	BIOSLOG(bios, "0x%04X: MagicLookupValue: 0x%02X, TMDSAddr: 0x%02X, "
+		      "Mask: 0x%02X, Data: 0x%02X\n",
+		offset, mlv, tmdsaddr, mask, data);
+
+	reg = get_tmds_index_reg(bios->dev, mlv);
+	if (!reg)
+		return 0;
+
+	bios_wr32(bios, reg,
+		  tmdsaddr | NV_PRAMDAC_FP_TMDS_CONTROL_WRITE_DISABLE);
+	value = (bios_rd32(bios, reg + 4) & mask) | data;
+	bios_wr32(bios, reg + 4, value);
+	bios_wr32(bios, reg, tmdsaddr);
+
+	return 5;
+}
+
+static int
+init_zm_tmds_group(struct nvbios *bios, uint16_t offset,
+		   struct init_exec *iexec)
+{
+	/*
+	 * INIT_ZM_TMDS_GROUP   opcode: 0x50 ('P')	(non-canon name)
+	 *
+	 * offset      (8 bit): opcode
+	 * offset + 1  (8 bit): magic lookup value
+	 * offset + 2  (8 bit): count
+	 * offset + 3  (8 bit): addr 1
+	 * offset + 4  (8 bit): data 1
+	 * ...
+	 *
+	 * For each of "count" TMDS address and data pairs write "data n" to
+	 * "addr n".  "magic lookup value" determines which TMDS base address
+	 * register is used -- see get_tmds_index_reg()
+	 */
+
+	uint8_t mlv = bios->data[offset + 1];
+	uint8_t count = bios->data[offset + 2];
+	int len = 3 + count * 2;
+	uint32_t reg;
+	int i;
+
+	if (!iexec->execute)
+		return len;
+
+	BIOSLOG(bios, "0x%04X: MagicLookupValue: 0x%02X, Count: 0x%02X\n",
+		offset, mlv, count);
+
+	reg = get_tmds_index_reg(bios->dev, mlv);
+	if (!reg)
+		return 0;
+
+	for (i = 0; i < count; i++) {
+		uint8_t tmdsaddr = bios->data[offset + 3 + i * 2];
+		uint8_t tmdsdata = bios->data[offset + 4 + i * 2];
+
+		bios_wr32(bios, reg + 4, tmdsdata);
+		bios_wr32(bios, reg, tmdsaddr);
+	}
+
+	return len;
+}
+
+static int
+init_cr_idx_adr_latch(struct nvbios *bios, uint16_t offset,
+		      struct init_exec *iexec)
+{
+	/*
+	 * INIT_CR_INDEX_ADDRESS_LATCHED   opcode: 0x51 ('Q')
+	 *
+	 * offset      (8 bit): opcode
+	 * offset + 1  (8 bit): CRTC index1
+	 * offset + 2  (8 bit): CRTC index2
+	 * offset + 3  (8 bit): baseaddr
+	 * offset + 4  (8 bit): count
+	 * offset + 5  (8 bit): data 1
+	 * ...
+	 *
+	 * For each of "count" address and data pairs, write "baseaddr + n" to
+	 * "CRTC index1" and "data n" to "CRTC index2"
+	 * Once complete, restore initial value read from "CRTC index1"
+	 */
+	uint8_t crtcindex1 = bios->data[offset + 1];
+	uint8_t crtcindex2 = bios->data[offset + 2];
+	uint8_t baseaddr = bios->data[offset + 3];
+	uint8_t count = bios->data[offset + 4];
+	int len = 5 + count;
+	uint8_t oldaddr, data;
+	int i;
+
+	if (!iexec->execute)
+		return len;
+
+	BIOSLOG(bios, "0x%04X: Index1: 0x%02X, Index2: 0x%02X, "
+		      "BaseAddr: 0x%02X, Count: 0x%02X\n",
+		offset, crtcindex1, crtcindex2, baseaddr, count);
+
+	oldaddr = bios_idxprt_rd(bios, NV_CIO_CRX__COLOR, crtcindex1);
+
+	for (i = 0; i < count; i++) {
+		bios_idxprt_wr(bios, NV_CIO_CRX__COLOR, crtcindex1,
+				     baseaddr + i);
+		data = bios->data[offset + 5 + i];
+		bios_idxprt_wr(bios, NV_CIO_CRX__COLOR, crtcindex2, data);
+	}
+
+	bios_idxprt_wr(bios, NV_CIO_CRX__COLOR, crtcindex1, oldaddr);
+
+	return len;
+}
+
+static int
+init_cr(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_CR   opcode: 0x52 ('R')
+	 *
+	 * offset      (8  bit): opcode
+	 * offset + 1  (8  bit): CRTC index
+	 * offset + 2  (8  bit): mask
+	 * offset + 3  (8  bit): data
+	 *
+	 * Assign the value of at "CRTC index" ANDed with mask and ORed with
+	 * data back to "CRTC index"
+	 */
+
+	uint8_t crtcindex = bios->data[offset + 1];
+	uint8_t mask = bios->data[offset + 2];
+	uint8_t data = bios->data[offset + 3];
+	uint8_t value;
+
+	if (!iexec->execute)
+		return 4;
+
+	BIOSLOG(bios, "0x%04X: Index: 0x%02X, Mask: 0x%02X, Data: 0x%02X\n",
+		offset, crtcindex, mask, data);
+
+	value  = bios_idxprt_rd(bios, NV_CIO_CRX__COLOR, crtcindex) & mask;
+	value |= data;
+	bios_idxprt_wr(bios, NV_CIO_CRX__COLOR, crtcindex, value);
+
+	return 4;
+}
+
+static int
+init_zm_cr(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_ZM_CR   opcode: 0x53 ('S')
+	 *
+	 * offset      (8 bit): opcode
+	 * offset + 1  (8 bit): CRTC index
+	 * offset + 2  (8 bit): value
+	 *
+	 * Assign "value" to CRTC register with index "CRTC index".
+	 */
+
+	uint8_t crtcindex = ROM32(bios->data[offset + 1]);
+	uint8_t data = bios->data[offset + 2];
+
+	if (!iexec->execute)
+		return 3;
+
+	bios_idxprt_wr(bios, NV_CIO_CRX__COLOR, crtcindex, data);
+
+	return 3;
+}
+
+static int
+init_zm_cr_group(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_ZM_CR_GROUP   opcode: 0x54 ('T')
+	 *
+	 * offset      (8 bit): opcode
+	 * offset + 1  (8 bit): count
+	 * offset + 2  (8 bit): CRTC index 1
+	 * offset + 3  (8 bit): value 1
+	 * ...
+	 *
+	 * For "count", assign "value n" to CRTC register with index
+	 * "CRTC index n".
+	 */
+
+	uint8_t count = bios->data[offset + 1];
+	int len = 2 + count * 2;
+	int i;
+
+	if (!iexec->execute)
+		return len;
+
+	for (i = 0; i < count; i++)
+		init_zm_cr(bios, offset + 2 + 2 * i - 1, iexec);
+
+	return len;
+}
+
+static int
+init_condition_time(struct nvbios *bios, uint16_t offset,
+		    struct init_exec *iexec)
+{
+	/*
+	 * INIT_CONDITION_TIME   opcode: 0x56 ('V')
+	 *
+	 * offset      (8 bit): opcode
+	 * offset + 1  (8 bit): condition number
+	 * offset + 2  (8 bit): retries / 50
+	 *
+	 * Check condition "condition number" in the condition table.
+	 * Bios code then sleeps for 2ms if the condition is not met, and
+	 * repeats up to "retries" times, but on one C51 this has proved
+	 * insufficient.  In mmiotraces the driver sleeps for 20ms, so we do
+	 * this, and bail after "retries" times, or 2s, whichever is less.
+	 * If still not met after retries, clear execution flag for this table.
+	 */
+
+	uint8_t cond = bios->data[offset + 1];
+	uint16_t retries = bios->data[offset + 2] * 50;
+	unsigned cnt;
+
+	if (!iexec->execute)
+		return 3;
+
+	if (retries > 100)
+		retries = 100;
+
+	BIOSLOG(bios, "0x%04X: Condition: 0x%02X, Retries: 0x%02X\n",
+		offset, cond, retries);
+
+	if (!bios->execute) /* avoid 2s delays when "faking" execution */
+		retries = 1;
+
+	for (cnt = 0; cnt < retries; cnt++) {
+		if (bios_condition_met(bios, offset, cond)) {
+			BIOSLOG(bios, "0x%04X: Condition met, continuing\n",
+								offset);
+			break;
+		} else {
+			BIOSLOG(bios, "0x%04X: "
+				"Condition not met, sleeping for 20ms\n",
+								offset);
+			msleep(20);
+		}
+	}
+
+	if (!bios_condition_met(bios, offset, cond)) {
+		NV_WARN(bios->dev,
+			"0x%04X: Condition still not met after %dms, "
+			"skipping following opcodes\n", offset, 20 * retries);
+		iexec->execute = false;
+	}
+
+	return 3;
+}
+
+static int
+init_zm_reg_sequence(struct nvbios *bios, uint16_t offset,
+		     struct init_exec *iexec)
+{
+	/*
+	 * INIT_ZM_REG_SEQUENCE   opcode: 0x58 ('X')
+	 *
+	 * offset      (8  bit): opcode
+	 * offset + 1  (32 bit): base register
+	 * offset + 5  (8  bit): count
+	 * offset + 6  (32 bit): value 1
+	 * ...
+	 *
+	 * Starting at offset + 6 there are "count" 32 bit values.
+	 * For "count" iterations set "base register" + 4 * current_iteration
+	 * to "value current_iteration"
+	 */
+
+	uint32_t basereg = ROM32(bios->data[offset + 1]);
+	uint32_t count = bios->data[offset + 5];
+	int len = 6 + count * 4;
+	int i;
+
+	if (!iexec->execute)
+		return len;
+
+	BIOSLOG(bios, "0x%04X: BaseReg: 0x%08X, Count: 0x%02X\n",
+		offset, basereg, count);
+
+	for (i = 0; i < count; i++) {
+		uint32_t reg = basereg + i * 4;
+		uint32_t data = ROM32(bios->data[offset + 6 + i * 4]);
+
+		bios_wr32(bios, reg, data);
+	}
+
+	return len;
+}
+
+static int
+init_sub_direct(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_SUB_DIRECT   opcode: 0x5B ('[')
+	 *
+	 * offset      (8  bit): opcode
+	 * offset + 1  (16 bit): subroutine offset (in bios)
+	 *
+	 * Calls a subroutine that will execute commands until INIT_DONE
+	 * is found.
+	 */
+
+	uint16_t sub_offset = ROM16(bios->data[offset + 1]);
+
+	if (!iexec->execute)
+		return 3;
+
+	BIOSLOG(bios, "0x%04X: Executing subroutine at 0x%04X\n",
+		offset, sub_offset);
+
+	parse_init_table(bios, sub_offset, iexec);
+
+	BIOSLOG(bios, "0x%04X: End of 0x%04X subroutine\n", offset, sub_offset);
+
+	return 3;
+}
+
+static int
+init_copy_nv_reg(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_COPY_NV_REG   opcode: 0x5F ('_')
+	 *
+	 * offset      (8  bit): opcode
+	 * offset + 1  (32 bit): src reg
+	 * offset + 5  (8  bit): shift
+	 * offset + 6  (32 bit): src mask
+	 * offset + 10 (32 bit): xor
+	 * offset + 14 (32 bit): dst reg
+	 * offset + 18 (32 bit): dst mask
+	 *
+	 * Shift REGVAL("src reg") right by (signed) "shift", AND result with
+	 * "src mask", then XOR with "xor". Write this OR'd with
+	 * (REGVAL("dst reg") AND'd with "dst mask") to "dst reg"
+	 */
+
+	uint32_t srcreg = *((uint32_t *)(&bios->data[offset + 1]));
+	uint8_t shift = bios->data[offset + 5];
+	uint32_t srcmask = *((uint32_t *)(&bios->data[offset + 6]));
+	uint32_t xor = *((uint32_t *)(&bios->data[offset + 10]));
+	uint32_t dstreg = *((uint32_t *)(&bios->data[offset + 14]));
+	uint32_t dstmask = *((uint32_t *)(&bios->data[offset + 18]));
+	uint32_t srcvalue, dstvalue;
+
+	if (!iexec->execute)
+		return 22;
+
+	BIOSLOG(bios, "0x%04X: SrcReg: 0x%08X, Shift: 0x%02X, SrcMask: 0x%08X, "
+		      "Xor: 0x%08X, DstReg: 0x%08X, DstMask: 0x%08X\n",
+		offset, srcreg, shift, srcmask, xor, dstreg, dstmask);
+
+	srcvalue = bios_rd32(bios, srcreg);
+
+	if (shift < 0x80)
+		srcvalue >>= shift;
+	else
+		srcvalue <<= (0x100 - shift);
+
+	srcvalue = (srcvalue & srcmask) ^ xor;
+
+	dstvalue = bios_rd32(bios, dstreg) & dstmask;
+
+	bios_wr32(bios, dstreg, dstvalue | srcvalue);
+
+	return 22;
+}
+
+static int
+init_zm_index_io(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_ZM_INDEX_IO   opcode: 0x62 ('b')
+	 *
+	 * offset      (8  bit): opcode
+	 * offset + 1  (16 bit): CRTC port
+	 * offset + 3  (8  bit): CRTC index
+	 * offset + 4  (8  bit): data
+	 *
+	 * Write "data" to index "CRTC index" of "CRTC port"
+	 */
+	uint16_t crtcport = ROM16(bios->data[offset + 1]);
+	uint8_t crtcindex = bios->data[offset + 3];
+	uint8_t data = bios->data[offset + 4];
+
+	if (!iexec->execute)
+		return 5;
+
+	bios_idxprt_wr(bios, crtcport, crtcindex, data);
+
+	return 5;
+}
+
+static int
+init_compute_mem(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_COMPUTE_MEM   opcode: 0x63 ('c')
+	 *
+	 * offset      (8 bit): opcode
+	 *
+	 * This opcode is meant to set NV_PFB_CFG0 (0x100200) appropriately so
+	 * that the hardware can correctly calculate how much VRAM it has
+	 * (and subsequently report that value in NV_PFB_CSTATUS (0x10020C))
+	 *
+	 * The implementation of this opcode in general consists of two parts:
+	 * 1) determination of the memory bus width
+	 * 2) determination of how many of the card's RAM pads have ICs attached
+	 *
+	 * 1) is done by a cunning combination of writes to offsets 0x1c and
+	 * 0x3c in the framebuffer, and seeing whether the written values are
+	 * read back correctly. This then affects bits 4-7 of NV_PFB_CFG0
+	 *
+	 * 2) is done by a cunning combination of writes to an offset slightly
+	 * less than the maximum memory reported by NV_PFB_CSTATUS, then seeing
+	 * if the test pattern can be read back. This then affects bits 12-15 of
+	 * NV_PFB_CFG0
+	 *
+	 * In this context a "cunning combination" may include multiple reads
+	 * and writes to varying locations, often alternating the test pattern
+	 * and 0, doubtless to make sure buffers are filled, residual charges
+	 * on tracks are removed etc.
+	 *
+	 * Unfortunately, the "cunning combination"s mentioned above, and the
+	 * changes to the bits in NV_PFB_CFG0 differ with nearly every bios
+	 * trace I have.
+	 *
+	 * Therefore, we cheat and assume the value of NV_PFB_CFG0 with which
+	 * we started was correct, and use that instead
+	 */
+
+	/* no iexec->execute check by design */
+
+	/*
+	 * This appears to be a NOP on G8x chipsets, both io logs of the VBIOS
+	 * and kmmio traces of the binary driver POSTing the card show nothing
+	 * being done for this opcode.  why is it still listed in the table?!
+	 */
+
+	struct drm_nouveau_private *dev_priv = bios->dev->dev_private;
+
+	if (dev_priv->card_type >= NV_40)
+		return 1;
+
+	/*
+	 * On every card I've seen, this step gets done for us earlier in
+	 * the init scripts
+	uint8_t crdata = bios_idxprt_rd(dev, NV_VIO_SRX, 0x01);
+	bios_idxprt_wr(dev, NV_VIO_SRX, 0x01, crdata | 0x20);
+	 */
+
+	/*
+	 * This also has probably been done in the scripts, but an mmio trace of
+	 * s3 resume shows nvidia doing it anyway (unlike the NV_VIO_SRX write)
+	 */
+	bios_wr32(bios, NV_PFB_REFCTRL, NV_PFB_REFCTRL_VALID_1);
+
+	/* write back the saved configuration value */
+	bios_wr32(bios, NV_PFB_CFG0, bios->state.saved_nv_pfb_cfg0);
+
+	return 1;
+}
+
+static int
+init_reset(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_RESET   opcode: 0x65 ('e')
+	 *
+	 * offset      (8  bit): opcode
+	 * offset + 1  (32 bit): register
+	 * offset + 5  (32 bit): value1
+	 * offset + 9  (32 bit): value2
+	 *
+	 * Assign "value1" to "register", then assign "value2" to "register"
+	 */
+
+	uint32_t reg = ROM32(bios->data[offset + 1]);
+	uint32_t value1 = ROM32(bios->data[offset + 5]);
+	uint32_t value2 = ROM32(bios->data[offset + 9]);
+	uint32_t pci_nv_19, pci_nv_20;
+
+	/* no iexec->execute check by design */
+
+	pci_nv_19 = bios_rd32(bios, NV_PBUS_PCI_NV_19);
+	bios_wr32(bios, NV_PBUS_PCI_NV_19, 0);
+	bios_wr32(bios, reg, value1);
+
+	udelay(10);
+
+	bios_wr32(bios, reg, value2);
+	bios_wr32(bios, NV_PBUS_PCI_NV_19, pci_nv_19);
+
+	pci_nv_20 = bios_rd32(bios, NV_PBUS_PCI_NV_20);
+	pci_nv_20 &= ~NV_PBUS_PCI_NV_20_ROM_SHADOW_ENABLED;	/* 0xfffffffe */
+	bios_wr32(bios, NV_PBUS_PCI_NV_20, pci_nv_20);
+
+	return 13;
+}
+
+static int
+init_configure_mem(struct nvbios *bios, uint16_t offset,
+		   struct init_exec *iexec)
+{
+	/*
+	 * INIT_CONFIGURE_MEM   opcode: 0x66 ('f')
+	 *
+	 * offset      (8 bit): opcode
+	 *
+	 * Equivalent to INIT_DONE on bios version 3 or greater.
+	 * For early bios versions, sets up the memory registers, using values
+	 * taken from the memory init table
+	 */
+
+	/* no iexec->execute check by design */
+
+	uint16_t meminitoffs = bios->legacy.mem_init_tbl_ptr + MEM_INIT_SIZE * (bios_idxprt_rd(bios, NV_CIO_CRX__COLOR, NV_CIO_CRE_SCRATCH4__INDEX) >> 4);
+	uint16_t seqtbloffs = bios->legacy.sdr_seq_tbl_ptr, meminitdata = meminitoffs + 6;
+	uint32_t reg, data;
+
+	if (bios->major_version > 2)
+		return 0;
+
+	bios_idxprt_wr(bios, NV_VIO_SRX, NV_VIO_SR_CLOCK_INDEX, bios_idxprt_rd(
+		       bios, NV_VIO_SRX, NV_VIO_SR_CLOCK_INDEX) | 0x20);
+
+	if (bios->data[meminitoffs] & 1)
+		seqtbloffs = bios->legacy.ddr_seq_tbl_ptr;
+
+	for (reg = ROM32(bios->data[seqtbloffs]);
+	     reg != 0xffffffff;
+	     reg = ROM32(bios->data[seqtbloffs += 4])) {
+
+		switch (reg) {
+		case NV_PFB_PRE:
+			data = NV_PFB_PRE_CMD_PRECHARGE;
+			break;
+		case NV_PFB_PAD:
+			data = NV_PFB_PAD_CKE_NORMAL;
+			break;
+		case NV_PFB_REF:
+			data = NV_PFB_REF_CMD_REFRESH;
+			break;
+		default:
+			data = ROM32(bios->data[meminitdata]);
+			meminitdata += 4;
+			if (data == 0xffffffff)
+				continue;
+		}
+
+		bios_wr32(bios, reg, data);
+	}
+
+	return 1;
+}
+
+static int
+init_configure_clk(struct nvbios *bios, uint16_t offset,
+		   struct init_exec *iexec)
+{
+	/*
+	 * INIT_CONFIGURE_CLK   opcode: 0x67 ('g')
+	 *
+	 * offset      (8 bit): opcode
+	 *
+	 * Equivalent to INIT_DONE on bios version 3 or greater.
+	 * For early bios versions, sets up the NVClk and MClk PLLs, using
+	 * values taken from the memory init table
+	 */
+
+	/* no iexec->execute check by design */
+
+	uint16_t meminitoffs = bios->legacy.mem_init_tbl_ptr + MEM_INIT_SIZE * (bios_idxprt_rd(bios, NV_CIO_CRX__COLOR, NV_CIO_CRE_SCRATCH4__INDEX) >> 4);
+	int clock;
+
+	if (bios->major_version > 2)
+		return 0;
+
+	clock = ROM16(bios->data[meminitoffs + 4]) * 10;
+	setPLL(bios, NV_PRAMDAC_NVPLL_COEFF, clock);
+
+	clock = ROM16(bios->data[meminitoffs + 2]) * 10;
+	if (bios->data[meminitoffs] & 1) /* DDR */
+		clock *= 2;
+	setPLL(bios, NV_PRAMDAC_MPLL_COEFF, clock);
+
+	return 1;
+}
+
+static int
+init_configure_preinit(struct nvbios *bios, uint16_t offset,
+		       struct init_exec *iexec)
+{
+	/*
+	 * INIT_CONFIGURE_PREINIT   opcode: 0x68 ('h')
+	 *
+	 * offset      (8 bit): opcode
+	 *
+	 * Equivalent to INIT_DONE on bios version 3 or greater.
+	 * For early bios versions, does early init, loading ram and crystal
+	 * configuration from straps into CR3C
+	 */
+
+	/* no iexec->execute check by design */
+
+	uint32_t straps = bios_rd32(bios, NV_PEXTDEV_BOOT_0);
+	uint8_t cr3c = ((straps << 2) & 0xf0) | (straps & (1 << 6));
+
+	if (bios->major_version > 2)
+		return 0;
+
+	bios_idxprt_wr(bios, NV_CIO_CRX__COLOR,
+			     NV_CIO_CRE_SCRATCH4__INDEX, cr3c);
+
+	return 1;
+}
+
+static int
+init_io(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_IO   opcode: 0x69 ('i')
+	 *
+	 * offset      (8  bit): opcode
+	 * offset + 1  (16 bit): CRTC port
+	 * offset + 3  (8  bit): mask
+	 * offset + 4  (8  bit): data
+	 *
+	 * Assign ((IOVAL("crtc port") & "mask") | "data") to "crtc port"
+	 */
+
+	struct drm_nouveau_private *dev_priv = bios->dev->dev_private;
+	uint16_t crtcport = ROM16(bios->data[offset + 1]);
+	uint8_t mask = bios->data[offset + 3];
+	uint8_t data = bios->data[offset + 4];
+
+	if (!iexec->execute)
+		return 5;
+
+	BIOSLOG(bios, "0x%04X: Port: 0x%04X, Mask: 0x%02X, Data: 0x%02X\n",
+		offset, crtcport, mask, data);
+
+	/*
+	 * I have no idea what this does, but NVIDIA do this magic sequence
+	 * in the places where this INIT_IO happens..
+	 */
+	if (dev_priv->card_type >= NV_50 && crtcport == 0x3c3 && data == 1) {
+		int i;
+
+		bios_wr32(bios, 0x614100, (bios_rd32(
+			  bios, 0x614100) & 0x0fffffff) | 0x00800000);
+
+		bios_wr32(bios, 0x00e18c, bios_rd32(
+			  bios, 0x00e18c) | 0x00020000);
+
+		bios_wr32(bios, 0x614900, (bios_rd32(
+			  bios, 0x614900) & 0x0fffffff) | 0x00800000);
+
+		bios_wr32(bios, 0x000200, bios_rd32(
+			  bios, 0x000200) & ~0x40000000);
+
+		mdelay(10);
+
+		bios_wr32(bios, 0x00e18c, bios_rd32(
+			  bios, 0x00e18c) & ~0x00020000);
+
+		bios_wr32(bios, 0x000200, bios_rd32(
+			  bios, 0x000200) | 0x40000000);
+
+		bios_wr32(bios, 0x614100, 0x00800018);
+		bios_wr32(bios, 0x614900, 0x00800018);
+
+		mdelay(10);
+
+		bios_wr32(bios, 0x614100, 0x10000018);
+		bios_wr32(bios, 0x614900, 0x10000018);
+
+		for (i = 0; i < 3; i++)
+			bios_wr32(bios, 0x614280 + (i*0x800), bios_rd32(
+				  bios, 0x614280 + (i*0x800)) & 0xf0f0f0f0);
+
+		for (i = 0; i < 2; i++)
+			bios_wr32(bios, 0x614300 + (i*0x800), bios_rd32(
+				  bios, 0x614300 + (i*0x800)) & 0xfffff0f0);
+
+		for (i = 0; i < 3; i++)
+			bios_wr32(bios, 0x614380 + (i*0x800), bios_rd32(
+				  bios, 0x614380 + (i*0x800)) & 0xfffff0f0);
+
+		for (i = 0; i < 2; i++)
+			bios_wr32(bios, 0x614200 + (i*0x800), bios_rd32(
+				  bios, 0x614200 + (i*0x800)) & 0xfffffff0);
+
+		for (i = 0; i < 2; i++)
+			bios_wr32(bios, 0x614108 + (i*0x800), bios_rd32(
+				  bios, 0x614108 + (i*0x800)) & 0x0fffffff);
+		return 5;
+	}
+
+	bios_port_wr(bios, crtcport, (bios_port_rd(bios, crtcport) & mask) |
+									data);
+	return 5;
+}
+
+static int
+init_sub(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_SUB   opcode: 0x6B ('k')
+	 *
+	 * offset      (8 bit): opcode
+	 * offset + 1  (8 bit): script number
+	 *
+	 * Execute script number "script number", as a subroutine
+	 */
+
+	uint8_t sub = bios->data[offset + 1];
+
+	if (!iexec->execute)
+		return 2;
+
+	BIOSLOG(bios, "0x%04X: Calling script %d\n", offset, sub);
+
+	parse_init_table(bios,
+			 ROM16(bios->data[bios->init_script_tbls_ptr + sub * 2]),
+			 iexec);
+
+	BIOSLOG(bios, "0x%04X: End of script %d\n", offset, sub);
+
+	return 2;
+}
+
+static int
+init_ram_condition(struct nvbios *bios, uint16_t offset,
+		   struct init_exec *iexec)
+{
+	/*
+	 * INIT_RAM_CONDITION   opcode: 0x6D ('m')
+	 *
+	 * offset      (8 bit): opcode
+	 * offset + 1  (8 bit): mask
+	 * offset + 2  (8 bit): cmpval
+	 *
+	 * Test if (NV_PFB_BOOT_0 & "mask") equals "cmpval".
+	 * If condition not met skip subsequent opcodes until condition is
+	 * inverted (INIT_NOT), or we hit INIT_RESUME
+	 */
+
+	uint8_t mask = bios->data[offset + 1];
+	uint8_t cmpval = bios->data[offset + 2];
+	uint8_t data;
+
+	if (!iexec->execute)
+		return 3;
+
+	data = bios_rd32(bios, NV_PFB_BOOT_0) & mask;
+
+	BIOSLOG(bios, "0x%04X: Checking if 0x%08X equals 0x%08X\n",
+		offset, data, cmpval);
+
+	if (data == cmpval)
+		BIOSLOG(bios, "0x%04X: Condition fulfilled -- continuing to execute\n", offset);
+	else {
+		BIOSLOG(bios, "0x%04X: Condition not fulfilled -- skipping following commands\n", offset);
+		iexec->execute = false;
+	}
+
+	return 3;
+}
+
+static int
+init_nv_reg(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_NV_REG   opcode: 0x6E ('n')
+	 *
+	 * offset      (8  bit): opcode
+	 * offset + 1  (32 bit): register
+	 * offset + 5  (32 bit): mask
+	 * offset + 9  (32 bit): data
+	 *
+	 * Assign ((REGVAL("register") & "mask") | "data") to "register"
+	 */
+
+	uint32_t reg = ROM32(bios->data[offset + 1]);
+	uint32_t mask = ROM32(bios->data[offset + 5]);
+	uint32_t data = ROM32(bios->data[offset + 9]);
+
+	if (!iexec->execute)
+		return 13;
+
+	BIOSLOG(bios, "0x%04X: Reg: 0x%08X, Mask: 0x%08X, Data: 0x%08X\n",
+		offset, reg, mask, data);
+
+	bios_wr32(bios, reg, (bios_rd32(bios, reg) & mask) | data);
+
+	return 13;
+}
+
+static int
+init_macro(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_MACRO   opcode: 0x6F ('o')
+	 *
+	 * offset      (8 bit): opcode
+	 * offset + 1  (8 bit): macro number
+	 *
+	 * Look up macro index "macro number" in the macro index table.
+	 * The macro index table entry has 1 byte for the index in the macro
+	 * table, and 1 byte for the number of times to repeat the macro.
+	 * The macro table entry has 4 bytes for the register address and
+	 * 4 bytes for the value to write to that register
+	 */
+
+	uint8_t macro_index_tbl_idx = bios->data[offset + 1];
+	uint16_t tmp = bios->macro_index_tbl_ptr + (macro_index_tbl_idx * MACRO_INDEX_SIZE);
+	uint8_t macro_tbl_idx = bios->data[tmp];
+	uint8_t count = bios->data[tmp + 1];
+	uint32_t reg, data;
+	int i;
+
+	if (!iexec->execute)
+		return 2;
+
+	BIOSLOG(bios, "0x%04X: Macro: 0x%02X, MacroTableIndex: 0x%02X, "
+		      "Count: 0x%02X\n",
+		offset, macro_index_tbl_idx, macro_tbl_idx, count);
+
+	for (i = 0; i < count; i++) {
+		uint16_t macroentryptr = bios->macro_tbl_ptr + (macro_tbl_idx + i) * MACRO_SIZE;
+
+		reg = ROM32(bios->data[macroentryptr]);
+		data = ROM32(bios->data[macroentryptr + 4]);
+
+		bios_wr32(bios, reg, data);
+	}
+
+	return 2;
+}
+
+static int
+init_done(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_DONE   opcode: 0x71 ('q')
+	 *
+	 * offset      (8  bit): opcode
+	 *
+	 * End the current script
+	 */
+
+	/* mild retval abuse to stop parsing this table */
+	return 0;
+}
+
+static int
+init_resume(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_RESUME   opcode: 0x72 ('r')
+	 *
+	 * offset      (8  bit): opcode
+	 *
+	 * End the current execute / no-execute condition
+	 */
+
+	if (iexec->execute)
+		return 1;
+
+	iexec->execute = true;
+	BIOSLOG(bios, "0x%04X: ---- Executing following commands ----\n", offset);
+
+	return 1;
+}
+
+static int
+init_time(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_TIME   opcode: 0x74 ('t')
+	 *
+	 * offset      (8  bit): opcode
+	 * offset + 1  (16 bit): time
+	 *
+	 * Sleep for "time" microseconds.
+	 */
+
+	unsigned time = ROM16(bios->data[offset + 1]);
+
+	if (!iexec->execute)
+		return 3;
+
+	BIOSLOG(bios, "0x%04X: Sleeping for 0x%04X microseconds\n",
+		offset, time);
+
+	if (time < 1000)
+		udelay(time);
+	else
+		msleep((time + 900) / 1000);
+
+	return 3;
+}
+
+static int
+init_condition(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_CONDITION   opcode: 0x75 ('u')
+	 *
+	 * offset      (8 bit): opcode
+	 * offset + 1  (8 bit): condition number
+	 *
+	 * Check condition "condition number" in the condition table.
+	 * If condition not met skip subsequent opcodes until condition is
+	 * inverted (INIT_NOT), or we hit INIT_RESUME
+	 */
+
+	uint8_t cond = bios->data[offset + 1];
+
+	if (!iexec->execute)
+		return 2;
+
+	BIOSLOG(bios, "0x%04X: Condition: 0x%02X\n", offset, cond);
+
+	if (bios_condition_met(bios, offset, cond))
+		BIOSLOG(bios, "0x%04X: Condition fulfilled -- continuing to execute\n", offset);
+	else {
+		BIOSLOG(bios, "0x%04X: Condition not fulfilled -- skipping following commands\n", offset);
+		iexec->execute = false;
+	}
+
+	return 2;
+}
+
+static int
+init_io_condition(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_IO_CONDITION  opcode: 0x76
+	 *
+	 * offset      (8 bit): opcode
+	 * offset + 1  (8 bit): condition number
+	 *
+	 * Check condition "condition number" in the io condition table.
+	 * If condition not met skip subsequent opcodes until condition is
+	 * inverted (INIT_NOT), or we hit INIT_RESUME
+	 */
+
+	uint8_t cond = bios->data[offset + 1];
+
+	if (!iexec->execute)
+		return 2;
+
+	BIOSLOG(bios, "0x%04X: IO condition: 0x%02X\n", offset, cond);
+
+	if (io_condition_met(bios, offset, cond))
+		BIOSLOG(bios, "0x%04X: Condition fulfilled -- continuing to execute\n", offset);
+	else {
+		BIOSLOG(bios, "0x%04X: Condition not fulfilled -- skipping following commands\n", offset);
+		iexec->execute = false;
+	}
+
+	return 2;
+}
+
+static int
+init_index_io(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_INDEX_IO   opcode: 0x78 ('x')
+	 *
+	 * offset      (8  bit): opcode
+	 * offset + 1  (16 bit): CRTC port
+	 * offset + 3  (8  bit): CRTC index
+	 * offset + 4  (8  bit): mask
+	 * offset + 5  (8  bit): data
+	 *
+	 * Read value at index "CRTC index" on "CRTC port", AND with "mask",
+	 * OR with "data", write-back
+	 */
+
+	uint16_t crtcport = ROM16(bios->data[offset + 1]);
+	uint8_t crtcindex = bios->data[offset + 3];
+	uint8_t mask = bios->data[offset + 4];
+	uint8_t data = bios->data[offset + 5];
+	uint8_t value;
+
+	if (!iexec->execute)
+		return 6;
+
+	BIOSLOG(bios, "0x%04X: Port: 0x%04X, Index: 0x%02X, Mask: 0x%02X, "
+		      "Data: 0x%02X\n",
+		offset, crtcport, crtcindex, mask, data);
+
+	value = (bios_idxprt_rd(bios, crtcport, crtcindex) & mask) | data;
+	bios_idxprt_wr(bios, crtcport, crtcindex, value);
+
+	return 6;
+}
+
+static int
+init_pll(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_PLL   opcode: 0x79 ('y')
+	 *
+	 * offset      (8  bit): opcode
+	 * offset + 1  (32 bit): register
+	 * offset + 5  (16 bit): freq
+	 *
+	 * Set PLL register "register" to coefficients for frequency (10kHz)
+	 * "freq"
+	 */
+
+	uint32_t reg = ROM32(bios->data[offset + 1]);
+	uint16_t freq = ROM16(bios->data[offset + 5]);
+
+	if (!iexec->execute)
+		return 7;
+
+	BIOSLOG(bios, "0x%04X: Reg: 0x%08X, Freq: %d0kHz\n", offset, reg, freq);
+
+	setPLL(bios, reg, freq * 10);
+
+	return 7;
+}
+
+static int
+init_zm_reg(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_ZM_REG   opcode: 0x7A ('z')
+	 *
+	 * offset      (8  bit): opcode
+	 * offset + 1  (32 bit): register
+	 * offset + 5  (32 bit): value
+	 *
+	 * Assign "value" to "register"
+	 */
+
+	uint32_t reg = ROM32(bios->data[offset + 1]);
+	uint32_t value = ROM32(bios->data[offset + 5]);
+
+	if (!iexec->execute)
+		return 9;
+
+	if (reg == 0x000200)
+		value |= 1;
+
+	bios_wr32(bios, reg, value);
+
+	return 9;
+}
+
+static int
+init_ram_restrict_pll(struct nvbios *bios, uint16_t offset,
+		      struct init_exec *iexec)
+{
+	/*
+	 * INIT_RAM_RESTRICT_PLL   opcode: 0x87 ('')
+	 *
+	 * offset      (8 bit): opcode
+	 * offset + 1  (8 bit): PLL type
+	 * offset + 2 (32 bit): frequency 0
+	 *
+	 * Uses the RAMCFG strap of PEXTDEV_BOOT as an index into the table at
+	 * ram_restrict_table_ptr.  The value read from there is used to select
+	 * a frequency from the table starting at 'frequency 0' to be
+	 * programmed into the PLL corresponding to 'type'.
+	 *
+	 * The PLL limits table on cards using this opcode has a mapping of
+	 * 'type' to the relevant registers.
+	 */
+
+	struct drm_device *dev = bios->dev;
+	uint32_t strap = (bios_rd32(bios, NV_PEXTDEV_BOOT_0) & 0x0000003c) >> 2;
+	uint8_t index = bios->data[bios->ram_restrict_tbl_ptr + strap];
+	uint8_t type = bios->data[offset + 1];
+	uint32_t freq = ROM32(bios->data[offset + 2 + (index * 4)]);
+	uint8_t *pll_limits = &bios->data[bios->pll_limit_tbl_ptr], *entry;
+	int len = 2 + bios->ram_restrict_group_count * 4;
+	int i;
+
+	if (!iexec->execute)
+		return len;
+
+	if (!bios->pll_limit_tbl_ptr || (pll_limits[0] & 0xf0) != 0x30) {
+		NV_ERROR(dev, "PLL limits table not version 3.x\n");
+		return len; /* deliberate, allow default clocks to remain */
+	}
+
+	entry = pll_limits + pll_limits[1];
+	for (i = 0; i < pll_limits[3]; i++, entry += pll_limits[2]) {
+		if (entry[0] == type) {
+			uint32_t reg = ROM32(entry[3]);
+
+			BIOSLOG(bios, "0x%04X: "
+				      "Type %02x Reg 0x%08x Freq %dKHz\n",
+				offset, type, reg, freq);
+
+			setPLL(bios, reg, freq);
+			return len;
+		}
+	}
+
+	NV_ERROR(dev, "PLL type 0x%02x not found in PLL limits table", type);
+	return len;
+}
+
+static int
+init_8c(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_8C   opcode: 0x8C ('')
+	 *
+	 * NOP so far....
+	 *
+	 */
+
+	return 1;
+}
+
+static int
+init_8d(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_8D   opcode: 0x8D ('')
+	 *
+	 * NOP so far....
+	 *
+	 */
+
+	return 1;
+}
+
+static int
+init_gpio(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_GPIO   opcode: 0x8E ('')
+	 *
+	 * offset      (8 bit): opcode
+	 *
+	 * Loop over all entries in the DCB GPIO table, and initialise
+	 * each GPIO according to various values listed in each entry
+	 */
+
+	const uint32_t nv50_gpio_reg[4] = { 0xe104, 0xe108, 0xe280, 0xe284 };
+	const uint32_t nv50_gpio_ctl[2] = { 0xe100, 0xe28c };
+	const uint8_t *gpio_table = &bios->data[bios->dcb.gpio_table_ptr];
+	const uint8_t *gpio_entry;
+	int i;
+
+	if (!iexec->execute)
+		return 1;
+
+	if (bios->dcb.version != 0x40) {
+		NV_ERROR(bios->dev, "DCB table not version 4.0\n");
+		return 0;
+	}
+
+	if (!bios->dcb.gpio_table_ptr) {
+		NV_WARN(bios->dev, "Invalid pointer to INIT_8E table\n");
+		return 0;
+	}
+
+	gpio_entry = gpio_table + gpio_table[1];
+	for (i = 0; i < gpio_table[2]; i++, gpio_entry += gpio_table[3]) {
+		uint32_t entry = ROM32(gpio_entry[0]), r, s, v;
+		int line = (entry & 0x0000001f);
+
+		BIOSLOG(bios, "0x%04X: Entry: 0x%08X\n", offset, entry);
+
+		if ((entry & 0x0000ff00) == 0x0000ff00)
+			continue;
+
+		r = nv50_gpio_reg[line >> 3];
+		s = (line & 0x07) << 2;
+		v = bios_rd32(bios, r) & ~(0x00000003 << s);
+		if (entry & 0x01000000)
+			v |= (((entry & 0x60000000) >> 29) ^ 2) << s;
+		else
+			v |= (((entry & 0x18000000) >> 27) ^ 2) << s;
+		bios_wr32(bios, r, v);
+
+		r = nv50_gpio_ctl[line >> 4];
+		s = (line & 0x0f);
+		v = bios_rd32(bios, r) & ~(0x00010001 << s);
+		switch ((entry & 0x06000000) >> 25) {
+		case 1:
+			v |= (0x00000001 << s);
+			break;
+		case 2:
+			v |= (0x00010000 << s);
+			break;
+		default:
+			break;
+		}
+		bios_wr32(bios, r, v);
+	}
+
+	return 1;
+}
+
+static int
+init_ram_restrict_zm_reg_group(struct nvbios *bios, uint16_t offset,
+			       struct init_exec *iexec)
+{
+	/*
+	 * INIT_RAM_RESTRICT_ZM_REG_GROUP   opcode: 0x8F ('')
+	 *
+	 * offset      (8  bit): opcode
+	 * offset + 1  (32 bit): reg
+	 * offset + 5  (8  bit): regincrement
+	 * offset + 6  (8  bit): count
+	 * offset + 7  (32 bit): value 1,1
+	 * ...
+	 *
+	 * Use the RAMCFG strap of PEXTDEV_BOOT as an index into the table at
+	 * ram_restrict_table_ptr. The value read from here is 'n', and
+	 * "value 1,n" gets written to "reg". This repeats "count" times and on
+	 * each iteration 'm', "reg" increases by "regincrement" and
+	 * "value m,n" is used. The extent of n is limited by a number read
+	 * from the 'M' BIT table, herein called "blocklen"
+	 */
+
+	uint32_t reg = ROM32(bios->data[offset + 1]);
+	uint8_t regincrement = bios->data[offset + 5];
+	uint8_t count = bios->data[offset + 6];
+	uint32_t strap_ramcfg, data;
+	/* previously set by 'M' BIT table */
+	uint16_t blocklen = bios->ram_restrict_group_count * 4;
+	int len = 7 + count * blocklen;
+	uint8_t index;
+	int i;
+
+
+	if (!iexec->execute)
+		return len;
+
+	if (!blocklen) {
+		NV_ERROR(bios->dev,
+			 "0x%04X: Zero block length - has the M table "
+			 "been parsed?\n", offset);
+		return 0;
+	}
+
+	strap_ramcfg = (bios_rd32(bios, NV_PEXTDEV_BOOT_0) >> 2) & 0xf;
+	index = bios->data[bios->ram_restrict_tbl_ptr + strap_ramcfg];
+
+	BIOSLOG(bios, "0x%04X: Reg: 0x%08X, RegIncrement: 0x%02X, "
+		      "Count: 0x%02X, StrapRamCfg: 0x%02X, Index: 0x%02X\n",
+		offset, reg, regincrement, count, strap_ramcfg, index);
+
+	for (i = 0; i < count; i++) {
+		data = ROM32(bios->data[offset + 7 + index * 4 + blocklen * i]);
+
+		bios_wr32(bios, reg, data);
+
+		reg += regincrement;
+	}
+
+	return len;
+}
+
+static int
+init_copy_zm_reg(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_COPY_ZM_REG   opcode: 0x90 ('')
+	 *
+	 * offset      (8  bit): opcode
+	 * offset + 1  (32 bit): src reg
+	 * offset + 5  (32 bit): dst reg
+	 *
+	 * Put contents of "src reg" into "dst reg"
+	 */
+
+	uint32_t srcreg = ROM32(bios->data[offset + 1]);
+	uint32_t dstreg = ROM32(bios->data[offset + 5]);
+
+	if (!iexec->execute)
+		return 9;
+
+	bios_wr32(bios, dstreg, bios_rd32(bios, srcreg));
+
+	return 9;
+}
+
+static int
+init_zm_reg_group_addr_latched(struct nvbios *bios, uint16_t offset,
+			       struct init_exec *iexec)
+{
+	/*
+	 * INIT_ZM_REG_GROUP_ADDRESS_LATCHED   opcode: 0x91 ('')
+	 *
+	 * offset      (8  bit): opcode
+	 * offset + 1  (32 bit): dst reg
+	 * offset + 5  (8  bit): count
+	 * offset + 6  (32 bit): data 1
+	 * ...
+	 *
+	 * For each of "count" values write "data n" to "dst reg"
+	 */
+
+	uint32_t reg = ROM32(bios->data[offset + 1]);
+	uint8_t count = bios->data[offset + 5];
+	int len = 6 + count * 4;
+	int i;
+
+	if (!iexec->execute)
+		return len;
+
+	for (i = 0; i < count; i++) {
+		uint32_t data = ROM32(bios->data[offset + 6 + 4 * i]);
+		bios_wr32(bios, reg, data);
+	}
+
+	return len;
+}
+
+static int
+init_reserved(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_RESERVED   opcode: 0x92 ('')
+	 *
+	 * offset      (8 bit): opcode
+	 *
+	 * Seemingly does nothing
+	 */
+
+	return 1;
+}
+
+static int
+init_96(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_96   opcode: 0x96 ('')
+	 *
+	 * offset      (8  bit): opcode
+	 * offset + 1  (32 bit): sreg
+	 * offset + 5  (8  bit): sshift
+	 * offset + 6  (8  bit): smask
+	 * offset + 7  (8  bit): index
+	 * offset + 8  (32 bit): reg
+	 * offset + 12 (32 bit): mask
+	 * offset + 16 (8  bit): shift
+	 *
+	 */
+
+	uint16_t xlatptr = bios->init96_tbl_ptr + (bios->data[offset + 7] * 2);
+	uint32_t reg = ROM32(bios->data[offset + 8]);
+	uint32_t mask = ROM32(bios->data[offset + 12]);
+	uint32_t val;
+
+	val = bios_rd32(bios, ROM32(bios->data[offset + 1]));
+	if (bios->data[offset + 5] < 0x80)
+		val >>= bios->data[offset + 5];
+	else
+		val <<= (0x100 - bios->data[offset + 5]);
+	val &= bios->data[offset + 6];
+
+	val   = bios->data[ROM16(bios->data[xlatptr]) + val];
+	val <<= bios->data[offset + 16];
+
+	if (!iexec->execute)
+		return 17;
+
+	bios_wr32(bios, reg, (bios_rd32(bios, reg) & mask) | val);
+	return 17;
+}
+
+static int
+init_97(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_97   opcode: 0x97 ('')
+	 *
+	 * offset      (8  bit): opcode
+	 * offset + 1  (32 bit): register
+	 * offset + 5  (32 bit): mask
+	 * offset + 9  (32 bit): value
+	 *
+	 * Adds "value" to "register" preserving the fields specified
+	 * by "mask"
+	 */
+
+	uint32_t reg = ROM32(bios->data[offset + 1]);
+	uint32_t mask = ROM32(bios->data[offset + 5]);
+	uint32_t add = ROM32(bios->data[offset + 9]);
+	uint32_t val;
+
+	val = bios_rd32(bios, reg);
+	val = (val & mask) | ((val + add) & ~mask);
+
+	if (!iexec->execute)
+		return 13;
+
+	bios_wr32(bios, reg, val);
+	return 13;
+}
+
+static int
+init_auxch(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_AUXCH   opcode: 0x98 ('')
+	 *
+	 * offset      (8  bit): opcode
+	 * offset + 1  (32 bit): address
+	 * offset + 5  (8  bit): count
+	 * offset + 6  (8  bit): mask 0
+	 * offset + 7  (8  bit): data 0
+	 *  ...
+	 *
+	 */
+
+	struct drm_device *dev = bios->dev;
+	struct nouveau_i2c_chan *auxch;
+	uint32_t addr = ROM32(bios->data[offset + 1]);
+	uint8_t count = bios->data[offset + 5];
+	int len = 6 + count * 2;
+	int ret, i;
+
+	if (!bios->display.output) {
+		NV_ERROR(dev, "INIT_AUXCH: no active output\n");
+		return 0;
+	}
+
+	auxch = init_i2c_device_find(dev, bios->display.output->i2c_index);
+	if (!auxch) {
+		NV_ERROR(dev, "INIT_AUXCH: couldn't get auxch %d\n",
+			 bios->display.output->i2c_index);
+		return 0;
+	}
+
+	if (!iexec->execute)
+		return len;
+
+	offset += 6;
+	for (i = 0; i < count; i++, offset += 2) {
+		uint8_t data;
+
+		ret = nouveau_dp_auxch(auxch, 9, addr, &data, 1);
+		if (ret) {
+			NV_ERROR(dev, "INIT_AUXCH: rd auxch fail %d\n", ret);
+			return 0;
+		}
+
+		data &= bios->data[offset + 0];
+		data |= bios->data[offset + 1];
+
+		ret = nouveau_dp_auxch(auxch, 8, addr, &data, 1);
+		if (ret) {
+			NV_ERROR(dev, "INIT_AUXCH: wr auxch fail %d\n", ret);
+			return 0;
+		}
+	}
+
+	return len;
+}
+
+static int
+init_zm_auxch(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
+{
+	/*
+	 * INIT_ZM_AUXCH   opcode: 0x99 ('')
+	 *
+	 * offset      (8  bit): opcode
+	 * offset + 1  (32 bit): address
+	 * offset + 5  (8  bit): count
+	 * offset + 6  (8  bit): data 0
+	 *  ...
+	 *
+	 */
+
+	struct drm_device *dev = bios->dev;
+	struct nouveau_i2c_chan *auxch;
+	uint32_t addr = ROM32(bios->data[offset + 1]);
+	uint8_t count = bios->data[offset + 5];
+	int len = 6 + count;
+	int ret, i;
+
+	if (!bios->display.output) {
+		NV_ERROR(dev, "INIT_ZM_AUXCH: no active output\n");
+		return 0;
+	}
+
+	auxch = init_i2c_device_find(dev, bios->display.output->i2c_index);
+	if (!auxch) {
+		NV_ERROR(dev, "INIT_ZM_AUXCH: couldn't get auxch %d\n",
+			 bios->display.output->i2c_index);
+		return 0;
+	}
+
+	if (!iexec->execute)
+		return len;
+
+	offset += 6;
+	for (i = 0; i < count; i++, offset++) {
+		ret = nouveau_dp_auxch(auxch, 8, addr, &bios->data[offset], 1);
+		if (ret) {
+			NV_ERROR(dev, "INIT_ZM_AUXCH: wr auxch fail %d\n", ret);
+			return 0;
+		}
+	}
+
+	return len;
+}
+
+static struct init_tbl_entry itbl_entry[] = {
+	/* command name                       , id  , length  , offset  , mult    , command handler                 */
+	/* INIT_PROG (0x31, 15, 10, 4) removed due to no example of use */
+	{ "INIT_IO_RESTRICT_PROG"             , 0x32, init_io_restrict_prog           },
+	{ "INIT_REPEAT"                       , 0x33, init_repeat                     },
+	{ "INIT_IO_RESTRICT_PLL"              , 0x34, init_io_restrict_pll            },
+	{ "INIT_END_REPEAT"                   , 0x36, init_end_repeat                 },
+	{ "INIT_COPY"                         , 0x37, init_copy                       },
+	{ "INIT_NOT"                          , 0x38, init_not                        },
+	{ "INIT_IO_FLAG_CONDITION"            , 0x39, init_io_flag_condition          },
+	{ "INIT_INDEX_ADDRESS_LATCHED"        , 0x49, init_idx_addr_latched           },
+	{ "INIT_IO_RESTRICT_PLL2"             , 0x4A, init_io_restrict_pll2           },
+	{ "INIT_PLL2"                         , 0x4B, init_pll2                       },
+	{ "INIT_I2C_BYTE"                     , 0x4C, init_i2c_byte                   },
+	{ "INIT_ZM_I2C_BYTE"                  , 0x4D, init_zm_i2c_byte                },
+	{ "INIT_ZM_I2C"                       , 0x4E, init_zm_i2c                     },
+	{ "INIT_TMDS"                         , 0x4F, init_tmds                       },
+	{ "INIT_ZM_TMDS_GROUP"                , 0x50, init_zm_tmds_group              },
+	{ "INIT_CR_INDEX_ADDRESS_LATCHED"     , 0x51, init_cr_idx_adr_latch           },
+	{ "INIT_CR"                           , 0x52, init_cr                         },
+	{ "INIT_ZM_CR"                        , 0x53, init_zm_cr                      },
+	{ "INIT_ZM_CR_GROUP"                  , 0x54, init_zm_cr_group                },
+	{ "INIT_CONDITION_TIME"               , 0x56, init_condition_time             },
+	{ "INIT_ZM_REG_SEQUENCE"              , 0x58, init_zm_reg_sequence            },
+	/* INIT_INDIRECT_REG (0x5A, 7, 0, 0) removed due to no example of use */
+	{ "INIT_SUB_DIRECT"                   , 0x5B, init_sub_direct                 },
+	{ "INIT_COPY_NV_REG"                  , 0x5F, init_copy_nv_reg                },
+	{ "INIT_ZM_INDEX_IO"                  , 0x62, init_zm_index_io                },
+	{ "INIT_COMPUTE_MEM"                  , 0x63, init_compute_mem                },
+	{ "INIT_RESET"                        , 0x65, init_reset                      },
+	{ "INIT_CONFIGURE_MEM"                , 0x66, init_configure_mem              },
+	{ "INIT_CONFIGURE_CLK"                , 0x67, init_configure_clk              },
+	{ "INIT_CONFIGURE_PREINIT"            , 0x68, init_configure_preinit          },
+	{ "INIT_IO"                           , 0x69, init_io                         },
+	{ "INIT_SUB"                          , 0x6B, init_sub                        },
+	{ "INIT_RAM_CONDITION"                , 0x6D, init_ram_condition              },
+	{ "INIT_NV_REG"                       , 0x6E, init_nv_reg                     },
+	{ "INIT_MACRO"                        , 0x6F, init_macro                      },
+	{ "INIT_DONE"                         , 0x71, init_done                       },
+	{ "INIT_RESUME"                       , 0x72, init_resume                     },
+	/* INIT_RAM_CONDITION2 (0x73, 9, 0, 0) removed due to no example of use */
+	{ "INIT_TIME"                         , 0x74, init_time                       },
+	{ "INIT_CONDITION"                    , 0x75, init_condition                  },
+	{ "INIT_IO_CONDITION"                 , 0x76, init_io_condition               },
+	{ "INIT_INDEX_IO"                     , 0x78, init_index_io                   },
+	{ "INIT_PLL"                          , 0x79, init_pll                        },
+	{ "INIT_ZM_REG"                       , 0x7A, init_zm_reg                     },
+	{ "INIT_RAM_RESTRICT_PLL"             , 0x87, init_ram_restrict_pll           },
+	{ "INIT_8C"                           , 0x8C, init_8c                         },
+	{ "INIT_8D"                           , 0x8D, init_8d                         },
+	{ "INIT_GPIO"                         , 0x8E, init_gpio                       },
+	{ "INIT_RAM_RESTRICT_ZM_REG_GROUP"    , 0x8F, init_ram_restrict_zm_reg_group  },
+	{ "INIT_COPY_ZM_REG"                  , 0x90, init_copy_zm_reg                },
+	{ "INIT_ZM_REG_GROUP_ADDRESS_LATCHED" , 0x91, init_zm_reg_group_addr_latched  },
+	{ "INIT_RESERVED"                     , 0x92, init_reserved                   },
+	{ "INIT_96"                           , 0x96, init_96                         },
+	{ "INIT_97"                           , 0x97, init_97                         },
+	{ "INIT_AUXCH"                        , 0x98, init_auxch                      },
+	{ "INIT_ZM_AUXCH"                     , 0x99, init_zm_auxch                   },
+	{ NULL                                , 0   , NULL                            }
+};
+
+#define MAX_TABLE_OPS 1000
+
+static int
+parse_init_table(struct nvbios *bios, unsigned int offset,
+		 struct init_exec *iexec)
+{
+	/*
+	 * Parses all commands in an init table.
+	 *
+	 * We start out executing all commands found in the init table. Some
+	 * opcodes may change the status of iexec->execute to SKIP, which will
+	 * cause the following opcodes to perform no operation until the value
+	 * is changed back to EXECUTE.
+	 */
+
+	int count = 0, i, res;
+	uint8_t id;
+
+	/*
+	 * Loop until INIT_DONE causes us to break out of the loop
+	 * (or until offset > bios length just in case... )
+	 * (and no more than MAX_TABLE_OPS iterations, just in case... )
+	 */
+	while ((offset < bios->length) && (count++ < MAX_TABLE_OPS)) {
+		id = bios->data[offset];
+
+		/* Find matching id in itbl_entry */
+		for (i = 0; itbl_entry[i].name && (itbl_entry[i].id != id); i++)
+			;
+
+		if (itbl_entry[i].name) {
+			BIOSLOG(bios, "0x%04X: [ (0x%02X) - %s ]\n",
+				offset, itbl_entry[i].id, itbl_entry[i].name);
+
+			/* execute eventual command handler */
+			res = (*itbl_entry[i].handler)(bios, offset, iexec);
+			if (!res)
+				break;
+			/*
+			 * Add the offset of the current command including all data
+			 * of that command. The offset will then be pointing on the
+			 * next op code.
+			 */
+			offset += res;
+		} else {
+			NV_ERROR(bios->dev,
+				 "0x%04X: Init table command not found: "
+				 "0x%02X\n", offset, id);
+			return -ENOENT;
+		}
+	}
+
+	if (offset >= bios->length)
+		NV_WARN(bios->dev,
+			"Offset 0x%04X greater than known bios image length.  "
+			"Corrupt image?\n", offset);
+	if (count >= MAX_TABLE_OPS)
+		NV_WARN(bios->dev,
+			"More than %d opcodes to a table is unlikely, "
+			"is the bios image corrupt?\n", MAX_TABLE_OPS);
+
+	return 0;
+}
+
+static void
+parse_init_tables(struct nvbios *bios)
+{
+	/* Loops and calls parse_init_table() for each present table. */
+
+	int i = 0;
+	uint16_t table;
+	struct init_exec iexec = {true, false};
+
+	if (bios->old_style_init) {
+		if (bios->init_script_tbls_ptr)
+			parse_init_table(bios, bios->init_script_tbls_ptr, &iexec);
+		if (bios->extra_init_script_tbl_ptr)
+			parse_init_table(bios, bios->extra_init_script_tbl_ptr, &iexec);
+
+		return;
+	}
+
+	while ((table = ROM16(bios->data[bios->init_script_tbls_ptr + i]))) {
+		NV_INFO(bios->dev,
+			"Parsing VBIOS init table %d at offset 0x%04X\n",
+			i / 2, table);
+		BIOSLOG(bios, "0x%04X: ------ Executing following commands ------\n", table);
+
+		parse_init_table(bios, table, &iexec);
+		i += 2;
+	}
+}
+
+static uint16_t clkcmptable(struct nvbios *bios, uint16_t clktable, int pxclk)
+{
+	int compare_record_len, i = 0;
+	uint16_t compareclk, scriptptr = 0;
+
+	if (bios->major_version < 5) /* pre BIT */
+		compare_record_len = 3;
+	else
+		compare_record_len = 4;
+
+	do {
+		compareclk = ROM16(bios->data[clktable + compare_record_len * i]);
+		if (pxclk >= compareclk * 10) {
+			if (bios->major_version < 5) {
+				uint8_t tmdssub = bios->data[clktable + 2 + compare_record_len * i];
+				scriptptr = ROM16(bios->data[bios->init_script_tbls_ptr + tmdssub * 2]);
+			} else
+				scriptptr = ROM16(bios->data[clktable + 2 + compare_record_len * i]);
+			break;
+		}
+		i++;
+	} while (compareclk);
+
+	return scriptptr;
+}
+
+static void
+run_digital_op_script(struct drm_device *dev, uint16_t scriptptr,
+		      struct dcb_entry *dcbent, int head, bool dl)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nvbios *bios = &dev_priv->vbios;
+	struct init_exec iexec = {true, false};
+
+	NV_TRACE(dev, "0x%04X: Parsing digital output script table\n",
+		 scriptptr);
+	bios_idxprt_wr(bios, NV_CIO_CRX__COLOR, NV_CIO_CRE_44,
+		       head ? NV_CIO_CRE_44_HEADB : NV_CIO_CRE_44_HEADA);
+	/* note: if dcb entries have been merged, index may be misleading */
+	NVWriteVgaCrtc5758(dev, head, 0, dcbent->index);
+	parse_init_table(bios, scriptptr, &iexec);
+
+	nv04_dfp_bind_head(dev, dcbent, head, dl);
+}
+
+static int call_lvds_manufacturer_script(struct drm_device *dev, struct dcb_entry *dcbent, int head, enum LVDS_script script)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nvbios *bios = &dev_priv->vbios;
+	uint8_t sub = bios->data[bios->fp.xlated_entry + script] + (bios->fp.link_c_increment && dcbent->or & OUTPUT_C ? 1 : 0);
+	uint16_t scriptofs = ROM16(bios->data[bios->init_script_tbls_ptr + sub * 2]);
+
+	if (!bios->fp.xlated_entry || !sub || !scriptofs)
+		return -EINVAL;
+
+	run_digital_op_script(dev, scriptofs, dcbent, head, bios->fp.dual_link);
+
+	if (script == LVDS_PANEL_OFF) {
+		/* off-on delay in ms */
+		msleep(ROM16(bios->data[bios->fp.xlated_entry + 7]));
+	}
+#ifdef __powerpc__
+	/* Powerbook specific quirks */
+	if ((dev->pci_device & 0xffff) == 0x0179 ||
+	    (dev->pci_device & 0xffff) == 0x0189 ||
+	    (dev->pci_device & 0xffff) == 0x0329) {
+		if (script == LVDS_RESET) {
+			nv_write_tmds(dev, dcbent->or, 0, 0x02, 0x72);
+
+		} else if (script == LVDS_PANEL_ON) {
+			bios_wr32(bios, NV_PBUS_DEBUG_DUALHEAD_CTL,
+				  bios_rd32(bios, NV_PBUS_DEBUG_DUALHEAD_CTL)
+				  | (1 << 31));
+			bios_wr32(bios, NV_PCRTC_GPIO_EXT,
+				  bios_rd32(bios, NV_PCRTC_GPIO_EXT) | 1);
+
+		} else if (script == LVDS_PANEL_OFF) {
+			bios_wr32(bios, NV_PBUS_DEBUG_DUALHEAD_CTL,
+				  bios_rd32(bios, NV_PBUS_DEBUG_DUALHEAD_CTL)
+				  & ~(1 << 31));
+			bios_wr32(bios, NV_PCRTC_GPIO_EXT,
+				  bios_rd32(bios, NV_PCRTC_GPIO_EXT) & ~3);
+		}
+	}
+#endif
+
+	return 0;
+}
+
+static int run_lvds_table(struct drm_device *dev, struct dcb_entry *dcbent, int head, enum LVDS_script script, int pxclk)
+{
+	/*
+	 * The BIT LVDS table's header has the information to setup the
+	 * necessary registers. Following the standard 4 byte header are:
+	 * A bitmask byte and a dual-link transition pxclk value for use in
+	 * selecting the init script when not using straps; 4 script pointers
+	 * for panel power, selected by output and on/off; and 8 table pointers
+	 * for panel init, the needed one determined by output, and bits in the
+	 * conf byte. These tables are similar to the TMDS tables, consisting
+	 * of a list of pxclks and script pointers.
+	 */
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nvbios *bios = &dev_priv->vbios;
+	unsigned int outputset = (dcbent->or == 4) ? 1 : 0;
+	uint16_t scriptptr = 0, clktable;
+	uint8_t clktableptr = 0;
+
+	/*
+	 * For now we assume version 3.0 table - g80 support will need some
+	 * changes
+	 */
+
+	switch (script) {
+	case LVDS_INIT:
+		return -ENOSYS;
+	case LVDS_BACKLIGHT_ON:
+	case LVDS_PANEL_ON:
+		scriptptr = ROM16(bios->data[bios->fp.lvdsmanufacturerpointer + 7 + outputset * 2]);
+		break;
+	case LVDS_BACKLIGHT_OFF:
+	case LVDS_PANEL_OFF:
+		scriptptr = ROM16(bios->data[bios->fp.lvdsmanufacturerpointer + 11 + outputset * 2]);
+		break;
+	case LVDS_RESET:
+		if (dcbent->lvdsconf.use_straps_for_mode) {
+			if (bios->fp.dual_link)
+				clktableptr += 2;
+			if (bios->fp.BITbit1)
+				clktableptr++;
+		} else {
+			/* using EDID */
+			uint8_t fallback = bios->data[bios->fp.lvdsmanufacturerpointer + 4];
+			int fallbackcmpval = (dcbent->or == 4) ? 4 : 1;
+
+			if (bios->fp.dual_link) {
+				clktableptr += 2;
+				fallbackcmpval *= 2;
+			}
+			if (fallbackcmpval & fallback)
+				clktableptr++;
+		}
+
+		/* adding outputset * 8 may not be correct */
+		clktable = ROM16(bios->data[bios->fp.lvdsmanufacturerpointer + 15 + clktableptr * 2 + outputset * 8]);
+		if (!clktable) {
+			NV_ERROR(dev, "Pixel clock comparison table not found\n");
+			return -ENOENT;
+		}
+		scriptptr = clkcmptable(bios, clktable, pxclk);
+	}
+
+	if (!scriptptr) {
+		NV_ERROR(dev, "LVDS output init script not found\n");
+		return -ENOENT;
+	}
+	run_digital_op_script(dev, scriptptr, dcbent, head, bios->fp.dual_link);
+
+	return 0;
+}
+
+int call_lvds_script(struct drm_device *dev, struct dcb_entry *dcbent, int head, enum LVDS_script script, int pxclk)
+{
+	/*
+	 * LVDS operations are multiplexed in an effort to present a single API
+	 * which works with two vastly differing underlying structures.
+	 * This acts as the demux
+	 */
+
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nvbios *bios = &dev_priv->vbios;
+	uint8_t lvds_ver = bios->data[bios->fp.lvdsmanufacturerpointer];
+	uint32_t sel_clk_binding, sel_clk;
+	int ret;
+
+	if (bios->fp.last_script_invoc == (script << 1 | head) || !lvds_ver ||
+	    (lvds_ver >= 0x30 && script == LVDS_INIT))
+		return 0;
+
+	if (!bios->fp.lvds_init_run) {
+		bios->fp.lvds_init_run = true;
+		call_lvds_script(dev, dcbent, head, LVDS_INIT, pxclk);
+	}
+
+	if (script == LVDS_PANEL_ON && bios->fp.reset_after_pclk_change)
+		call_lvds_script(dev, dcbent, head, LVDS_RESET, pxclk);
+	if (script == LVDS_RESET && bios->fp.power_off_for_reset)
+		call_lvds_script(dev, dcbent, head, LVDS_PANEL_OFF, pxclk);
+
+	NV_TRACE(dev, "Calling LVDS script %d:\n", script);
+
+	/* don't let script change pll->head binding */
+	sel_clk_binding = bios_rd32(bios, NV_PRAMDAC_SEL_CLK) & 0x50000;
+
+	if (lvds_ver < 0x30)
+		ret = call_lvds_manufacturer_script(dev, dcbent, head, script);
+	else
+		ret = run_lvds_table(dev, dcbent, head, script, pxclk);
+
+	bios->fp.last_script_invoc = (script << 1 | head);
+
+	sel_clk = NVReadRAMDAC(dev, 0, NV_PRAMDAC_SEL_CLK) & ~0x50000;
+	NVWriteRAMDAC(dev, 0, NV_PRAMDAC_SEL_CLK, sel_clk | sel_clk_binding);
+	/* some scripts set a value in NV_PBUS_POWERCTRL_2 and break video overlay */
+	nvWriteMC(dev, NV_PBUS_POWERCTRL_2, 0);
+
+	return ret;
+}
+
+struct lvdstableheader {
+	uint8_t lvds_ver, headerlen, recordlen;
+};
+
+static int parse_lvds_manufacturer_table_header(struct drm_device *dev, struct nvbios *bios, struct lvdstableheader *lth)
+{
+	/*
+	 * BMP version (0xa) LVDS table has a simple header of version and
+	 * record length. The BIT LVDS table has the typical BIT table header:
+	 * version byte, header length byte, record length byte, and a byte for
+	 * the maximum number of records that can be held in the table.
+	 */
+
+	uint8_t lvds_ver, headerlen, recordlen;
+
+	memset(lth, 0, sizeof(struct lvdstableheader));
+
+	if (bios->fp.lvdsmanufacturerpointer == 0x0) {
+		NV_ERROR(dev, "Pointer to LVDS manufacturer table invalid\n");
+		return -EINVAL;
+	}
+
+	lvds_ver = bios->data[bios->fp.lvdsmanufacturerpointer];
+
+	switch (lvds_ver) {
+	case 0x0a:	/* pre NV40 */
+		headerlen = 2;
+		recordlen = bios->data[bios->fp.lvdsmanufacturerpointer + 1];
+		break;
+	case 0x30:	/* NV4x */
+		headerlen = bios->data[bios->fp.lvdsmanufacturerpointer + 1];
+		if (headerlen < 0x1f) {
+			NV_ERROR(dev, "LVDS table header not understood\n");
+			return -EINVAL;
+		}
+		recordlen = bios->data[bios->fp.lvdsmanufacturerpointer + 2];
+		break;
+	case 0x40:	/* G80/G90 */
+		headerlen = bios->data[bios->fp.lvdsmanufacturerpointer + 1];
+		if (headerlen < 0x7) {
+			NV_ERROR(dev, "LVDS table header not understood\n");
+			return -EINVAL;
+		}
+		recordlen = bios->data[bios->fp.lvdsmanufacturerpointer + 2];
+		break;
+	default:
+		NV_ERROR(dev,
+			 "LVDS table revision %d.%d not currently supported\n",
+			 lvds_ver >> 4, lvds_ver & 0xf);
+		return -ENOSYS;
+	}
+
+	lth->lvds_ver = lvds_ver;
+	lth->headerlen = headerlen;
+	lth->recordlen = recordlen;
+
+	return 0;
+}
+
+static int
+get_fp_strap(struct drm_device *dev, struct nvbios *bios)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	/*
+	 * The fp strap is normally dictated by the "User Strap" in
+	 * PEXTDEV_BOOT_0[20:16], but on BMP cards when bit 2 of the
+	 * Internal_Flags struct at 0x48 is set, the user strap gets overriden
+	 * by the PCI subsystem ID during POST, but not before the previous user
+	 * strap has been committed to CR58 for CR57=0xf on head A, which may be
+	 * read and used instead
+	 */
+
+	if (bios->major_version < 5 && bios->data[0x48] & 0x4)
+		return NVReadVgaCrtc5758(dev, 0, 0xf) & 0xf;
+
+	if (dev_priv->card_type >= NV_50)
+		return (bios_rd32(bios, NV_PEXTDEV_BOOT_0) >> 24) & 0xf;
+	else
+		return (bios_rd32(bios, NV_PEXTDEV_BOOT_0) >> 16) & 0xf;
+}
+
+static int parse_fp_mode_table(struct drm_device *dev, struct nvbios *bios)
+{
+	uint8_t *fptable;
+	uint8_t fptable_ver, headerlen = 0, recordlen, fpentries = 0xf, fpindex;
+	int ret, ofs, fpstrapping;
+	struct lvdstableheader lth;
+
+	if (bios->fp.fptablepointer == 0x0) {
+		/* Apple cards don't have the fp table; the laptops use DDC */
+		/* The table is also missing on some x86 IGPs */
+#ifndef __powerpc__
+		NV_ERROR(dev, "Pointer to flat panel table invalid\n");
+#endif
+		bios->digital_min_front_porch = 0x4b;
+		return 0;
+	}
+
+	fptable = &bios->data[bios->fp.fptablepointer];
+	fptable_ver = fptable[0];
+
+	switch (fptable_ver) {
+	/*
+	 * BMP version 0x5.0x11 BIOSen have version 1 like tables, but no
+	 * version field, and miss one of the spread spectrum/PWM bytes.
+	 * This could affect early GF2Go parts (not seen any appropriate ROMs
+	 * though). Here we assume that a version of 0x05 matches this case
+	 * (combining with a BMP version check would be better), as the
+	 * common case for the panel type field is 0x0005, and that is in
+	 * fact what we are reading the first byte of.
+	 */
+	case 0x05:	/* some NV10, 11, 15, 16 */
+		recordlen = 42;
+		ofs = -1;
+		break;
+	case 0x10:	/* some NV15/16, and NV11+ */
+		recordlen = 44;
+		ofs = 0;
+		break;
+	case 0x20:	/* NV40+ */
+		headerlen = fptable[1];
+		recordlen = fptable[2];
+		fpentries = fptable[3];
+		/*
+		 * fptable[4] is the minimum
+		 * RAMDAC_FP_HCRTC -> RAMDAC_FP_HSYNC_START gap
+		 */
+		bios->digital_min_front_porch = fptable[4];
+		ofs = -7;
+		break;
+	default:
+		NV_ERROR(dev,
+			 "FP table revision %d.%d not currently supported\n",
+			 fptable_ver >> 4, fptable_ver & 0xf);
+		return -ENOSYS;
+	}
+
+	if (!bios->is_mobile) /* !mobile only needs digital_min_front_porch */
+		return 0;
+
+	ret = parse_lvds_manufacturer_table_header(dev, bios, &lth);
+	if (ret)
+		return ret;
+
+	if (lth.lvds_ver == 0x30 || lth.lvds_ver == 0x40) {
+		bios->fp.fpxlatetableptr = bios->fp.lvdsmanufacturerpointer +
+							lth.headerlen + 1;
+		bios->fp.xlatwidth = lth.recordlen;
+	}
+	if (bios->fp.fpxlatetableptr == 0x0) {
+		NV_ERROR(dev, "Pointer to flat panel xlat table invalid\n");
+		return -EINVAL;
+	}
+
+	fpstrapping = get_fp_strap(dev, bios);
+
+	fpindex = bios->data[bios->fp.fpxlatetableptr +
+					fpstrapping * bios->fp.xlatwidth];
+
+	if (fpindex > fpentries) {
+		NV_ERROR(dev, "Bad flat panel table index\n");
+		return -ENOENT;
+	}
+
+	/* nv4x cards need both a strap value and fpindex of 0xf to use DDC */
+	if (lth.lvds_ver > 0x10)
+		bios->fp_no_ddc = fpstrapping != 0xf || fpindex != 0xf;
+
+	/*
+	 * If either the strap or xlated fpindex value are 0xf there is no
+	 * panel using a strap-derived bios mode present.  this condition
+	 * includes, but is different from, the DDC panel indicator above
+	 */
+	if (fpstrapping == 0xf || fpindex == 0xf)
+		return 0;
+
+	bios->fp.mode_ptr = bios->fp.fptablepointer + headerlen +
+			    recordlen * fpindex + ofs;
+
+	NV_TRACE(dev, "BIOS FP mode: %dx%d (%dkHz pixel clock)\n",
+		 ROM16(bios->data[bios->fp.mode_ptr + 11]) + 1,
+		 ROM16(bios->data[bios->fp.mode_ptr + 25]) + 1,
+		 ROM16(bios->data[bios->fp.mode_ptr + 7]) * 10);
+
+	return 0;
+}
+
+bool nouveau_bios_fp_mode(struct drm_device *dev, struct drm_display_mode *mode)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nvbios *bios = &dev_priv->vbios;
+	uint8_t *mode_entry = &bios->data[bios->fp.mode_ptr];
+
+	if (!mode)	/* just checking whether we can produce a mode */
+		return bios->fp.mode_ptr;
+
+	memset(mode, 0, sizeof(struct drm_display_mode));
+	/*
+	 * For version 1.0 (version in byte 0):
+	 * bytes 1-2 are "panel type", including bits on whether Colour/mono,
+	 * single/dual link, and type (TFT etc.)
+	 * bytes 3-6 are bits per colour in RGBX
+	 */
+	mode->clock = ROM16(mode_entry[7]) * 10;
+	/* bytes 9-10 is HActive */
+	mode->hdisplay = ROM16(mode_entry[11]) + 1;
+	/*
+	 * bytes 13-14 is HValid Start
+	 * bytes 15-16 is HValid End
+	 */
+	mode->hsync_start = ROM16(mode_entry[17]) + 1;
+	mode->hsync_end = ROM16(mode_entry[19]) + 1;
+	mode->htotal = ROM16(mode_entry[21]) + 1;
+	/* bytes 23-24, 27-30 similarly, but vertical */
+	mode->vdisplay = ROM16(mode_entry[25]) + 1;
+	mode->vsync_start = ROM16(mode_entry[31]) + 1;
+	mode->vsync_end = ROM16(mode_entry[33]) + 1;
+	mode->vtotal = ROM16(mode_entry[35]) + 1;
+	mode->flags |= (mode_entry[37] & 0x10) ?
+			DRM_MODE_FLAG_PHSYNC : DRM_MODE_FLAG_NHSYNC;
+	mode->flags |= (mode_entry[37] & 0x1) ?
+			DRM_MODE_FLAG_PVSYNC : DRM_MODE_FLAG_NVSYNC;
+	/*
+	 * bytes 38-39 relate to spread spectrum settings
+	 * bytes 40-43 are something to do with PWM
+	 */
+
+	mode->status = MODE_OK;
+	mode->type = DRM_MODE_TYPE_DRIVER | DRM_MODE_TYPE_PREFERRED;
+	drm_mode_set_name(mode);
+	return bios->fp.mode_ptr;
+}
+
+int nouveau_bios_parse_lvds_table(struct drm_device *dev, int pxclk, bool *dl, bool *if_is_24bit)
+{
+	/*
+	 * The LVDS table header is (mostly) described in
+	 * parse_lvds_manufacturer_table_header(): the BIT header additionally
+	 * contains the dual-link transition pxclk (in 10s kHz), at byte 5 - if
+	 * straps are not being used for the panel, this specifies the frequency
+	 * at which modes should be set up in the dual link style.
+	 *
+	 * Following the header, the BMP (ver 0xa) table has several records,
+	 * indexed by a separate xlat table, indexed in turn by the fp strap in
+	 * EXTDEV_BOOT. Each record had a config byte, followed by 6 script
+	 * numbers for use by INIT_SUB which controlled panel init and power,
+	 * and finally a dword of ms to sleep between power off and on
+	 * operations.
+	 *
+	 * In the BIT versions, the table following the header serves as an
+	 * integrated config and xlat table: the records in the table are
+	 * indexed by the FP strap nibble in EXTDEV_BOOT, and each record has
+	 * two bytes - the first as a config byte, the second for indexing the
+	 * fp mode table pointed to by the BIT 'D' table
+	 *
+	 * DDC is not used until after card init, so selecting the correct table
+	 * entry and setting the dual link flag for EDID equipped panels,
+	 * requiring tests against the native-mode pixel clock, cannot be done
+	 * until later, when this function should be called with non-zero pxclk
+	 */
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nvbios *bios = &dev_priv->vbios;
+	int fpstrapping = get_fp_strap(dev, bios), lvdsmanufacturerindex = 0;
+	struct lvdstableheader lth;
+	uint16_t lvdsofs;
+	int ret, chip_version = bios->chip_version;
+
+	ret = parse_lvds_manufacturer_table_header(dev, bios, &lth);
+	if (ret)
+		return ret;
+
+	switch (lth.lvds_ver) {
+	case 0x0a:	/* pre NV40 */
+		lvdsmanufacturerindex = bios->data[
+					bios->fp.fpxlatemanufacturertableptr +
+					fpstrapping];
+
+		/* we're done if this isn't the EDID panel case */
+		if (!pxclk)
+			break;
+
+		if (chip_version < 0x25) {
+			/* nv17 behaviour
+			 *
+			 * It seems the old style lvds script pointer is reused
+			 * to select 18/24 bit colour depth for EDID panels.
+			 */
+			lvdsmanufacturerindex =
+				(bios->legacy.lvds_single_a_script_ptr & 1) ?
+									2 : 0;
+			if (pxclk >= bios->fp.duallink_transition_clk)
+				lvdsmanufacturerindex++;
+		} else if (chip_version < 0x30) {
+			/* nv28 behaviour (off-chip encoder)
+			 *
+			 * nv28 does a complex dance of first using byte 121 of
+			 * the EDID to choose the lvdsmanufacturerindex, then
+			 * later attempting to match the EDID manufacturer and
+			 * product IDs in a table (signature 'pidt' (panel id
+			 * table?)), setting an lvdsmanufacturerindex of 0 and
+			 * an fp strap of the match index (or 0xf if none)
+			 */
+			lvdsmanufacturerindex = 0;
+		} else {
+			/* nv31, nv34 behaviour */
+			lvdsmanufacturerindex = 0;
+			if (pxclk >= bios->fp.duallink_transition_clk)
+				lvdsmanufacturerindex = 2;
+			if (pxclk >= 140000)
+				lvdsmanufacturerindex = 3;
+		}
+
+		/*
+		 * nvidia set the high nibble of (cr57=f, cr58) to
+		 * lvdsmanufacturerindex in this case; we don't
+		 */
+		break;
+	case 0x30:	/* NV4x */
+	case 0x40:	/* G80/G90 */
+		lvdsmanufacturerindex = fpstrapping;
+		break;
+	default:
+		NV_ERROR(dev, "LVDS table revision not currently supported\n");
+		return -ENOSYS;
+	}
+
+	lvdsofs = bios->fp.xlated_entry = bios->fp.lvdsmanufacturerpointer + lth.headerlen + lth.recordlen * lvdsmanufacturerindex;
+	switch (lth.lvds_ver) {
+	case 0x0a:
+		bios->fp.power_off_for_reset = bios->data[lvdsofs] & 1;
+		bios->fp.reset_after_pclk_change = bios->data[lvdsofs] & 2;
+		bios->fp.dual_link = bios->data[lvdsofs] & 4;
+		bios->fp.link_c_increment = bios->data[lvdsofs] & 8;
+		*if_is_24bit = bios->data[lvdsofs] & 16;
+		break;
+	case 0x30:
+		/*
+		 * My money would be on there being a 24 bit interface bit in
+		 * this table, but I have no example of a laptop bios with a
+		 * 24 bit panel to confirm that. Hence we shout loudly if any
+		 * bit other than bit 0 is set (I've not even seen bit 1)
+		 */
+		if (bios->data[lvdsofs] > 1)
+			NV_ERROR(dev,
+				 "You have a very unusual laptop display; please report it\n");
+		/*
+		 * No sign of the "power off for reset" or "reset for panel
+		 * on" bits, but it's safer to assume we should
+		 */
+		bios->fp.power_off_for_reset = true;
+		bios->fp.reset_after_pclk_change = true;
+		/*
+		 * It's ok lvdsofs is wrong for nv4x edid case; dual_link is
+		 * over-written, and BITbit1 isn't used
+		 */
+		bios->fp.dual_link = bios->data[lvdsofs] & 1;
+		bios->fp.BITbit1 = bios->data[lvdsofs] & 2;
+		bios->fp.duallink_transition_clk = ROM16(bios->data[bios->fp.lvdsmanufacturerpointer + 5]) * 10;
+		break;
+	case 0x40:
+		bios->fp.dual_link = bios->data[lvdsofs] & 1;
+		bios->fp.if_is_24bit = bios->data[lvdsofs] & 2;
+		bios->fp.strapless_is_24bit = bios->data[bios->fp.lvdsmanufacturerpointer + 4];
+		bios->fp.duallink_transition_clk = ROM16(bios->data[bios->fp.lvdsmanufacturerpointer + 5]) * 10;
+		break;
+	}
+
+	/* set dual_link flag for EDID case */
+	if (pxclk && (chip_version < 0x25 || chip_version > 0x28))
+		bios->fp.dual_link = (pxclk >= bios->fp.duallink_transition_clk);
+
+	*dl = bios->fp.dual_link;
+
+	return 0;
+}
+
+static uint8_t *
+bios_output_config_match(struct drm_device *dev, struct dcb_entry *dcbent,
+			 uint16_t record, int record_len, int record_nr)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nvbios *bios = &dev_priv->vbios;
+	uint32_t entry;
+	uint16_t table;
+	int i, v;
+
+	for (i = 0; i < record_nr; i++, record += record_len) {
+		table = ROM16(bios->data[record]);
+		if (!table)
+			continue;
+		entry = ROM32(bios->data[table]);
+
+		v = (entry & 0x000f0000) >> 16;
+		if (!(v & dcbent->or))
+			continue;
+
+		v = (entry & 0x000000f0) >> 4;
+		if (v != dcbent->location)
+			continue;
+
+		v = (entry & 0x0000000f);
+		if (v != dcbent->type)
+			continue;
+
+		return &bios->data[table];
+	}
+
+	return NULL;
+}
+
+void *
+nouveau_bios_dp_table(struct drm_device *dev, struct dcb_entry *dcbent,
+		      int *length)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nvbios *bios = &dev_priv->vbios;
+	uint8_t *table;
+
+	if (!bios->display.dp_table_ptr) {
+		NV_ERROR(dev, "No pointer to DisplayPort table\n");
+		return NULL;
+	}
+	table = &bios->data[bios->display.dp_table_ptr];
+
+	if (table[0] != 0x20 && table[0] != 0x21) {
+		NV_ERROR(dev, "DisplayPort table version 0x%02x unknown\n",
+			 table[0]);
+		return NULL;
+	}
+
+	*length = table[4];
+	return bios_output_config_match(dev, dcbent,
+					bios->display.dp_table_ptr + table[1],
+					table[2], table[3]);
+}
+
+int
+nouveau_bios_run_display_table(struct drm_device *dev, struct dcb_entry *dcbent,
+			       uint32_t sub, int pxclk)
+{
+	/*
+	 * The display script table is located by the BIT 'U' table.
+	 *
+	 * It contains an array of pointers to various tables describing
+	 * a particular output type.  The first 32-bits of the output
+	 * tables contains similar information to a DCB entry, and is
+	 * used to decide whether that particular table is suitable for
+	 * the output you want to access.
+	 *
+	 * The "record header length" field here seems to indicate the
+	 * offset of the first configuration entry in the output tables.
+	 * This is 10 on most cards I've seen, but 12 has been witnessed
+	 * on DP cards, and there's another script pointer within the
+	 * header.
+	 *
+	 * offset + 0   ( 8 bits): version
+	 * offset + 1   ( 8 bits): header length
+	 * offset + 2   ( 8 bits): record length
+	 * offset + 3   ( 8 bits): number of records
+	 * offset + 4   ( 8 bits): record header length
+	 * offset + 5   (16 bits): pointer to first output script table
+	 */
+
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nvbios *bios = &dev_priv->vbios;
+	uint8_t *table = &bios->data[bios->display.script_table_ptr];
+	uint8_t *otable = NULL;
+	uint16_t script;
+	int i = 0;
+
+	if (!bios->display.script_table_ptr) {
+		NV_ERROR(dev, "No pointer to output script table\n");
+		return 1;
+	}
+
+	/*
+	 * Nothing useful has been in any of the pre-2.0 tables I've seen,
+	 * so until they are, we really don't need to care.
+	 */
+	if (table[0] < 0x20)
+		return 1;
+
+	if (table[0] != 0x20 && table[0] != 0x21) {
+		NV_ERROR(dev, "Output script table version 0x%02x unknown\n",
+			 table[0]);
+		return 1;
+	}
+
+	/*
+	 * The output script tables describing a particular output type
+	 * look as follows:
+	 *
+	 * offset + 0   (32 bits): output this table matches (hash of DCB)
+	 * offset + 4   ( 8 bits): unknown
+	 * offset + 5   ( 8 bits): number of configurations
+	 * offset + 6   (16 bits): pointer to some script
+	 * offset + 8   (16 bits): pointer to some script
+	 *
+	 * headerlen == 10
+	 * offset + 10           : configuration 0
+	 *
+	 * headerlen == 12
+	 * offset + 10           : pointer to some script
+	 * offset + 12           : configuration 0
+	 *
+	 * Each config entry is as follows:
+	 *
+	 * offset + 0   (16 bits): unknown, assumed to be a match value
+	 * offset + 2   (16 bits): pointer to script table (clock set?)
+	 * offset + 4   (16 bits): pointer to script table (reset?)
+	 *
+	 * There doesn't appear to be a count value to say how many
+	 * entries exist in each script table, instead, a 0 value in
+	 * the first 16-bit word seems to indicate both the end of the
+	 * list and the default entry.  The second 16-bit word in the
+	 * script tables is a pointer to the script to execute.
+	 */
+
+	NV_DEBUG_KMS(dev, "Searching for output entry for %d %d %d\n",
+			dcbent->type, dcbent->location, dcbent->or);
+	otable = bios_output_config_match(dev, dcbent, table[1] +
+					  bios->display.script_table_ptr,
+					  table[2], table[3]);
+	if (!otable) {
+		NV_ERROR(dev, "Couldn't find matching output script table\n");
+		return 1;
+	}
+
+	if (pxclk < -2 || pxclk > 0) {
+		/* Try to find matching script table entry */
+		for (i = 0; i < otable[5]; i++) {
+			if (ROM16(otable[table[4] + i*6]) == sub)
+				break;
+		}
+
+		if (i == otable[5]) {
+			NV_ERROR(dev, "Table 0x%04x not found for %d/%d, "
+				      "using first\n",
+				 sub, dcbent->type, dcbent->or);
+			i = 0;
+		}
+	}
+
+	if (pxclk == 0) {
+		script = ROM16(otable[6]);
+		if (!script) {
+			NV_DEBUG_KMS(dev, "output script 0 not found\n");
+			return 1;
+		}
+
+		NV_TRACE(dev, "0x%04X: parsing output script 0\n", script);
+		nouveau_bios_run_init_table(dev, script, dcbent);
+	} else
+	if (pxclk == -1) {
+		script = ROM16(otable[8]);
+		if (!script) {
+			NV_DEBUG_KMS(dev, "output script 1 not found\n");
+			return 1;
+		}
+
+		NV_TRACE(dev, "0x%04X: parsing output script 1\n", script);
+		nouveau_bios_run_init_table(dev, script, dcbent);
+	} else
+	if (pxclk == -2) {
+		if (table[4] >= 12)
+			script = ROM16(otable[10]);
+		else
+			script = 0;
+		if (!script) {
+			NV_DEBUG_KMS(dev, "output script 2 not found\n");
+			return 1;
+		}
+
+		NV_TRACE(dev, "0x%04X: parsing output script 2\n", script);
+		nouveau_bios_run_init_table(dev, script, dcbent);
+	} else
+	if (pxclk > 0) {
+		script = ROM16(otable[table[4] + i*6 + 2]);
+		if (script)
+			script = clkcmptable(bios, script, pxclk);
+		if (!script) {
+			NV_ERROR(dev, "clock script 0 not found\n");
+			return 1;
+		}
+
+		NV_TRACE(dev, "0x%04X: parsing clock script 0\n", script);
+		nouveau_bios_run_init_table(dev, script, dcbent);
+	} else
+	if (pxclk < 0) {
+		script = ROM16(otable[table[4] + i*6 + 4]);
+		if (script)
+			script = clkcmptable(bios, script, -pxclk);
+		if (!script) {
+			NV_DEBUG_KMS(dev, "clock script 1 not found\n");
+			return 1;
+		}
+
+		NV_TRACE(dev, "0x%04X: parsing clock script 1\n", script);
+		nouveau_bios_run_init_table(dev, script, dcbent);
+	}
+
+	return 0;
+}
+
+
+int run_tmds_table(struct drm_device *dev, struct dcb_entry *dcbent, int head, int pxclk)
+{
+	/*
+	 * the pxclk parameter is in kHz
+	 *
+	 * This runs the TMDS regs setting code found on BIT bios cards
+	 *
+	 * For ffs(or) == 1 use the first table, for ffs(or) == 2 and
+	 * ffs(or) == 3, use the second.
+	 */
+
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nvbios *bios = &dev_priv->vbios;
+	int cv = bios->chip_version;
+	uint16_t clktable = 0, scriptptr;
+	uint32_t sel_clk_binding, sel_clk;
+
+	/* pre-nv17 off-chip tmds uses scripts, post nv17 doesn't */
+	if (cv >= 0x17 && cv != 0x1a && cv != 0x20 &&
+	    dcbent->location != DCB_LOC_ON_CHIP)
+		return 0;
+
+	switch (ffs(dcbent->or)) {
+	case 1:
+		clktable = bios->tmds.output0_script_ptr;
+		break;
+	case 2:
+	case 3:
+		clktable = bios->tmds.output1_script_ptr;
+		break;
+	}
+
+	if (!clktable) {
+		NV_ERROR(dev, "Pixel clock comparison table not found\n");
+		return -EINVAL;
+	}
+
+	scriptptr = clkcmptable(bios, clktable, pxclk);
+
+	if (!scriptptr) {
+		NV_ERROR(dev, "TMDS output init script not found\n");
+		return -ENOENT;
+	}
+
+	/* don't let script change pll->head binding */
+	sel_clk_binding = bios_rd32(bios, NV_PRAMDAC_SEL_CLK) & 0x50000;
+	run_digital_op_script(dev, scriptptr, dcbent, head, pxclk >= 165000);
+	sel_clk = NVReadRAMDAC(dev, 0, NV_PRAMDAC_SEL_CLK) & ~0x50000;
+	NVWriteRAMDAC(dev, 0, NV_PRAMDAC_SEL_CLK, sel_clk | sel_clk_binding);
+
+	return 0;
+}
+
+int get_pll_limits(struct drm_device *dev, uint32_t limit_match, struct pll_lims *pll_lim)
+{
+	/*
+	 * PLL limits table
+	 *
+	 * Version 0x10: NV30, NV31
+	 * One byte header (version), one record of 24 bytes
+	 * Version 0x11: NV36 - Not implemented
+	 * Seems to have same record style as 0x10, but 3 records rather than 1
+	 * Version 0x20: Found on Geforce 6 cards
+	 * Trivial 4 byte BIT header. 31 (0x1f) byte record length
+	 * Version 0x21: Found on Geforce 7, 8 and some Geforce 6 cards
+	 * 5 byte header, fifth byte of unknown purpose. 35 (0x23) byte record
+	 * length in general, some (integrated) have an extra configuration byte
+	 * Version 0x30: Found on Geforce 8, separates the register mapping
+	 * from the limits tables.
+	 */
+
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nvbios *bios = &dev_priv->vbios;
+	int cv = bios->chip_version, pllindex = 0;
+	uint8_t pll_lim_ver = 0, headerlen = 0, recordlen = 0, entries = 0;
+	uint32_t crystal_strap_mask, crystal_straps;
+
+	if (!bios->pll_limit_tbl_ptr) {
+		if (cv == 0x30 || cv == 0x31 || cv == 0x35 || cv == 0x36 ||
+		    cv >= 0x40) {
+			NV_ERROR(dev, "Pointer to PLL limits table invalid\n");
+			return -EINVAL;
+		}
+	} else
+		pll_lim_ver = bios->data[bios->pll_limit_tbl_ptr];
+
+	crystal_strap_mask = 1 << 6;
+	/* open coded dev->twoHeads test */
+	if (cv > 0x10 && cv != 0x15 && cv != 0x1a && cv != 0x20)
+		crystal_strap_mask |= 1 << 22;
+	crystal_straps = nvReadEXTDEV(dev, NV_PEXTDEV_BOOT_0) &
+							crystal_strap_mask;
+
+	switch (pll_lim_ver) {
+	/*
+	 * We use version 0 to indicate a pre limit table bios (single stage
+	 * pll) and load the hard coded limits instead.
+	 */
+	case 0:
+		break;
+	case 0x10:
+	case 0x11:
+		/*
+		 * Strictly v0x11 has 3 entries, but the last two don't seem
+		 * to get used.
+		 */
+		headerlen = 1;
+		recordlen = 0x18;
+		entries = 1;
+		pllindex = 0;
+		break;
+	case 0x20:
+	case 0x21:
+	case 0x30:
+	case 0x40:
+		headerlen = bios->data[bios->pll_limit_tbl_ptr + 1];
+		recordlen = bios->data[bios->pll_limit_tbl_ptr + 2];
+		entries = bios->data[bios->pll_limit_tbl_ptr + 3];
+		break;
+	default:
+		NV_ERROR(dev, "PLL limits table revision 0x%X not currently "
+				"supported\n", pll_lim_ver);
+		return -ENOSYS;
+	}
+
+	/* initialize all members to zero */
+	memset(pll_lim, 0, sizeof(struct pll_lims));
+
+	if (pll_lim_ver == 0x10 || pll_lim_ver == 0x11) {
+		uint8_t *pll_rec = &bios->data[bios->pll_limit_tbl_ptr + headerlen + recordlen * pllindex];
+
+		pll_lim->vco1.minfreq = ROM32(pll_rec[0]);
+		pll_lim->vco1.maxfreq = ROM32(pll_rec[4]);
+		pll_lim->vco2.minfreq = ROM32(pll_rec[8]);
+		pll_lim->vco2.maxfreq = ROM32(pll_rec[12]);
+		pll_lim->vco1.min_inputfreq = ROM32(pll_rec[16]);
+		pll_lim->vco2.min_inputfreq = ROM32(pll_rec[20]);
+		pll_lim->vco1.max_inputfreq = pll_lim->vco2.max_inputfreq = INT_MAX;
+
+		/* these values taken from nv30/31/36 */
+		pll_lim->vco1.min_n = 0x1;
+		if (cv == 0x36)
+			pll_lim->vco1.min_n = 0x5;
+		pll_lim->vco1.max_n = 0xff;
+		pll_lim->vco1.min_m = 0x1;
+		pll_lim->vco1.max_m = 0xd;
+		pll_lim->vco2.min_n = 0x4;
+		/*
+		 * On nv30, 31, 36 (i.e. all cards with two stage PLLs with this
+		 * table version (apart from nv35)), N2 is compared to
+		 * maxN2 (0x46) and 10 * maxM2 (0x4), so set maxN2 to 0x28 and
+		 * save a comparison
+		 */
+		pll_lim->vco2.max_n = 0x28;
+		if (cv == 0x30 || cv == 0x35)
+			/* only 5 bits available for N2 on nv30/35 */
+			pll_lim->vco2.max_n = 0x1f;
+		pll_lim->vco2.min_m = 0x1;
+		pll_lim->vco2.max_m = 0x4;
+		pll_lim->max_log2p = 0x7;
+		pll_lim->max_usable_log2p = 0x6;
+	} else if (pll_lim_ver == 0x20 || pll_lim_ver == 0x21) {
+		uint16_t plloffs = bios->pll_limit_tbl_ptr + headerlen;
+		uint32_t reg = 0; /* default match */
+		uint8_t *pll_rec;
+		int i;
+
+		/*
+		 * First entry is default match, if nothing better. warn if
+		 * reg field nonzero
+		 */
+		if (ROM32(bios->data[plloffs]))
+			NV_WARN(dev, "Default PLL limit entry has non-zero "
+				       "register field\n");
+
+		if (limit_match > MAX_PLL_TYPES)
+			/* we've been passed a reg as the match */
+			reg = limit_match;
+		else /* limit match is a pll type */
+			for (i = 1; i < entries && !reg; i++) {
+				uint32_t cmpreg = ROM32(bios->data[plloffs + recordlen * i]);
+
+				if (limit_match == NVPLL &&
+				    (cmpreg == NV_PRAMDAC_NVPLL_COEFF || cmpreg == 0x4000))
+					reg = cmpreg;
+				if (limit_match == MPLL &&
+				    (cmpreg == NV_PRAMDAC_MPLL_COEFF || cmpreg == 0x4020))
+					reg = cmpreg;
+				if (limit_match == VPLL1 &&
+				    (cmpreg == NV_PRAMDAC_VPLL_COEFF || cmpreg == 0x4010))
+					reg = cmpreg;
+				if (limit_match == VPLL2 &&
+				    (cmpreg == NV_RAMDAC_VPLL2 || cmpreg == 0x4018))
+					reg = cmpreg;
+			}
+
+		for (i = 1; i < entries; i++)
+			if (ROM32(bios->data[plloffs + recordlen * i]) == reg) {
+				pllindex = i;
+				break;
+			}
+
+		pll_rec = &bios->data[plloffs + recordlen * pllindex];
+
+		BIOSLOG(bios, "Loading PLL limits for reg 0x%08x\n",
+			pllindex ? reg : 0);
+
+		/*
+		 * Frequencies are stored in tables in MHz, kHz are more
+		 * useful, so we convert.
+		 */
+
+		/* What output frequencies can each VCO generate? */
+		pll_lim->vco1.minfreq = ROM16(pll_rec[4]) * 1000;
+		pll_lim->vco1.maxfreq = ROM16(pll_rec[6]) * 1000;
+		pll_lim->vco2.minfreq = ROM16(pll_rec[8]) * 1000;
+		pll_lim->vco2.maxfreq = ROM16(pll_rec[10]) * 1000;
+
+		/* What input frequencies they accept (past the m-divider)? */
+		pll_lim->vco1.min_inputfreq = ROM16(pll_rec[12]) * 1000;
+		pll_lim->vco2.min_inputfreq = ROM16(pll_rec[14]) * 1000;
+		pll_lim->vco1.max_inputfreq = ROM16(pll_rec[16]) * 1000;
+		pll_lim->vco2.max_inputfreq = ROM16(pll_rec[18]) * 1000;
+
+		/* What values are accepted as multiplier and divider? */
+		pll_lim->vco1.min_n = pll_rec[20];
+		pll_lim->vco1.max_n = pll_rec[21];
+		pll_lim->vco1.min_m = pll_rec[22];
+		pll_lim->vco1.max_m = pll_rec[23];
+		pll_lim->vco2.min_n = pll_rec[24];
+		pll_lim->vco2.max_n = pll_rec[25];
+		pll_lim->vco2.min_m = pll_rec[26];
+		pll_lim->vco2.max_m = pll_rec[27];
+
+		pll_lim->max_usable_log2p = pll_lim->max_log2p = pll_rec[29];
+		if (pll_lim->max_log2p > 0x7)
+			/* pll decoding in nv_hw.c assumes never > 7 */
+			NV_WARN(dev, "Max log2 P value greater than 7 (%d)\n",
+				pll_lim->max_log2p);
+		if (cv < 0x60)
+			pll_lim->max_usable_log2p = 0x6;
+		pll_lim->log2p_bias = pll_rec[30];
+
+		if (recordlen > 0x22)
+			pll_lim->refclk = ROM32(pll_rec[31]);
+
+		if (recordlen > 0x23 && pll_rec[35])
+			NV_WARN(dev,
+				"Bits set in PLL configuration byte (%x)\n",
+				pll_rec[35]);
+
+		/* C51 special not seen elsewhere */
+		if (cv == 0x51 && !pll_lim->refclk) {
+			uint32_t sel_clk = bios_rd32(bios, NV_PRAMDAC_SEL_CLK);
+
+			if (((limit_match == NV_PRAMDAC_VPLL_COEFF || limit_match == VPLL1) && sel_clk & 0x20) ||
+			    ((limit_match == NV_RAMDAC_VPLL2 || limit_match == VPLL2) && sel_clk & 0x80)) {
+				if (bios_idxprt_rd(bios, NV_CIO_CRX__COLOR, NV_CIO_CRE_CHIP_ID_INDEX) < 0xa3)
+					pll_lim->refclk = 200000;
+				else
+					pll_lim->refclk = 25000;
+			}
+		}
+	} else if (pll_lim_ver == 0x30) { /* ver 0x30 */
+		uint8_t *entry = &bios->data[bios->pll_limit_tbl_ptr + headerlen];
+		uint8_t *record = NULL;
+		int i;
+
+		BIOSLOG(bios, "Loading PLL limits for register 0x%08x\n",
+			limit_match);
+
+		for (i = 0; i < entries; i++, entry += recordlen) {
+			if (ROM32(entry[3]) == limit_match) {
+				record = &bios->data[ROM16(entry[1])];
+				break;
+			}
+		}
+
+		if (!record) {
+			NV_ERROR(dev, "Register 0x%08x not found in PLL "
+				 "limits table", limit_match);
+			return -ENOENT;
+		}
+
+		pll_lim->vco1.minfreq = ROM16(record[0]) * 1000;
+		pll_lim->vco1.maxfreq = ROM16(record[2]) * 1000;
+		pll_lim->vco2.minfreq = ROM16(record[4]) * 1000;
+		pll_lim->vco2.maxfreq = ROM16(record[6]) * 1000;
+		pll_lim->vco1.min_inputfreq = ROM16(record[8]) * 1000;
+		pll_lim->vco2.min_inputfreq = ROM16(record[10]) * 1000;
+		pll_lim->vco1.max_inputfreq = ROM16(record[12]) * 1000;
+		pll_lim->vco2.max_inputfreq = ROM16(record[14]) * 1000;
+		pll_lim->vco1.min_n = record[16];
+		pll_lim->vco1.max_n = record[17];
+		pll_lim->vco1.min_m = record[18];
+		pll_lim->vco1.max_m = record[19];
+		pll_lim->vco2.min_n = record[20];
+		pll_lim->vco2.max_n = record[21];
+		pll_lim->vco2.min_m = record[22];
+		pll_lim->vco2.max_m = record[23];
+		pll_lim->max_usable_log2p = pll_lim->max_log2p = record[25];
+		pll_lim->log2p_bias = record[27];
+		pll_lim->refclk = ROM32(record[28]);
+	} else if (pll_lim_ver) { /* ver 0x40 */
+		uint8_t *entry = &bios->data[bios->pll_limit_tbl_ptr + headerlen];
+		uint8_t *record = NULL;
+		int i;
+
+		BIOSLOG(bios, "Loading PLL limits for register 0x%08x\n",
+			limit_match);
+
+		for (i = 0; i < entries; i++, entry += recordlen) {
+			if (ROM32(entry[3]) == limit_match) {
+				record = &bios->data[ROM16(entry[1])];
+				break;
+			}
+		}
+
+		if (!record) {
+			NV_ERROR(dev, "Register 0x%08x not found in PLL "
+				 "limits table", limit_match);
+			return -ENOENT;
+		}
+
+		pll_lim->vco1.minfreq = ROM16(record[0]) * 1000;
+		pll_lim->vco1.maxfreq = ROM16(record[2]) * 1000;
+		pll_lim->vco1.min_inputfreq = ROM16(record[4]) * 1000;
+		pll_lim->vco1.max_inputfreq = ROM16(record[6]) * 1000;
+		pll_lim->vco1.min_m = record[8];
+		pll_lim->vco1.max_m = record[9];
+		pll_lim->vco1.min_n = record[10];
+		pll_lim->vco1.max_n = record[11];
+		pll_lim->min_p = record[12];
+		pll_lim->max_p = record[13];
+		/* where did this go to?? */
+		if (limit_match == 0x00614100 || limit_match == 0x00614900)
+			pll_lim->refclk = 27000;
+		else
+			pll_lim->refclk = 100000;
+	}
+
+	/*
+	 * By now any valid limit table ought to have set a max frequency for
+	 * vco1, so if it's zero it's either a pre limit table bios, or one
+	 * with an empty limit table (seen on nv18)
+	 */
+	if (!pll_lim->vco1.maxfreq) {
+		pll_lim->vco1.minfreq = bios->fminvco;
+		pll_lim->vco1.maxfreq = bios->fmaxvco;
+		pll_lim->vco1.min_inputfreq = 0;
+		pll_lim->vco1.max_inputfreq = INT_MAX;
+		pll_lim->vco1.min_n = 0x1;
+		pll_lim->vco1.max_n = 0xff;
+		pll_lim->vco1.min_m = 0x1;
+		if (crystal_straps == 0) {
+			/* nv05 does this, nv11 doesn't, nv10 unknown */
+			if (cv < 0x11)
+				pll_lim->vco1.min_m = 0x7;
+			pll_lim->vco1.max_m = 0xd;
+		} else {
+			if (cv < 0x11)
+				pll_lim->vco1.min_m = 0x8;
+			pll_lim->vco1.max_m = 0xe;
+		}
+		if (cv < 0x17 || cv == 0x1a || cv == 0x20)
+			pll_lim->max_log2p = 4;
+		else
+			pll_lim->max_log2p = 5;
+		pll_lim->max_usable_log2p = pll_lim->max_log2p;
+	}
+
+	if (!pll_lim->refclk)
+		switch (crystal_straps) {
+		case 0:
+			pll_lim->refclk = 13500;
+			break;
+		case (1 << 6):
+			pll_lim->refclk = 14318;
+			break;
+		case (1 << 22):
+			pll_lim->refclk = 27000;
+			break;
+		case (1 << 22 | 1 << 6):
+			pll_lim->refclk = 25000;
+			break;
+		}
+
+#if 0 /* for easy debugging */
+	ErrorF("pll.vco1.minfreq: %d\n", pll_lim->vco1.minfreq);
+	ErrorF("pll.vco1.maxfreq: %d\n", pll_lim->vco1.maxfreq);
+	ErrorF("pll.vco2.minfreq: %d\n", pll_lim->vco2.minfreq);
+	ErrorF("pll.vco2.maxfreq: %d\n", pll_lim->vco2.maxfreq);
+
+	ErrorF("pll.vco1.min_inputfreq: %d\n", pll_lim->vco1.min_inputfreq);
+	ErrorF("pll.vco1.max_inputfreq: %d\n", pll_lim->vco1.max_inputfreq);
+	ErrorF("pll.vco2.min_inputfreq: %d\n", pll_lim->vco2.min_inputfreq);
+	ErrorF("pll.vco2.max_inputfreq: %d\n", pll_lim->vco2.max_inputfreq);
+
+	ErrorF("pll.vco1.min_n: %d\n", pll_lim->vco1.min_n);
+	ErrorF("pll.vco1.max_n: %d\n", pll_lim->vco1.max_n);
+	ErrorF("pll.vco1.min_m: %d\n", pll_lim->vco1.min_m);
+	ErrorF("pll.vco1.max_m: %d\n", pll_lim->vco1.max_m);
+	ErrorF("pll.vco2.min_n: %d\n", pll_lim->vco2.min_n);
+	ErrorF("pll.vco2.max_n: %d\n", pll_lim->vco2.max_n);
+	ErrorF("pll.vco2.min_m: %d\n", pll_lim->vco2.min_m);
+	ErrorF("pll.vco2.max_m: %d\n", pll_lim->vco2.max_m);
+
+	ErrorF("pll.max_log2p: %d\n", pll_lim->max_log2p);
+	ErrorF("pll.log2p_bias: %d\n", pll_lim->log2p_bias);
+
+	ErrorF("pll.refclk: %d\n", pll_lim->refclk);
+#endif
+
+	return 0;
+}
+
+static void parse_bios_version(struct drm_device *dev, struct nvbios *bios, uint16_t offset)
+{
+	/*
+	 * offset + 0  (8 bits): Micro version
+	 * offset + 1  (8 bits): Minor version
+	 * offset + 2  (8 bits): Chip version
+	 * offset + 3  (8 bits): Major version
+	 */
+
+	bios->major_version = bios->data[offset + 3];
+	bios->chip_version = bios->data[offset + 2];
+	NV_TRACE(dev, "Bios version %02x.%02x.%02x.%02x\n",
+		 bios->data[offset + 3], bios->data[offset + 2],
+		 bios->data[offset + 1], bios->data[offset]);
+}
+
+static void parse_script_table_pointers(struct nvbios *bios, uint16_t offset)
+{
+	/*
+	 * Parses the init table segment for pointers used in script execution.
+	 *
+	 * offset + 0  (16 bits): init script tables pointer
+	 * offset + 2  (16 bits): macro index table pointer
+	 * offset + 4  (16 bits): macro table pointer
+	 * offset + 6  (16 bits): condition table pointer
+	 * offset + 8  (16 bits): io condition table pointer
+	 * offset + 10 (16 bits): io flag condition table pointer
+	 * offset + 12 (16 bits): init function table pointer
+	 */
+
+	bios->init_script_tbls_ptr = ROM16(bios->data[offset]);
+	bios->macro_index_tbl_ptr = ROM16(bios->data[offset + 2]);
+	bios->macro_tbl_ptr = ROM16(bios->data[offset + 4]);
+	bios->condition_tbl_ptr = ROM16(bios->data[offset + 6]);
+	bios->io_condition_tbl_ptr = ROM16(bios->data[offset + 8]);
+	bios->io_flag_condition_tbl_ptr = ROM16(bios->data[offset + 10]);
+	bios->init_function_tbl_ptr = ROM16(bios->data[offset + 12]);
+}
+
+static int parse_bit_A_tbl_entry(struct drm_device *dev, struct nvbios *bios, struct bit_entry *bitentry)
+{
+	/*
+	 * Parses the load detect values for g80 cards.
+	 *
+	 * offset + 0 (16 bits): loadval table pointer
+	 */
+
+	uint16_t load_table_ptr;
+	uint8_t version, headerlen, entrylen, num_entries;
+
+	if (bitentry->length != 3) {
+		NV_ERROR(dev, "Do not understand BIT A table\n");
+		return -EINVAL;
+	}
+
+	load_table_ptr = ROM16(bios->data[bitentry->offset]);
+
+	if (load_table_ptr == 0x0) {
+		NV_ERROR(dev, "Pointer to BIT loadval table invalid\n");
+		return -EINVAL;
+	}
+
+	version = bios->data[load_table_ptr];
+
+	if (version != 0x10) {
+		NV_ERROR(dev, "BIT loadval table version %d.%d not supported\n",
+			 version >> 4, version & 0xF);
+		return -ENOSYS;
+	}
+
+	headerlen = bios->data[load_table_ptr + 1];
+	entrylen = bios->data[load_table_ptr + 2];
+	num_entries = bios->data[load_table_ptr + 3];
+
+	if (headerlen != 4 || entrylen != 4 || num_entries != 2) {
+		NV_ERROR(dev, "Do not understand BIT loadval table\n");
+		return -EINVAL;
+	}
+
+	/* First entry is normal dac, 2nd tv-out perhaps? */
+	bios->dactestval = ROM32(bios->data[load_table_ptr + headerlen]) & 0x3ff;
+
+	return 0;
+}
+
+static int parse_bit_C_tbl_entry(struct drm_device *dev, struct nvbios *bios, struct bit_entry *bitentry)
+{
+	/*
+	 * offset + 8  (16 bits): PLL limits table pointer
+	 *
+	 * There's more in here, but that's unknown.
+	 */
+
+	if (bitentry->length < 10) {
+		NV_ERROR(dev, "Do not understand BIT C table\n");
+		return -EINVAL;
+	}
+
+	bios->pll_limit_tbl_ptr = ROM16(bios->data[bitentry->offset + 8]);
+
+	return 0;
+}
+
+static int parse_bit_display_tbl_entry(struct drm_device *dev, struct nvbios *bios, struct bit_entry *bitentry)
+{
+	/*
+	 * Parses the flat panel table segment that the bit entry points to.
+	 * Starting at bitentry->offset:
+	 *
+	 * offset + 0  (16 bits): ??? table pointer - seems to have 18 byte
+	 * records beginning with a freq.
+	 * offset + 2  (16 bits): mode table pointer
+	 */
+
+	if (bitentry->length != 4) {
+		NV_ERROR(dev, "Do not understand BIT display table\n");
+		return -EINVAL;
+	}
+
+	bios->fp.fptablepointer = ROM16(bios->data[bitentry->offset + 2]);
+
+	return 0;
+}
+
+static int parse_bit_init_tbl_entry(struct drm_device *dev, struct nvbios *bios, struct bit_entry *bitentry)
+{
+	/*
+	 * Parses the init table segment that the bit entry points to.
+	 *
+	 * See parse_script_table_pointers for layout
+	 */
+
+	if (bitentry->length < 14) {
+		NV_ERROR(dev, "Do not understand init table\n");
+		return -EINVAL;
+	}
+
+	parse_script_table_pointers(bios, bitentry->offset);
+
+	if (bitentry->length >= 16)
+		bios->some_script_ptr = ROM16(bios->data[bitentry->offset + 14]);
+	if (bitentry->length >= 18)
+		bios->init96_tbl_ptr = ROM16(bios->data[bitentry->offset + 16]);
+
+	return 0;
+}
+
+static int parse_bit_i_tbl_entry(struct drm_device *dev, struct nvbios *bios, struct bit_entry *bitentry)
+{
+	/*
+	 * BIT 'i' (info?) table
+	 *
+	 * offset + 0  (32 bits): BIOS version dword (as in B table)
+	 * offset + 5  (8  bits): BIOS feature byte (same as for BMP?)
+	 * offset + 13 (16 bits): pointer to table containing DAC load
+	 * detection comparison values
+	 *
+	 * There's other things in the table, purpose unknown
+	 */
+
+	uint16_t daccmpoffset;
+	uint8_t dacver, dacheaderlen;
+
+	if (bitentry->length < 6) {
+		NV_ERROR(dev, "BIT i table too short for needed information\n");
+		return -EINVAL;
+	}
+
+	parse_bios_version(dev, bios, bitentry->offset);
+
+	/*
+	 * bit 4 seems to indicate a mobile bios (doesn't suffer from BMP's
+	 * Quadro identity crisis), other bits possibly as for BMP feature byte
+	 */
+	bios->feature_byte = bios->data[bitentry->offset + 5];
+	bios->is_mobile = bios->feature_byte & FEATURE_MOBILE;
+
+	if (bitentry->length < 15) {
+		NV_WARN(dev, "BIT i table not long enough for DAC load "
+			       "detection comparison table\n");
+		return -EINVAL;
+	}
+
+	daccmpoffset = ROM16(bios->data[bitentry->offset + 13]);
+
+	/* doesn't exist on g80 */
+	if (!daccmpoffset)
+		return 0;
+
+	/*
+	 * The first value in the table, following the header, is the
+	 * comparison value, the second entry is a comparison value for
+	 * TV load detection.
+	 */
+
+	dacver = bios->data[daccmpoffset];
+	dacheaderlen = bios->data[daccmpoffset + 1];
+
+	if (dacver != 0x00 && dacver != 0x10) {
+		NV_WARN(dev, "DAC load detection comparison table version "
+			       "%d.%d not known\n", dacver >> 4, dacver & 0xf);
+		return -ENOSYS;
+	}
+
+	bios->dactestval = ROM32(bios->data[daccmpoffset + dacheaderlen]);
+	bios->tvdactestval = ROM32(bios->data[daccmpoffset + dacheaderlen + 4]);
+
+	return 0;
+}
+
+static int parse_bit_lvds_tbl_entry(struct drm_device *dev, struct nvbios *bios, struct bit_entry *bitentry)
+{
+	/*
+	 * Parses the LVDS table segment that the bit entry points to.
+	 * Starting at bitentry->offset:
+	 *
+	 * offset + 0  (16 bits): LVDS strap xlate table pointer
+	 */
+
+	if (bitentry->length != 2) {
+		NV_ERROR(dev, "Do not understand BIT LVDS table\n");
+		return -EINVAL;
+	}
+
+	/*
+	 * No idea if it's still called the LVDS manufacturer table, but
+	 * the concept's close enough.
+	 */
+	bios->fp.lvdsmanufacturerpointer = ROM16(bios->data[bitentry->offset]);
+
+	return 0;
+}
+
+static int
+parse_bit_M_tbl_entry(struct drm_device *dev, struct nvbios *bios,
+		      struct bit_entry *bitentry)
+{
+	/*
+	 * offset + 2  (8  bits): number of options in an
+	 * 	INIT_RAM_RESTRICT_ZM_REG_GROUP opcode option set
+	 * offset + 3  (16 bits): pointer to strap xlate table for RAM
+	 * 	restrict option selection
+	 *
+	 * There's a bunch of bits in this table other than the RAM restrict
+	 * stuff that we don't use - their use currently unknown
+	 */
+
+	/*
+	 * Older bios versions don't have a sufficiently long table for
+	 * what we want
+	 */
+	if (bitentry->length < 0x5)
+		return 0;
+
+	if (bitentry->id[1] < 2) {
+		bios->ram_restrict_group_count = bios->data[bitentry->offset + 2];
+		bios->ram_restrict_tbl_ptr = ROM16(bios->data[bitentry->offset + 3]);
+	} else {
+		bios->ram_restrict_group_count = bios->data[bitentry->offset + 0];
+		bios->ram_restrict_tbl_ptr = ROM16(bios->data[bitentry->offset + 1]);
+	}
+
+	return 0;
+}
+
+static int parse_bit_tmds_tbl_entry(struct drm_device *dev, struct nvbios *bios, struct bit_entry *bitentry)
+{
+	/*
+	 * Parses the pointer to the TMDS table
+	 *
+	 * Starting at bitentry->offset:
+	 *
+	 * offset + 0  (16 bits): TMDS table pointer
+	 *
+	 * The TMDS table is typically found just before the DCB table, with a
+	 * characteristic signature of 0x11,0x13 (1.1 being version, 0x13 being
+	 * length?)
+	 *
+	 * At offset +7 is a pointer to a script, which I don't know how to
+	 * run yet.
+	 * At offset +9 is a pointer to another script, likewise
+	 * Offset +11 has a pointer to a table where the first word is a pxclk
+	 * frequency and the second word a pointer to a script, which should be
+	 * run if the comparison pxclk frequency is less than the pxclk desired.
+	 * This repeats for decreasing comparison frequencies
+	 * Offset +13 has a pointer to a similar table
+	 * The selection of table (and possibly +7/+9 script) is dictated by
+	 * "or" from the DCB.
+	 */
+
+	uint16_t tmdstableptr, script1, script2;
+
+	if (bitentry->length != 2) {
+		NV_ERROR(dev, "Do not understand BIT TMDS table\n");
+		return -EINVAL;
+	}
+
+	tmdstableptr = ROM16(bios->data[bitentry->offset]);
+
+	if (tmdstableptr == 0x0) {
+		NV_ERROR(dev, "Pointer to TMDS table invalid\n");
+		return -EINVAL;
+	}
+
+	/* nv50+ has v2.0, but we don't parse it atm */
+	if (bios->data[tmdstableptr] != 0x11) {
+		NV_WARN(dev,
+			"TMDS table revision %d.%d not currently supported\n",
+			bios->data[tmdstableptr] >> 4, bios->data[tmdstableptr] & 0xf);
+		return -ENOSYS;
+	}
+
+	/*
+	 * These two scripts are odd: they don't seem to get run even when
+	 * they are not stubbed.
+	 */
+	script1 = ROM16(bios->data[tmdstableptr + 7]);
+	script2 = ROM16(bios->data[tmdstableptr + 9]);
+	if (bios->data[script1] != 'q' || bios->data[script2] != 'q')
+		NV_WARN(dev, "TMDS table script pointers not stubbed\n");
+
+	bios->tmds.output0_script_ptr = ROM16(bios->data[tmdstableptr + 11]);
+	bios->tmds.output1_script_ptr = ROM16(bios->data[tmdstableptr + 13]);
+
+	return 0;
+}
+
+static int
+parse_bit_U_tbl_entry(struct drm_device *dev, struct nvbios *bios,
+		      struct bit_entry *bitentry)
+{
+	/*
+	 * Parses the pointer to the G80 output script tables
+	 *
+	 * Starting at bitentry->offset:
+	 *
+	 * offset + 0  (16 bits): output script table pointer
+	 */
+
+	uint16_t outputscripttableptr;
+
+	if (bitentry->length != 3) {
+		NV_ERROR(dev, "Do not understand BIT U table\n");
+		return -EINVAL;
+	}
+
+	outputscripttableptr = ROM16(bios->data[bitentry->offset]);
+	bios->display.script_table_ptr = outputscripttableptr;
+	return 0;
+}
+
+static int
+parse_bit_displayport_tbl_entry(struct drm_device *dev, struct nvbios *bios,
+				struct bit_entry *bitentry)
+{
+	bios->display.dp_table_ptr = ROM16(bios->data[bitentry->offset]);
+	return 0;
+}
+
+struct bit_table {
+	const char id;
+	int (* const parse_fn)(struct drm_device *, struct nvbios *, struct bit_entry *);
+};
+
+#define BIT_TABLE(id, funcid) ((struct bit_table){ id, parse_bit_##funcid##_tbl_entry })
+
+static int
+parse_bit_table(struct nvbios *bios, const uint16_t bitoffset,
+		struct bit_table *table)
+{
+	struct drm_device *dev = bios->dev;
+	uint8_t maxentries = bios->data[bitoffset + 4];
+	int i, offset;
+	struct bit_entry bitentry;
+
+	for (i = 0, offset = bitoffset + 6; i < maxentries; i++, offset += 6) {
+		bitentry.id[0] = bios->data[offset];
+
+		if (bitentry.id[0] != table->id)
+			continue;
+
+		bitentry.id[1] = bios->data[offset + 1];
+		bitentry.length = ROM16(bios->data[offset + 2]);
+		bitentry.offset = ROM16(bios->data[offset + 4]);
+
+		return table->parse_fn(dev, bios, &bitentry);
+	}
+
+	NV_INFO(dev, "BIT table '%c' not found\n", table->id);
+	return -ENOSYS;
+}
+
+static int
+parse_bit_structure(struct nvbios *bios, const uint16_t bitoffset)
+{
+	int ret;
+
+	/*
+	 * The only restriction on parsing order currently is having 'i' first
+	 * for use of bios->*_version or bios->feature_byte while parsing;
+	 * functions shouldn't be actually *doing* anything apart from pulling
+	 * data from the image into the bios struct, thus no interdependencies
+	 */
+	ret = parse_bit_table(bios, bitoffset, &BIT_TABLE('i', i));
+	if (ret) /* info? */
+		return ret;
+	if (bios->major_version >= 0x60) /* g80+ */
+		parse_bit_table(bios, bitoffset, &BIT_TABLE('A', A));
+	ret = parse_bit_table(bios, bitoffset, &BIT_TABLE('C', C));
+	if (ret)
+		return ret;
+	parse_bit_table(bios, bitoffset, &BIT_TABLE('D', display));
+	ret = parse_bit_table(bios, bitoffset, &BIT_TABLE('I', init));
+	if (ret)
+		return ret;
+	parse_bit_table(bios, bitoffset, &BIT_TABLE('M', M)); /* memory? */
+	parse_bit_table(bios, bitoffset, &BIT_TABLE('L', lvds));
+	parse_bit_table(bios, bitoffset, &BIT_TABLE('T', tmds));
+	parse_bit_table(bios, bitoffset, &BIT_TABLE('U', U));
+	parse_bit_table(bios, bitoffset, &BIT_TABLE('d', displayport));
+
+	return 0;
+}
+
+static int parse_bmp_structure(struct drm_device *dev, struct nvbios *bios, unsigned int offset)
+{
+	/*
+	 * Parses the BMP structure for useful things, but does not act on them
+	 *
+	 * offset +   5: BMP major version
+	 * offset +   6: BMP minor version
+	 * offset +   9: BMP feature byte
+	 * offset +  10: BCD encoded BIOS version
+	 *
+	 * offset +  18: init script table pointer (for bios versions < 5.10h)
+	 * offset +  20: extra init script table pointer (for bios
+	 * versions < 5.10h)
+	 *
+	 * offset +  24: memory init table pointer (used on early bios versions)
+	 * offset +  26: SDR memory sequencing setup data table
+	 * offset +  28: DDR memory sequencing setup data table
+	 *
+	 * offset +  54: index of I2C CRTC pair to use for CRT output
+	 * offset +  55: index of I2C CRTC pair to use for TV output
+	 * offset +  56: index of I2C CRTC pair to use for flat panel output
+	 * offset +  58: write CRTC index for I2C pair 0
+	 * offset +  59: read CRTC index for I2C pair 0
+	 * offset +  60: write CRTC index for I2C pair 1
+	 * offset +  61: read CRTC index for I2C pair 1
+	 *
+	 * offset +  67: maximum internal PLL frequency (single stage PLL)
+	 * offset +  71: minimum internal PLL frequency (single stage PLL)
+	 *
+	 * offset +  75: script table pointers, as described in
+	 * parse_script_table_pointers
+	 *
+	 * offset +  89: TMDS single link output A table pointer
+	 * offset +  91: TMDS single link output B table pointer
+	 * offset +  95: LVDS single link output A table pointer
+	 * offset + 105: flat panel timings table pointer
+	 * offset + 107: flat panel strapping translation table pointer
+	 * offset + 117: LVDS manufacturer panel config table pointer
+	 * offset + 119: LVDS manufacturer strapping translation table pointer
+	 *
+	 * offset + 142: PLL limits table pointer
+	 *
+	 * offset + 156: minimum pixel clock for LVDS dual link
+	 */
+
+	uint8_t *bmp = &bios->data[offset], bmp_version_major, bmp_version_minor;
+	uint16_t bmplength;
+	uint16_t legacy_scripts_offset, legacy_i2c_offset;
+
+	/* load needed defaults in case we can't parse this info */
+	bios->dcb.i2c[0].write = NV_CIO_CRE_DDC_WR__INDEX;
+	bios->dcb.i2c[0].read = NV_CIO_CRE_DDC_STATUS__INDEX;
+	bios->dcb.i2c[1].write = NV_CIO_CRE_DDC0_WR__INDEX;
+	bios->dcb.i2c[1].read = NV_CIO_CRE_DDC0_STATUS__INDEX;
+	bios->digital_min_front_porch = 0x4b;
+	bios->fmaxvco = 256000;
+	bios->fminvco = 128000;
+	bios->fp.duallink_transition_clk = 90000;
+
+	bmp_version_major = bmp[5];
+	bmp_version_minor = bmp[6];
+
+	NV_TRACE(dev, "BMP version %d.%d\n",
+		 bmp_version_major, bmp_version_minor);
+
+	/*
+	 * Make sure that 0x36 is blank and can't be mistaken for a DCB
+	 * pointer on early versions
+	 */
+	if (bmp_version_major < 5)
+		*(uint16_t *)&bios->data[0x36] = 0;
+
+	/*
+	 * Seems that the minor version was 1 for all major versions prior
+	 * to 5. Version 6 could theoretically exist, but I suspect BIT
+	 * happened instead.
+	 */
+	if ((bmp_version_major < 5 && bmp_version_minor != 1) || bmp_version_major > 5) {
+		NV_ERROR(dev, "You have an unsupported BMP version. "
+				"Please send in your bios\n");
+		return -ENOSYS;
+	}
+
+	if (bmp_version_major == 0)
+		/* nothing that's currently useful in this version */
+		return 0;
+	else if (bmp_version_major == 1)
+		bmplength = 44; /* exact for 1.01 */
+	else if (bmp_version_major == 2)
+		bmplength = 48; /* exact for 2.01 */
+	else if (bmp_version_major == 3)
+		bmplength = 54;
+		/* guessed - mem init tables added in this version */
+	else if (bmp_version_major == 4 || bmp_version_minor < 0x1)
+		/* don't know if 5.0 exists... */
+		bmplength = 62;
+		/* guessed - BMP I2C indices added in version 4*/
+	else if (bmp_version_minor < 0x6)
+		bmplength = 67; /* exact for 5.01 */
+	else if (bmp_version_minor < 0x10)
+		bmplength = 75; /* exact for 5.06 */
+	else if (bmp_version_minor == 0x10)
+		bmplength = 89; /* exact for 5.10h */
+	else if (bmp_version_minor < 0x14)
+		bmplength = 118; /* exact for 5.11h */
+	else if (bmp_version_minor < 0x24)
+		/*
+		 * Not sure of version where pll limits came in;
+		 * certainly exist by 0x24 though.
+		 */
+		/* length not exact: this is long enough to get lvds members */
+		bmplength = 123;
+	else if (bmp_version_minor < 0x27)
+		/*
+		 * Length not exact: this is long enough to get pll limit
+		 * member
+		 */
+		bmplength = 144;
+	else
+		/*
+		 * Length not exact: this is long enough to get dual link
+		 * transition clock.
+		 */
+		bmplength = 158;
+
+	/* checksum */
+	if (nv_cksum(bmp, 8)) {
+		NV_ERROR(dev, "Bad BMP checksum\n");
+		return -EINVAL;
+	}
+
+	/*
+	 * Bit 4 seems to indicate either a mobile bios or a quadro card --
+	 * mobile behaviour consistent (nv11+), quadro only seen nv18gl-nv36gl
+	 * (not nv10gl), bit 5 that the flat panel tables are present, and
+	 * bit 6 a tv bios.
+	 */
+	bios->feature_byte = bmp[9];
+
+	parse_bios_version(dev, bios, offset + 10);
+
+	if (bmp_version_major < 5 || bmp_version_minor < 0x10)
+		bios->old_style_init = true;
+	legacy_scripts_offset = 18;
+	if (bmp_version_major < 2)
+		legacy_scripts_offset -= 4;
+	bios->init_script_tbls_ptr = ROM16(bmp[legacy_scripts_offset]);
+	bios->extra_init_script_tbl_ptr = ROM16(bmp[legacy_scripts_offset + 2]);
+
+	if (bmp_version_major > 2) {	/* appears in BMP 3 */
+		bios->legacy.mem_init_tbl_ptr = ROM16(bmp[24]);
+		bios->legacy.sdr_seq_tbl_ptr = ROM16(bmp[26]);
+		bios->legacy.ddr_seq_tbl_ptr = ROM16(bmp[28]);
+	}
+
+	legacy_i2c_offset = 0x48;	/* BMP version 2 & 3 */
+	if (bmplength > 61)
+		legacy_i2c_offset = offset + 54;
+	bios->legacy.i2c_indices.crt = bios->data[legacy_i2c_offset];
+	bios->legacy.i2c_indices.tv = bios->data[legacy_i2c_offset + 1];
+	bios->legacy.i2c_indices.panel = bios->data[legacy_i2c_offset + 2];
+	bios->dcb.i2c[0].write = bios->data[legacy_i2c_offset + 4];
+	bios->dcb.i2c[0].read = bios->data[legacy_i2c_offset + 5];
+	bios->dcb.i2c[1].write = bios->data[legacy_i2c_offset + 6];
+	bios->dcb.i2c[1].read = bios->data[legacy_i2c_offset + 7];
+
+	if (bmplength > 74) {
+		bios->fmaxvco = ROM32(bmp[67]);
+		bios->fminvco = ROM32(bmp[71]);
+	}
+	if (bmplength > 88)
+		parse_script_table_pointers(bios, offset + 75);
+	if (bmplength > 94) {
+		bios->tmds.output0_script_ptr = ROM16(bmp[89]);
+		bios->tmds.output1_script_ptr = ROM16(bmp[91]);
+		/*
+		 * Never observed in use with lvds scripts, but is reused for
+		 * 18/24 bit panel interface default for EDID equipped panels
+		 * (if_is_24bit not set directly to avoid any oscillation).
+		 */
+		bios->legacy.lvds_single_a_script_ptr = ROM16(bmp[95]);
+	}
+	if (bmplength > 108) {
+		bios->fp.fptablepointer = ROM16(bmp[105]);
+		bios->fp.fpxlatetableptr = ROM16(bmp[107]);
+		bios->fp.xlatwidth = 1;
+	}
+	if (bmplength > 120) {
+		bios->fp.lvdsmanufacturerpointer = ROM16(bmp[117]);
+		bios->fp.fpxlatemanufacturertableptr = ROM16(bmp[119]);
+	}
+	if (bmplength > 143)
+		bios->pll_limit_tbl_ptr = ROM16(bmp[142]);
+
+	if (bmplength > 157)
+		bios->fp.duallink_transition_clk = ROM16(bmp[156]) * 10;
+
+	return 0;
+}
+
+static uint16_t findstr(uint8_t *data, int n, const uint8_t *str, int len)
+{
+	int i, j;
+
+	for (i = 0; i <= (n - len); i++) {
+		for (j = 0; j < len; j++)
+			if (data[i + j] != str[j])
+				break;
+		if (j == len)
+			return i;
+	}
+
+	return 0;
+}
+
+static int
+read_dcb_i2c_entry(struct drm_device *dev, int dcb_version, uint8_t *i2ctable, int index, struct dcb_i2c_entry *i2c)
+{
+	uint8_t dcb_i2c_ver = dcb_version, headerlen = 0, entry_len = 4;
+	int i2c_entries = DCB_MAX_NUM_I2C_ENTRIES;
+	int recordoffset = 0, rdofs = 1, wrofs = 0;
+	uint8_t port_type = 0;
+
+	if (!i2ctable)
+		return -EINVAL;
+
+	if (dcb_version >= 0x30) {
+		if (i2ctable[0] != dcb_version) /* necessary? */
+			NV_WARN(dev,
+				"DCB I2C table version mismatch (%02X vs %02X)\n",
+				i2ctable[0], dcb_version);
+		dcb_i2c_ver = i2ctable[0];
+		headerlen = i2ctable[1];
+		if (i2ctable[2] <= DCB_MAX_NUM_I2C_ENTRIES)
+			i2c_entries = i2ctable[2];
+		else
+			NV_WARN(dev,
+				"DCB I2C table has more entries than indexable "
+				"(%d entries, max %d)\n", i2ctable[2],
+				DCB_MAX_NUM_I2C_ENTRIES);
+		entry_len = i2ctable[3];
+		/* [4] is i2c_default_indices, read in parse_dcb_table() */
+	}
+	/*
+	 * It's your own fault if you call this function on a DCB 1.1 BIOS --
+	 * the test below is for DCB 1.2
+	 */
+	if (dcb_version < 0x14) {
+		recordoffset = 2;
+		rdofs = 0;
+		wrofs = 1;
+	}
+
+	if (index == 0xf)
+		return 0;
+	if (index >= i2c_entries) {
+		NV_ERROR(dev, "DCB I2C index too big (%d >= %d)\n",
+			 index, i2ctable[2]);
+		return -ENOENT;
+	}
+	if (i2ctable[headerlen + entry_len * index + 3] == 0xff) {
+		NV_ERROR(dev, "DCB I2C entry invalid\n");
+		return -EINVAL;
+	}
+
+	if (dcb_i2c_ver >= 0x30) {
+		port_type = i2ctable[headerlen + recordoffset + 3 + entry_len * index];
+
+		/*
+		 * Fixup for chips using same address offset for read and
+		 * write.
+		 */
+		if (port_type == 4)	/* seen on C51 */
+			rdofs = wrofs = 1;
+		if (port_type >= 5)	/* G80+ */
+			rdofs = wrofs = 0;
+	}
+
+	if (dcb_i2c_ver >= 0x40 && port_type != 5 && port_type != 6)
+		NV_WARN(dev, "DCB I2C table has port type %d\n", port_type);
+
+	i2c->port_type = port_type;
+	i2c->read = i2ctable[headerlen + recordoffset + rdofs + entry_len * index];
+	i2c->write = i2ctable[headerlen + recordoffset + wrofs + entry_len * index];
+
+	return 0;
+}
+
+static struct dcb_gpio_entry *
+new_gpio_entry(struct nvbios *bios)
+{
+	struct dcb_gpio_table *gpio = &bios->dcb.gpio;
+
+	return &gpio->entry[gpio->entries++];
+}
+
+struct dcb_gpio_entry *
+nouveau_bios_gpio_entry(struct drm_device *dev, enum dcb_gpio_tag tag)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nvbios *bios = &dev_priv->vbios;
+	int i;
+
+	for (i = 0; i < bios->dcb.gpio.entries; i++) {
+		if (bios->dcb.gpio.entry[i].tag != tag)
+			continue;
+
+		return &bios->dcb.gpio.entry[i];
+	}
+
+	return NULL;
+}
+
+static void
+parse_dcb30_gpio_entry(struct nvbios *bios, uint16_t offset)
+{
+	struct dcb_gpio_entry *gpio;
+	uint16_t ent = ROM16(bios->data[offset]);
+	uint8_t line = ent & 0x1f,
+		tag = ent >> 5 & 0x3f,
+		flags = ent >> 11 & 0x1f;
+
+	if (tag == 0x3f)
+		return;
+
+	gpio = new_gpio_entry(bios);
+
+	gpio->tag = tag;
+	gpio->line = line;
+	gpio->invert = flags != 4;
+}
+
+static void
+parse_dcb40_gpio_entry(struct nvbios *bios, uint16_t offset)
+{
+	struct dcb_gpio_entry *gpio;
+	uint32_t ent = ROM32(bios->data[offset]);
+	uint8_t line = ent & 0x1f,
+		tag = ent >> 8 & 0xff;
+
+	if (tag == 0xff)
+		return;
+
+	gpio = new_gpio_entry(bios);
+
+	/* Currently unused, we may need more fields parsed at some
+	 * point. */
+	gpio->tag = tag;
+	gpio->line = line;
+}
+
+static void
+parse_dcb_gpio_table(struct nvbios *bios)
+{
+	struct drm_device *dev = bios->dev;
+	uint16_t gpio_table_ptr = bios->dcb.gpio_table_ptr;
+	uint8_t *gpio_table = &bios->data[gpio_table_ptr];
+	int header_len = gpio_table[1],
+	    entries = gpio_table[2],
+	    entry_len = gpio_table[3];
+	void (*parse_entry)(struct nvbios *, uint16_t) = NULL;
+	int i;
+
+	if (bios->dcb.version >= 0x40) {
+		if (gpio_table_ptr && entry_len != 4) {
+			NV_WARN(dev, "Invalid DCB GPIO table entry length.\n");
+			return;
+		}
+
+		parse_entry = parse_dcb40_gpio_entry;
+
+	} else if (bios->dcb.version >= 0x30) {
+		if (gpio_table_ptr && entry_len != 2) {
+			NV_WARN(dev, "Invalid DCB GPIO table entry length.\n");
+			return;
+		}
+
+		parse_entry = parse_dcb30_gpio_entry;
+
+	} else if (bios->dcb.version >= 0x22) {
+		/*
+		 * DCBs older than v3.0 don't really have a GPIO
+		 * table, instead they keep some GPIO info at fixed
+		 * locations.
+		 */
+		uint16_t dcbptr = ROM16(bios->data[0x36]);
+		uint8_t *tvdac_gpio = &bios->data[dcbptr - 5];
+
+		if (tvdac_gpio[0] & 1) {
+			struct dcb_gpio_entry *gpio = new_gpio_entry(bios);
+
+			gpio->tag = DCB_GPIO_TVDAC0;
+			gpio->line = tvdac_gpio[1] >> 4;
+			gpio->invert = tvdac_gpio[0] & 2;
+		}
+	}
+
+	if (!gpio_table_ptr)
+		return;
+
+	if (entries > DCB_MAX_NUM_GPIO_ENTRIES) {
+		NV_WARN(dev, "Too many entries in the DCB GPIO table.\n");
+		entries = DCB_MAX_NUM_GPIO_ENTRIES;
+	}
+
+	for (i = 0; i < entries; i++)
+		parse_entry(bios, gpio_table_ptr + header_len + entry_len * i);
+}
+
+struct dcb_connector_table_entry *
+nouveau_bios_connector_entry(struct drm_device *dev, int index)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nvbios *bios = &dev_priv->vbios;
+	struct dcb_connector_table_entry *cte;
+
+	if (index >= bios->dcb.connector.entries)
+		return NULL;
+
+	cte = &bios->dcb.connector.entry[index];
+	if (cte->type == 0xff)
+		return NULL;
+
+	return cte;
+}
+
+static enum dcb_connector_type
+divine_connector_type(struct nvbios *bios, int index)
+{
+	struct dcb_table *dcb = &bios->dcb;
+	unsigned encoders = 0, type = DCB_CONNECTOR_NONE;
+	int i;
+
+	for (i = 0; i < dcb->entries; i++) {
+		if (dcb->entry[i].connector == index)
+			encoders |= (1 << dcb->entry[i].type);
+	}
+
+	if (encoders & (1 << OUTPUT_DP)) {
+		if (encoders & (1 << OUTPUT_TMDS))
+			type = DCB_CONNECTOR_DP;
+		else
+			type = DCB_CONNECTOR_eDP;
+	} else
+	if (encoders & (1 << OUTPUT_TMDS)) {
+		if (encoders & (1 << OUTPUT_ANALOG))
+			type = DCB_CONNECTOR_DVI_I;
+		else
+			type = DCB_CONNECTOR_DVI_D;
+	} else
+	if (encoders & (1 << OUTPUT_ANALOG)) {
+		type = DCB_CONNECTOR_VGA;
+	} else
+	if (encoders & (1 << OUTPUT_LVDS)) {
+		type = DCB_CONNECTOR_LVDS;
+	} else
+	if (encoders & (1 << OUTPUT_TV)) {
+		type = DCB_CONNECTOR_TV_0;
+	}
+
+	return type;
+}
+
+static void
+parse_dcb_connector_table(struct nvbios *bios)
+{
+	struct drm_device *dev = bios->dev;
+	struct dcb_connector_table *ct = &bios->dcb.connector;
+	struct dcb_connector_table_entry *cte;
+	uint8_t *conntab = &bios->data[bios->dcb.connector_table_ptr];
+	uint8_t *entry;
+	int i;
+
+	if (!bios->dcb.connector_table_ptr) {
+		NV_DEBUG_KMS(dev, "No DCB connector table present\n");
+		return;
+	}
+
+	NV_INFO(dev, "DCB connector table: VHER 0x%02x %d %d %d\n",
+		conntab[0], conntab[1], conntab[2], conntab[3]);
+	if ((conntab[0] != 0x30 && conntab[0] != 0x40) ||
+	    (conntab[3] != 2 && conntab[3] != 4)) {
+		NV_ERROR(dev, "  Unknown!  Please report.\n");
+		return;
+	}
+
+	ct->entries = conntab[2];
+
+	entry = conntab + conntab[1];
+	cte = &ct->entry[0];
+	for (i = 0; i < conntab[2]; i++, entry += conntab[3], cte++) {
+		if (conntab[3] == 2)
+			cte->entry = ROM16(entry[0]);
+		else
+			cte->entry = ROM32(entry[0]);
+
+		cte->type  = (cte->entry & 0x000000ff) >> 0;
+		cte->index = (cte->entry & 0x00000f00) >> 8;
+		switch (cte->entry & 0x00033000) {
+		case 0x00001000:
+			cte->gpio_tag = 0x07;
+			break;
+		case 0x00002000:
+			cte->gpio_tag = 0x08;
+			break;
+		case 0x00010000:
+			cte->gpio_tag = 0x51;
+			break;
+		case 0x00020000:
+			cte->gpio_tag = 0x52;
+			break;
+		default:
+			cte->gpio_tag = 0xff;
+			break;
+		}
+
+		if (cte->type == 0xff)
+			continue;
+
+		NV_INFO(dev, "  %d: 0x%08x: type 0x%02x idx %d tag 0x%02x\n",
+			i, cte->entry, cte->type, cte->index, cte->gpio_tag);
+
+		/* check for known types, fallback to guessing the type
+		 * from attached encoders if we hit an unknown.
+		 */
+		switch (cte->type) {
+		case DCB_CONNECTOR_VGA:
+		case DCB_CONNECTOR_TV_0:
+		case DCB_CONNECTOR_TV_1:
+		case DCB_CONNECTOR_TV_3:
+		case DCB_CONNECTOR_DVI_I:
+		case DCB_CONNECTOR_DVI_D:
+		case DCB_CONNECTOR_LVDS:
+		case DCB_CONNECTOR_DP:
+		case DCB_CONNECTOR_eDP:
+		case DCB_CONNECTOR_HDMI_0:
+		case DCB_CONNECTOR_HDMI_1:
+			break;
+		default:
+			cte->type = divine_connector_type(bios, cte->index);
+			NV_WARN(dev, "unknown type, using 0x%02x", cte->type);
+			break;
+		}
+
+	}
+}
+
+static struct dcb_entry *new_dcb_entry(struct dcb_table *dcb)
+{
+	struct dcb_entry *entry = &dcb->entry[dcb->entries];
+
+	memset(entry, 0, sizeof(struct dcb_entry));
+	entry->index = dcb->entries++;
+
+	return entry;
+}
+
+static void fabricate_vga_output(struct dcb_table *dcb, int i2c, int heads)
+{
+	struct dcb_entry *entry = new_dcb_entry(dcb);
+
+	entry->type = 0;
+	entry->i2c_index = i2c;
+	entry->heads = heads;
+	entry->location = DCB_LOC_ON_CHIP;
+	/* "or" mostly unused in early gen crt modesetting, 0 is fine */
+}
+
+static void fabricate_dvi_i_output(struct dcb_table *dcb, bool twoHeads)
+{
+	struct dcb_entry *entry = new_dcb_entry(dcb);
+
+	entry->type = 2;
+	entry->i2c_index = LEGACY_I2C_PANEL;
+	entry->heads = twoHeads ? 3 : 1;
+	entry->location = !DCB_LOC_ON_CHIP;	/* ie OFF CHIP */
+	entry->or = 1;	/* means |0x10 gets set on CRE_LCD__INDEX */
+	entry->duallink_possible = false; /* SiI164 and co. are single link */
+
+#if 0
+	/*
+	 * For dvi-a either crtc probably works, but my card appears to only
+	 * support dvi-d.  "nvidia" still attempts to program it for dvi-a,
+	 * doing the full fp output setup (program 0x6808.. fp dimension regs,
+	 * setting 0x680848 to 0x10000111 to enable, maybe setting 0x680880);
+	 * the monitor picks up the mode res ok and lights up, but no pixel
+	 * data appears, so the board manufacturer probably connected up the
+	 * sync lines, but missed the video traces / components
+	 *
+	 * with this introduction, dvi-a left as an exercise for the reader.
+	 */
+	fabricate_vga_output(dcb, LEGACY_I2C_PANEL, entry->heads);
+#endif
+}
+
+static void fabricate_tv_output(struct dcb_table *dcb, bool twoHeads)
+{
+	struct dcb_entry *entry = new_dcb_entry(dcb);
+
+	entry->type = 1;
+	entry->i2c_index = LEGACY_I2C_TV;
+	entry->heads = twoHeads ? 3 : 1;
+	entry->location = !DCB_LOC_ON_CHIP;	/* ie OFF CHIP */
+}
+
+static bool
+parse_dcb20_entry(struct drm_device *dev, struct dcb_table *dcb,
+		  uint32_t conn, uint32_t conf, struct dcb_entry *entry)
+{
+	entry->type = conn & 0xf;
+	entry->i2c_index = (conn >> 4) & 0xf;
+	entry->heads = (conn >> 8) & 0xf;
+	if (dcb->version >= 0x40)
+		entry->connector = (conn >> 12) & 0xf;
+	entry->bus = (conn >> 16) & 0xf;
+	entry->location = (conn >> 20) & 0x3;
+	entry->or = (conn >> 24) & 0xf;
+	/*
+	 * Normal entries consist of a single bit, but dual link has the
+	 * next most significant bit set too
+	 */
+	entry->duallink_possible =
+			((1 << (ffs(entry->or) - 1)) * 3 == entry->or);
+
+	switch (entry->type) {
+	case OUTPUT_ANALOG:
+		/*
+		 * Although the rest of a CRT conf dword is usually
+		 * zeros, mac biosen have stuff there so we must mask
+		 */
+		entry->crtconf.maxfreq = (dcb->version < 0x30) ?
+					 (conf & 0xffff) * 10 :
+					 (conf & 0xff) * 10000;
+		break;
+	case OUTPUT_LVDS:
+		{
+		uint32_t mask;
+		if (conf & 0x1)
+			entry->lvdsconf.use_straps_for_mode = true;
+		if (dcb->version < 0x22) {
+			mask = ~0xd;
+			/*
+			 * The laptop in bug 14567 lies and claims to not use
+			 * straps when it does, so assume all DCB 2.0 laptops
+			 * use straps, until a broken EDID using one is produced
+			 */
+			entry->lvdsconf.use_straps_for_mode = true;
+			/*
+			 * Both 0x4 and 0x8 show up in v2.0 tables; assume they
+			 * mean the same thing (probably wrong, but might work)
+			 */
+			if (conf & 0x4 || conf & 0x8)
+				entry->lvdsconf.use_power_scripts = true;
+		} else {
+			mask = ~0x5;
+			if (conf & 0x4)
+				entry->lvdsconf.use_power_scripts = true;
+		}
+		if (conf & mask) {
+			/*
+			 * Until we even try to use these on G8x, it's
+			 * useless reporting unknown bits.  They all are.
+			 */
+			if (dcb->version >= 0x40)
+				break;
+
+			NV_ERROR(dev, "Unknown LVDS configuration bits, "
+				      "please report\n");
+		}
+		break;
+		}
+	case OUTPUT_TV:
+	{
+		if (dcb->version >= 0x30)
+			entry->tvconf.has_component_output = conf & (0x8 << 4);
+		else
+			entry->tvconf.has_component_output = false;
+
+		break;
+	}
+	case OUTPUT_DP:
+		entry->dpconf.sor.link = (conf & 0x00000030) >> 4;
+		entry->dpconf.link_bw = (conf & 0x00e00000) >> 21;
+		switch ((conf & 0x0f000000) >> 24) {
+		case 0xf:
+			entry->dpconf.link_nr = 4;
+			break;
+		case 0x3:
+			entry->dpconf.link_nr = 2;
+			break;
+		default:
+			entry->dpconf.link_nr = 1;
+			break;
+		}
+		break;
+	case OUTPUT_TMDS:
+		entry->tmdsconf.sor.link = (conf & 0x00000030) >> 4;
+		break;
+	case 0xe:
+		/* weird g80 mobile type that "nv" treats as a terminator */
+		dcb->entries--;
+		return false;
+	default:
+		break;
+	}
+
+	/* unsure what DCB version introduces this, 3.0? */
+	if (conf & 0x100000)
+		entry->i2c_upper_default = true;
+
+	return true;
+}
+
+static bool
+parse_dcb15_entry(struct drm_device *dev, struct dcb_table *dcb,
+		  uint32_t conn, uint32_t conf, struct dcb_entry *entry)
+{
+	switch (conn & 0x0000000f) {
+	case 0:
+		entry->type = OUTPUT_ANALOG;
+		break;
+	case 1:
+		entry->type = OUTPUT_TV;
+		break;
+	case 2:
+	case 3:
+		entry->type = OUTPUT_LVDS;
+		break;
+	case 4:
+		switch ((conn & 0x000000f0) >> 4) {
+		case 0:
+			entry->type = OUTPUT_TMDS;
+			break;
+		case 1:
+			entry->type = OUTPUT_LVDS;
+			break;
+		default:
+			NV_ERROR(dev, "Unknown DCB subtype 4/%d\n",
+				 (conn & 0x000000f0) >> 4);
+			return false;
+		}
+		break;
+	default:
+		NV_ERROR(dev, "Unknown DCB type %d\n", conn & 0x0000000f);
+		return false;
+	}
+
+	entry->i2c_index = (conn & 0x0003c000) >> 14;
+	entry->heads = ((conn & 0x001c0000) >> 18) + 1;
+	entry->or = entry->heads; /* same as heads, hopefully safe enough */
+	entry->location = (conn & 0x01e00000) >> 21;
+	entry->bus = (conn & 0x0e000000) >> 25;
+	entry->duallink_possible = false;
+
+	switch (entry->type) {
+	case OUTPUT_ANALOG:
+		entry->crtconf.maxfreq = (conf & 0xffff) * 10;
+		break;
+	case OUTPUT_TV:
+		entry->tvconf.has_component_output = false;
+		break;
+	case OUTPUT_TMDS:
+		/*
+		 * Invent a DVI-A output, by copying the fields of the DVI-D
+		 * output; reported to work by math_b on an NV20(!).
+		 */
+		fabricate_vga_output(dcb, entry->i2c_index, entry->heads);
+		break;
+	case OUTPUT_LVDS:
+		if ((conn & 0x00003f00) != 0x10)
+			entry->lvdsconf.use_straps_for_mode = true;
+		entry->lvdsconf.use_power_scripts = true;
+		break;
+	default:
+		break;
+	}
+
+	return true;
+}
+
+static bool parse_dcb_entry(struct drm_device *dev, struct dcb_table *dcb,
+			    uint32_t conn, uint32_t conf)
+{
+	struct dcb_entry *entry = new_dcb_entry(dcb);
+	bool ret;
+
+	if (dcb->version >= 0x20)
+		ret = parse_dcb20_entry(dev, dcb, conn, conf, entry);
+	else
+		ret = parse_dcb15_entry(dev, dcb, conn, conf, entry);
+	if (!ret)
+		return ret;
+
+	read_dcb_i2c_entry(dev, dcb->version, dcb->i2c_table,
+			   entry->i2c_index, &dcb->i2c[entry->i2c_index]);
+
+	return true;
+}
+
+static
+void merge_like_dcb_entries(struct drm_device *dev, struct dcb_table *dcb)
+{
+	/*
+	 * DCB v2.0 lists each output combination separately.
+	 * Here we merge compatible entries to have fewer outputs, with
+	 * more options
+	 */
+
+	int i, newentries = 0;
+
+	for (i = 0; i < dcb->entries; i++) {
+		struct dcb_entry *ient = &dcb->entry[i];
+		int j;
+
+		for (j = i + 1; j < dcb->entries; j++) {
+			struct dcb_entry *jent = &dcb->entry[j];
+
+			if (jent->type == 100) /* already merged entry */
+				continue;
+
+			/* merge heads field when all other fields the same */
+			if (jent->i2c_index == ient->i2c_index &&
+			    jent->type == ient->type &&
+			    jent->location == ient->location &&
+			    jent->or == ient->or) {
+				NV_TRACE(dev, "Merging DCB entries %d and %d\n",
+					 i, j);
+				ient->heads |= jent->heads;
+				jent->type = 100; /* dummy value */
+			}
+		}
+	}
+
+	/* Compact entries merged into others out of dcb */
+	for (i = 0; i < dcb->entries; i++) {
+		if (dcb->entry[i].type == 100)
+			continue;
+
+		if (newentries != i) {
+			dcb->entry[newentries] = dcb->entry[i];
+			dcb->entry[newentries].index = newentries;
+		}
+		newentries++;
+	}
+
+	dcb->entries = newentries;
+}
+
+static int
+parse_dcb_table(struct drm_device *dev, struct nvbios *bios, bool twoHeads)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct dcb_table *dcb = &bios->dcb;
+	uint16_t dcbptr = 0, i2ctabptr = 0;
+	uint8_t *dcbtable;
+	uint8_t headerlen = 0x4, entries = DCB_MAX_NUM_ENTRIES;
+	bool configblock = true;
+	int recordlength = 8, confofs = 4;
+	int i;
+
+	/* get the offset from 0x36 */
+	if (dev_priv->card_type > NV_04) {
+		dcbptr = ROM16(bios->data[0x36]);
+		if (dcbptr == 0x0000)
+			NV_WARN(dev, "No output data (DCB) found in BIOS\n");
+	}
+
+	/* this situation likely means a really old card, pre DCB */
+	if (dcbptr == 0x0) {
+		NV_INFO(dev, "Assuming a CRT output exists\n");
+		fabricate_vga_output(dcb, LEGACY_I2C_CRT, 1);
+
+		if (nv04_tv_identify(dev, bios->legacy.i2c_indices.tv) >= 0)
+			fabricate_tv_output(dcb, twoHeads);
+
+		return 0;
+	}
+
+	dcbtable = &bios->data[dcbptr];
+
+	/* get DCB version */
+	dcb->version = dcbtable[0];
+	NV_TRACE(dev, "Found Display Configuration Block version %d.%d\n",
+		 dcb->version >> 4, dcb->version & 0xf);
+
+	if (dcb->version >= 0x20) { /* NV17+ */
+		uint32_t sig;
+
+		if (dcb->version >= 0x30) { /* NV40+ */
+			headerlen = dcbtable[1];
+			entries = dcbtable[2];
+			recordlength = dcbtable[3];
+			i2ctabptr = ROM16(dcbtable[4]);
+			sig = ROM32(dcbtable[6]);
+			dcb->gpio_table_ptr = ROM16(dcbtable[10]);
+			dcb->connector_table_ptr = ROM16(dcbtable[20]);
+		} else {
+			i2ctabptr = ROM16(dcbtable[2]);
+			sig = ROM32(dcbtable[4]);
+			headerlen = 8;
+		}
+
+		if (sig != 0x4edcbdcb) {
+			NV_ERROR(dev, "Bad Display Configuration Block "
+					"signature (%08X)\n", sig);
+			return -EINVAL;
+		}
+	} else if (dcb->version >= 0x15) { /* some NV11 and NV20 */
+		char sig[8] = { 0 };
+
+		strncpy(sig, (char *)&dcbtable[-7], 7);
+		i2ctabptr = ROM16(dcbtable[2]);
+		recordlength = 10;
+		confofs = 6;
+
+		if (strcmp(sig, "DEV_REC")) {
+			NV_ERROR(dev, "Bad Display Configuration Block "
+					"signature (%s)\n", sig);
+			return -EINVAL;
+		}
+	} else {
+		/*
+		 * v1.4 (some NV15/16, NV11+) seems the same as v1.5, but always
+		 * has the same single (crt) entry, even when tv-out present, so
+		 * the conclusion is this version cannot really be used.
+		 * v1.2 tables (some NV6/10, and NV15+) normally have the same
+		 * 5 entries, which are not specific to the card and so no use.
+		 * v1.2 does have an I2C table that read_dcb_i2c_table can
+		 * handle, but cards exist (nv11 in #14821) with a bad i2c table
+		 * pointer, so use the indices parsed in parse_bmp_structure.
+		 * v1.1 (NV5+, maybe some NV4) is entirely unhelpful
+		 */
+		NV_TRACEWARN(dev, "No useful information in BIOS output table; "
+				  "adding all possible outputs\n");
+		fabricate_vga_output(dcb, LEGACY_I2C_CRT, 1);
+
+		/*
+		 * Attempt to detect TV before DVI because the test
+		 * for the former is more accurate and it rules the
+		 * latter out.
+		 */
+		if (nv04_tv_identify(dev,
+				     bios->legacy.i2c_indices.tv) >= 0)
+			fabricate_tv_output(dcb, twoHeads);
+
+		else if (bios->tmds.output0_script_ptr ||
+			 bios->tmds.output1_script_ptr)
+			fabricate_dvi_i_output(dcb, twoHeads);
+
+		return 0;
+	}
+
+	if (!i2ctabptr)
+		NV_WARN(dev, "No pointer to DCB I2C port table\n");
+	else {
+		dcb->i2c_table = &bios->data[i2ctabptr];
+		if (dcb->version >= 0x30)
+			dcb->i2c_default_indices = dcb->i2c_table[4];
+	}
+
+	if (entries > DCB_MAX_NUM_ENTRIES)
+		entries = DCB_MAX_NUM_ENTRIES;
+
+	for (i = 0; i < entries; i++) {
+		uint32_t connection, config = 0;
+
+		connection = ROM32(dcbtable[headerlen + recordlength * i]);
+		if (configblock)
+			config = ROM32(dcbtable[headerlen + confofs + recordlength * i]);
+
+		/* seen on an NV11 with DCB v1.5 */
+		if (connection == 0x00000000)
+			break;
+
+		/* seen on an NV17 with DCB v2.0 */
+		if (connection == 0xffffffff)
+			break;
+
+		if ((connection & 0x0000000f) == 0x0000000f)
+			continue;
+
+		NV_TRACEWARN(dev, "Raw DCB entry %d: %08x %08x\n",
+			     dcb->entries, connection, config);
+
+		if (!parse_dcb_entry(dev, dcb, connection, config))
+			break;
+	}
+
+	/*
+	 * apart for v2.1+ not being known for requiring merging, this
+	 * guarantees dcbent->index is the index of the entry in the rom image
+	 */
+	if (dcb->version < 0x21)
+		merge_like_dcb_entries(dev, dcb);
+
+	if (!dcb->entries)
+		return -ENXIO;
+
+	parse_dcb_gpio_table(bios);
+	parse_dcb_connector_table(bios);
+	return 0;
+}
+
+static void
+fixup_legacy_connector(struct nvbios *bios)
+{
+	struct dcb_table *dcb = &bios->dcb;
+	int i, i2c, i2c_conn[DCB_MAX_NUM_I2C_ENTRIES] = { };
+
+	/*
+	 * DCB 3.0 also has the table in most cases, but there are some cards
+	 * where the table is filled with stub entries, and the DCB entriy
+	 * indices are all 0.  We don't need the connector indices on pre-G80
+	 * chips (yet?) so limit the use to DCB 4.0 and above.
+	 */
+	if (dcb->version >= 0x40)
+		return;
+
+	dcb->connector.entries = 0;
+
+	/*
+	 * No known connector info before v3.0, so make it up.  the rule here
+	 * is: anything on the same i2c bus is considered to be on the same
+	 * connector.  any output without an associated i2c bus is assigned
+	 * its own unique connector index.
+	 */
+	for (i = 0; i < dcb->entries; i++) {
+		/*
+		 * Ignore the I2C index for on-chip TV-out, as there
+		 * are cards with bogus values (nv31m in bug 23212),
+		 * and it's otherwise useless.
+		 */
+		if (dcb->entry[i].type == OUTPUT_TV &&
+		    dcb->entry[i].location == DCB_LOC_ON_CHIP)
+			dcb->entry[i].i2c_index = 0xf;
+		i2c = dcb->entry[i].i2c_index;
+
+		if (i2c_conn[i2c]) {
+			dcb->entry[i].connector = i2c_conn[i2c] - 1;
+			continue;
+		}
+
+		dcb->entry[i].connector = dcb->connector.entries++;
+		if (i2c != 0xf)
+			i2c_conn[i2c] = dcb->connector.entries;
+	}
+
+	/* Fake the connector table as well as just connector indices */
+	for (i = 0; i < dcb->connector.entries; i++) {
+		dcb->connector.entry[i].index = i;
+		dcb->connector.entry[i].type = divine_connector_type(bios, i);
+		dcb->connector.entry[i].gpio_tag = 0xff;
+	}
+}
+
+static void
+fixup_legacy_i2c(struct nvbios *bios)
+{
+	struct dcb_table *dcb = &bios->dcb;
+	int i;
+
+	for (i = 0; i < dcb->entries; i++) {
+		if (dcb->entry[i].i2c_index == LEGACY_I2C_CRT)
+			dcb->entry[i].i2c_index = bios->legacy.i2c_indices.crt;
+		if (dcb->entry[i].i2c_index == LEGACY_I2C_PANEL)
+			dcb->entry[i].i2c_index = bios->legacy.i2c_indices.panel;
+		if (dcb->entry[i].i2c_index == LEGACY_I2C_TV)
+			dcb->entry[i].i2c_index = bios->legacy.i2c_indices.tv;
+	}
+}
+
+static int load_nv17_hwsq_ucode_entry(struct drm_device *dev, struct nvbios *bios, uint16_t hwsq_offset, int entry)
+{
+	/*
+	 * The header following the "HWSQ" signature has the number of entries,
+	 * and the entry size
+	 *
+	 * An entry consists of a dword to write to the sequencer control reg
+	 * (0x00001304), followed by the ucode bytes, written sequentially,
+	 * starting at reg 0x00001400
+	 */
+
+	uint8_t bytes_to_write;
+	uint16_t hwsq_entry_offset;
+	int i;
+
+	if (bios->data[hwsq_offset] <= entry) {
+		NV_ERROR(dev, "Too few entries in HW sequencer table for "
+				"requested entry\n");
+		return -ENOENT;
+	}
+
+	bytes_to_write = bios->data[hwsq_offset + 1];
+
+	if (bytes_to_write != 36) {
+		NV_ERROR(dev, "Unknown HW sequencer entry size\n");
+		return -EINVAL;
+	}
+
+	NV_TRACE(dev, "Loading NV17 power sequencing microcode\n");
+
+	hwsq_entry_offset = hwsq_offset + 2 + entry * bytes_to_write;
+
+	/* set sequencer control */
+	bios_wr32(bios, 0x00001304, ROM32(bios->data[hwsq_entry_offset]));
+	bytes_to_write -= 4;
+
+	/* write ucode */
+	for (i = 0; i < bytes_to_write; i += 4)
+		bios_wr32(bios, 0x00001400 + i, ROM32(bios->data[hwsq_entry_offset + i + 4]));
+
+	/* twiddle NV_PBUS_DEBUG_4 */
+	bios_wr32(bios, NV_PBUS_DEBUG_4, bios_rd32(bios, NV_PBUS_DEBUG_4) | 0x18);
+
+	return 0;
+}
+
+static int load_nv17_hw_sequencer_ucode(struct drm_device *dev,
+					struct nvbios *bios)
+{
+	/*
+	 * BMP based cards, from NV17, need a microcode loading to correctly
+	 * control the GPIO etc for LVDS panels
+	 *
+	 * BIT based cards seem to do this directly in the init scripts
+	 *
+	 * The microcode entries are found by the "HWSQ" signature.
+	 */
+
+	const uint8_t hwsq_signature[] = { 'H', 'W', 'S', 'Q' };
+	const int sz = sizeof(hwsq_signature);
+	int hwsq_offset;
+
+	hwsq_offset = findstr(bios->data, bios->length, hwsq_signature, sz);
+	if (!hwsq_offset)
+		return 0;
+
+	/* always use entry 0? */
+	return load_nv17_hwsq_ucode_entry(dev, bios, hwsq_offset + sz, 0);
+}
+
+uint8_t *nouveau_bios_embedded_edid(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nvbios *bios = &dev_priv->vbios;
+	const uint8_t edid_sig[] = {
+			0x00, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0x00 };
+	uint16_t offset = 0;
+	uint16_t newoffset;
+	int searchlen = NV_PROM_SIZE;
+
+	if (bios->fp.edid)
+		return bios->fp.edid;
+
+	while (searchlen) {
+		newoffset = findstr(&bios->data[offset], searchlen,
+								edid_sig, 8);
+		if (!newoffset)
+			return NULL;
+		offset += newoffset;
+		if (!nv_cksum(&bios->data[offset], EDID1_LEN))
+			break;
+
+		searchlen -= offset;
+		offset++;
+	}
+
+	NV_TRACE(dev, "Found EDID in BIOS\n");
+
+	return bios->fp.edid = &bios->data[offset];
+}
+
+void
+nouveau_bios_run_init_table(struct drm_device *dev, uint16_t table,
+			    struct dcb_entry *dcbent)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nvbios *bios = &dev_priv->vbios;
+	struct init_exec iexec = { true, false };
+
+	mutex_lock(&bios->lock);
+	bios->display.output = dcbent;
+	parse_init_table(bios, table, &iexec);
+	bios->display.output = NULL;
+	mutex_unlock(&bios->lock);
+}
+
+static bool NVInitVBIOS(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nvbios *bios = &dev_priv->vbios;
+
+	memset(bios, 0, sizeof(struct nvbios));
+	mutex_init(&bios->lock);
+	bios->dev = dev;
+
+	if (!NVShadowVBIOS(dev, bios->data))
+		return false;
+
+	bios->length = NV_PROM_SIZE;
+	return true;
+}
+
+static int nouveau_parse_vbios_struct(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nvbios *bios = &dev_priv->vbios;
+	const uint8_t bit_signature[] = { 0xff, 0xb8, 'B', 'I', 'T' };
+	const uint8_t bmp_signature[] = { 0xff, 0x7f, 'N', 'V', 0x0 };
+	int offset;
+
+	offset = findstr(bios->data, bios->length,
+					bit_signature, sizeof(bit_signature));
+	if (offset) {
+		NV_TRACE(dev, "BIT BIOS found\n");
+		return parse_bit_structure(bios, offset + 6);
+	}
+
+	offset = findstr(bios->data, bios->length,
+					bmp_signature, sizeof(bmp_signature));
+	if (offset) {
+		NV_TRACE(dev, "BMP BIOS found\n");
+		return parse_bmp_structure(dev, bios, offset);
+	}
+
+	NV_ERROR(dev, "No known BIOS signature found\n");
+	return -ENODEV;
+}
+
+int
+nouveau_run_vbios_init(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nvbios *bios = &dev_priv->vbios;
+	int i, ret = 0;
+
+	NVLockVgaCrtcs(dev, false);
+	if (nv_two_heads(dev))
+		NVSetOwner(dev, bios->state.crtchead);
+
+	if (bios->major_version < 5)	/* BMP only */
+		load_nv17_hw_sequencer_ucode(dev, bios);
+
+	if (bios->execute) {
+		bios->fp.last_script_invoc = 0;
+		bios->fp.lvds_init_run = false;
+	}
+
+	parse_init_tables(bios);
+
+	/*
+	 * Runs some additional script seen on G8x VBIOSen.  The VBIOS'
+	 * parser will run this right after the init tables, the binary
+	 * driver appears to run it at some point later.
+	 */
+	if (bios->some_script_ptr) {
+		struct init_exec iexec = {true, false};
+
+		NV_INFO(dev, "Parsing VBIOS init table at offset 0x%04X\n",
+			bios->some_script_ptr);
+		parse_init_table(bios, bios->some_script_ptr, &iexec);
+	}
+
+	if (dev_priv->card_type >= NV_50) {
+		for (i = 0; i < bios->dcb.entries; i++) {
+			nouveau_bios_run_display_table(dev,
+						       &bios->dcb.entry[i],
+						       0, 0);
+		}
+	}
+
+	NVLockVgaCrtcs(dev, true);
+
+	return ret;
+}
+
+static void
+nouveau_bios_i2c_devices_takedown(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nvbios *bios = &dev_priv->vbios;
+	struct dcb_i2c_entry *entry;
+	int i;
+
+	entry = &bios->dcb.i2c[0];
+	for (i = 0; i < DCB_MAX_NUM_I2C_ENTRIES; i++, entry++)
+		nouveau_i2c_fini(dev, entry);
+}
+
+int
+nouveau_bios_init(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nvbios *bios = &dev_priv->vbios;
+	uint32_t saved_nv_pextdev_boot_0;
+	bool was_locked;
+	int ret;
+
+	if (!NVInitVBIOS(dev))
+		return -ENODEV;
+
+	ret = nouveau_parse_vbios_struct(dev);
+	if (ret)
+		return ret;
+
+	ret = parse_dcb_table(dev, bios, nv_two_heads(dev));
+	if (ret)
+		return ret;
+
+	fixup_legacy_i2c(bios);
+	fixup_legacy_connector(bios);
+
+	if (!bios->major_version)	/* we don't run version 0 bios */
+		return 0;
+
+	/* these will need remembering across a suspend */
+	saved_nv_pextdev_boot_0 = bios_rd32(bios, NV_PEXTDEV_BOOT_0);
+	bios->state.saved_nv_pfb_cfg0 = bios_rd32(bios, NV_PFB_CFG0);
+
+	/* init script execution disabled */
+	bios->execute = false;
+
+	/* ... unless card isn't POSTed already */
+	if (dev_priv->card_type >= NV_10 &&
+	    NVReadVgaCrtc(dev, 0, 0x00) == 0 &&
+	    NVReadVgaCrtc(dev, 0, 0x1a) == 0) {
+		NV_INFO(dev, "Adaptor not initialised\n");
+		if (dev_priv->card_type < NV_50) {
+			NV_ERROR(dev, "Unable to POST this chipset\n");
+			return -ENODEV;
+		}
+
+		NV_INFO(dev, "Running VBIOS init tables\n");
+		bios->execute = true;
+	}
+
+	bios_wr32(bios, NV_PEXTDEV_BOOT_0, saved_nv_pextdev_boot_0);
+
+	ret = nouveau_run_vbios_init(dev);
+	if (ret)
+		return ret;
+
+	/* feature_byte on BMP is poor, but init always sets CR4B */
+	was_locked = NVLockVgaCrtcs(dev, false);
+	if (bios->major_version < 5)
+		bios->is_mobile = NVReadVgaCrtc(dev, 0, NV_CIO_CRE_4B) & 0x40;
+
+	/* all BIT systems need p_f_m_t for digital_min_front_porch */
+	if (bios->is_mobile || bios->major_version >= 5)
+		ret = parse_fp_mode_table(dev, bios);
+	NVLockVgaCrtcs(dev, was_locked);
+
+	/* allow subsequent scripts to execute */
+	bios->execute = true;
+
+	return 0;
+}
+
+void
+nouveau_bios_takedown(struct drm_device *dev)
+{
+	nouveau_bios_i2c_devices_takedown(dev);
+}
diff --git a/drivers/gpu/drm/nouveau/nouveau_bios.h b/drivers/gpu/drm/nouveau/nouveau_bios.h
new file mode 100644
index 000000000000..9f688aa9a655
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_bios.h
@@ -0,0 +1,298 @@
+/*
+ * Copyright 2007-2008 Nouveau Project
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef __NOUVEAU_BIOS_H__
+#define __NOUVEAU_BIOS_H__
+
+#include "nvreg.h"
+#include "nouveau_i2c.h"
+
+#define DCB_MAX_NUM_ENTRIES 16
+#define DCB_MAX_NUM_I2C_ENTRIES 16
+#define DCB_MAX_NUM_GPIO_ENTRIES 32
+#define DCB_MAX_NUM_CONNECTOR_ENTRIES 16
+
+#define DCB_LOC_ON_CHIP 0
+
+struct dcb_i2c_entry {
+	uint8_t port_type;
+	uint8_t read, write;
+	struct nouveau_i2c_chan *chan;
+};
+
+enum dcb_gpio_tag {
+	DCB_GPIO_TVDAC0 = 0xc,
+	DCB_GPIO_TVDAC1 = 0x2d,
+};
+
+struct dcb_gpio_entry {
+	enum dcb_gpio_tag tag;
+	int line;
+	bool invert;
+};
+
+struct dcb_gpio_table {
+	int entries;
+	struct dcb_gpio_entry entry[DCB_MAX_NUM_GPIO_ENTRIES];
+};
+
+enum dcb_connector_type {
+	DCB_CONNECTOR_VGA = 0x00,
+	DCB_CONNECTOR_TV_0 = 0x10,
+	DCB_CONNECTOR_TV_1 = 0x11,
+	DCB_CONNECTOR_TV_3 = 0x13,
+	DCB_CONNECTOR_DVI_I = 0x30,
+	DCB_CONNECTOR_DVI_D = 0x31,
+	DCB_CONNECTOR_LVDS = 0x40,
+	DCB_CONNECTOR_DP = 0x46,
+	DCB_CONNECTOR_eDP = 0x47,
+	DCB_CONNECTOR_HDMI_0 = 0x60,
+	DCB_CONNECTOR_HDMI_1 = 0x61,
+	DCB_CONNECTOR_NONE = 0xff
+};
+
+struct dcb_connector_table_entry {
+	uint32_t entry;
+	enum dcb_connector_type type;
+	uint8_t index;
+	uint8_t gpio_tag;
+};
+
+struct dcb_connector_table {
+	int entries;
+	struct dcb_connector_table_entry entry[DCB_MAX_NUM_CONNECTOR_ENTRIES];
+};
+
+enum dcb_type {
+	OUTPUT_ANALOG = 0,
+	OUTPUT_TV = 1,
+	OUTPUT_TMDS = 2,
+	OUTPUT_LVDS = 3,
+	OUTPUT_DP = 6,
+	OUTPUT_ANY = -1
+};
+
+struct dcb_entry {
+	int index;	/* may not be raw dcb index if merging has happened */
+	enum dcb_type type;
+	uint8_t i2c_index;
+	uint8_t heads;
+	uint8_t connector;
+	uint8_t bus;
+	uint8_t location;
+	uint8_t or;
+	bool duallink_possible;
+	union {
+		struct sor_conf {
+			int link;
+		} sorconf;
+		struct {
+			int maxfreq;
+		} crtconf;
+		struct {
+			struct sor_conf sor;
+			bool use_straps_for_mode;
+			bool use_power_scripts;
+		} lvdsconf;
+		struct {
+			bool has_component_output;
+		} tvconf;
+		struct {
+			struct sor_conf sor;
+			int link_nr;
+			int link_bw;
+		} dpconf;
+		struct {
+			struct sor_conf sor;
+		} tmdsconf;
+	};
+	bool i2c_upper_default;
+};
+
+struct dcb_table {
+	uint8_t version;
+
+	int entries;
+	struct dcb_entry entry[DCB_MAX_NUM_ENTRIES];
+
+	uint8_t *i2c_table;
+	uint8_t i2c_default_indices;
+	struct dcb_i2c_entry i2c[DCB_MAX_NUM_I2C_ENTRIES];
+
+	uint16_t gpio_table_ptr;
+	struct dcb_gpio_table gpio;
+	uint16_t connector_table_ptr;
+	struct dcb_connector_table connector;
+};
+
+enum nouveau_or {
+	OUTPUT_A = (1 << 0),
+	OUTPUT_B = (1 << 1),
+	OUTPUT_C = (1 << 2)
+};
+
+enum LVDS_script {
+	/* Order *does* matter here */
+	LVDS_INIT = 1,
+	LVDS_RESET,
+	LVDS_BACKLIGHT_ON,
+	LVDS_BACKLIGHT_OFF,
+	LVDS_PANEL_ON,
+	LVDS_PANEL_OFF
+};
+
+/* changing these requires matching changes to reg tables in nv_get_clock */
+#define MAX_PLL_TYPES	4
+enum pll_types {
+	NVPLL,
+	MPLL,
+	VPLL1,
+	VPLL2
+};
+
+struct pll_lims {
+	struct {
+		int minfreq;
+		int maxfreq;
+		int min_inputfreq;
+		int max_inputfreq;
+
+		uint8_t min_m;
+		uint8_t max_m;
+		uint8_t min_n;
+		uint8_t max_n;
+	} vco1, vco2;
+
+	uint8_t max_log2p;
+	/*
+	 * for most pre nv50 cards setting a log2P of 7 (the common max_log2p
+	 * value) is no different to 6 (at least for vplls) so allowing the MNP
+	 * calc to use 7 causes the generated clock to be out by a factor of 2.
+	 * however, max_log2p cannot be fixed-up during parsing as the
+	 * unmodified max_log2p value is still needed for setting mplls, hence
+	 * an additional max_usable_log2p member
+	 */
+	uint8_t max_usable_log2p;
+	uint8_t log2p_bias;
+
+	uint8_t min_p;
+	uint8_t max_p;
+
+	int refclk;
+};
+
+struct nvbios {
+	struct drm_device *dev;
+
+	uint8_t chip_version;
+
+	uint32_t dactestval;
+	uint32_t tvdactestval;
+	uint8_t digital_min_front_porch;
+	bool fp_no_ddc;
+
+	struct mutex lock;
+
+	uint8_t data[NV_PROM_SIZE];
+	unsigned int length;
+	bool execute;
+
+	uint8_t major_version;
+	uint8_t feature_byte;
+	bool is_mobile;
+
+	uint32_t fmaxvco, fminvco;
+
+	bool old_style_init;
+	uint16_t init_script_tbls_ptr;
+	uint16_t extra_init_script_tbl_ptr;
+	uint16_t macro_index_tbl_ptr;
+	uint16_t macro_tbl_ptr;
+	uint16_t condition_tbl_ptr;
+	uint16_t io_condition_tbl_ptr;
+	uint16_t io_flag_condition_tbl_ptr;
+	uint16_t init_function_tbl_ptr;
+
+	uint16_t pll_limit_tbl_ptr;
+	uint16_t ram_restrict_tbl_ptr;
+	uint8_t ram_restrict_group_count;
+
+	uint16_t some_script_ptr; /* BIT I + 14 */
+	uint16_t init96_tbl_ptr; /* BIT I + 16 */
+
+	struct dcb_table dcb;
+
+	struct {
+		int crtchead;
+		/* these need remembering across suspend */
+		uint32_t saved_nv_pfb_cfg0;
+	} state;
+
+	struct {
+		struct dcb_entry *output;
+		uint16_t script_table_ptr;
+		uint16_t dp_table_ptr;
+	} display;
+
+	struct {
+		uint16_t fptablepointer;	/* also used by tmds */
+		uint16_t fpxlatetableptr;
+		int xlatwidth;
+		uint16_t lvdsmanufacturerpointer;
+		uint16_t fpxlatemanufacturertableptr;
+		uint16_t mode_ptr;
+		uint16_t xlated_entry;
+		bool power_off_for_reset;
+		bool reset_after_pclk_change;
+		bool dual_link;
+		bool link_c_increment;
+		bool BITbit1;
+		bool if_is_24bit;
+		int duallink_transition_clk;
+		uint8_t strapless_is_24bit;
+		uint8_t *edid;
+
+		/* will need resetting after suspend */
+		int last_script_invoc;
+		bool lvds_init_run;
+	} fp;
+
+	struct {
+		uint16_t output0_script_ptr;
+		uint16_t output1_script_ptr;
+	} tmds;
+
+	struct {
+		uint16_t mem_init_tbl_ptr;
+		uint16_t sdr_seq_tbl_ptr;
+		uint16_t ddr_seq_tbl_ptr;
+
+		struct {
+			uint8_t crt, tv, panel;
+		} i2c_indices;
+
+		uint16_t lvds_single_a_script_ptr;
+	} legacy;
+};
+
+#endif
diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c
new file mode 100644
index 000000000000..028719fddf76
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
@@ -0,0 +1,773 @@
+/*
+ * Copyright 2007 Dave Airlied
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+/*
+ * Authors: Dave Airlied <airlied@linux.ie>
+ *	    Ben Skeggs   <darktama@iinet.net.au>
+ *	    Jeremy Kolb  <jkolb@brandeis.edu>
+ */
+
+#include "drmP.h"
+
+#include "nouveau_drm.h"
+#include "nouveau_drv.h"
+#include "nouveau_dma.h"
+
+#include <linux/log2.h>
+
+static void
+nouveau_bo_del_ttm(struct ttm_buffer_object *bo)
+{
+	struct drm_nouveau_private *dev_priv = nouveau_bdev(bo->bdev);
+	struct drm_device *dev = dev_priv->dev;
+	struct nouveau_bo *nvbo = nouveau_bo(bo);
+
+	ttm_bo_kunmap(&nvbo->kmap);
+
+	if (unlikely(nvbo->gem))
+		DRM_ERROR("bo %p still attached to GEM object\n", bo);
+
+	if (nvbo->tile)
+		nv10_mem_expire_tiling(dev, nvbo->tile, NULL);
+
+	spin_lock(&dev_priv->ttm.bo_list_lock);
+	list_del(&nvbo->head);
+	spin_unlock(&dev_priv->ttm.bo_list_lock);
+	kfree(nvbo);
+}
+
+static void
+nouveau_bo_fixup_align(struct drm_device *dev,
+		       uint32_t tile_mode, uint32_t tile_flags,
+		       int *align, int *size)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	/*
+	 * Some of the tile_flags have a periodic structure of N*4096 bytes,
+	 * align to to that as well as the page size. Align the size to the
+	 * appropriate boundaries. This does imply that sizes are rounded up
+	 * 3-7 pages, so be aware of this and do not waste memory by allocating
+	 * many small buffers.
+	 */
+	if (dev_priv->card_type == NV_50) {
+		uint32_t block_size = nouveau_mem_fb_amount(dev) >> 15;
+		int i;
+
+		switch (tile_flags) {
+		case 0x1800:
+		case 0x2800:
+		case 0x4800:
+		case 0x7a00:
+			if (is_power_of_2(block_size)) {
+				for (i = 1; i < 10; i++) {
+					*align = 12 * i * block_size;
+					if (!(*align % 65536))
+						break;
+				}
+			} else {
+				for (i = 1; i < 10; i++) {
+					*align = 8 * i * block_size;
+					if (!(*align % 65536))
+						break;
+				}
+			}
+			*size = roundup(*size, *align);
+			break;
+		default:
+			break;
+		}
+
+	} else {
+		if (tile_mode) {
+			if (dev_priv->chipset >= 0x40) {
+				*align = 65536;
+				*size = roundup(*size, 64 * tile_mode);
+
+			} else if (dev_priv->chipset >= 0x30) {
+				*align = 32768;
+				*size = roundup(*size, 64 * tile_mode);
+
+			} else if (dev_priv->chipset >= 0x20) {
+				*align = 16384;
+				*size = roundup(*size, 64 * tile_mode);
+
+			} else if (dev_priv->chipset >= 0x10) {
+				*align = 16384;
+				*size = roundup(*size, 32 * tile_mode);
+			}
+		}
+	}
+
+	/* ALIGN works only on powers of two. */
+	*size = roundup(*size, PAGE_SIZE);
+
+	if (dev_priv->card_type == NV_50) {
+		*size = roundup(*size, 65536);
+		*align = max(65536, *align);
+	}
+}
+
+int
+nouveau_bo_new(struct drm_device *dev, struct nouveau_channel *chan,
+	       int size, int align, uint32_t flags, uint32_t tile_mode,
+	       uint32_t tile_flags, bool no_vm, bool mappable,
+	       struct nouveau_bo **pnvbo)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_bo *nvbo;
+	int ret = 0;
+
+	nvbo = kzalloc(sizeof(struct nouveau_bo), GFP_KERNEL);
+	if (!nvbo)
+		return -ENOMEM;
+	INIT_LIST_HEAD(&nvbo->head);
+	INIT_LIST_HEAD(&nvbo->entry);
+	nvbo->mappable = mappable;
+	nvbo->no_vm = no_vm;
+	nvbo->tile_mode = tile_mode;
+	nvbo->tile_flags = tile_flags;
+
+	nouveau_bo_fixup_align(dev, tile_mode, tile_flags, &align, &size);
+	align >>= PAGE_SHIFT;
+
+	nvbo->placement.fpfn = 0;
+	nvbo->placement.lpfn = mappable ? dev_priv->fb_mappable_pages : 0;
+	nouveau_bo_placement_set(nvbo, flags);
+
+	nvbo->channel = chan;
+	ret = ttm_bo_init(&dev_priv->ttm.bdev, &nvbo->bo, size,
+			  ttm_bo_type_device, &nvbo->placement, align, 0,
+			  false, NULL, size, nouveau_bo_del_ttm);
+	nvbo->channel = NULL;
+	if (ret) {
+		/* ttm will call nouveau_bo_del_ttm if it fails.. */
+		return ret;
+	}
+
+	spin_lock(&dev_priv->ttm.bo_list_lock);
+	list_add_tail(&nvbo->head, &dev_priv->ttm.bo_list);
+	spin_unlock(&dev_priv->ttm.bo_list_lock);
+	*pnvbo = nvbo;
+	return 0;
+}
+
+void
+nouveau_bo_placement_set(struct nouveau_bo *nvbo, uint32_t memtype)
+{
+	int n = 0;
+
+	if (memtype & TTM_PL_FLAG_VRAM)
+		nvbo->placements[n++] = TTM_PL_FLAG_VRAM | TTM_PL_MASK_CACHING;
+	if (memtype & TTM_PL_FLAG_TT)
+		nvbo->placements[n++] = TTM_PL_FLAG_TT | TTM_PL_MASK_CACHING;
+	if (memtype & TTM_PL_FLAG_SYSTEM)
+		nvbo->placements[n++] = TTM_PL_FLAG_SYSTEM | TTM_PL_MASK_CACHING;
+	nvbo->placement.placement = nvbo->placements;
+	nvbo->placement.busy_placement = nvbo->placements;
+	nvbo->placement.num_placement = n;
+	nvbo->placement.num_busy_placement = n;
+
+	if (nvbo->pin_refcnt) {
+		while (n--)
+			nvbo->placements[n] |= TTM_PL_FLAG_NO_EVICT;
+	}
+}
+
+int
+nouveau_bo_pin(struct nouveau_bo *nvbo, uint32_t memtype)
+{
+	struct drm_nouveau_private *dev_priv = nouveau_bdev(nvbo->bo.bdev);
+	struct ttm_buffer_object *bo = &nvbo->bo;
+	int ret, i;
+
+	if (nvbo->pin_refcnt && !(memtype & (1 << bo->mem.mem_type))) {
+		NV_ERROR(nouveau_bdev(bo->bdev)->dev,
+			 "bo %p pinned elsewhere: 0x%08x vs 0x%08x\n", bo,
+			 1 << bo->mem.mem_type, memtype);
+		return -EINVAL;
+	}
+
+	if (nvbo->pin_refcnt++)
+		return 0;
+
+	ret = ttm_bo_reserve(bo, false, false, false, 0);
+	if (ret)
+		goto out;
+
+	nouveau_bo_placement_set(nvbo, memtype);
+	for (i = 0; i < nvbo->placement.num_placement; i++)
+		nvbo->placements[i] |= TTM_PL_FLAG_NO_EVICT;
+
+	ret = ttm_bo_validate(bo, &nvbo->placement, false, false);
+	if (ret == 0) {
+		switch (bo->mem.mem_type) {
+		case TTM_PL_VRAM:
+			dev_priv->fb_aper_free -= bo->mem.size;
+			break;
+		case TTM_PL_TT:
+			dev_priv->gart_info.aper_free -= bo->mem.size;
+			break;
+		default:
+			break;
+		}
+	}
+	ttm_bo_unreserve(bo);
+out:
+	if (unlikely(ret))
+		nvbo->pin_refcnt--;
+	return ret;
+}
+
+int
+nouveau_bo_unpin(struct nouveau_bo *nvbo)
+{
+	struct drm_nouveau_private *dev_priv = nouveau_bdev(nvbo->bo.bdev);
+	struct ttm_buffer_object *bo = &nvbo->bo;
+	int ret, i;
+
+	if (--nvbo->pin_refcnt)
+		return 0;
+
+	ret = ttm_bo_reserve(bo, false, false, false, 0);
+	if (ret)
+		return ret;
+
+	for (i = 0; i < nvbo->placement.num_placement; i++)
+		nvbo->placements[i] &= ~TTM_PL_FLAG_NO_EVICT;
+
+	ret = ttm_bo_validate(bo, &nvbo->placement, false, false);
+	if (ret == 0) {
+		switch (bo->mem.mem_type) {
+		case TTM_PL_VRAM:
+			dev_priv->fb_aper_free += bo->mem.size;
+			break;
+		case TTM_PL_TT:
+			dev_priv->gart_info.aper_free += bo->mem.size;
+			break;
+		default:
+			break;
+		}
+	}
+
+	ttm_bo_unreserve(bo);
+	return ret;
+}
+
+int
+nouveau_bo_map(struct nouveau_bo *nvbo)
+{
+	int ret;
+
+	ret = ttm_bo_reserve(&nvbo->bo, false, false, false, 0);
+	if (ret)
+		return ret;
+
+	ret = ttm_bo_kmap(&nvbo->bo, 0, nvbo->bo.mem.num_pages, &nvbo->kmap);
+	ttm_bo_unreserve(&nvbo->bo);
+	return ret;
+}
+
+void
+nouveau_bo_unmap(struct nouveau_bo *nvbo)
+{
+	ttm_bo_kunmap(&nvbo->kmap);
+}
+
+u16
+nouveau_bo_rd16(struct nouveau_bo *nvbo, unsigned index)
+{
+	bool is_iomem;
+	u16 *mem = ttm_kmap_obj_virtual(&nvbo->kmap, &is_iomem);
+	mem = &mem[index];
+	if (is_iomem)
+		return ioread16_native((void __force __iomem *)mem);
+	else
+		return *mem;
+}
+
+void
+nouveau_bo_wr16(struct nouveau_bo *nvbo, unsigned index, u16 val)
+{
+	bool is_iomem;
+	u16 *mem = ttm_kmap_obj_virtual(&nvbo->kmap, &is_iomem);
+	mem = &mem[index];
+	if (is_iomem)
+		iowrite16_native(val, (void __force __iomem *)mem);
+	else
+		*mem = val;
+}
+
+u32
+nouveau_bo_rd32(struct nouveau_bo *nvbo, unsigned index)
+{
+	bool is_iomem;
+	u32 *mem = ttm_kmap_obj_virtual(&nvbo->kmap, &is_iomem);
+	mem = &mem[index];
+	if (is_iomem)
+		return ioread32_native((void __force __iomem *)mem);
+	else
+		return *mem;
+}
+
+void
+nouveau_bo_wr32(struct nouveau_bo *nvbo, unsigned index, u32 val)
+{
+	bool is_iomem;
+	u32 *mem = ttm_kmap_obj_virtual(&nvbo->kmap, &is_iomem);
+	mem = &mem[index];
+	if (is_iomem)
+		iowrite32_native(val, (void __force __iomem *)mem);
+	else
+		*mem = val;
+}
+
+static struct ttm_backend *
+nouveau_bo_create_ttm_backend_entry(struct ttm_bo_device *bdev)
+{
+	struct drm_nouveau_private *dev_priv = nouveau_bdev(bdev);
+	struct drm_device *dev = dev_priv->dev;
+
+	switch (dev_priv->gart_info.type) {
+#if __OS_HAS_AGP
+	case NOUVEAU_GART_AGP:
+		return ttm_agp_backend_init(bdev, dev->agp->bridge);
+#endif
+	case NOUVEAU_GART_SGDMA:
+		return nouveau_sgdma_init_ttm(dev);
+	default:
+		NV_ERROR(dev, "Unknown GART type %d\n",
+			 dev_priv->gart_info.type);
+		break;
+	}
+
+	return NULL;
+}
+
+static int
+nouveau_bo_invalidate_caches(struct ttm_bo_device *bdev, uint32_t flags)
+{
+	/* We'll do this from user space. */
+	return 0;
+}
+
+static int
+nouveau_bo_init_mem_type(struct ttm_bo_device *bdev, uint32_t type,
+			 struct ttm_mem_type_manager *man)
+{
+	struct drm_nouveau_private *dev_priv = nouveau_bdev(bdev);
+	struct drm_device *dev = dev_priv->dev;
+
+	switch (type) {
+	case TTM_PL_SYSTEM:
+		man->flags = TTM_MEMTYPE_FLAG_MAPPABLE;
+		man->available_caching = TTM_PL_MASK_CACHING;
+		man->default_caching = TTM_PL_FLAG_CACHED;
+		break;
+	case TTM_PL_VRAM:
+		man->flags = TTM_MEMTYPE_FLAG_FIXED |
+			     TTM_MEMTYPE_FLAG_MAPPABLE |
+			     TTM_MEMTYPE_FLAG_NEEDS_IOREMAP;
+		man->available_caching = TTM_PL_FLAG_UNCACHED |
+					 TTM_PL_FLAG_WC;
+		man->default_caching = TTM_PL_FLAG_WC;
+
+		man->io_addr = NULL;
+		man->io_offset = drm_get_resource_start(dev, 1);
+		man->io_size = drm_get_resource_len(dev, 1);
+		if (man->io_size > nouveau_mem_fb_amount(dev))
+			man->io_size = nouveau_mem_fb_amount(dev);
+
+		man->gpu_offset = dev_priv->vm_vram_base;
+		break;
+	case TTM_PL_TT:
+		switch (dev_priv->gart_info.type) {
+		case NOUVEAU_GART_AGP:
+			man->flags = TTM_MEMTYPE_FLAG_MAPPABLE |
+				     TTM_MEMTYPE_FLAG_NEEDS_IOREMAP;
+			man->available_caching = TTM_PL_FLAG_UNCACHED;
+			man->default_caching = TTM_PL_FLAG_UNCACHED;
+			break;
+		case NOUVEAU_GART_SGDMA:
+			man->flags = TTM_MEMTYPE_FLAG_MAPPABLE |
+				     TTM_MEMTYPE_FLAG_CMA;
+			man->available_caching = TTM_PL_MASK_CACHING;
+			man->default_caching = TTM_PL_FLAG_CACHED;
+			break;
+		default:
+			NV_ERROR(dev, "Unknown GART type: %d\n",
+				 dev_priv->gart_info.type);
+			return -EINVAL;
+		}
+
+		man->io_offset  = dev_priv->gart_info.aper_base;
+		man->io_size    = dev_priv->gart_info.aper_size;
+		man->io_addr   = NULL;
+		man->gpu_offset = dev_priv->vm_gart_base;
+		break;
+	default:
+		NV_ERROR(dev, "Unsupported memory type %u\n", (unsigned)type);
+		return -EINVAL;
+	}
+	return 0;
+}
+
+static void
+nouveau_bo_evict_flags(struct ttm_buffer_object *bo, struct ttm_placement *pl)
+{
+	struct nouveau_bo *nvbo = nouveau_bo(bo);
+
+	switch (bo->mem.mem_type) {
+	case TTM_PL_VRAM:
+		nouveau_bo_placement_set(nvbo, TTM_PL_FLAG_TT |
+					 TTM_PL_FLAG_SYSTEM);
+		break;
+	default:
+		nouveau_bo_placement_set(nvbo, TTM_PL_FLAG_SYSTEM);
+		break;
+	}
+
+	*pl = nvbo->placement;
+}
+
+
+/* GPU-assisted copy using NV_MEMORY_TO_MEMORY_FORMAT, can access
+ * TTM_PL_{VRAM,TT} directly.
+ */
+
+static int
+nouveau_bo_move_accel_cleanup(struct nouveau_channel *chan,
+			      struct nouveau_bo *nvbo, bool evict, bool no_wait,
+			      struct ttm_mem_reg *new_mem)
+{
+	struct nouveau_fence *fence = NULL;
+	int ret;
+
+	ret = nouveau_fence_new(chan, &fence, true);
+	if (ret)
+		return ret;
+
+	ret = ttm_bo_move_accel_cleanup(&nvbo->bo, fence, NULL,
+					evict, no_wait, new_mem);
+	if (nvbo->channel && nvbo->channel != chan)
+		ret = nouveau_fence_wait(fence, NULL, false, false);
+	nouveau_fence_unref((void *)&fence);
+	return ret;
+}
+
+static inline uint32_t
+nouveau_bo_mem_ctxdma(struct nouveau_bo *nvbo, struct nouveau_channel *chan,
+		      struct ttm_mem_reg *mem)
+{
+	if (chan == nouveau_bdev(nvbo->bo.bdev)->channel) {
+		if (mem->mem_type == TTM_PL_TT)
+			return NvDmaGART;
+		return NvDmaVRAM;
+	}
+
+	if (mem->mem_type == TTM_PL_TT)
+		return chan->gart_handle;
+	return chan->vram_handle;
+}
+
+static int
+nouveau_bo_move_m2mf(struct ttm_buffer_object *bo, int evict, bool intr,
+		     int no_wait, struct ttm_mem_reg *new_mem)
+{
+	struct nouveau_bo *nvbo = nouveau_bo(bo);
+	struct drm_nouveau_private *dev_priv = nouveau_bdev(bo->bdev);
+	struct ttm_mem_reg *old_mem = &bo->mem;
+	struct nouveau_channel *chan;
+	uint64_t src_offset, dst_offset;
+	uint32_t page_count;
+	int ret;
+
+	chan = nvbo->channel;
+	if (!chan || nvbo->tile_flags || nvbo->no_vm)
+		chan = dev_priv->channel;
+
+	src_offset = old_mem->mm_node->start << PAGE_SHIFT;
+	dst_offset = new_mem->mm_node->start << PAGE_SHIFT;
+	if (chan != dev_priv->channel) {
+		if (old_mem->mem_type == TTM_PL_TT)
+			src_offset += dev_priv->vm_gart_base;
+		else
+			src_offset += dev_priv->vm_vram_base;
+
+		if (new_mem->mem_type == TTM_PL_TT)
+			dst_offset += dev_priv->vm_gart_base;
+		else
+			dst_offset += dev_priv->vm_vram_base;
+	}
+
+	ret = RING_SPACE(chan, 3);
+	if (ret)
+		return ret;
+	BEGIN_RING(chan, NvSubM2MF, NV_MEMORY_TO_MEMORY_FORMAT_DMA_SOURCE, 2);
+	OUT_RING(chan, nouveau_bo_mem_ctxdma(nvbo, chan, old_mem));
+	OUT_RING(chan, nouveau_bo_mem_ctxdma(nvbo, chan, new_mem));
+
+	if (dev_priv->card_type >= NV_50) {
+		ret = RING_SPACE(chan, 4);
+		if (ret)
+			return ret;
+		BEGIN_RING(chan, NvSubM2MF, 0x0200, 1);
+		OUT_RING(chan, 1);
+		BEGIN_RING(chan, NvSubM2MF, 0x021c, 1);
+		OUT_RING(chan, 1);
+	}
+
+	page_count = new_mem->num_pages;
+	while (page_count) {
+		int line_count = (page_count > 2047) ? 2047 : page_count;
+
+		if (dev_priv->card_type >= NV_50) {
+			ret = RING_SPACE(chan, 3);
+			if (ret)
+				return ret;
+			BEGIN_RING(chan, NvSubM2MF, 0x0238, 2);
+			OUT_RING(chan, upper_32_bits(src_offset));
+			OUT_RING(chan, upper_32_bits(dst_offset));
+		}
+		ret = RING_SPACE(chan, 11);
+		if (ret)
+			return ret;
+		BEGIN_RING(chan, NvSubM2MF,
+				 NV_MEMORY_TO_MEMORY_FORMAT_OFFSET_IN, 8);
+		OUT_RING(chan, lower_32_bits(src_offset));
+		OUT_RING(chan, lower_32_bits(dst_offset));
+		OUT_RING(chan, PAGE_SIZE); /* src_pitch */
+		OUT_RING(chan, PAGE_SIZE); /* dst_pitch */
+		OUT_RING(chan, PAGE_SIZE); /* line_length */
+		OUT_RING(chan, line_count);
+		OUT_RING(chan, (1<<8)|(1<<0));
+		OUT_RING(chan, 0);
+		BEGIN_RING(chan, NvSubM2MF, NV_MEMORY_TO_MEMORY_FORMAT_NOP, 1);
+		OUT_RING(chan, 0);
+
+		page_count -= line_count;
+		src_offset += (PAGE_SIZE * line_count);
+		dst_offset += (PAGE_SIZE * line_count);
+	}
+
+	return nouveau_bo_move_accel_cleanup(chan, nvbo, evict, no_wait, new_mem);
+}
+
+static int
+nouveau_bo_move_flipd(struct ttm_buffer_object *bo, bool evict, bool intr,
+		      bool no_wait, struct ttm_mem_reg *new_mem)
+{
+	u32 placement_memtype = TTM_PL_FLAG_TT | TTM_PL_MASK_CACHING;
+	struct ttm_placement placement;
+	struct ttm_mem_reg tmp_mem;
+	int ret;
+
+	placement.fpfn = placement.lpfn = 0;
+	placement.num_placement = placement.num_busy_placement = 1;
+	placement.placement = placement.busy_placement = &placement_memtype;
+
+	tmp_mem = *new_mem;
+	tmp_mem.mm_node = NULL;
+	ret = ttm_bo_mem_space(bo, &placement, &tmp_mem, intr, no_wait);
+	if (ret)
+		return ret;
+
+	ret = ttm_tt_bind(bo->ttm, &tmp_mem);
+	if (ret)
+		goto out;
+
+	ret = nouveau_bo_move_m2mf(bo, true, intr, no_wait, &tmp_mem);
+	if (ret)
+		goto out;
+
+	ret = ttm_bo_move_ttm(bo, evict, no_wait, new_mem);
+out:
+	if (tmp_mem.mm_node) {
+		spin_lock(&bo->bdev->glob->lru_lock);
+		drm_mm_put_block(tmp_mem.mm_node);
+		spin_unlock(&bo->bdev->glob->lru_lock);
+	}
+
+	return ret;
+}
+
+static int
+nouveau_bo_move_flips(struct ttm_buffer_object *bo, bool evict, bool intr,
+		      bool no_wait, struct ttm_mem_reg *new_mem)
+{
+	u32 placement_memtype = TTM_PL_FLAG_TT | TTM_PL_MASK_CACHING;
+	struct ttm_placement placement;
+	struct ttm_mem_reg tmp_mem;
+	int ret;
+
+	placement.fpfn = placement.lpfn = 0;
+	placement.num_placement = placement.num_busy_placement = 1;
+	placement.placement = placement.busy_placement = &placement_memtype;
+
+	tmp_mem = *new_mem;
+	tmp_mem.mm_node = NULL;
+	ret = ttm_bo_mem_space(bo, &placement, &tmp_mem, intr, no_wait);
+	if (ret)
+		return ret;
+
+	ret = ttm_bo_move_ttm(bo, evict, no_wait, &tmp_mem);
+	if (ret)
+		goto out;
+
+	ret = nouveau_bo_move_m2mf(bo, evict, intr, no_wait, new_mem);
+	if (ret)
+		goto out;
+
+out:
+	if (tmp_mem.mm_node) {
+		spin_lock(&bo->bdev->glob->lru_lock);
+		drm_mm_put_block(tmp_mem.mm_node);
+		spin_unlock(&bo->bdev->glob->lru_lock);
+	}
+
+	return ret;
+}
+
+static int
+nouveau_bo_vm_bind(struct ttm_buffer_object *bo, struct ttm_mem_reg *new_mem,
+		   struct nouveau_tile_reg **new_tile)
+{
+	struct drm_nouveau_private *dev_priv = nouveau_bdev(bo->bdev);
+	struct drm_device *dev = dev_priv->dev;
+	struct nouveau_bo *nvbo = nouveau_bo(bo);
+	uint64_t offset;
+	int ret;
+
+	if (nvbo->no_vm || new_mem->mem_type != TTM_PL_VRAM) {
+		/* Nothing to do. */
+		*new_tile = NULL;
+		return 0;
+	}
+
+	offset = new_mem->mm_node->start << PAGE_SHIFT;
+
+	if (dev_priv->card_type == NV_50) {
+		ret = nv50_mem_vm_bind_linear(dev,
+					      offset + dev_priv->vm_vram_base,
+					      new_mem->size, nvbo->tile_flags,
+					      offset);
+		if (ret)
+			return ret;
+
+	} else if (dev_priv->card_type >= NV_10) {
+		*new_tile = nv10_mem_set_tiling(dev, offset, new_mem->size,
+						nvbo->tile_mode);
+	}
+
+	return 0;
+}
+
+static void
+nouveau_bo_vm_cleanup(struct ttm_buffer_object *bo,
+		      struct nouveau_tile_reg *new_tile,
+		      struct nouveau_tile_reg **old_tile)
+{
+	struct drm_nouveau_private *dev_priv = nouveau_bdev(bo->bdev);
+	struct drm_device *dev = dev_priv->dev;
+
+	if (dev_priv->card_type >= NV_10 &&
+	    dev_priv->card_type < NV_50) {
+		if (*old_tile)
+			nv10_mem_expire_tiling(dev, *old_tile, bo->sync_obj);
+
+		*old_tile = new_tile;
+	}
+}
+
+static int
+nouveau_bo_move(struct ttm_buffer_object *bo, bool evict, bool intr,
+		bool no_wait, struct ttm_mem_reg *new_mem)
+{
+	struct drm_nouveau_private *dev_priv = nouveau_bdev(bo->bdev);
+	struct nouveau_bo *nvbo = nouveau_bo(bo);
+	struct ttm_mem_reg *old_mem = &bo->mem;
+	struct nouveau_tile_reg *new_tile = NULL;
+	int ret = 0;
+
+	ret = nouveau_bo_vm_bind(bo, new_mem, &new_tile);
+	if (ret)
+		return ret;
+
+	/* Software copy if the card isn't up and running yet. */
+	if (dev_priv->init_state != NOUVEAU_CARD_INIT_DONE ||
+	    !dev_priv->channel) {
+		ret = ttm_bo_move_memcpy(bo, evict, no_wait, new_mem);
+		goto out;
+	}
+
+	/* Fake bo copy. */
+	if (old_mem->mem_type == TTM_PL_SYSTEM && !bo->ttm) {
+		BUG_ON(bo->mem.mm_node != NULL);
+		bo->mem = *new_mem;
+		new_mem->mm_node = NULL;
+		goto out;
+	}
+
+	/* Hardware assisted copy. */
+	if (new_mem->mem_type == TTM_PL_SYSTEM)
+		ret = nouveau_bo_move_flipd(bo, evict, intr, no_wait, new_mem);
+	else if (old_mem->mem_type == TTM_PL_SYSTEM)
+		ret = nouveau_bo_move_flips(bo, evict, intr, no_wait, new_mem);
+	else
+		ret = nouveau_bo_move_m2mf(bo, evict, intr, no_wait, new_mem);
+
+	if (!ret)
+		goto out;
+
+	/* Fallback to software copy. */
+	ret = ttm_bo_move_memcpy(bo, evict, no_wait, new_mem);
+
+out:
+	if (ret)
+		nouveau_bo_vm_cleanup(bo, NULL, &new_tile);
+	else
+		nouveau_bo_vm_cleanup(bo, new_tile, &nvbo->tile);
+
+	return ret;
+}
+
+static int
+nouveau_bo_verify_access(struct ttm_buffer_object *bo, struct file *filp)
+{
+	return 0;
+}
+
+struct ttm_bo_driver nouveau_bo_driver = {
+	.create_ttm_backend_entry = nouveau_bo_create_ttm_backend_entry,
+	.invalidate_caches = nouveau_bo_invalidate_caches,
+	.init_mem_type = nouveau_bo_init_mem_type,
+	.evict_flags = nouveau_bo_evict_flags,
+	.move = nouveau_bo_move,
+	.verify_access = nouveau_bo_verify_access,
+	.sync_obj_signaled = nouveau_fence_signalled,
+	.sync_obj_wait = nouveau_fence_wait,
+	.sync_obj_flush = nouveau_fence_flush,
+	.sync_obj_unref = nouveau_fence_unref,
+	.sync_obj_ref = nouveau_fence_ref,
+};
+
diff --git a/drivers/gpu/drm/nouveau/nouveau_calc.c b/drivers/gpu/drm/nouveau/nouveau_calc.c
new file mode 100644
index 000000000000..88f9bc0941eb
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_calc.c
@@ -0,0 +1,478 @@
+/*
+ * Copyright 1993-2003 NVIDIA, Corporation
+ * Copyright 2007-2009 Stuart Bennett
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF
+ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include "drmP.h"
+#include "nouveau_drv.h"
+#include "nouveau_hw.h"
+
+/****************************************************************************\
+*                                                                            *
+* The video arbitration routines calculate some "magic" numbers.  Fixes      *
+* the snow seen when accessing the framebuffer without it.                   *
+* It just works (I hope).                                                    *
+*                                                                            *
+\****************************************************************************/
+
+struct nv_fifo_info {
+	int lwm;
+	int burst;
+};
+
+struct nv_sim_state {
+	int pclk_khz;
+	int mclk_khz;
+	int nvclk_khz;
+	int bpp;
+	int mem_page_miss;
+	int mem_latency;
+	int memory_type;
+	int memory_width;
+	int two_heads;
+};
+
+static void
+nv04_calc_arb(struct nv_fifo_info *fifo, struct nv_sim_state *arb)
+{
+	int pagemiss, cas, width, bpp;
+	int nvclks, mclks, pclks, crtpagemiss;
+	int found, mclk_extra, mclk_loop, cbs, m1, p1;
+	int mclk_freq, pclk_freq, nvclk_freq;
+	int us_m, us_n, us_p, crtc_drain_rate;
+	int cpm_us, us_crt, clwm;
+
+	pclk_freq = arb->pclk_khz;
+	mclk_freq = arb->mclk_khz;
+	nvclk_freq = arb->nvclk_khz;
+	pagemiss = arb->mem_page_miss;
+	cas = arb->mem_latency;
+	width = arb->memory_width >> 6;
+	bpp = arb->bpp;
+	cbs = 128;
+
+	pclks = 2;
+	nvclks = 10;
+	mclks = 13 + cas;
+	mclk_extra = 3;
+	found = 0;
+
+	while (!found) {
+		found = 1;
+
+		mclk_loop = mclks + mclk_extra;
+		us_m = mclk_loop * 1000 * 1000 / mclk_freq;
+		us_n = nvclks * 1000 * 1000 / nvclk_freq;
+		us_p = nvclks * 1000 * 1000 / pclk_freq;
+
+		crtc_drain_rate = pclk_freq * bpp / 8;
+		crtpagemiss = 2;
+		crtpagemiss += 1;
+		cpm_us = crtpagemiss * pagemiss * 1000 * 1000 / mclk_freq;
+		us_crt = cpm_us + us_m + us_n + us_p;
+		clwm = us_crt * crtc_drain_rate / (1000 * 1000);
+		clwm++;
+
+		m1 = clwm + cbs - 512;
+		p1 = m1 * pclk_freq / mclk_freq;
+		p1 = p1 * bpp / 8;
+		if ((p1 < m1 && m1 > 0) || clwm > 519) {
+			found = !mclk_extra;
+			mclk_extra--;
+		}
+		if (clwm < 384)
+			clwm = 384;
+
+		fifo->lwm = clwm;
+		fifo->burst = cbs;
+	}
+}
+
+static void
+nv10_calc_arb(struct nv_fifo_info *fifo, struct nv_sim_state *arb)
+{
+	int fill_rate, drain_rate;
+	int pclks, nvclks, mclks, xclks;
+	int pclk_freq, nvclk_freq, mclk_freq;
+	int fill_lat, extra_lat;
+	int max_burst_o, max_burst_l;
+	int fifo_len, min_lwm, max_lwm;
+	const int burst_lat = 80; /* Maximum allowable latency due
+				   * to the CRTC FIFO burst. (ns) */
+
+	pclk_freq = arb->pclk_khz;
+	nvclk_freq = arb->nvclk_khz;
+	mclk_freq = arb->mclk_khz;
+
+	fill_rate = mclk_freq * arb->memory_width / 8; /* kB/s */
+	drain_rate = pclk_freq * arb->bpp / 8; /* kB/s */
+
+	fifo_len = arb->two_heads ? 1536 : 1024; /* B */
+
+	/* Fixed FIFO refill latency. */
+
+	pclks = 4;	/* lwm detect. */
+
+	nvclks = 3	/* lwm -> sync. */
+		+ 2	/* fbi bus cycles (1 req + 1 busy) */
+		+ 1	/* 2 edge sync.  may be very close to edge so
+			 * just put one. */
+		+ 1	/* fbi_d_rdv_n */
+		+ 1	/* Fbi_d_rdata */
+		+ 1;	/* crtfifo load */
+
+	mclks = 1	/* 2 edge sync.  may be very close to edge so
+			 * just put one. */
+		+ 1	/* arb_hp_req */
+		+ 5	/* tiling pipeline */
+		+ 2	/* latency fifo */
+		+ 2	/* memory request to fbio block */
+		+ 7;	/* data returned from fbio block */
+
+	/* Need to accumulate 256 bits for read */
+	mclks += (arb->memory_type == 0 ? 2 : 1)
+		* arb->memory_width / 32;
+
+	fill_lat = mclks * 1000 * 1000 / mclk_freq   /* minimum mclk latency */
+		+ nvclks * 1000 * 1000 / nvclk_freq  /* nvclk latency */
+		+ pclks * 1000 * 1000 / pclk_freq;   /* pclk latency */
+
+	/* Conditional FIFO refill latency. */
+
+	xclks = 2 * arb->mem_page_miss + mclks /* Extra latency due to
+						* the overlay. */
+		+ 2 * arb->mem_page_miss       /* Extra pagemiss latency. */
+		+ (arb->bpp == 32 ? 8 : 4);    /* Margin of error. */
+
+	extra_lat = xclks * 1000 * 1000 / mclk_freq;
+
+	if (arb->two_heads)
+		/* Account for another CRTC. */
+		extra_lat += fill_lat + extra_lat + burst_lat;
+
+	/* FIFO burst */
+
+	/* Max burst not leading to overflows. */
+	max_burst_o = (1 + fifo_len - extra_lat * drain_rate / (1000 * 1000))
+		* (fill_rate / 1000) / ((fill_rate - drain_rate) / 1000);
+	fifo->burst = min(max_burst_o, 1024);
+
+	/* Max burst value with an acceptable latency. */
+	max_burst_l = burst_lat * fill_rate / (1000 * 1000);
+	fifo->burst = min(max_burst_l, fifo->burst);
+
+	fifo->burst = rounddown_pow_of_two(fifo->burst);
+
+	/* FIFO low watermark */
+
+	min_lwm = (fill_lat + extra_lat) * drain_rate / (1000 * 1000) + 1;
+	max_lwm = fifo_len - fifo->burst
+		+ fill_lat * drain_rate / (1000 * 1000)
+		+ fifo->burst * drain_rate / fill_rate;
+
+	fifo->lwm = min_lwm + 10 * (max_lwm - min_lwm) / 100; /* Empirical. */
+}
+
+static void
+nv04_update_arb(struct drm_device *dev, int VClk, int bpp,
+		int *burst, int *lwm)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nv_fifo_info fifo_data;
+	struct nv_sim_state sim_data;
+	int MClk = nouveau_hw_get_clock(dev, MPLL);
+	int NVClk = nouveau_hw_get_clock(dev, NVPLL);
+	uint32_t cfg1 = nvReadFB(dev, NV_PFB_CFG1);
+
+	sim_data.pclk_khz = VClk;
+	sim_data.mclk_khz = MClk;
+	sim_data.nvclk_khz = NVClk;
+	sim_data.bpp = bpp;
+	sim_data.two_heads = nv_two_heads(dev);
+	if ((dev->pci_device & 0xffff) == 0x01a0 /*CHIPSET_NFORCE*/ ||
+	    (dev->pci_device & 0xffff) == 0x01f0 /*CHIPSET_NFORCE2*/) {
+		uint32_t type;
+
+		pci_read_config_dword(pci_get_bus_and_slot(0, 1), 0x7c, &type);
+
+		sim_data.memory_type = (type >> 12) & 1;
+		sim_data.memory_width = 64;
+		sim_data.mem_latency = 3;
+		sim_data.mem_page_miss = 10;
+	} else {
+		sim_data.memory_type = nvReadFB(dev, NV_PFB_CFG0) & 0x1;
+		sim_data.memory_width = (nvReadEXTDEV(dev, NV_PEXTDEV_BOOT_0) & 0x10) ? 128 : 64;
+		sim_data.mem_latency = cfg1 & 0xf;
+		sim_data.mem_page_miss = ((cfg1 >> 4) & 0xf) + ((cfg1 >> 31) & 0x1);
+	}
+
+	if (dev_priv->card_type == NV_04)
+		nv04_calc_arb(&fifo_data, &sim_data);
+	else
+		nv10_calc_arb(&fifo_data, &sim_data);
+
+	*burst = ilog2(fifo_data.burst >> 4);
+	*lwm = fifo_data.lwm >> 3;
+}
+
+static void
+nv30_update_arb(int *burst, int *lwm)
+{
+	unsigned int fifo_size, burst_size, graphics_lwm;
+
+	fifo_size = 2048;
+	burst_size = 512;
+	graphics_lwm = fifo_size - burst_size;
+
+	*burst = ilog2(burst_size >> 5);
+	*lwm = graphics_lwm >> 3;
+}
+
+void
+nouveau_calc_arb(struct drm_device *dev, int vclk, int bpp, int *burst, int *lwm)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	if (dev_priv->card_type < NV_30)
+		nv04_update_arb(dev, vclk, bpp, burst, lwm);
+	else if ((dev->pci_device & 0xfff0) == 0x0240 /*CHIPSET_C51*/ ||
+		 (dev->pci_device & 0xfff0) == 0x03d0 /*CHIPSET_C512*/) {
+		*burst = 128;
+		*lwm = 0x0480;
+	} else
+		nv30_update_arb(burst, lwm);
+}
+
+static int
+getMNP_single(struct drm_device *dev, struct pll_lims *pll_lim, int clk,
+	      struct nouveau_pll_vals *bestpv)
+{
+	/* Find M, N and P for a single stage PLL
+	 *
+	 * Note that some bioses (NV3x) have lookup tables of precomputed MNP
+	 * values, but we're too lazy to use those atm
+	 *
+	 * "clk" parameter in kHz
+	 * returns calculated clock
+	 */
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	int cv = dev_priv->vbios.chip_version;
+	int minvco = pll_lim->vco1.minfreq, maxvco = pll_lim->vco1.maxfreq;
+	int minM = pll_lim->vco1.min_m, maxM = pll_lim->vco1.max_m;
+	int minN = pll_lim->vco1.min_n, maxN = pll_lim->vco1.max_n;
+	int minU = pll_lim->vco1.min_inputfreq;
+	int maxU = pll_lim->vco1.max_inputfreq;
+	int minP = pll_lim->max_p ? pll_lim->min_p : 0;
+	int maxP = pll_lim->max_p ? pll_lim->max_p : pll_lim->max_usable_log2p;
+	int crystal = pll_lim->refclk;
+	int M, N, thisP, P;
+	int clkP, calcclk;
+	int delta, bestdelta = INT_MAX;
+	int bestclk = 0;
+
+	/* this division verified for nv20, nv18, nv28 (Haiku), and nv34 */
+	/* possibly correlated with introduction of 27MHz crystal */
+	if (dev_priv->card_type < NV_50) {
+		if (cv < 0x17 || cv == 0x1a || cv == 0x20) {
+			if (clk > 250000)
+				maxM = 6;
+			if (clk > 340000)
+				maxM = 2;
+		} else if (cv < 0x40) {
+			if (clk > 150000)
+				maxM = 6;
+			if (clk > 200000)
+				maxM = 4;
+			if (clk > 340000)
+				maxM = 2;
+		}
+	}
+
+	P = pll_lim->max_p ? maxP : (1 << maxP);
+	if ((clk * P) < minvco) {
+		minvco = clk * maxP;
+		maxvco = minvco * 2;
+	}
+
+	if (clk + clk/200 > maxvco)	/* +0.5% */
+		maxvco = clk + clk/200;
+
+	/* NV34 goes maxlog2P->0, NV20 goes 0->maxlog2P */
+	for (thisP = minP; thisP <= maxP; thisP++) {
+		P = pll_lim->max_p ? thisP : (1 << thisP);
+		clkP = clk * P;
+
+		if (clkP < minvco)
+			continue;
+		if (clkP > maxvco)
+			return bestclk;
+
+		for (M = minM; M <= maxM; M++) {
+			if (crystal/M < minU)
+				return bestclk;
+			if (crystal/M > maxU)
+				continue;
+
+			/* add crystal/2 to round better */
+			N = (clkP * M + crystal/2) / crystal;
+
+			if (N < minN)
+				continue;
+			if (N > maxN)
+				break;
+
+			/* more rounding additions */
+			calcclk = ((N * crystal + P/2) / P + M/2) / M;
+			delta = abs(calcclk - clk);
+			/* we do an exhaustive search rather than terminating
+			 * on an optimality condition...
+			 */
+			if (delta < bestdelta) {
+				bestdelta = delta;
+				bestclk = calcclk;
+				bestpv->N1 = N;
+				bestpv->M1 = M;
+				bestpv->log2P = thisP;
+				if (delta == 0)	/* except this one */
+					return bestclk;
+			}
+		}
+	}
+
+	return bestclk;
+}
+
+static int
+getMNP_double(struct drm_device *dev, struct pll_lims *pll_lim, int clk,
+	      struct nouveau_pll_vals *bestpv)
+{
+	/* Find M, N and P for a two stage PLL
+	 *
+	 * Note that some bioses (NV30+) have lookup tables of precomputed MNP
+	 * values, but we're too lazy to use those atm
+	 *
+	 * "clk" parameter in kHz
+	 * returns calculated clock
+	 */
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	int chip_version = dev_priv->vbios.chip_version;
+	int minvco1 = pll_lim->vco1.minfreq, maxvco1 = pll_lim->vco1.maxfreq;
+	int minvco2 = pll_lim->vco2.minfreq, maxvco2 = pll_lim->vco2.maxfreq;
+	int minU1 = pll_lim->vco1.min_inputfreq, minU2 = pll_lim->vco2.min_inputfreq;
+	int maxU1 = pll_lim->vco1.max_inputfreq, maxU2 = pll_lim->vco2.max_inputfreq;
+	int minM1 = pll_lim->vco1.min_m, maxM1 = pll_lim->vco1.max_m;
+	int minN1 = pll_lim->vco1.min_n, maxN1 = pll_lim->vco1.max_n;
+	int minM2 = pll_lim->vco2.min_m, maxM2 = pll_lim->vco2.max_m;
+	int minN2 = pll_lim->vco2.min_n, maxN2 = pll_lim->vco2.max_n;
+	int maxlog2P = pll_lim->max_usable_log2p;
+	int crystal = pll_lim->refclk;
+	bool fixedgain2 = (minM2 == maxM2 && minN2 == maxN2);
+	int M1, N1, M2, N2, log2P;
+	int clkP, calcclk1, calcclk2, calcclkout;
+	int delta, bestdelta = INT_MAX;
+	int bestclk = 0;
+
+	int vco2 = (maxvco2 - maxvco2/200) / 2;
+	for (log2P = 0; clk && log2P < maxlog2P && clk <= (vco2 >> log2P); log2P++)
+		;
+	clkP = clk << log2P;
+
+	if (maxvco2 < clk + clk/200)	/* +0.5% */
+		maxvco2 = clk + clk/200;
+
+	for (M1 = minM1; M1 <= maxM1; M1++) {
+		if (crystal/M1 < minU1)
+			return bestclk;
+		if (crystal/M1 > maxU1)
+			continue;
+
+		for (N1 = minN1; N1 <= maxN1; N1++) {
+			calcclk1 = crystal * N1 / M1;
+			if (calcclk1 < minvco1)
+				continue;
+			if (calcclk1 > maxvco1)
+				break;
+
+			for (M2 = minM2; M2 <= maxM2; M2++) {
+				if (calcclk1/M2 < minU2)
+					break;
+				if (calcclk1/M2 > maxU2)
+					continue;
+
+				/* add calcclk1/2 to round better */
+				N2 = (clkP * M2 + calcclk1/2) / calcclk1;
+				if (N2 < minN2)
+					continue;
+				if (N2 > maxN2)
+					break;
+
+				if (!fixedgain2) {
+					if (chip_version < 0x60)
+						if (N2/M2 < 4 || N2/M2 > 10)
+							continue;
+
+					calcclk2 = calcclk1 * N2 / M2;
+					if (calcclk2 < minvco2)
+						break;
+					if (calcclk2 > maxvco2)
+						continue;
+				} else
+					calcclk2 = calcclk1;
+
+				calcclkout = calcclk2 >> log2P;
+				delta = abs(calcclkout - clk);
+				/* we do an exhaustive search rather than terminating
+				 * on an optimality condition...
+				 */
+				if (delta < bestdelta) {
+					bestdelta = delta;
+					bestclk = calcclkout;
+					bestpv->N1 = N1;
+					bestpv->M1 = M1;
+					bestpv->N2 = N2;
+					bestpv->M2 = M2;
+					bestpv->log2P = log2P;
+					if (delta == 0)	/* except this one */
+						return bestclk;
+				}
+			}
+		}
+	}
+
+	return bestclk;
+}
+
+int
+nouveau_calc_pll_mnp(struct drm_device *dev, struct pll_lims *pll_lim, int clk,
+		     struct nouveau_pll_vals *pv)
+{
+	int outclk;
+
+	if (!pll_lim->vco2.maxfreq)
+		outclk = getMNP_single(dev, pll_lim, clk, pv);
+	else
+		outclk = getMNP_double(dev, pll_lim, clk, pv);
+
+	if (!outclk)
+		NV_ERROR(dev, "Could not find a compatible set of PLL values\n");
+
+	return outclk;
+}
diff --git a/drivers/gpu/drm/nouveau/nouveau_channel.c b/drivers/gpu/drm/nouveau/nouveau_channel.c
new file mode 100644
index 000000000000..6dfb425cbae9
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_channel.c
@@ -0,0 +1,449 @@
+/*
+ * Copyright 2005-2006 Stephane Marchesin
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include "drmP.h"
+#include "drm.h"
+#include "nouveau_drv.h"
+#include "nouveau_drm.h"
+#include "nouveau_dma.h"
+
+static int
+nouveau_channel_pushbuf_ctxdma_init(struct nouveau_channel *chan)
+{
+	struct drm_device *dev = chan->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_bo *pb = chan->pushbuf_bo;
+	struct nouveau_gpuobj *pushbuf = NULL;
+	int ret;
+
+	if (dev_priv->card_type >= NV_50) {
+		ret = nouveau_gpuobj_dma_new(chan, NV_CLASS_DMA_IN_MEMORY, 0,
+					     dev_priv->vm_end, NV_DMA_ACCESS_RO,
+					     NV_DMA_TARGET_AGP, &pushbuf);
+		chan->pushbuf_base = pb->bo.offset;
+	} else
+	if (pb->bo.mem.mem_type == TTM_PL_TT) {
+		ret = nouveau_gpuobj_gart_dma_new(chan, 0,
+						  dev_priv->gart_info.aper_size,
+						  NV_DMA_ACCESS_RO, &pushbuf,
+						  NULL);
+		chan->pushbuf_base = pb->bo.mem.mm_node->start << PAGE_SHIFT;
+	} else
+	if (dev_priv->card_type != NV_04) {
+		ret = nouveau_gpuobj_dma_new(chan, NV_CLASS_DMA_IN_MEMORY, 0,
+					     dev_priv->fb_available_size,
+					     NV_DMA_ACCESS_RO,
+					     NV_DMA_TARGET_VIDMEM, &pushbuf);
+		chan->pushbuf_base = pb->bo.mem.mm_node->start << PAGE_SHIFT;
+	} else {
+		/* NV04 cmdbuf hack, from original ddx.. not sure of it's
+		 * exact reason for existing :)  PCI access to cmdbuf in
+		 * VRAM.
+		 */
+		ret = nouveau_gpuobj_dma_new(chan, NV_CLASS_DMA_IN_MEMORY,
+					     drm_get_resource_start(dev, 1),
+					     dev_priv->fb_available_size,
+					     NV_DMA_ACCESS_RO,
+					     NV_DMA_TARGET_PCI, &pushbuf);
+		chan->pushbuf_base = pb->bo.mem.mm_node->start << PAGE_SHIFT;
+	}
+
+	ret = nouveau_gpuobj_ref_add(dev, chan, 0, pushbuf, &chan->pushbuf);
+	if (ret) {
+		NV_ERROR(dev, "Error referencing pushbuf ctxdma: %d\n", ret);
+		if (pushbuf != dev_priv->gart_info.sg_ctxdma)
+			nouveau_gpuobj_del(dev, &pushbuf);
+		return ret;
+	}
+
+	return 0;
+}
+
+static struct nouveau_bo *
+nouveau_channel_user_pushbuf_alloc(struct drm_device *dev)
+{
+	struct nouveau_bo *pushbuf = NULL;
+	int location, ret;
+
+	if (nouveau_vram_pushbuf)
+		location = TTM_PL_FLAG_VRAM;
+	else
+		location = TTM_PL_FLAG_TT;
+
+	ret = nouveau_bo_new(dev, NULL, 65536, 0, location, 0, 0x0000, false,
+			     true, &pushbuf);
+	if (ret) {
+		NV_ERROR(dev, "error allocating DMA push buffer: %d\n", ret);
+		return NULL;
+	}
+
+	ret = nouveau_bo_pin(pushbuf, location);
+	if (ret) {
+		NV_ERROR(dev, "error pinning DMA push buffer: %d\n", ret);
+		nouveau_bo_ref(NULL, &pushbuf);
+		return NULL;
+	}
+
+	return pushbuf;
+}
+
+/* allocates and initializes a fifo for user space consumption */
+int
+nouveau_channel_alloc(struct drm_device *dev, struct nouveau_channel **chan_ret,
+		      struct drm_file *file_priv,
+		      uint32_t vram_handle, uint32_t tt_handle)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
+	struct nouveau_fifo_engine *pfifo = &dev_priv->engine.fifo;
+	struct nouveau_channel *chan;
+	int channel, user;
+	int ret;
+
+	/*
+	 * Alright, here is the full story
+	 * Nvidia cards have multiple hw fifo contexts (praise them for that,
+	 * no complicated crash-prone context switches)
+	 * We allocate a new context for each app and let it write to it
+	 * directly (woo, full userspace command submission !)
+	 * When there are no more contexts, you lost
+	 */
+	for (channel = 0; channel < pfifo->channels; channel++) {
+		if (dev_priv->fifos[channel] == NULL)
+			break;
+	}
+
+	/* no more fifos. you lost. */
+	if (channel == pfifo->channels)
+		return -EINVAL;
+
+	dev_priv->fifos[channel] = kzalloc(sizeof(struct nouveau_channel),
+					   GFP_KERNEL);
+	if (!dev_priv->fifos[channel])
+		return -ENOMEM;
+	dev_priv->fifo_alloc_count++;
+	chan = dev_priv->fifos[channel];
+	INIT_LIST_HEAD(&chan->nvsw.vbl_wait);
+	INIT_LIST_HEAD(&chan->fence.pending);
+	chan->dev = dev;
+	chan->id = channel;
+	chan->file_priv = file_priv;
+	chan->vram_handle = vram_handle;
+	chan->gart_handle = tt_handle;
+
+	NV_INFO(dev, "Allocating FIFO number %d\n", channel);
+
+	/* Allocate DMA push buffer */
+	chan->pushbuf_bo = nouveau_channel_user_pushbuf_alloc(dev);
+	if (!chan->pushbuf_bo) {
+		ret = -ENOMEM;
+		NV_ERROR(dev, "pushbuf %d\n", ret);
+		nouveau_channel_free(chan);
+		return ret;
+	}
+
+	nouveau_dma_pre_init(chan);
+
+	/* Locate channel's user control regs */
+	if (dev_priv->card_type < NV_40)
+		user = NV03_USER(channel);
+	else
+	if (dev_priv->card_type < NV_50)
+		user = NV40_USER(channel);
+	else
+		user = NV50_USER(channel);
+
+	chan->user = ioremap(pci_resource_start(dev->pdev, 0) + user,
+								PAGE_SIZE);
+	if (!chan->user) {
+		NV_ERROR(dev, "ioremap of regs failed.\n");
+		nouveau_channel_free(chan);
+		return -ENOMEM;
+	}
+	chan->user_put = 0x40;
+	chan->user_get = 0x44;
+
+	/* Allocate space for per-channel fixed notifier memory */
+	ret = nouveau_notifier_init_channel(chan);
+	if (ret) {
+		NV_ERROR(dev, "ntfy %d\n", ret);
+		nouveau_channel_free(chan);
+		return ret;
+	}
+
+	/* Setup channel's default objects */
+	ret = nouveau_gpuobj_channel_init(chan, vram_handle, tt_handle);
+	if (ret) {
+		NV_ERROR(dev, "gpuobj %d\n", ret);
+		nouveau_channel_free(chan);
+		return ret;
+	}
+
+	/* Create a dma object for the push buffer */
+	ret = nouveau_channel_pushbuf_ctxdma_init(chan);
+	if (ret) {
+		NV_ERROR(dev, "pbctxdma %d\n", ret);
+		nouveau_channel_free(chan);
+		return ret;
+	}
+
+	/* disable the fifo caches */
+	pfifo->reassign(dev, false);
+
+	/* Create a graphics context for new channel */
+	ret = pgraph->create_context(chan);
+	if (ret) {
+		nouveau_channel_free(chan);
+		return ret;
+	}
+
+	/* Construct inital RAMFC for new channel */
+	ret = pfifo->create_context(chan);
+	if (ret) {
+		nouveau_channel_free(chan);
+		return ret;
+	}
+
+	pfifo->reassign(dev, true);
+
+	ret = nouveau_dma_init(chan);
+	if (!ret)
+		ret = nouveau_fence_init(chan);
+	if (ret) {
+		nouveau_channel_free(chan);
+		return ret;
+	}
+
+	nouveau_debugfs_channel_init(chan);
+
+	NV_INFO(dev, "%s: initialised FIFO %d\n", __func__, channel);
+	*chan_ret = chan;
+	return 0;
+}
+
+/* stops a fifo */
+void
+nouveau_channel_free(struct nouveau_channel *chan)
+{
+	struct drm_device *dev = chan->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
+	struct nouveau_fifo_engine *pfifo = &dev_priv->engine.fifo;
+	unsigned long flags;
+	int ret;
+
+	NV_INFO(dev, "%s: freeing fifo %d\n", __func__, chan->id);
+
+	nouveau_debugfs_channel_fini(chan);
+
+	/* Give outstanding push buffers a chance to complete */
+	spin_lock_irqsave(&chan->fence.lock, flags);
+	nouveau_fence_update(chan);
+	spin_unlock_irqrestore(&chan->fence.lock, flags);
+	if (chan->fence.sequence != chan->fence.sequence_ack) {
+		struct nouveau_fence *fence = NULL;
+
+		ret = nouveau_fence_new(chan, &fence, true);
+		if (ret == 0) {
+			ret = nouveau_fence_wait(fence, NULL, false, false);
+			nouveau_fence_unref((void *)&fence);
+		}
+
+		if (ret)
+			NV_ERROR(dev, "Failed to idle channel %d.\n", chan->id);
+	}
+
+	/* Ensure all outstanding fences are signaled.  They should be if the
+	 * above attempts at idling were OK, but if we failed this'll tell TTM
+	 * we're done with the buffers.
+	 */
+	nouveau_fence_fini(chan);
+
+	/* This will prevent pfifo from switching channels. */
+	pfifo->reassign(dev, false);
+
+	/* We want to give pgraph a chance to idle and get rid of all potential
+	 * errors. We need to do this before the lock, otherwise the irq handler
+	 * is unable to process them.
+	 */
+	if (pgraph->channel(dev) == chan)
+		nouveau_wait_for_idle(dev);
+
+	spin_lock_irqsave(&dev_priv->context_switch_lock, flags);
+
+	pgraph->fifo_access(dev, false);
+	if (pgraph->channel(dev) == chan)
+		pgraph->unload_context(dev);
+	pgraph->destroy_context(chan);
+	pgraph->fifo_access(dev, true);
+
+	if (pfifo->channel_id(dev) == chan->id) {
+		pfifo->disable(dev);
+		pfifo->unload_context(dev);
+		pfifo->enable(dev);
+	}
+	pfifo->destroy_context(chan);
+
+	pfifo->reassign(dev, true);
+
+	spin_unlock_irqrestore(&dev_priv->context_switch_lock, flags);
+
+	/* Release the channel's resources */
+	nouveau_gpuobj_ref_del(dev, &chan->pushbuf);
+	if (chan->pushbuf_bo) {
+		nouveau_bo_unpin(chan->pushbuf_bo);
+		nouveau_bo_ref(NULL, &chan->pushbuf_bo);
+	}
+	nouveau_gpuobj_channel_takedown(chan);
+	nouveau_notifier_takedown_channel(chan);
+	if (chan->user)
+		iounmap(chan->user);
+
+	dev_priv->fifos[chan->id] = NULL;
+	dev_priv->fifo_alloc_count--;
+	kfree(chan);
+}
+
+/* cleans up all the fifos from file_priv */
+void
+nouveau_channel_cleanup(struct drm_device *dev, struct drm_file *file_priv)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_engine *engine = &dev_priv->engine;
+	int i;
+
+	NV_DEBUG(dev, "clearing FIFO enables from file_priv\n");
+	for (i = 0; i < engine->fifo.channels; i++) {
+		struct nouveau_channel *chan = dev_priv->fifos[i];
+
+		if (chan && chan->file_priv == file_priv)
+			nouveau_channel_free(chan);
+	}
+}
+
+int
+nouveau_channel_owner(struct drm_device *dev, struct drm_file *file_priv,
+		      int channel)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_engine *engine = &dev_priv->engine;
+
+	if (channel >= engine->fifo.channels)
+		return 0;
+	if (dev_priv->fifos[channel] == NULL)
+		return 0;
+
+	return (dev_priv->fifos[channel]->file_priv == file_priv);
+}
+
+/***********************************
+ * ioctls wrapping the functions
+ ***********************************/
+
+static int
+nouveau_ioctl_fifo_alloc(struct drm_device *dev, void *data,
+			 struct drm_file *file_priv)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct drm_nouveau_channel_alloc *init = data;
+	struct nouveau_channel *chan;
+	int ret;
+
+	NOUVEAU_CHECK_INITIALISED_WITH_RETURN;
+
+	if (dev_priv->engine.graph.accel_blocked)
+		return -ENODEV;
+
+	if (init->fb_ctxdma_handle == ~0 || init->tt_ctxdma_handle == ~0)
+		return -EINVAL;
+
+	ret = nouveau_channel_alloc(dev, &chan, file_priv,
+				    init->fb_ctxdma_handle,
+				    init->tt_ctxdma_handle);
+	if (ret)
+		return ret;
+	init->channel  = chan->id;
+
+	if (chan->dma.ib_max)
+		init->pushbuf_domains = NOUVEAU_GEM_DOMAIN_VRAM |
+					NOUVEAU_GEM_DOMAIN_GART;
+	else if (chan->pushbuf_bo->bo.mem.mem_type == TTM_PL_VRAM)
+		init->pushbuf_domains = NOUVEAU_GEM_DOMAIN_VRAM;
+	else
+		init->pushbuf_domains = NOUVEAU_GEM_DOMAIN_GART;
+
+	init->subchan[0].handle = NvM2MF;
+	if (dev_priv->card_type < NV_50)
+		init->subchan[0].grclass = 0x0039;
+	else
+		init->subchan[0].grclass = 0x5039;
+	init->subchan[1].handle = NvSw;
+	init->subchan[1].grclass = NV_SW;
+	init->nr_subchan = 2;
+
+	/* Named memory object area */
+	ret = drm_gem_handle_create(file_priv, chan->notifier_bo->gem,
+				    &init->notifier_handle);
+	if (ret) {
+		nouveau_channel_free(chan);
+		return ret;
+	}
+
+	return 0;
+}
+
+static int
+nouveau_ioctl_fifo_free(struct drm_device *dev, void *data,
+			struct drm_file *file_priv)
+{
+	struct drm_nouveau_channel_free *cfree = data;
+	struct nouveau_channel *chan;
+
+	NOUVEAU_CHECK_INITIALISED_WITH_RETURN;
+	NOUVEAU_GET_USER_CHANNEL_WITH_RETURN(cfree->channel, file_priv, chan);
+
+	nouveau_channel_free(chan);
+	return 0;
+}
+
+/***********************************
+ * finally, the ioctl table
+ ***********************************/
+
+struct drm_ioctl_desc nouveau_ioctls[] = {
+	DRM_IOCTL_DEF(DRM_NOUVEAU_GETPARAM, nouveau_ioctl_getparam, DRM_AUTH),
+	DRM_IOCTL_DEF(DRM_NOUVEAU_SETPARAM, nouveau_ioctl_setparam, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY),
+	DRM_IOCTL_DEF(DRM_NOUVEAU_CHANNEL_ALLOC, nouveau_ioctl_fifo_alloc, DRM_AUTH),
+	DRM_IOCTL_DEF(DRM_NOUVEAU_CHANNEL_FREE, nouveau_ioctl_fifo_free, DRM_AUTH),
+	DRM_IOCTL_DEF(DRM_NOUVEAU_GROBJ_ALLOC, nouveau_ioctl_grobj_alloc, DRM_AUTH),
+	DRM_IOCTL_DEF(DRM_NOUVEAU_NOTIFIEROBJ_ALLOC, nouveau_ioctl_notifier_alloc, DRM_AUTH),
+	DRM_IOCTL_DEF(DRM_NOUVEAU_GPUOBJ_FREE, nouveau_ioctl_gpuobj_free, DRM_AUTH),
+	DRM_IOCTL_DEF(DRM_NOUVEAU_GEM_NEW, nouveau_gem_ioctl_new, DRM_AUTH),
+	DRM_IOCTL_DEF(DRM_NOUVEAU_GEM_PUSHBUF, nouveau_gem_ioctl_pushbuf, DRM_AUTH),
+	DRM_IOCTL_DEF(DRM_NOUVEAU_GEM_CPU_PREP, nouveau_gem_ioctl_cpu_prep, DRM_AUTH),
+	DRM_IOCTL_DEF(DRM_NOUVEAU_GEM_CPU_FINI, nouveau_gem_ioctl_cpu_fini, DRM_AUTH),
+	DRM_IOCTL_DEF(DRM_NOUVEAU_GEM_INFO, nouveau_gem_ioctl_info, DRM_AUTH),
+};
+
+int nouveau_max_ioctl = DRM_ARRAY_SIZE(nouveau_ioctls);
diff --git a/drivers/gpu/drm/nouveau/nouveau_connector.c b/drivers/gpu/drm/nouveau/nouveau_connector.c
new file mode 100644
index 000000000000..24327f468c4b
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_connector.c
@@ -0,0 +1,881 @@
+/*
+ * Copyright (C) 2008 Maarten Maathuis.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <acpi/button.h>
+
+#include "drmP.h"
+#include "drm_edid.h"
+#include "drm_crtc_helper.h"
+
+#include "nouveau_reg.h"
+#include "nouveau_drv.h"
+#include "nouveau_encoder.h"
+#include "nouveau_crtc.h"
+#include "nouveau_connector.h"
+#include "nouveau_hw.h"
+
+static inline struct drm_encoder_slave_funcs *
+get_slave_funcs(struct nouveau_encoder *enc)
+{
+	return to_encoder_slave(to_drm_encoder(enc))->slave_funcs;
+}
+
+static struct nouveau_encoder *
+find_encoder_by_type(struct drm_connector *connector, int type)
+{
+	struct drm_device *dev = connector->dev;
+	struct nouveau_encoder *nv_encoder;
+	struct drm_mode_object *obj;
+	int i, id;
+
+	for (i = 0; i < DRM_CONNECTOR_MAX_ENCODER; i++) {
+		id = connector->encoder_ids[i];
+		if (!id)
+			break;
+
+		obj = drm_mode_object_find(dev, id, DRM_MODE_OBJECT_ENCODER);
+		if (!obj)
+			continue;
+		nv_encoder = nouveau_encoder(obj_to_encoder(obj));
+
+		if (type == OUTPUT_ANY || nv_encoder->dcb->type == type)
+			return nv_encoder;
+	}
+
+	return NULL;
+}
+
+struct nouveau_connector *
+nouveau_encoder_connector_get(struct nouveau_encoder *encoder)
+{
+	struct drm_device *dev = to_drm_encoder(encoder)->dev;
+	struct drm_connector *drm_connector;
+
+	list_for_each_entry(drm_connector, &dev->mode_config.connector_list, head) {
+		if (drm_connector->encoder == to_drm_encoder(encoder))
+			return nouveau_connector(drm_connector);
+	}
+
+	return NULL;
+}
+
+
+static void
+nouveau_connector_destroy(struct drm_connector *drm_connector)
+{
+	struct nouveau_connector *nv_connector =
+		nouveau_connector(drm_connector);
+	struct drm_device *dev;
+
+	if (!nv_connector)
+		return;
+
+	dev = nv_connector->base.dev;
+	NV_DEBUG_KMS(dev, "\n");
+
+	kfree(nv_connector->edid);
+	drm_sysfs_connector_remove(drm_connector);
+	drm_connector_cleanup(drm_connector);
+	kfree(drm_connector);
+}
+
+static void
+nouveau_connector_ddc_prepare(struct drm_connector *connector, int *flags)
+{
+	struct drm_nouveau_private *dev_priv = connector->dev->dev_private;
+
+	if (dev_priv->card_type >= NV_50)
+		return;
+
+	*flags = 0;
+	if (NVLockVgaCrtcs(dev_priv->dev, false))
+		*flags |= 1;
+	if (nv_heads_tied(dev_priv->dev))
+		*flags |= 2;
+
+	if (*flags & 2)
+		NVSetOwner(dev_priv->dev, 0); /* necessary? */
+}
+
+static void
+nouveau_connector_ddc_finish(struct drm_connector *connector, int flags)
+{
+	struct drm_nouveau_private *dev_priv = connector->dev->dev_private;
+
+	if (dev_priv->card_type >= NV_50)
+		return;
+
+	if (flags & 2)
+		NVSetOwner(dev_priv->dev, 4);
+	if (flags & 1)
+		NVLockVgaCrtcs(dev_priv->dev, true);
+}
+
+static struct nouveau_i2c_chan *
+nouveau_connector_ddc_detect(struct drm_connector *connector,
+			     struct nouveau_encoder **pnv_encoder)
+{
+	struct drm_device *dev = connector->dev;
+	uint8_t out_buf[] = { 0x0, 0x0}, buf[2];
+	int ret, flags, i;
+
+	struct i2c_msg msgs[] = {
+		{
+			.addr = 0x50,
+			.flags = 0,
+			.len = 1,
+			.buf = out_buf,
+		},
+		{
+			.addr = 0x50,
+			.flags = I2C_M_RD,
+			.len = 1,
+			.buf = buf,
+		}
+	};
+
+	for (i = 0; i < DRM_CONNECTOR_MAX_ENCODER; i++) {
+		struct nouveau_i2c_chan *i2c = NULL;
+		struct nouveau_encoder *nv_encoder;
+		struct drm_mode_object *obj;
+		int id;
+
+		id = connector->encoder_ids[i];
+		if (!id)
+			break;
+
+		obj = drm_mode_object_find(dev, id, DRM_MODE_OBJECT_ENCODER);
+		if (!obj)
+			continue;
+		nv_encoder = nouveau_encoder(obj_to_encoder(obj));
+
+		if (nv_encoder->dcb->i2c_index < 0xf)
+			i2c = nouveau_i2c_find(dev, nv_encoder->dcb->i2c_index);
+		if (!i2c)
+			continue;
+
+		nouveau_connector_ddc_prepare(connector, &flags);
+		ret = i2c_transfer(&i2c->adapter, msgs, 2);
+		nouveau_connector_ddc_finish(connector, flags);
+
+		if (ret == 2) {
+			*pnv_encoder = nv_encoder;
+			return i2c;
+		}
+	}
+
+	return NULL;
+}
+
+static void
+nouveau_connector_set_encoder(struct drm_connector *connector,
+			      struct nouveau_encoder *nv_encoder)
+{
+	struct nouveau_connector *nv_connector = nouveau_connector(connector);
+	struct drm_nouveau_private *dev_priv = connector->dev->dev_private;
+	struct drm_device *dev = connector->dev;
+
+	if (nv_connector->detected_encoder == nv_encoder)
+		return;
+	nv_connector->detected_encoder = nv_encoder;
+
+	if (nv_encoder->dcb->type == OUTPUT_LVDS ||
+	    nv_encoder->dcb->type == OUTPUT_TMDS) {
+		connector->doublescan_allowed = false;
+		connector->interlace_allowed = false;
+	} else {
+		connector->doublescan_allowed = true;
+		if (dev_priv->card_type == NV_20 ||
+		   (dev_priv->card_type == NV_10 &&
+		    (dev->pci_device & 0x0ff0) != 0x0100 &&
+		    (dev->pci_device & 0x0ff0) != 0x0150))
+			/* HW is broken */
+			connector->interlace_allowed = false;
+		else
+			connector->interlace_allowed = true;
+	}
+
+	if (nv_connector->dcb->type == DCB_CONNECTOR_DVI_I) {
+		drm_connector_property_set_value(connector,
+			dev->mode_config.dvi_i_subconnector_property,
+			nv_encoder->dcb->type == OUTPUT_TMDS ?
+			DRM_MODE_SUBCONNECTOR_DVID :
+			DRM_MODE_SUBCONNECTOR_DVIA);
+	}
+}
+
+static enum drm_connector_status
+nouveau_connector_detect(struct drm_connector *connector)
+{
+	struct drm_device *dev = connector->dev;
+	struct nouveau_connector *nv_connector = nouveau_connector(connector);
+	struct nouveau_encoder *nv_encoder = NULL;
+	struct nouveau_i2c_chan *i2c;
+	int type, flags;
+
+	if (nv_connector->dcb->type == DCB_CONNECTOR_LVDS)
+		nv_encoder = find_encoder_by_type(connector, OUTPUT_LVDS);
+	if (nv_encoder && nv_connector->native_mode) {
+		unsigned status = connector_status_connected;
+
+#ifdef CONFIG_ACPI
+		if (!nouveau_ignorelid && !acpi_lid_open())
+			status = connector_status_unknown;
+#endif
+		nouveau_connector_set_encoder(connector, nv_encoder);
+		return status;
+	}
+
+	/* Cleanup the previous EDID block. */
+	if (nv_connector->edid) {
+		drm_mode_connector_update_edid_property(connector, NULL);
+		kfree(nv_connector->edid);
+		nv_connector->edid = NULL;
+	}
+
+	i2c = nouveau_connector_ddc_detect(connector, &nv_encoder);
+	if (i2c) {
+		nouveau_connector_ddc_prepare(connector, &flags);
+		nv_connector->edid = drm_get_edid(connector, &i2c->adapter);
+		nouveau_connector_ddc_finish(connector, flags);
+		drm_mode_connector_update_edid_property(connector,
+							nv_connector->edid);
+		if (!nv_connector->edid) {
+			NV_ERROR(dev, "DDC responded, but no EDID for %s\n",
+				 drm_get_connector_name(connector));
+			goto detect_analog;
+		}
+
+		if (nv_encoder->dcb->type == OUTPUT_DP &&
+		    !nouveau_dp_detect(to_drm_encoder(nv_encoder))) {
+			NV_ERROR(dev, "Detected %s, but failed init\n",
+				 drm_get_connector_name(connector));
+			return connector_status_disconnected;
+		}
+
+		/* Override encoder type for DVI-I based on whether EDID
+		 * says the display is digital or analog, both use the
+		 * same i2c channel so the value returned from ddc_detect
+		 * isn't necessarily correct.
+		 */
+		if (nv_connector->dcb->type == DCB_CONNECTOR_DVI_I) {
+			if (nv_connector->edid->input & DRM_EDID_INPUT_DIGITAL)
+				type = OUTPUT_TMDS;
+			else
+				type = OUTPUT_ANALOG;
+
+			nv_encoder = find_encoder_by_type(connector, type);
+			if (!nv_encoder) {
+				NV_ERROR(dev, "Detected %d encoder on %s, "
+					      "but no object!\n", type,
+					 drm_get_connector_name(connector));
+				return connector_status_disconnected;
+			}
+		}
+
+		nouveau_connector_set_encoder(connector, nv_encoder);
+		return connector_status_connected;
+	}
+
+detect_analog:
+	nv_encoder = find_encoder_by_type(connector, OUTPUT_ANALOG);
+	if (!nv_encoder)
+		nv_encoder = find_encoder_by_type(connector, OUTPUT_TV);
+	if (nv_encoder) {
+		struct drm_encoder *encoder = to_drm_encoder(nv_encoder);
+		struct drm_encoder_helper_funcs *helper =
+						encoder->helper_private;
+
+		if (helper->detect(encoder, connector) ==
+						connector_status_connected) {
+			nouveau_connector_set_encoder(connector, nv_encoder);
+			return connector_status_connected;
+		}
+
+	}
+
+	return connector_status_disconnected;
+}
+
+static void
+nouveau_connector_force(struct drm_connector *connector)
+{
+	struct nouveau_connector *nv_connector = nouveau_connector(connector);
+	struct nouveau_encoder *nv_encoder;
+	int type;
+
+	if (nv_connector->dcb->type == DCB_CONNECTOR_DVI_I) {
+		if (connector->force == DRM_FORCE_ON_DIGITAL)
+			type = OUTPUT_TMDS;
+		else
+			type = OUTPUT_ANALOG;
+	} else
+		type = OUTPUT_ANY;
+
+	nv_encoder = find_encoder_by_type(connector, type);
+	if (!nv_encoder) {
+		NV_ERROR(connector->dev, "can't find encoder to force %s on!\n",
+			 drm_get_connector_name(connector));
+		connector->status = connector_status_disconnected;
+		return;
+	}
+
+	nouveau_connector_set_encoder(connector, nv_encoder);
+}
+
+static int
+nouveau_connector_set_property(struct drm_connector *connector,
+			       struct drm_property *property, uint64_t value)
+{
+	struct nouveau_connector *nv_connector = nouveau_connector(connector);
+	struct nouveau_encoder *nv_encoder = nv_connector->detected_encoder;
+	struct drm_device *dev = connector->dev;
+	int ret;
+
+	/* Scaling mode */
+	if (property == dev->mode_config.scaling_mode_property) {
+		struct nouveau_crtc *nv_crtc = NULL;
+		bool modeset = false;
+
+		switch (value) {
+		case DRM_MODE_SCALE_NONE:
+		case DRM_MODE_SCALE_FULLSCREEN:
+		case DRM_MODE_SCALE_CENTER:
+		case DRM_MODE_SCALE_ASPECT:
+			break;
+		default:
+			return -EINVAL;
+		}
+
+		/* LVDS always needs gpu scaling */
+		if (nv_connector->dcb->type == DCB_CONNECTOR_LVDS &&
+		    value == DRM_MODE_SCALE_NONE)
+			return -EINVAL;
+
+		/* Changing between GPU and panel scaling requires a full
+		 * modeset
+		 */
+		if ((nv_connector->scaling_mode == DRM_MODE_SCALE_NONE) ||
+		    (value == DRM_MODE_SCALE_NONE))
+			modeset = true;
+		nv_connector->scaling_mode = value;
+
+		if (connector->encoder && connector->encoder->crtc)
+			nv_crtc = nouveau_crtc(connector->encoder->crtc);
+		if (!nv_crtc)
+			return 0;
+
+		if (modeset || !nv_crtc->set_scale) {
+			ret = drm_crtc_helper_set_mode(&nv_crtc->base,
+							&nv_crtc->base.mode,
+							nv_crtc->base.x,
+							nv_crtc->base.y, NULL);
+			if (!ret)
+				return -EINVAL;
+		} else {
+			ret = nv_crtc->set_scale(nv_crtc, value, true);
+			if (ret)
+				return ret;
+		}
+
+		return 0;
+	}
+
+	/* Dithering */
+	if (property == dev->mode_config.dithering_mode_property) {
+		struct nouveau_crtc *nv_crtc = NULL;
+
+		if (value == DRM_MODE_DITHERING_ON)
+			nv_connector->use_dithering = true;
+		else
+			nv_connector->use_dithering = false;
+
+		if (connector->encoder && connector->encoder->crtc)
+			nv_crtc = nouveau_crtc(connector->encoder->crtc);
+
+		if (!nv_crtc || !nv_crtc->set_dither)
+			return 0;
+
+		return nv_crtc->set_dither(nv_crtc, nv_connector->use_dithering,
+					   true);
+	}
+
+	if (nv_encoder && nv_encoder->dcb->type == OUTPUT_TV)
+		return get_slave_funcs(nv_encoder)->
+			set_property(to_drm_encoder(nv_encoder), connector, property, value);
+
+	return -EINVAL;
+}
+
+static struct drm_display_mode *
+nouveau_connector_native_mode(struct nouveau_connector *connector)
+{
+	struct drm_device *dev = connector->base.dev;
+	struct drm_display_mode *mode, *largest = NULL;
+	int high_w = 0, high_h = 0, high_v = 0;
+
+	/* Use preferred mode if there is one.. */
+	list_for_each_entry(mode, &connector->base.probed_modes, head) {
+		if (mode->type & DRM_MODE_TYPE_PREFERRED) {
+			NV_DEBUG_KMS(dev, "native mode from preferred\n");
+			return drm_mode_duplicate(dev, mode);
+		}
+	}
+
+	/* Otherwise, take the resolution with the largest width, then height,
+	 * then vertical refresh
+	 */
+	list_for_each_entry(mode, &connector->base.probed_modes, head) {
+		if (mode->hdisplay < high_w)
+			continue;
+
+		if (mode->hdisplay == high_w && mode->vdisplay < high_h)
+			continue;
+
+		if (mode->hdisplay == high_w && mode->vdisplay == high_h &&
+		    mode->vrefresh < high_v)
+			continue;
+
+		high_w = mode->hdisplay;
+		high_h = mode->vdisplay;
+		high_v = mode->vrefresh;
+		largest = mode;
+	}
+
+	NV_DEBUG_KMS(dev, "native mode from largest: %dx%d@%d\n",
+		      high_w, high_h, high_v);
+	return largest ? drm_mode_duplicate(dev, largest) : NULL;
+}
+
+struct moderec {
+	int hdisplay;
+	int vdisplay;
+};
+
+static struct moderec scaler_modes[] = {
+	{ 1920, 1200 },
+	{ 1920, 1080 },
+	{ 1680, 1050 },
+	{ 1600, 1200 },
+	{ 1400, 1050 },
+	{ 1280, 1024 },
+	{ 1280, 960 },
+	{ 1152, 864 },
+	{ 1024, 768 },
+	{ 800, 600 },
+	{ 720, 400 },
+	{ 640, 480 },
+	{ 640, 400 },
+	{ 640, 350 },
+	{}
+};
+
+static int
+nouveau_connector_scaler_modes_add(struct drm_connector *connector)
+{
+	struct nouveau_connector *nv_connector = nouveau_connector(connector);
+	struct drm_display_mode *native = nv_connector->native_mode, *m;
+	struct drm_device *dev = connector->dev;
+	struct moderec *mode = &scaler_modes[0];
+	int modes = 0;
+
+	if (!native)
+		return 0;
+
+	while (mode->hdisplay) {
+		if (mode->hdisplay <= native->hdisplay &&
+		    mode->vdisplay <= native->vdisplay) {
+			m = drm_cvt_mode(dev, mode->hdisplay, mode->vdisplay,
+					 drm_mode_vrefresh(native), false,
+					 false, false);
+			if (!m)
+				continue;
+
+			m->type |= DRM_MODE_TYPE_DRIVER;
+
+			drm_mode_probed_add(connector, m);
+			modes++;
+		}
+
+		mode++;
+	}
+
+	return modes;
+}
+
+static int
+nouveau_connector_get_modes(struct drm_connector *connector)
+{
+	struct drm_device *dev = connector->dev;
+	struct nouveau_connector *nv_connector = nouveau_connector(connector);
+	struct nouveau_encoder *nv_encoder = nv_connector->detected_encoder;
+	int ret = 0;
+
+	/* If we're not LVDS, destroy the previous native mode, the attached
+	 * monitor could have changed.
+	 */
+	if (nv_connector->dcb->type != DCB_CONNECTOR_LVDS &&
+	    nv_connector->native_mode) {
+		drm_mode_destroy(dev, nv_connector->native_mode);
+		nv_connector->native_mode = NULL;
+	}
+
+	if (nv_connector->edid)
+		ret = drm_add_edid_modes(connector, nv_connector->edid);
+
+	/* Find the native mode if this is a digital panel, if we didn't
+	 * find any modes through DDC previously add the native mode to
+	 * the list of modes.
+	 */
+	if (!nv_connector->native_mode)
+		nv_connector->native_mode =
+			nouveau_connector_native_mode(nv_connector);
+	if (ret == 0 && nv_connector->native_mode) {
+		struct drm_display_mode *mode;
+
+		mode = drm_mode_duplicate(dev, nv_connector->native_mode);
+		drm_mode_probed_add(connector, mode);
+		ret = 1;
+	}
+
+	if (nv_encoder->dcb->type == OUTPUT_TV)
+		ret = get_slave_funcs(nv_encoder)->
+			get_modes(to_drm_encoder(nv_encoder), connector);
+
+	if (nv_encoder->dcb->type == OUTPUT_LVDS)
+		ret += nouveau_connector_scaler_modes_add(connector);
+
+	return ret;
+}
+
+static int
+nouveau_connector_mode_valid(struct drm_connector *connector,
+			     struct drm_display_mode *mode)
+{
+	struct drm_nouveau_private *dev_priv = connector->dev->dev_private;
+	struct nouveau_connector *nv_connector = nouveau_connector(connector);
+	struct nouveau_encoder *nv_encoder = nv_connector->detected_encoder;
+	unsigned min_clock = 25000, max_clock = min_clock;
+	unsigned clock = mode->clock;
+
+	switch (nv_encoder->dcb->type) {
+	case OUTPUT_LVDS:
+		BUG_ON(!nv_connector->native_mode);
+		if (mode->hdisplay > nv_connector->native_mode->hdisplay ||
+		    mode->vdisplay > nv_connector->native_mode->vdisplay)
+			return MODE_PANEL;
+
+		min_clock = 0;
+		max_clock = 400000;
+		break;
+	case OUTPUT_TMDS:
+		if ((dev_priv->card_type >= NV_50 && !nouveau_duallink) ||
+		    (dev_priv->card_type < NV_50 &&
+		     !nv_encoder->dcb->duallink_possible))
+			max_clock = 165000;
+		else
+			max_clock = 330000;
+		break;
+	case OUTPUT_ANALOG:
+		max_clock = nv_encoder->dcb->crtconf.maxfreq;
+		if (!max_clock)
+			max_clock = 350000;
+		break;
+	case OUTPUT_TV:
+		return get_slave_funcs(nv_encoder)->
+			mode_valid(to_drm_encoder(nv_encoder), mode);
+	case OUTPUT_DP:
+		if (nv_encoder->dp.link_bw == DP_LINK_BW_2_7)
+			max_clock = nv_encoder->dp.link_nr * 270000;
+		else
+			max_clock = nv_encoder->dp.link_nr * 162000;
+
+		clock *= 3;
+		break;
+	default:
+		BUG_ON(1);
+		return MODE_BAD;
+	}
+
+	if (clock < min_clock)
+		return MODE_CLOCK_LOW;
+
+	if (clock > max_clock)
+		return MODE_CLOCK_HIGH;
+
+	return MODE_OK;
+}
+
+static struct drm_encoder *
+nouveau_connector_best_encoder(struct drm_connector *connector)
+{
+	struct nouveau_connector *nv_connector = nouveau_connector(connector);
+
+	if (nv_connector->detected_encoder)
+		return to_drm_encoder(nv_connector->detected_encoder);
+
+	return NULL;
+}
+
+static const struct drm_connector_helper_funcs
+nouveau_connector_helper_funcs = {
+	.get_modes = nouveau_connector_get_modes,
+	.mode_valid = nouveau_connector_mode_valid,
+	.best_encoder = nouveau_connector_best_encoder,
+};
+
+static const struct drm_connector_funcs
+nouveau_connector_funcs = {
+	.dpms = drm_helper_connector_dpms,
+	.save = NULL,
+	.restore = NULL,
+	.detect = nouveau_connector_detect,
+	.destroy = nouveau_connector_destroy,
+	.fill_modes = drm_helper_probe_single_connector_modes,
+	.set_property = nouveau_connector_set_property,
+	.force = nouveau_connector_force
+};
+
+static int
+nouveau_connector_create_lvds(struct drm_device *dev,
+			      struct drm_connector *connector)
+{
+	struct nouveau_connector *nv_connector = nouveau_connector(connector);
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_i2c_chan *i2c = NULL;
+	struct nouveau_encoder *nv_encoder;
+	struct drm_display_mode native, *mode, *temp;
+	bool dummy, if_is_24bit = false;
+	int ret, flags;
+
+	nv_encoder = find_encoder_by_type(connector, OUTPUT_LVDS);
+	if (!nv_encoder)
+		return -ENODEV;
+
+	ret = nouveau_bios_parse_lvds_table(dev, 0, &dummy, &if_is_24bit);
+	if (ret) {
+		NV_ERROR(dev, "Error parsing LVDS table, disabling LVDS\n");
+		return ret;
+	}
+	nv_connector->use_dithering = !if_is_24bit;
+
+	/* Firstly try getting EDID over DDC, if allowed and I2C channel
+	 * is available.
+	 */
+	if (!dev_priv->vbios.fp_no_ddc && nv_encoder->dcb->i2c_index < 0xf)
+		i2c = nouveau_i2c_find(dev, nv_encoder->dcb->i2c_index);
+
+	if (i2c) {
+		nouveau_connector_ddc_prepare(connector, &flags);
+		nv_connector->edid = drm_get_edid(connector, &i2c->adapter);
+		nouveau_connector_ddc_finish(connector, flags);
+	}
+
+	/* If no EDID found above, and the VBIOS indicates a hardcoded
+	 * modeline is avalilable for the panel, set it as the panel's
+	 * native mode and exit.
+	 */
+	if (!nv_connector->edid && nouveau_bios_fp_mode(dev, &native) &&
+	     (nv_encoder->dcb->lvdsconf.use_straps_for_mode ||
+	      dev_priv->vbios.fp_no_ddc)) {
+		nv_connector->native_mode = drm_mode_duplicate(dev, &native);
+		goto out;
+	}
+
+	/* Still nothing, some VBIOS images have a hardcoded EDID block
+	 * stored for the panel stored in them.
+	 */
+	if (!nv_connector->edid && !nv_connector->native_mode &&
+	    !dev_priv->vbios.fp_no_ddc) {
+		struct edid *edid =
+			(struct edid *)nouveau_bios_embedded_edid(dev);
+		if (edid) {
+			nv_connector->edid = kmalloc(EDID_LENGTH, GFP_KERNEL);
+			*(nv_connector->edid) = *edid;
+		}
+	}
+
+	if (!nv_connector->edid)
+		goto out;
+
+	/* We didn't find/use a panel mode from the VBIOS, so parse the EDID
+	 * block and look for the preferred mode there.
+	 */
+	ret = drm_add_edid_modes(connector, nv_connector->edid);
+	if (ret == 0)
+		goto out;
+	nv_connector->detected_encoder = nv_encoder;
+	nv_connector->native_mode = nouveau_connector_native_mode(nv_connector);
+	list_for_each_entry_safe(mode, temp, &connector->probed_modes, head)
+		drm_mode_remove(connector, mode);
+
+out:
+	if (!nv_connector->native_mode) {
+		NV_ERROR(dev, "LVDS present in DCB table, but couldn't "
+			      "determine its native mode.  Disabling.\n");
+		return -ENODEV;
+	}
+
+	drm_mode_connector_update_edid_property(connector, nv_connector->edid);
+	return 0;
+}
+
+int
+nouveau_connector_create(struct drm_device *dev,
+			 struct dcb_connector_table_entry *dcb)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_connector *nv_connector = NULL;
+	struct drm_connector *connector;
+	struct drm_encoder *encoder;
+	int ret, type;
+
+	NV_DEBUG_KMS(dev, "\n");
+
+	switch (dcb->type) {
+	case DCB_CONNECTOR_NONE:
+		return 0;
+	case DCB_CONNECTOR_VGA:
+		NV_INFO(dev, "Detected a VGA connector\n");
+		type = DRM_MODE_CONNECTOR_VGA;
+		break;
+	case DCB_CONNECTOR_TV_0:
+	case DCB_CONNECTOR_TV_1:
+	case DCB_CONNECTOR_TV_3:
+		NV_INFO(dev, "Detected a TV connector\n");
+		type = DRM_MODE_CONNECTOR_TV;
+		break;
+	case DCB_CONNECTOR_DVI_I:
+		NV_INFO(dev, "Detected a DVI-I connector\n");
+		type = DRM_MODE_CONNECTOR_DVII;
+		break;
+	case DCB_CONNECTOR_DVI_D:
+		NV_INFO(dev, "Detected a DVI-D connector\n");
+		type = DRM_MODE_CONNECTOR_DVID;
+		break;
+	case DCB_CONNECTOR_HDMI_0:
+	case DCB_CONNECTOR_HDMI_1:
+		NV_INFO(dev, "Detected a HDMI connector\n");
+		type = DRM_MODE_CONNECTOR_HDMIA;
+		break;
+	case DCB_CONNECTOR_LVDS:
+		NV_INFO(dev, "Detected a LVDS connector\n");
+		type = DRM_MODE_CONNECTOR_LVDS;
+		break;
+	case DCB_CONNECTOR_DP:
+		NV_INFO(dev, "Detected a DisplayPort connector\n");
+		type = DRM_MODE_CONNECTOR_DisplayPort;
+		break;
+	case DCB_CONNECTOR_eDP:
+		NV_INFO(dev, "Detected an eDP connector\n");
+		type = DRM_MODE_CONNECTOR_eDP;
+		break;
+	default:
+		NV_ERROR(dev, "unknown connector type: 0x%02x!!\n", dcb->type);
+		return -EINVAL;
+	}
+
+	nv_connector = kzalloc(sizeof(*nv_connector), GFP_KERNEL);
+	if (!nv_connector)
+		return -ENOMEM;
+	nv_connector->dcb = dcb;
+	connector = &nv_connector->base;
+
+	/* defaults, will get overridden in detect() */
+	connector->interlace_allowed = false;
+	connector->doublescan_allowed = false;
+
+	drm_connector_init(dev, connector, &nouveau_connector_funcs, type);
+	drm_connector_helper_add(connector, &nouveau_connector_helper_funcs);
+
+	/* attach encoders */
+	list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) {
+		struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+
+		if (nv_encoder->dcb->connector != dcb->index)
+			continue;
+
+		if (get_slave_funcs(nv_encoder))
+			get_slave_funcs(nv_encoder)->create_resources(encoder, connector);
+
+		drm_mode_connector_attach_encoder(connector, encoder);
+	}
+
+	if (!connector->encoder_ids[0]) {
+		NV_WARN(dev, "  no encoders, ignoring\n");
+		drm_connector_cleanup(connector);
+		kfree(connector);
+		return 0;
+	}
+
+	/* Init DVI-I specific properties */
+	if (dcb->type == DCB_CONNECTOR_DVI_I) {
+		drm_mode_create_dvi_i_properties(dev);
+		drm_connector_attach_property(connector, dev->mode_config.dvi_i_subconnector_property, 0);
+		drm_connector_attach_property(connector, dev->mode_config.dvi_i_select_subconnector_property, 0);
+	}
+
+	if (dcb->type != DCB_CONNECTOR_LVDS)
+		nv_connector->use_dithering = false;
+
+	switch (dcb->type) {
+	case DCB_CONNECTOR_VGA:
+		if (dev_priv->card_type >= NV_50) {
+			drm_connector_attach_property(connector,
+					dev->mode_config.scaling_mode_property,
+					nv_connector->scaling_mode);
+		}
+		/* fall-through */
+	case DCB_CONNECTOR_TV_0:
+	case DCB_CONNECTOR_TV_1:
+	case DCB_CONNECTOR_TV_3:
+		nv_connector->scaling_mode = DRM_MODE_SCALE_NONE;
+		break;
+	default:
+		nv_connector->scaling_mode = DRM_MODE_SCALE_FULLSCREEN;
+
+		drm_connector_attach_property(connector,
+				dev->mode_config.scaling_mode_property,
+				nv_connector->scaling_mode);
+		drm_connector_attach_property(connector,
+				dev->mode_config.dithering_mode_property,
+				nv_connector->use_dithering ?
+				DRM_MODE_DITHERING_ON : DRM_MODE_DITHERING_OFF);
+		break;
+	}
+
+	drm_sysfs_connector_add(connector);
+
+	if (dcb->type == DCB_CONNECTOR_LVDS) {
+		ret = nouveau_connector_create_lvds(dev, connector);
+		if (ret) {
+			connector->funcs->destroy(connector);
+			return ret;
+		}
+	}
+
+	return 0;
+}
diff --git a/drivers/gpu/drm/nouveau/nouveau_connector.h b/drivers/gpu/drm/nouveau/nouveau_connector.h
new file mode 100644
index 000000000000..4ef38abc2d9c
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_connector.h
@@ -0,0 +1,55 @@
+/*
+ * Copyright (C) 2008 Maarten Maathuis.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __NOUVEAU_CONNECTOR_H__
+#define __NOUVEAU_CONNECTOR_H__
+
+#include "drm_edid.h"
+#include "nouveau_i2c.h"
+
+struct nouveau_connector {
+	struct drm_connector base;
+
+	struct dcb_connector_table_entry *dcb;
+
+	int scaling_mode;
+	bool use_dithering;
+
+	struct nouveau_encoder *detected_encoder;
+	struct edid *edid;
+	struct drm_display_mode *native_mode;
+};
+
+static inline struct nouveau_connector *nouveau_connector(
+						struct drm_connector *con)
+{
+	return container_of(con, struct nouveau_connector, base);
+}
+
+int nouveau_connector_create(struct drm_device *,
+			     struct dcb_connector_table_entry *);
+
+#endif /* __NOUVEAU_CONNECTOR_H__ */
diff --git a/drivers/gpu/drm/nouveau/nouveau_crtc.h b/drivers/gpu/drm/nouveau/nouveau_crtc.h
new file mode 100644
index 000000000000..49fa7b2d257e
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_crtc.h
@@ -0,0 +1,95 @@
+/*
+ * Copyright (C) 2008 Maarten Maathuis.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __NOUVEAU_CRTC_H__
+#define __NOUVEAU_CRTC_H__
+
+struct nouveau_crtc {
+	struct drm_crtc base;
+
+	int index;
+
+	struct drm_display_mode *mode;
+
+	uint32_t dpms_saved_fp_control;
+	uint32_t fp_users;
+	int saturation;
+	int sharpness;
+	int last_dpms;
+
+	struct {
+		int cpp;
+		bool blanked;
+		uint32_t offset;
+		uint32_t tile_flags;
+	} fb;
+
+	struct {
+		struct nouveau_bo *nvbo;
+		bool visible;
+		uint32_t offset;
+		void (*set_offset)(struct nouveau_crtc *, uint32_t offset);
+		void (*set_pos)(struct nouveau_crtc *, int x, int y);
+		void (*hide)(struct nouveau_crtc *, bool update);
+		void (*show)(struct nouveau_crtc *, bool update);
+	} cursor;
+
+	struct {
+		struct nouveau_bo *nvbo;
+		uint16_t r[256];
+		uint16_t g[256];
+		uint16_t b[256];
+		int depth;
+	} lut;
+
+	int (*set_dither)(struct nouveau_crtc *crtc, bool on, bool update);
+	int (*set_scale)(struct nouveau_crtc *crtc, int mode, bool update);
+};
+
+static inline struct nouveau_crtc *nouveau_crtc(struct drm_crtc *crtc)
+{
+	return container_of(crtc, struct nouveau_crtc, base);
+}
+
+static inline struct drm_crtc *to_drm_crtc(struct nouveau_crtc *crtc)
+{
+	return &crtc->base;
+}
+
+int nv50_crtc_create(struct drm_device *dev, int index);
+int nv50_cursor_init(struct nouveau_crtc *);
+void nv50_cursor_fini(struct nouveau_crtc *);
+int nv50_crtc_cursor_set(struct drm_crtc *drm_crtc, struct drm_file *file_priv,
+			 uint32_t buffer_handle, uint32_t width,
+			 uint32_t height);
+int nv50_crtc_cursor_move(struct drm_crtc *drm_crtc, int x, int y);
+
+int nv04_cursor_init(struct nouveau_crtc *);
+
+struct nouveau_connector *
+nouveau_crtc_connector_get(struct nouveau_crtc *crtc);
+
+#endif /* __NOUVEAU_CRTC_H__ */
diff --git a/drivers/gpu/drm/nouveau/nouveau_debugfs.c b/drivers/gpu/drm/nouveau/nouveau_debugfs.c
new file mode 100644
index 000000000000..8ff9ef5d4b47
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_debugfs.c
@@ -0,0 +1,179 @@
+/*
+ * Copyright (C) 2009 Red Hat <bskeggs@redhat.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+/*
+ * Authors:
+ *  Ben Skeggs <bskeggs@redhat.com>
+ */
+
+#include <linux/debugfs.h>
+
+#include "drmP.h"
+#include "nouveau_drv.h"
+
+static int
+nouveau_debugfs_channel_info(struct seq_file *m, void *data)
+{
+	struct drm_info_node *node = (struct drm_info_node *) m->private;
+	struct nouveau_channel *chan = node->info_ent->data;
+
+	seq_printf(m, "channel id    : %d\n", chan->id);
+
+	seq_printf(m, "cpu fifo state:\n");
+	seq_printf(m, "          base: 0x%08x\n", chan->pushbuf_base);
+	seq_printf(m, "           max: 0x%08x\n", chan->dma.max << 2);
+	seq_printf(m, "           cur: 0x%08x\n", chan->dma.cur << 2);
+	seq_printf(m, "           put: 0x%08x\n", chan->dma.put << 2);
+	seq_printf(m, "          free: 0x%08x\n", chan->dma.free << 2);
+	if (chan->dma.ib_max) {
+		seq_printf(m, "        ib max: 0x%08x\n", chan->dma.ib_max);
+		seq_printf(m, "        ib put: 0x%08x\n", chan->dma.ib_put);
+		seq_printf(m, "       ib free: 0x%08x\n", chan->dma.ib_free);
+	}
+
+	seq_printf(m, "gpu fifo state:\n");
+	seq_printf(m, "           get: 0x%08x\n",
+					nvchan_rd32(chan, chan->user_get));
+	seq_printf(m, "           put: 0x%08x\n",
+					nvchan_rd32(chan, chan->user_put));
+	if (chan->dma.ib_max) {
+		seq_printf(m, "        ib get: 0x%08x\n",
+			   nvchan_rd32(chan, 0x88));
+		seq_printf(m, "        ib put: 0x%08x\n",
+			   nvchan_rd32(chan, 0x8c));
+	}
+
+	seq_printf(m, "last fence    : %d\n", chan->fence.sequence);
+	seq_printf(m, "last signalled: %d\n", chan->fence.sequence_ack);
+	return 0;
+}
+
+int
+nouveau_debugfs_channel_init(struct nouveau_channel *chan)
+{
+	struct drm_nouveau_private *dev_priv = chan->dev->dev_private;
+	struct drm_minor *minor = chan->dev->primary;
+	int ret;
+
+	if (!dev_priv->debugfs.channel_root) {
+		dev_priv->debugfs.channel_root =
+			debugfs_create_dir("channel", minor->debugfs_root);
+		if (!dev_priv->debugfs.channel_root)
+			return -ENOENT;
+	}
+
+	snprintf(chan->debugfs.name, 32, "%d", chan->id);
+	chan->debugfs.info.name = chan->debugfs.name;
+	chan->debugfs.info.show = nouveau_debugfs_channel_info;
+	chan->debugfs.info.driver_features = 0;
+	chan->debugfs.info.data = chan;
+
+	ret = drm_debugfs_create_files(&chan->debugfs.info, 1,
+				       dev_priv->debugfs.channel_root,
+				       chan->dev->primary);
+	if (ret == 0)
+		chan->debugfs.active = true;
+	return ret;
+}
+
+void
+nouveau_debugfs_channel_fini(struct nouveau_channel *chan)
+{
+	struct drm_nouveau_private *dev_priv = chan->dev->dev_private;
+
+	if (!chan->debugfs.active)
+		return;
+
+	drm_debugfs_remove_files(&chan->debugfs.info, 1, chan->dev->primary);
+	chan->debugfs.active = false;
+
+	if (chan == dev_priv->channel) {
+		debugfs_remove(dev_priv->debugfs.channel_root);
+		dev_priv->debugfs.channel_root = NULL;
+	}
+}
+
+static int
+nouveau_debugfs_chipset_info(struct seq_file *m, void *data)
+{
+	struct drm_info_node *node = (struct drm_info_node *) m->private;
+	struct drm_minor *minor = node->minor;
+	struct drm_device *dev = minor->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t ppci_0;
+
+	ppci_0 = nv_rd32(dev, dev_priv->chipset >= 0x40 ? 0x88000 : 0x1800);
+
+	seq_printf(m, "PMC_BOOT_0: 0x%08x\n", nv_rd32(dev, NV03_PMC_BOOT_0));
+	seq_printf(m, "PCI ID    : 0x%04x:0x%04x\n",
+		   ppci_0 & 0xffff, ppci_0 >> 16);
+	return 0;
+}
+
+static int
+nouveau_debugfs_memory_info(struct seq_file *m, void *data)
+{
+	struct drm_info_node *node = (struct drm_info_node *) m->private;
+	struct drm_minor *minor = node->minor;
+	struct drm_device *dev = minor->dev;
+
+	seq_printf(m, "VRAM total: %dKiB\n",
+		   (int)(nouveau_mem_fb_amount(dev) >> 10));
+	return 0;
+}
+
+static int
+nouveau_debugfs_vbios_image(struct seq_file *m, void *data)
+{
+	struct drm_info_node *node = (struct drm_info_node *) m->private;
+	struct drm_nouveau_private *dev_priv = node->minor->dev->dev_private;
+	int i;
+
+	for (i = 0; i < dev_priv->vbios.length; i++)
+		seq_printf(m, "%c", dev_priv->vbios.data[i]);
+	return 0;
+}
+
+static struct drm_info_list nouveau_debugfs_list[] = {
+	{ "chipset", nouveau_debugfs_chipset_info, 0, NULL },
+	{ "memory", nouveau_debugfs_memory_info, 0, NULL },
+	{ "vbios.rom", nouveau_debugfs_vbios_image, 0, NULL },
+};
+#define NOUVEAU_DEBUGFS_ENTRIES ARRAY_SIZE(nouveau_debugfs_list)
+
+int
+nouveau_debugfs_init(struct drm_minor *minor)
+{
+	drm_debugfs_create_files(nouveau_debugfs_list, NOUVEAU_DEBUGFS_ENTRIES,
+				 minor->debugfs_root, minor);
+	return 0;
+}
+
+void
+nouveau_debugfs_takedown(struct drm_minor *minor)
+{
+	drm_debugfs_remove_files(nouveau_debugfs_list, NOUVEAU_DEBUGFS_ENTRIES,
+				 minor);
+}
diff --git a/drivers/gpu/drm/nouveau/nouveau_display.c b/drivers/gpu/drm/nouveau/nouveau_display.c
new file mode 100644
index 000000000000..cf1c5c0a0abe
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_display.c
@@ -0,0 +1,112 @@
+/*
+ * Copyright (C) 2008 Maarten Maathuis.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "drmP.h"
+#include "drm_crtc_helper.h"
+#include "nouveau_drv.h"
+#include "nouveau_fb.h"
+#include "nouveau_fbcon.h"
+
+static void
+nouveau_user_framebuffer_destroy(struct drm_framebuffer *drm_fb)
+{
+	struct nouveau_framebuffer *fb = nouveau_framebuffer(drm_fb);
+	struct drm_device *dev = drm_fb->dev;
+
+	if (drm_fb->fbdev)
+		nouveau_fbcon_remove(dev, drm_fb);
+
+	if (fb->nvbo)
+		drm_gem_object_unreference_unlocked(fb->nvbo->gem);
+
+	drm_framebuffer_cleanup(drm_fb);
+	kfree(fb);
+}
+
+static int
+nouveau_user_framebuffer_create_handle(struct drm_framebuffer *drm_fb,
+				       struct drm_file *file_priv,
+				       unsigned int *handle)
+{
+	struct nouveau_framebuffer *fb = nouveau_framebuffer(drm_fb);
+
+	return drm_gem_handle_create(file_priv, fb->nvbo->gem, handle);
+}
+
+static const struct drm_framebuffer_funcs nouveau_framebuffer_funcs = {
+	.destroy = nouveau_user_framebuffer_destroy,
+	.create_handle = nouveau_user_framebuffer_create_handle,
+};
+
+struct drm_framebuffer *
+nouveau_framebuffer_create(struct drm_device *dev, struct nouveau_bo *nvbo,
+			   struct drm_mode_fb_cmd *mode_cmd)
+{
+	struct nouveau_framebuffer *fb;
+	int ret;
+
+	fb = kzalloc(sizeof(struct nouveau_framebuffer), GFP_KERNEL);
+	if (!fb)
+		return NULL;
+
+	ret = drm_framebuffer_init(dev, &fb->base, &nouveau_framebuffer_funcs);
+	if (ret) {
+		kfree(fb);
+		return NULL;
+	}
+
+	drm_helper_mode_fill_fb_struct(&fb->base, mode_cmd);
+
+	fb->nvbo = nvbo;
+	return &fb->base;
+}
+
+static struct drm_framebuffer *
+nouveau_user_framebuffer_create(struct drm_device *dev,
+				struct drm_file *file_priv,
+				struct drm_mode_fb_cmd *mode_cmd)
+{
+	struct drm_framebuffer *fb;
+	struct drm_gem_object *gem;
+
+	gem = drm_gem_object_lookup(dev, file_priv, mode_cmd->handle);
+	if (!gem)
+		return NULL;
+
+	fb = nouveau_framebuffer_create(dev, nouveau_gem_object(gem), mode_cmd);
+	if (!fb) {
+		drm_gem_object_unreference(gem);
+		return NULL;
+	}
+
+	return fb;
+}
+
+const struct drm_mode_config_funcs nouveau_mode_config_funcs = {
+	.fb_create = nouveau_user_framebuffer_create,
+	.fb_changed = nouveau_fbcon_probe,
+};
+
diff --git a/drivers/gpu/drm/nouveau/nouveau_dma.c b/drivers/gpu/drm/nouveau/nouveau_dma.c
new file mode 100644
index 000000000000..c8482a108a78
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_dma.c
@@ -0,0 +1,348 @@
+/*
+ * Copyright (C) 2007 Ben Skeggs.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "drmP.h"
+#include "drm.h"
+#include "nouveau_drv.h"
+#include "nouveau_dma.h"
+
+void
+nouveau_dma_pre_init(struct nouveau_channel *chan)
+{
+	struct drm_nouveau_private *dev_priv = chan->dev->dev_private;
+	struct nouveau_bo *pushbuf = chan->pushbuf_bo;
+
+	if (dev_priv->card_type == NV_50) {
+		const int ib_size = pushbuf->bo.mem.size / 2;
+
+		chan->dma.ib_base = (pushbuf->bo.mem.size - ib_size) >> 2;
+		chan->dma.ib_max = (ib_size / 8) - 1;
+		chan->dma.ib_put = 0;
+		chan->dma.ib_free = chan->dma.ib_max - chan->dma.ib_put;
+
+		chan->dma.max = (pushbuf->bo.mem.size - ib_size) >> 2;
+	} else {
+		chan->dma.max  = (pushbuf->bo.mem.size >> 2) - 2;
+	}
+
+	chan->dma.put  = 0;
+	chan->dma.cur  = chan->dma.put;
+	chan->dma.free = chan->dma.max - chan->dma.cur;
+}
+
+int
+nouveau_dma_init(struct nouveau_channel *chan)
+{
+	struct drm_device *dev = chan->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_gpuobj *m2mf = NULL;
+	struct nouveau_gpuobj *nvsw = NULL;
+	int ret, i;
+
+	/* Create NV_MEMORY_TO_MEMORY_FORMAT for buffer moves */
+	ret = nouveau_gpuobj_gr_new(chan, dev_priv->card_type < NV_50 ?
+				    0x0039 : 0x5039, &m2mf);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_ref_add(dev, chan, NvM2MF, m2mf, NULL);
+	if (ret)
+		return ret;
+
+	/* Create an NV_SW object for various sync purposes */
+	ret = nouveau_gpuobj_sw_new(chan, NV_SW, &nvsw);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_ref_add(dev, chan, NvSw, nvsw, NULL);
+	if (ret)
+		return ret;
+
+	/* NV_MEMORY_TO_MEMORY_FORMAT requires a notifier object */
+	ret = nouveau_notifier_alloc(chan, NvNotify0, 32, &chan->m2mf_ntfy);
+	if (ret)
+		return ret;
+
+	/* Map push buffer */
+	ret = nouveau_bo_map(chan->pushbuf_bo);
+	if (ret)
+		return ret;
+
+	/* Map M2MF notifier object - fbcon. */
+	if (drm_core_check_feature(dev, DRIVER_MODESET)) {
+		ret = nouveau_bo_map(chan->notifier_bo);
+		if (ret)
+			return ret;
+	}
+
+	/* Insert NOPS for NOUVEAU_DMA_SKIPS */
+	ret = RING_SPACE(chan, NOUVEAU_DMA_SKIPS);
+	if (ret)
+		return ret;
+
+	for (i = 0; i < NOUVEAU_DMA_SKIPS; i++)
+		OUT_RING(chan, 0);
+
+	/* Initialise NV_MEMORY_TO_MEMORY_FORMAT */
+	ret = RING_SPACE(chan, 4);
+	if (ret)
+		return ret;
+	BEGIN_RING(chan, NvSubM2MF, NV_MEMORY_TO_MEMORY_FORMAT_NAME, 1);
+	OUT_RING(chan, NvM2MF);
+	BEGIN_RING(chan, NvSubM2MF, NV_MEMORY_TO_MEMORY_FORMAT_DMA_NOTIFY, 1);
+	OUT_RING(chan, NvNotify0);
+
+	/* Initialise NV_SW */
+	ret = RING_SPACE(chan, 2);
+	if (ret)
+		return ret;
+	BEGIN_RING(chan, NvSubSw, 0, 1);
+	OUT_RING(chan, NvSw);
+
+	/* Sit back and pray the channel works.. */
+	FIRE_RING(chan);
+
+	return 0;
+}
+
+void
+OUT_RINGp(struct nouveau_channel *chan, const void *data, unsigned nr_dwords)
+{
+	bool is_iomem;
+	u32 *mem = ttm_kmap_obj_virtual(&chan->pushbuf_bo->kmap, &is_iomem);
+	mem = &mem[chan->dma.cur];
+	if (is_iomem)
+		memcpy_toio((void __force __iomem *)mem, data, nr_dwords * 4);
+	else
+		memcpy(mem, data, nr_dwords * 4);
+	chan->dma.cur += nr_dwords;
+}
+
+/* Fetch and adjust GPU GET pointer
+ *
+ * Returns:
+ *  value >= 0, the adjusted GET pointer
+ *  -EINVAL if GET pointer currently outside main push buffer
+ *  -EBUSY if timeout exceeded
+ */
+static inline int
+READ_GET(struct nouveau_channel *chan, uint32_t *prev_get, uint32_t *timeout)
+{
+	uint32_t val;
+
+	val = nvchan_rd32(chan, chan->user_get);
+
+	/* reset counter as long as GET is still advancing, this is
+	 * to avoid misdetecting a GPU lockup if the GPU happens to
+	 * just be processing an operation that takes a long time
+	 */
+	if (val != *prev_get) {
+		*prev_get = val;
+		*timeout = 0;
+	}
+
+	if ((++*timeout & 0xff) == 0) {
+		DRM_UDELAY(1);
+		if (*timeout > 100000)
+			return -EBUSY;
+	}
+
+	if (val < chan->pushbuf_base ||
+	    val > chan->pushbuf_base + (chan->dma.max << 2))
+		return -EINVAL;
+
+	return (val - chan->pushbuf_base) >> 2;
+}
+
+void
+nv50_dma_push(struct nouveau_channel *chan, struct nouveau_bo *bo,
+	      int delta, int length)
+{
+	struct nouveau_bo *pb = chan->pushbuf_bo;
+	uint64_t offset = bo->bo.offset + delta;
+	int ip = (chan->dma.ib_put * 2) + chan->dma.ib_base;
+
+	BUG_ON(chan->dma.ib_free < 1);
+	nouveau_bo_wr32(pb, ip++, lower_32_bits(offset));
+	nouveau_bo_wr32(pb, ip++, upper_32_bits(offset) | length << 8);
+
+	chan->dma.ib_put = (chan->dma.ib_put + 1) & chan->dma.ib_max;
+	nvchan_wr32(chan, 0x8c, chan->dma.ib_put);
+	chan->dma.ib_free--;
+}
+
+static int
+nv50_dma_push_wait(struct nouveau_channel *chan, int count)
+{
+	uint32_t cnt = 0, prev_get = 0;
+
+	while (chan->dma.ib_free < count) {
+		uint32_t get = nvchan_rd32(chan, 0x88);
+		if (get != prev_get) {
+			prev_get = get;
+			cnt = 0;
+		}
+
+		if ((++cnt & 0xff) == 0) {
+			DRM_UDELAY(1);
+			if (cnt > 100000)
+				return -EBUSY;
+		}
+
+		chan->dma.ib_free = get - chan->dma.ib_put;
+		if (chan->dma.ib_free <= 0)
+			chan->dma.ib_free += chan->dma.ib_max + 1;
+	}
+
+	return 0;
+}
+
+static int
+nv50_dma_wait(struct nouveau_channel *chan, int slots, int count)
+{
+	uint32_t cnt = 0, prev_get = 0;
+	int ret;
+
+	ret = nv50_dma_push_wait(chan, slots + 1);
+	if (unlikely(ret))
+		return ret;
+
+	while (chan->dma.free < count) {
+		int get = READ_GET(chan, &prev_get, &cnt);
+		if (unlikely(get < 0)) {
+			if (get == -EINVAL)
+				continue;
+
+			return get;
+		}
+
+		if (get <= chan->dma.cur) {
+			chan->dma.free = chan->dma.max - chan->dma.cur;
+			if (chan->dma.free >= count)
+				break;
+
+			FIRE_RING(chan);
+			do {
+				get = READ_GET(chan, &prev_get, &cnt);
+				if (unlikely(get < 0)) {
+					if (get == -EINVAL)
+						continue;
+					return get;
+				}
+			} while (get == 0);
+			chan->dma.cur = 0;
+			chan->dma.put = 0;
+		}
+
+		chan->dma.free = get - chan->dma.cur - 1;
+	}
+
+	return 0;
+}
+
+int
+nouveau_dma_wait(struct nouveau_channel *chan, int slots, int size)
+{
+	uint32_t prev_get = 0, cnt = 0;
+	int get;
+
+	if (chan->dma.ib_max)
+		return nv50_dma_wait(chan, slots, size);
+
+	while (chan->dma.free < size) {
+		get = READ_GET(chan, &prev_get, &cnt);
+		if (unlikely(get == -EBUSY))
+			return -EBUSY;
+
+		/* loop until we have a usable GET pointer.  the value
+		 * we read from the GPU may be outside the main ring if
+		 * PFIFO is processing a buffer called from the main ring,
+		 * discard these values until something sensible is seen.
+		 *
+		 * the other case we discard GET is while the GPU is fetching
+		 * from the SKIPS area, so the code below doesn't have to deal
+		 * with some fun corner cases.
+		 */
+		if (unlikely(get == -EINVAL) || get < NOUVEAU_DMA_SKIPS)
+			continue;
+
+		if (get <= chan->dma.cur) {
+			/* engine is fetching behind us, or is completely
+			 * idle (GET == PUT) so we have free space up until
+			 * the end of the push buffer
+			 *
+			 * we can only hit that path once per call due to
+			 * looping back to the beginning of the push buffer,
+			 * we'll hit the fetching-ahead-of-us path from that
+			 * point on.
+			 *
+			 * the *one* exception to that rule is if we read
+			 * GET==PUT, in which case the below conditional will
+			 * always succeed and break us out of the wait loop.
+			 */
+			chan->dma.free = chan->dma.max - chan->dma.cur;
+			if (chan->dma.free >= size)
+				break;
+
+			/* not enough space left at the end of the push buffer,
+			 * instruct the GPU to jump back to the start right
+			 * after processing the currently pending commands.
+			 */
+			OUT_RING(chan, chan->pushbuf_base | 0x20000000);
+
+			/* wait for GET to depart from the skips area.
+			 * prevents writing GET==PUT and causing a race
+			 * condition that causes us to think the GPU is
+			 * idle when it's not.
+			 */
+			do {
+				get = READ_GET(chan, &prev_get, &cnt);
+				if (unlikely(get == -EBUSY))
+					return -EBUSY;
+				if (unlikely(get == -EINVAL))
+					continue;
+			} while (get <= NOUVEAU_DMA_SKIPS);
+			WRITE_PUT(NOUVEAU_DMA_SKIPS);
+
+			/* we're now submitting commands at the start of
+			 * the push buffer.
+			 */
+			chan->dma.cur  =
+			chan->dma.put  = NOUVEAU_DMA_SKIPS;
+		}
+
+		/* engine fetching ahead of us, we have space up until the
+		 * current GET pointer.  the "- 1" is to ensure there's
+		 * space left to emit a jump back to the beginning of the
+		 * push buffer if we require it.  we can never get GET == PUT
+		 * here, so this is safe.
+		 */
+		chan->dma.free = get - chan->dma.cur - 1;
+	}
+
+	return 0;
+}
+
diff --git a/drivers/gpu/drm/nouveau/nouveau_dma.h b/drivers/gpu/drm/nouveau/nouveau_dma.h
new file mode 100644
index 000000000000..8b05c15866d5
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_dma.h
@@ -0,0 +1,166 @@
+/*
+ * Copyright (C) 2007 Ben Skeggs.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __NOUVEAU_DMA_H__
+#define __NOUVEAU_DMA_H__
+
+#ifndef NOUVEAU_DMA_DEBUG
+#define NOUVEAU_DMA_DEBUG 0
+#endif
+
+void nv50_dma_push(struct nouveau_channel *, struct nouveau_bo *,
+		   int delta, int length);
+
+/*
+ * There's a hw race condition where you can't jump to your PUT offset,
+ * to avoid this we jump to offset + SKIPS and fill the difference with
+ * NOPs.
+ *
+ * xf86-video-nv configures the DMA fetch size to 32 bytes, and uses
+ * a SKIPS value of 8.  Lets assume that the race condition is to do
+ * with writing into the fetch area, we configure a fetch size of 128
+ * bytes so we need a larger SKIPS value.
+ */
+#define NOUVEAU_DMA_SKIPS (128 / 4)
+
+/* Hardcoded object assignments to subchannels (subchannel id). */
+enum {
+	NvSubM2MF	= 0,
+	NvSubSw		= 1,
+	NvSub2D		= 2,
+	NvSubCtxSurf2D  = 2,
+	NvSubGdiRect    = 3,
+	NvSubImageBlit  = 4
+};
+
+/* Object handles. */
+enum {
+	NvM2MF		= 0x80000001,
+	NvDmaFB		= 0x80000002,
+	NvDmaTT		= 0x80000003,
+	NvDmaVRAM	= 0x80000004,
+	NvDmaGART	= 0x80000005,
+	NvNotify0       = 0x80000006,
+	Nv2D		= 0x80000007,
+	NvCtxSurf2D	= 0x80000008,
+	NvRop		= 0x80000009,
+	NvImagePatt	= 0x8000000a,
+	NvClipRect	= 0x8000000b,
+	NvGdiRect	= 0x8000000c,
+	NvImageBlit	= 0x8000000d,
+	NvSw		= 0x8000000e,
+
+	/* G80+ display objects */
+	NvEvoVRAM	= 0x01000000,
+	NvEvoFB16	= 0x01000001,
+	NvEvoFB32	= 0x01000002
+};
+
+#define NV_MEMORY_TO_MEMORY_FORMAT                                    0x00000039
+#define NV_MEMORY_TO_MEMORY_FORMAT_NAME                               0x00000000
+#define NV_MEMORY_TO_MEMORY_FORMAT_SET_REF                            0x00000050
+#define NV_MEMORY_TO_MEMORY_FORMAT_NOP                                0x00000100
+#define NV_MEMORY_TO_MEMORY_FORMAT_NOTIFY                             0x00000104
+#define NV_MEMORY_TO_MEMORY_FORMAT_NOTIFY_STYLE_WRITE                 0x00000000
+#define NV_MEMORY_TO_MEMORY_FORMAT_NOTIFY_STYLE_WRITE_LE_AWAKEN       0x00000001
+#define NV_MEMORY_TO_MEMORY_FORMAT_DMA_NOTIFY                         0x00000180
+#define NV_MEMORY_TO_MEMORY_FORMAT_DMA_SOURCE                         0x00000184
+#define NV_MEMORY_TO_MEMORY_FORMAT_OFFSET_IN                          0x0000030c
+
+#define NV50_MEMORY_TO_MEMORY_FORMAT                                  0x00005039
+#define NV50_MEMORY_TO_MEMORY_FORMAT_UNK200                           0x00000200
+#define NV50_MEMORY_TO_MEMORY_FORMAT_UNK21C                           0x0000021c
+#define NV50_MEMORY_TO_MEMORY_FORMAT_OFFSET_IN_HIGH                   0x00000238
+#define NV50_MEMORY_TO_MEMORY_FORMAT_OFFSET_OUT_HIGH                  0x0000023c
+
+static __must_check inline int
+RING_SPACE(struct nouveau_channel *chan, int size)
+{
+	int ret;
+
+	ret = nouveau_dma_wait(chan, 1, size);
+	if (ret)
+		return ret;
+
+	chan->dma.free -= size;
+	return 0;
+}
+
+static inline void
+OUT_RING(struct nouveau_channel *chan, int data)
+{
+	if (NOUVEAU_DMA_DEBUG) {
+		NV_INFO(chan->dev, "Ch%d/0x%08x: 0x%08x\n",
+			chan->id, chan->dma.cur << 2, data);
+	}
+
+	nouveau_bo_wr32(chan->pushbuf_bo, chan->dma.cur++, data);
+}
+
+extern void
+OUT_RINGp(struct nouveau_channel *chan, const void *data, unsigned nr_dwords);
+
+static inline void
+BEGIN_RING(struct nouveau_channel *chan, int subc, int mthd, int size)
+{
+	OUT_RING(chan, (subc << 13) | (size << 18) | mthd);
+}
+
+#define WRITE_PUT(val) do {                                                    \
+	DRM_MEMORYBARRIER();                                                   \
+	nouveau_bo_rd32(chan->pushbuf_bo, 0);                                  \
+	nvchan_wr32(chan, chan->user_put, ((val) << 2) + chan->pushbuf_base);  \
+} while (0)
+
+static inline void
+FIRE_RING(struct nouveau_channel *chan)
+{
+	if (NOUVEAU_DMA_DEBUG) {
+		NV_INFO(chan->dev, "Ch%d/0x%08x: PUSH!\n",
+			chan->id, chan->dma.cur << 2);
+	}
+
+	if (chan->dma.cur == chan->dma.put)
+		return;
+	chan->accel_done = true;
+
+	if (chan->dma.ib_max) {
+		nv50_dma_push(chan, chan->pushbuf_bo, chan->dma.put << 2,
+			      (chan->dma.cur - chan->dma.put) << 2);
+	} else {
+		WRITE_PUT(chan->dma.cur);
+	}
+
+	chan->dma.put = chan->dma.cur;
+}
+
+static inline void
+WIND_RING(struct nouveau_channel *chan)
+{
+	chan->dma.cur = chan->dma.put;
+}
+
+#endif
diff --git a/drivers/gpu/drm/nouveau/nouveau_dp.c b/drivers/gpu/drm/nouveau/nouveau_dp.c
new file mode 100644
index 000000000000..f954ad93e81f
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_dp.c
@@ -0,0 +1,575 @@
+/*
+ * Copyright 2009 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include "drmP.h"
+#include "nouveau_drv.h"
+#include "nouveau_i2c.h"
+#include "nouveau_encoder.h"
+
+static int
+auxch_rd(struct drm_encoder *encoder, int address, uint8_t *buf, int size)
+{
+	struct drm_device *dev = encoder->dev;
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	struct nouveau_i2c_chan *auxch;
+	int ret;
+
+	auxch = nouveau_i2c_find(dev, nv_encoder->dcb->i2c_index);
+	if (!auxch)
+		return -ENODEV;
+
+	ret = nouveau_dp_auxch(auxch, 9, address, buf, size);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static int
+auxch_wr(struct drm_encoder *encoder, int address, uint8_t *buf, int size)
+{
+	struct drm_device *dev = encoder->dev;
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	struct nouveau_i2c_chan *auxch;
+	int ret;
+
+	auxch = nouveau_i2c_find(dev, nv_encoder->dcb->i2c_index);
+	if (!auxch)
+		return -ENODEV;
+
+	ret = nouveau_dp_auxch(auxch, 8, address, buf, size);
+	return ret;
+}
+
+static int
+nouveau_dp_lane_count_set(struct drm_encoder *encoder, uint8_t cmd)
+{
+	struct drm_device *dev = encoder->dev;
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	uint32_t tmp;
+	int or = nv_encoder->or, link = !(nv_encoder->dcb->sorconf.link & 1);
+
+	tmp  = nv_rd32(dev, NV50_SOR_DP_CTRL(or, link));
+	tmp &= ~(NV50_SOR_DP_CTRL_ENHANCED_FRAME_ENABLED |
+		 NV50_SOR_DP_CTRL_LANE_MASK);
+	tmp |= ((1 << (cmd & DP_LANE_COUNT_MASK)) - 1) << 16;
+	if (cmd & DP_LANE_COUNT_ENHANCED_FRAME_EN)
+		tmp |= NV50_SOR_DP_CTRL_ENHANCED_FRAME_ENABLED;
+	nv_wr32(dev, NV50_SOR_DP_CTRL(or, link), tmp);
+
+	return auxch_wr(encoder, DP_LANE_COUNT_SET, &cmd, 1);
+}
+
+static int
+nouveau_dp_link_bw_set(struct drm_encoder *encoder, uint8_t cmd)
+{
+	struct drm_device *dev = encoder->dev;
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	uint32_t tmp;
+	int reg = 0x614300 + (nv_encoder->or * 0x800);
+
+	tmp  = nv_rd32(dev, reg);
+	tmp &= 0xfff3ffff;
+	if (cmd == DP_LINK_BW_2_7)
+		tmp |= 0x00040000;
+	nv_wr32(dev, reg, tmp);
+
+	return auxch_wr(encoder, DP_LINK_BW_SET, &cmd, 1);
+}
+
+static int
+nouveau_dp_link_train_set(struct drm_encoder *encoder, int pattern)
+{
+	struct drm_device *dev = encoder->dev;
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	uint32_t tmp;
+	uint8_t cmd;
+	int or = nv_encoder->or, link = !(nv_encoder->dcb->sorconf.link & 1);
+	int ret;
+
+	tmp  = nv_rd32(dev, NV50_SOR_DP_CTRL(or, link));
+	tmp &= ~NV50_SOR_DP_CTRL_TRAINING_PATTERN;
+	tmp |= (pattern << 24);
+	nv_wr32(dev, NV50_SOR_DP_CTRL(or, link), tmp);
+
+	ret = auxch_rd(encoder, DP_TRAINING_PATTERN_SET, &cmd, 1);
+	if (ret)
+		return ret;
+	cmd &= ~DP_TRAINING_PATTERN_MASK;
+	cmd |= (pattern & DP_TRAINING_PATTERN_MASK);
+	return auxch_wr(encoder, DP_TRAINING_PATTERN_SET, &cmd, 1);
+}
+
+static int
+nouveau_dp_max_voltage_swing(struct drm_encoder *encoder)
+{
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	struct drm_device *dev = encoder->dev;
+	struct bit_displayport_encoder_table_entry *dpse;
+	struct bit_displayport_encoder_table *dpe;
+	int i, dpe_headerlen, max_vs = 0;
+
+	dpe = nouveau_bios_dp_table(dev, nv_encoder->dcb, &dpe_headerlen);
+	if (!dpe)
+		return false;
+	dpse = (void *)((char *)dpe + dpe_headerlen);
+
+	for (i = 0; i < dpe_headerlen; i++, dpse++) {
+		if (dpse->vs_level > max_vs)
+			max_vs = dpse->vs_level;
+	}
+
+	return max_vs;
+}
+
+static int
+nouveau_dp_max_pre_emphasis(struct drm_encoder *encoder, int vs)
+{
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	struct drm_device *dev = encoder->dev;
+	struct bit_displayport_encoder_table_entry *dpse;
+	struct bit_displayport_encoder_table *dpe;
+	int i, dpe_headerlen, max_pre = 0;
+
+	dpe = nouveau_bios_dp_table(dev, nv_encoder->dcb, &dpe_headerlen);
+	if (!dpe)
+		return false;
+	dpse = (void *)((char *)dpe + dpe_headerlen);
+
+	for (i = 0; i < dpe_headerlen; i++, dpse++) {
+		if (dpse->vs_level != vs)
+			continue;
+
+		if (dpse->pre_level > max_pre)
+			max_pre = dpse->pre_level;
+	}
+
+	return max_pre;
+}
+
+static bool
+nouveau_dp_link_train_adjust(struct drm_encoder *encoder, uint8_t *config)
+{
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	struct drm_device *dev = encoder->dev;
+	struct bit_displayport_encoder_table_entry *dpse;
+	struct bit_displayport_encoder_table *dpe;
+	int ret, i, dpe_headerlen, vs = 0, pre = 0;
+	uint8_t request[2];
+
+	dpe = nouveau_bios_dp_table(dev, nv_encoder->dcb, &dpe_headerlen);
+	if (!dpe)
+		return false;
+	dpse = (void *)((char *)dpe + dpe_headerlen);
+
+	ret = auxch_rd(encoder, DP_ADJUST_REQUEST_LANE0_1, request, 2);
+	if (ret)
+		return false;
+
+	NV_DEBUG_KMS(dev, "\t\tadjust 0x%02x 0x%02x\n", request[0], request[1]);
+
+	/* Keep all lanes at the same level.. */
+	for (i = 0; i < nv_encoder->dp.link_nr; i++) {
+		int lane_req = (request[i >> 1] >> ((i & 1) << 2)) & 0xf;
+		int lane_vs = lane_req & 3;
+		int lane_pre = (lane_req >> 2) & 3;
+
+		if (lane_vs > vs)
+			vs = lane_vs;
+		if (lane_pre > pre)
+			pre = lane_pre;
+	}
+
+	if (vs >= nouveau_dp_max_voltage_swing(encoder)) {
+		vs  = nouveau_dp_max_voltage_swing(encoder);
+		vs |= 4;
+	}
+
+	if (pre >= nouveau_dp_max_pre_emphasis(encoder, vs & 3)) {
+		pre  = nouveau_dp_max_pre_emphasis(encoder, vs & 3);
+		pre |= 4;
+	}
+
+	/* Update the configuration for all lanes.. */
+	for (i = 0; i < nv_encoder->dp.link_nr; i++)
+		config[i] = (pre << 3) | vs;
+
+	return true;
+}
+
+static bool
+nouveau_dp_link_train_commit(struct drm_encoder *encoder, uint8_t *config)
+{
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	struct drm_device *dev = encoder->dev;
+	struct bit_displayport_encoder_table_entry *dpse;
+	struct bit_displayport_encoder_table *dpe;
+	int or = nv_encoder->or, link = !(nv_encoder->dcb->sorconf.link & 1);
+	int dpe_headerlen, ret, i;
+
+	NV_DEBUG_KMS(dev, "\t\tconfig 0x%02x 0x%02x 0x%02x 0x%02x\n",
+		 config[0], config[1], config[2], config[3]);
+
+	dpe = nouveau_bios_dp_table(dev, nv_encoder->dcb, &dpe_headerlen);
+	if (!dpe)
+		return false;
+	dpse = (void *)((char *)dpe + dpe_headerlen);
+
+	for (i = 0; i < dpe->record_nr; i++, dpse++) {
+		if (dpse->vs_level == (config[0] & 3) &&
+		    dpse->pre_level == ((config[0] >> 3) & 3))
+			break;
+	}
+	BUG_ON(i == dpe->record_nr);
+
+	for (i = 0; i < nv_encoder->dp.link_nr; i++) {
+		const int shift[4] = { 16, 8, 0, 24 };
+		uint32_t mask = 0xff << shift[i];
+		uint32_t reg0, reg1, reg2;
+
+		reg0  = nv_rd32(dev, NV50_SOR_DP_UNK118(or, link)) & ~mask;
+		reg0 |= (dpse->reg0 << shift[i]);
+		reg1  = nv_rd32(dev, NV50_SOR_DP_UNK120(or, link)) & ~mask;
+		reg1 |= (dpse->reg1 << shift[i]);
+		reg2  = nv_rd32(dev, NV50_SOR_DP_UNK130(or, link)) & 0xffff00ff;
+		reg2 |= (dpse->reg2 << 8);
+		nv_wr32(dev, NV50_SOR_DP_UNK118(or, link), reg0);
+		nv_wr32(dev, NV50_SOR_DP_UNK120(or, link), reg1);
+		nv_wr32(dev, NV50_SOR_DP_UNK130(or, link), reg2);
+	}
+
+	ret = auxch_wr(encoder, DP_TRAINING_LANE0_SET, config, 4);
+	if (ret)
+		return false;
+
+	return true;
+}
+
+bool
+nouveau_dp_link_train(struct drm_encoder *encoder)
+{
+	struct drm_device *dev = encoder->dev;
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	uint8_t config[4];
+	uint8_t status[3];
+	bool cr_done, cr_max_vs, eq_done;
+	int ret = 0, i, tries, voltage;
+
+	NV_DEBUG_KMS(dev, "link training!!\n");
+train:
+	cr_done = eq_done = false;
+
+	/* set link configuration */
+	NV_DEBUG_KMS(dev, "\tbegin train: bw %d, lanes %d\n",
+		 nv_encoder->dp.link_bw, nv_encoder->dp.link_nr);
+
+	ret = nouveau_dp_link_bw_set(encoder, nv_encoder->dp.link_bw);
+	if (ret)
+		return false;
+
+	config[0] = nv_encoder->dp.link_nr;
+	if (nv_encoder->dp.dpcd_version >= 0x11)
+		config[0] |= DP_LANE_COUNT_ENHANCED_FRAME_EN;
+
+	ret = nouveau_dp_lane_count_set(encoder, config[0]);
+	if (ret)
+		return false;
+
+	/* clock recovery */
+	NV_DEBUG_KMS(dev, "\tbegin cr\n");
+	ret = nouveau_dp_link_train_set(encoder, DP_TRAINING_PATTERN_1);
+	if (ret)
+		goto stop;
+
+	tries = 0;
+	voltage = -1;
+	memset(config, 0x00, sizeof(config));
+	for (;;) {
+		if (!nouveau_dp_link_train_commit(encoder, config))
+			break;
+
+		udelay(100);
+
+		ret = auxch_rd(encoder, DP_LANE0_1_STATUS, status, 2);
+		if (ret)
+			break;
+		NV_DEBUG_KMS(dev, "\t\tstatus: 0x%02x 0x%02x\n",
+			 status[0], status[1]);
+
+		cr_done = true;
+		cr_max_vs = false;
+		for (i = 0; i < nv_encoder->dp.link_nr; i++) {
+			int lane = (status[i >> 1] >> ((i & 1) * 4)) & 0xf;
+
+			if (!(lane & DP_LANE_CR_DONE)) {
+				cr_done = false;
+				if (config[i] & DP_TRAIN_MAX_PRE_EMPHASIS_REACHED)
+					cr_max_vs = true;
+				break;
+			}
+		}
+
+		if ((config[0] & DP_TRAIN_VOLTAGE_SWING_MASK) != voltage) {
+			voltage = config[0] & DP_TRAIN_VOLTAGE_SWING_MASK;
+			tries = 0;
+		}
+
+		if (cr_done || cr_max_vs || (++tries == 5))
+			break;
+
+		if (!nouveau_dp_link_train_adjust(encoder, config))
+			break;
+	}
+
+	if (!cr_done)
+		goto stop;
+
+	/* channel equalisation */
+	NV_DEBUG_KMS(dev, "\tbegin eq\n");
+	ret = nouveau_dp_link_train_set(encoder, DP_TRAINING_PATTERN_2);
+	if (ret)
+		goto stop;
+
+	for (tries = 0; tries <= 5; tries++) {
+		udelay(400);
+
+		ret = auxch_rd(encoder, DP_LANE0_1_STATUS, status, 3);
+		if (ret)
+			break;
+		NV_DEBUG_KMS(dev, "\t\tstatus: 0x%02x 0x%02x\n",
+			 status[0], status[1]);
+
+		eq_done = true;
+		if (!(status[2] & DP_INTERLANE_ALIGN_DONE))
+			eq_done = false;
+
+		for (i = 0; eq_done && i < nv_encoder->dp.link_nr; i++) {
+			int lane = (status[i >> 1] >> ((i & 1) * 4)) & 0xf;
+
+			if (!(lane & DP_LANE_CR_DONE)) {
+				cr_done = false;
+				break;
+			}
+
+			if (!(lane & DP_LANE_CHANNEL_EQ_DONE) ||
+			    !(lane & DP_LANE_SYMBOL_LOCKED)) {
+				eq_done = false;
+				break;
+			}
+		}
+
+		if (eq_done || !cr_done)
+			break;
+
+		if (!nouveau_dp_link_train_adjust(encoder, config) ||
+		    !nouveau_dp_link_train_commit(encoder, config))
+			break;
+	}
+
+stop:
+	/* end link training */
+	ret = nouveau_dp_link_train_set(encoder, DP_TRAINING_PATTERN_DISABLE);
+	if (ret)
+		return false;
+
+	/* retry at a lower setting, if possible */
+	if (!ret && !(eq_done && cr_done)) {
+		NV_DEBUG_KMS(dev, "\twe failed\n");
+		if (nv_encoder->dp.link_bw != DP_LINK_BW_1_62) {
+			NV_DEBUG_KMS(dev, "retry link training at low rate\n");
+			nv_encoder->dp.link_bw = DP_LINK_BW_1_62;
+			goto train;
+		}
+	}
+
+	return eq_done;
+}
+
+bool
+nouveau_dp_detect(struct drm_encoder *encoder)
+{
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	struct drm_device *dev = encoder->dev;
+	uint8_t dpcd[4];
+	int ret;
+
+	ret = auxch_rd(encoder, 0x0000, dpcd, 4);
+	if (ret)
+		return false;
+
+	NV_DEBUG_KMS(dev, "encoder: link_bw %d, link_nr %d\n"
+		      "display: link_bw %d, link_nr %d version 0x%02x\n",
+		 nv_encoder->dcb->dpconf.link_bw,
+		 nv_encoder->dcb->dpconf.link_nr,
+		 dpcd[1], dpcd[2] & 0x0f, dpcd[0]);
+
+	nv_encoder->dp.dpcd_version = dpcd[0];
+
+	nv_encoder->dp.link_bw = dpcd[1];
+	if (nv_encoder->dp.link_bw != DP_LINK_BW_1_62 &&
+	    !nv_encoder->dcb->dpconf.link_bw)
+		nv_encoder->dp.link_bw = DP_LINK_BW_1_62;
+
+	nv_encoder->dp.link_nr = dpcd[2] & 0xf;
+	if (nv_encoder->dp.link_nr > nv_encoder->dcb->dpconf.link_nr)
+		nv_encoder->dp.link_nr = nv_encoder->dcb->dpconf.link_nr;
+
+	return true;
+}
+
+int
+nouveau_dp_auxch(struct nouveau_i2c_chan *auxch, int cmd, int addr,
+		 uint8_t *data, int data_nr)
+{
+	struct drm_device *dev = auxch->dev;
+	uint32_t tmp, ctrl, stat = 0, data32[4] = {};
+	int ret = 0, i, index = auxch->rd;
+
+	NV_DEBUG_KMS(dev, "ch %d cmd %d addr 0x%x len %d\n", index, cmd, addr, data_nr);
+
+	tmp = nv_rd32(dev, NV50_AUXCH_CTRL(auxch->rd));
+	nv_wr32(dev, NV50_AUXCH_CTRL(auxch->rd), tmp | 0x00100000);
+	tmp = nv_rd32(dev, NV50_AUXCH_CTRL(auxch->rd));
+	if (!(tmp & 0x01000000)) {
+		NV_ERROR(dev, "expected bit 24 == 1, got 0x%08x\n", tmp);
+		ret = -EIO;
+		goto out;
+	}
+
+	for (i = 0; i < 3; i++) {
+		tmp = nv_rd32(dev, NV50_AUXCH_STAT(auxch->rd));
+		if (tmp & NV50_AUXCH_STAT_STATE_READY)
+			break;
+		udelay(100);
+	}
+
+	if (i == 3) {
+		ret = -EBUSY;
+		goto out;
+	}
+
+	if (!(cmd & 1)) {
+		memcpy(data32, data, data_nr);
+		for (i = 0; i < 4; i++) {
+			NV_DEBUG_KMS(dev, "wr %d: 0x%08x\n", i, data32[i]);
+			nv_wr32(dev, NV50_AUXCH_DATA_OUT(index, i), data32[i]);
+		}
+	}
+
+	nv_wr32(dev, NV50_AUXCH_ADDR(index), addr);
+	ctrl  = nv_rd32(dev, NV50_AUXCH_CTRL(index));
+	ctrl &= ~(NV50_AUXCH_CTRL_CMD | NV50_AUXCH_CTRL_LEN);
+	ctrl |= (cmd << NV50_AUXCH_CTRL_CMD_SHIFT);
+	ctrl |= ((data_nr - 1) << NV50_AUXCH_CTRL_LEN_SHIFT);
+
+	for (;;) {
+		nv_wr32(dev, NV50_AUXCH_CTRL(index), ctrl | 0x80000000);
+		nv_wr32(dev, NV50_AUXCH_CTRL(index), ctrl);
+		nv_wr32(dev, NV50_AUXCH_CTRL(index), ctrl | 0x00010000);
+		if (!nv_wait(NV50_AUXCH_CTRL(index), 0x00010000, 0x00000000)) {
+			NV_ERROR(dev, "expected bit 16 == 0, got 0x%08x\n",
+				 nv_rd32(dev, NV50_AUXCH_CTRL(index)));
+			ret = -EBUSY;
+			goto out;
+		}
+
+		udelay(400);
+
+		stat = nv_rd32(dev, NV50_AUXCH_STAT(index));
+		if ((stat & NV50_AUXCH_STAT_REPLY_AUX) !=
+			    NV50_AUXCH_STAT_REPLY_AUX_DEFER)
+			break;
+	}
+
+	if (cmd & 1) {
+		if ((stat & NV50_AUXCH_STAT_COUNT) != data_nr) {
+			ret = -EREMOTEIO;
+			goto out;
+		}
+
+		for (i = 0; i < 4; i++) {
+			data32[i] = nv_rd32(dev, NV50_AUXCH_DATA_IN(index, i));
+			NV_DEBUG_KMS(dev, "rd %d: 0x%08x\n", i, data32[i]);
+		}
+		memcpy(data, data32, data_nr);
+	}
+
+out:
+	tmp = nv_rd32(dev, NV50_AUXCH_CTRL(auxch->rd));
+	nv_wr32(dev, NV50_AUXCH_CTRL(auxch->rd), tmp & ~0x00100000);
+	tmp = nv_rd32(dev, NV50_AUXCH_CTRL(auxch->rd));
+	if (tmp & 0x01000000) {
+		NV_ERROR(dev, "expected bit 24 == 0, got 0x%08x\n", tmp);
+		ret = -EIO;
+	}
+
+	udelay(400);
+
+	return ret ? ret : (stat & NV50_AUXCH_STAT_REPLY);
+}
+
+int
+nouveau_dp_i2c_aux_ch(struct i2c_adapter *adapter, int mode,
+		      uint8_t write_byte, uint8_t *read_byte)
+{
+	struct i2c_algo_dp_aux_data *algo_data = adapter->algo_data;
+	struct nouveau_i2c_chan *auxch = (struct nouveau_i2c_chan *)adapter;
+	struct drm_device *dev = auxch->dev;
+	int ret = 0, cmd, addr = algo_data->address;
+	uint8_t *buf;
+
+	if (mode == MODE_I2C_READ) {
+		cmd = AUX_I2C_READ;
+		buf = read_byte;
+	} else {
+		cmd = (mode & MODE_I2C_READ) ? AUX_I2C_READ : AUX_I2C_WRITE;
+		buf = &write_byte;
+	}
+
+	if (!(mode & MODE_I2C_STOP))
+		cmd |= AUX_I2C_MOT;
+
+	if (mode & MODE_I2C_START)
+		return 1;
+
+	for (;;) {
+		ret = nouveau_dp_auxch(auxch, cmd, addr, buf, 1);
+		if (ret < 0)
+			return ret;
+
+		switch (ret & NV50_AUXCH_STAT_REPLY_I2C) {
+		case NV50_AUXCH_STAT_REPLY_I2C_ACK:
+			return 1;
+		case NV50_AUXCH_STAT_REPLY_I2C_NACK:
+			return -EREMOTEIO;
+		case NV50_AUXCH_STAT_REPLY_I2C_DEFER:
+			udelay(100);
+			break;
+		default:
+			NV_ERROR(dev, "invalid auxch status: 0x%08x\n", ret);
+			return -EREMOTEIO;
+		}
+	}
+}
+
diff --git a/drivers/gpu/drm/nouveau/nouveau_drv.c b/drivers/gpu/drm/nouveau/nouveau_drv.c
new file mode 100644
index 000000000000..30cc09e8a709
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_drv.c
@@ -0,0 +1,424 @@
+/*
+ * Copyright 2005 Stephane Marchesin.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include <linux/console.h>
+
+#include "drmP.h"
+#include "drm.h"
+#include "drm_crtc_helper.h"
+#include "nouveau_drv.h"
+#include "nouveau_hw.h"
+#include "nouveau_fb.h"
+#include "nouveau_fbcon.h"
+#include "nv50_display.h"
+
+#include "drm_pciids.h"
+
+MODULE_PARM_DESC(ctxfw, "Use external firmware blob for grctx init (NV40)");
+int nouveau_ctxfw = 0;
+module_param_named(ctxfw, nouveau_ctxfw, int, 0400);
+
+MODULE_PARM_DESC(noagp, "Disable AGP");
+int nouveau_noagp;
+module_param_named(noagp, nouveau_noagp, int, 0400);
+
+MODULE_PARM_DESC(modeset, "Enable kernel modesetting");
+static int nouveau_modeset = -1; /* kms */
+module_param_named(modeset, nouveau_modeset, int, 0400);
+
+MODULE_PARM_DESC(vbios, "Override default VBIOS location");
+char *nouveau_vbios;
+module_param_named(vbios, nouveau_vbios, charp, 0400);
+
+MODULE_PARM_DESC(vram_pushbuf, "Force DMA push buffers to be in VRAM");
+int nouveau_vram_pushbuf;
+module_param_named(vram_pushbuf, nouveau_vram_pushbuf, int, 0400);
+
+MODULE_PARM_DESC(vram_notify, "Force DMA notifiers to be in VRAM");
+int nouveau_vram_notify = 1;
+module_param_named(vram_notify, nouveau_vram_notify, int, 0400);
+
+MODULE_PARM_DESC(duallink, "Allow dual-link TMDS (>=GeForce 8)");
+int nouveau_duallink = 1;
+module_param_named(duallink, nouveau_duallink, int, 0400);
+
+MODULE_PARM_DESC(uscript_lvds, "LVDS output script table ID (>=GeForce 8)");
+int nouveau_uscript_lvds = -1;
+module_param_named(uscript_lvds, nouveau_uscript_lvds, int, 0400);
+
+MODULE_PARM_DESC(uscript_tmds, "TMDS output script table ID (>=GeForce 8)");
+int nouveau_uscript_tmds = -1;
+module_param_named(uscript_tmds, nouveau_uscript_tmds, int, 0400);
+
+MODULE_PARM_DESC(ignorelid, "Ignore ACPI lid status");
+int nouveau_ignorelid = 0;
+module_param_named(ignorelid, nouveau_ignorelid, int, 0400);
+
+MODULE_PARM_DESC(noaccel, "Disable all acceleration");
+int nouveau_noaccel = 0;
+module_param_named(noaccel, nouveau_noaccel, int, 0400);
+
+MODULE_PARM_DESC(nofbaccel, "Disable fbcon acceleration");
+int nouveau_nofbaccel = 0;
+module_param_named(nofbaccel, nouveau_nofbaccel, int, 0400);
+
+MODULE_PARM_DESC(tv_norm, "Default TV norm.\n"
+		 "\t\tSupported: PAL, PAL-M, PAL-N, PAL-Nc, NTSC-M, NTSC-J,\n"
+		 "\t\t\thd480i, hd480p, hd576i, hd576p, hd720p, hd1080i.\n"
+		 "\t\tDefault: PAL\n"
+		 "\t\t*NOTE* Ignored for cards with external TV encoders.");
+char *nouveau_tv_norm;
+module_param_named(tv_norm, nouveau_tv_norm, charp, 0400);
+
+MODULE_PARM_DESC(reg_debug, "Register access debug bitmask:\n"
+		"\t\t0x1 mc, 0x2 video, 0x4 fb, 0x8 extdev,\n"
+		"\t\t0x10 crtc, 0x20 ramdac, 0x40 vgacrtc, 0x80 rmvio,\n"
+		"\t\t0x100 vgaattr, 0x200 EVO (G80+). ");
+int nouveau_reg_debug;
+module_param_named(reg_debug, nouveau_reg_debug, int, 0600);
+
+int nouveau_fbpercrtc;
+#if 0
+module_param_named(fbpercrtc, nouveau_fbpercrtc, int, 0400);
+#endif
+
+static struct pci_device_id pciidlist[] = {
+	{
+		PCI_DEVICE(PCI_VENDOR_ID_NVIDIA, PCI_ANY_ID),
+		.class = PCI_BASE_CLASS_DISPLAY << 16,
+		.class_mask  = 0xff << 16,
+	},
+	{
+		PCI_DEVICE(PCI_VENDOR_ID_NVIDIA_SGS, PCI_ANY_ID),
+		.class = PCI_BASE_CLASS_DISPLAY << 16,
+		.class_mask  = 0xff << 16,
+	},
+	{}
+};
+
+MODULE_DEVICE_TABLE(pci, pciidlist);
+
+static struct drm_driver driver;
+
+static int __devinit
+nouveau_pci_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
+{
+	return drm_get_dev(pdev, ent, &driver);
+}
+
+static void
+nouveau_pci_remove(struct pci_dev *pdev)
+{
+	struct drm_device *dev = pci_get_drvdata(pdev);
+
+	drm_put_dev(dev);
+}
+
+int
+nouveau_pci_suspend(struct pci_dev *pdev, pm_message_t pm_state)
+{
+	struct drm_device *dev = pci_get_drvdata(pdev);
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_instmem_engine *pinstmem = &dev_priv->engine.instmem;
+	struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
+	struct nouveau_fifo_engine *pfifo = &dev_priv->engine.fifo;
+	struct nouveau_channel *chan;
+	struct drm_crtc *crtc;
+	uint32_t fbdev_flags;
+	int ret, i;
+
+	if (!drm_core_check_feature(dev, DRIVER_MODESET))
+		return -ENODEV;
+
+	if (pm_state.event == PM_EVENT_PRETHAW)
+		return 0;
+
+	fbdev_flags = dev_priv->fbdev_info->flags;
+	dev_priv->fbdev_info->flags |= FBINFO_HWACCEL_DISABLED;
+
+	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
+		struct nouveau_framebuffer *nouveau_fb;
+
+		nouveau_fb = nouveau_framebuffer(crtc->fb);
+		if (!nouveau_fb || !nouveau_fb->nvbo)
+			continue;
+
+		nouveau_bo_unpin(nouveau_fb->nvbo);
+	}
+
+	NV_INFO(dev, "Evicting buffers...\n");
+	ttm_bo_evict_mm(&dev_priv->ttm.bdev, TTM_PL_VRAM);
+
+	NV_INFO(dev, "Idling channels...\n");
+	for (i = 0; i < pfifo->channels; i++) {
+		struct nouveau_fence *fence = NULL;
+
+		chan = dev_priv->fifos[i];
+		if (!chan || (dev_priv->card_type >= NV_50 &&
+			      chan == dev_priv->fifos[0]))
+			continue;
+
+		ret = nouveau_fence_new(chan, &fence, true);
+		if (ret == 0) {
+			ret = nouveau_fence_wait(fence, NULL, false, false);
+			nouveau_fence_unref((void *)&fence);
+		}
+
+		if (ret) {
+			NV_ERROR(dev, "Failed to idle channel %d for suspend\n",
+				 chan->id);
+		}
+	}
+
+	pgraph->fifo_access(dev, false);
+	nouveau_wait_for_idle(dev);
+	pfifo->reassign(dev, false);
+	pfifo->disable(dev);
+	pfifo->unload_context(dev);
+	pgraph->unload_context(dev);
+
+	NV_INFO(dev, "Suspending GPU objects...\n");
+	ret = nouveau_gpuobj_suspend(dev);
+	if (ret) {
+		NV_ERROR(dev, "... failed: %d\n", ret);
+		goto out_abort;
+	}
+
+	ret = pinstmem->suspend(dev);
+	if (ret) {
+		NV_ERROR(dev, "... failed: %d\n", ret);
+		nouveau_gpuobj_suspend_cleanup(dev);
+		goto out_abort;
+	}
+
+	NV_INFO(dev, "And we're gone!\n");
+	pci_save_state(pdev);
+	if (pm_state.event == PM_EVENT_SUSPEND) {
+		pci_disable_device(pdev);
+		pci_set_power_state(pdev, PCI_D3hot);
+	}
+
+	acquire_console_sem();
+	fb_set_suspend(dev_priv->fbdev_info, 1);
+	release_console_sem();
+	dev_priv->fbdev_info->flags = fbdev_flags;
+	return 0;
+
+out_abort:
+	NV_INFO(dev, "Re-enabling acceleration..\n");
+	pfifo->enable(dev);
+	pfifo->reassign(dev, true);
+	pgraph->fifo_access(dev, true);
+	return ret;
+}
+
+int
+nouveau_pci_resume(struct pci_dev *pdev)
+{
+	struct drm_device *dev = pci_get_drvdata(pdev);
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_engine *engine = &dev_priv->engine;
+	struct drm_crtc *crtc;
+	uint32_t fbdev_flags;
+	int ret, i;
+
+	if (!drm_core_check_feature(dev, DRIVER_MODESET))
+		return -ENODEV;
+
+	fbdev_flags = dev_priv->fbdev_info->flags;
+	dev_priv->fbdev_info->flags |= FBINFO_HWACCEL_DISABLED;
+
+	NV_INFO(dev, "We're back, enabling device...\n");
+	pci_set_power_state(pdev, PCI_D0);
+	pci_restore_state(pdev);
+	if (pci_enable_device(pdev))
+		return -1;
+	pci_set_master(dev->pdev);
+
+	NV_INFO(dev, "POSTing device...\n");
+	ret = nouveau_run_vbios_init(dev);
+	if (ret)
+		return ret;
+
+	if (dev_priv->gart_info.type == NOUVEAU_GART_AGP) {
+		ret = nouveau_mem_init_agp(dev);
+		if (ret) {
+			NV_ERROR(dev, "error reinitialising AGP: %d\n", ret);
+			return ret;
+		}
+	}
+
+	NV_INFO(dev, "Reinitialising engines...\n");
+	engine->instmem.resume(dev);
+	engine->mc.init(dev);
+	engine->timer.init(dev);
+	engine->fb.init(dev);
+	engine->graph.init(dev);
+	engine->fifo.init(dev);
+
+	NV_INFO(dev, "Restoring GPU objects...\n");
+	nouveau_gpuobj_resume(dev);
+
+	nouveau_irq_postinstall(dev);
+
+	/* Re-write SKIPS, they'll have been lost over the suspend */
+	if (nouveau_vram_pushbuf) {
+		struct nouveau_channel *chan;
+		int j;
+
+		for (i = 0; i < dev_priv->engine.fifo.channels; i++) {
+			chan = dev_priv->fifos[i];
+			if (!chan || !chan->pushbuf_bo)
+				continue;
+
+			for (j = 0; j < NOUVEAU_DMA_SKIPS; j++)
+				nouveau_bo_wr32(chan->pushbuf_bo, i, 0);
+		}
+	}
+
+	NV_INFO(dev, "Restoring mode...\n");
+	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
+		struct nouveau_framebuffer *nouveau_fb;
+
+		nouveau_fb = nouveau_framebuffer(crtc->fb);
+		if (!nouveau_fb || !nouveau_fb->nvbo)
+			continue;
+
+		nouveau_bo_pin(nouveau_fb->nvbo, TTM_PL_FLAG_VRAM);
+	}
+
+	if (dev_priv->card_type < NV_50) {
+		nv04_display_restore(dev);
+		NVLockVgaCrtcs(dev, false);
+	} else
+		nv50_display_init(dev);
+
+	/* Force CLUT to get re-loaded during modeset */
+	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
+		struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+
+		nv_crtc->lut.depth = 0;
+	}
+
+	acquire_console_sem();
+	fb_set_suspend(dev_priv->fbdev_info, 0);
+	release_console_sem();
+
+	nouveau_fbcon_zfill(dev);
+
+	drm_helper_resume_force_mode(dev);
+	dev_priv->fbdev_info->flags = fbdev_flags;
+	return 0;
+}
+
+static struct drm_driver driver = {
+	.driver_features =
+		DRIVER_USE_AGP | DRIVER_PCI_DMA | DRIVER_SG |
+		DRIVER_HAVE_IRQ | DRIVER_IRQ_SHARED | DRIVER_GEM,
+	.load = nouveau_load,
+	.firstopen = nouveau_firstopen,
+	.lastclose = nouveau_lastclose,
+	.unload = nouveau_unload,
+	.preclose = nouveau_preclose,
+#if defined(CONFIG_DRM_NOUVEAU_DEBUG)
+	.debugfs_init = nouveau_debugfs_init,
+	.debugfs_cleanup = nouveau_debugfs_takedown,
+#endif
+	.irq_preinstall = nouveau_irq_preinstall,
+	.irq_postinstall = nouveau_irq_postinstall,
+	.irq_uninstall = nouveau_irq_uninstall,
+	.irq_handler = nouveau_irq_handler,
+	.reclaim_buffers = drm_core_reclaim_buffers,
+	.get_map_ofs = drm_core_get_map_ofs,
+	.get_reg_ofs = drm_core_get_reg_ofs,
+	.ioctls = nouveau_ioctls,
+	.fops = {
+		.owner = THIS_MODULE,
+		.open = drm_open,
+		.release = drm_release,
+		.unlocked_ioctl = drm_ioctl,
+		.mmap = nouveau_ttm_mmap,
+		.poll = drm_poll,
+		.fasync = drm_fasync,
+#if defined(CONFIG_COMPAT)
+		.compat_ioctl = nouveau_compat_ioctl,
+#endif
+	},
+	.pci_driver = {
+		.name = DRIVER_NAME,
+		.id_table = pciidlist,
+		.probe = nouveau_pci_probe,
+		.remove = nouveau_pci_remove,
+		.suspend = nouveau_pci_suspend,
+		.resume = nouveau_pci_resume
+	},
+
+	.gem_init_object = nouveau_gem_object_new,
+	.gem_free_object = nouveau_gem_object_del,
+
+	.name = DRIVER_NAME,
+	.desc = DRIVER_DESC,
+#ifdef GIT_REVISION
+	.date = GIT_REVISION,
+#else
+	.date = DRIVER_DATE,
+#endif
+	.major = DRIVER_MAJOR,
+	.minor = DRIVER_MINOR,
+	.patchlevel = DRIVER_PATCHLEVEL,
+};
+
+static int __init nouveau_init(void)
+{
+	driver.num_ioctls = nouveau_max_ioctl;
+
+	if (nouveau_modeset == -1) {
+#ifdef CONFIG_VGA_CONSOLE
+		if (vgacon_text_force())
+			nouveau_modeset = 0;
+		else
+#endif
+			nouveau_modeset = 1;
+	}
+
+	if (nouveau_modeset == 1) {
+		driver.driver_features |= DRIVER_MODESET;
+		nouveau_register_dsm_handler();
+	}
+
+	return drm_init(&driver);
+}
+
+static void __exit nouveau_exit(void)
+{
+	drm_exit(&driver);
+	nouveau_unregister_dsm_handler();
+}
+
+module_init(nouveau_init);
+module_exit(nouveau_exit);
+
+MODULE_AUTHOR(DRIVER_AUTHOR);
+MODULE_DESCRIPTION(DRIVER_DESC);
+MODULE_LICENSE("GPL and additional rights");
diff --git a/drivers/gpu/drm/nouveau/nouveau_drv.h b/drivers/gpu/drm/nouveau/nouveau_drv.h
new file mode 100644
index 000000000000..4b9aaf2a8d0f
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_drv.h
@@ -0,0 +1,1340 @@
+/*
+ * Copyright 2005 Stephane Marchesin.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef __NOUVEAU_DRV_H__
+#define __NOUVEAU_DRV_H__
+
+#define DRIVER_AUTHOR		"Stephane Marchesin"
+#define DRIVER_EMAIL		"dri-devel@lists.sourceforge.net"
+
+#define DRIVER_NAME		"nouveau"
+#define DRIVER_DESC		"nVidia Riva/TNT/GeForce"
+#define DRIVER_DATE		"20090420"
+
+#define DRIVER_MAJOR		0
+#define DRIVER_MINOR		0
+#define DRIVER_PATCHLEVEL	16
+
+#define NOUVEAU_FAMILY   0x0000FFFF
+#define NOUVEAU_FLAGS    0xFFFF0000
+
+#include "ttm/ttm_bo_api.h"
+#include "ttm/ttm_bo_driver.h"
+#include "ttm/ttm_placement.h"
+#include "ttm/ttm_memory.h"
+#include "ttm/ttm_module.h"
+
+struct nouveau_fpriv {
+	struct ttm_object_file *tfile;
+};
+
+#define DRM_FILE_PAGE_OFFSET (0x100000000ULL >> PAGE_SHIFT)
+
+#include "nouveau_drm.h"
+#include "nouveau_reg.h"
+#include "nouveau_bios.h"
+struct nouveau_grctx;
+
+#define MAX_NUM_DCB_ENTRIES 16
+
+#define NOUVEAU_MAX_CHANNEL_NR 128
+#define NOUVEAU_MAX_TILE_NR 15
+
+#define NV50_VM_MAX_VRAM (2*1024*1024*1024ULL)
+#define NV50_VM_BLOCK    (512*1024*1024ULL)
+#define NV50_VM_VRAM_NR  (NV50_VM_MAX_VRAM / NV50_VM_BLOCK)
+
+struct nouveau_tile_reg {
+	struct nouveau_fence *fence;
+	uint32_t addr;
+	uint32_t size;
+	bool used;
+};
+
+struct nouveau_bo {
+	struct ttm_buffer_object bo;
+	struct ttm_placement placement;
+	u32 placements[3];
+	struct ttm_bo_kmap_obj kmap;
+	struct list_head head;
+
+	/* protected by ttm_bo_reserve() */
+	struct drm_file *reserved_by;
+	struct list_head entry;
+	int pbbo_index;
+	bool validate_mapped;
+
+	struct nouveau_channel *channel;
+
+	bool mappable;
+	bool no_vm;
+
+	uint32_t tile_mode;
+	uint32_t tile_flags;
+	struct nouveau_tile_reg *tile;
+
+	struct drm_gem_object *gem;
+	struct drm_file *cpu_filp;
+	int pin_refcnt;
+};
+
+static inline struct nouveau_bo *
+nouveau_bo(struct ttm_buffer_object *bo)
+{
+	return container_of(bo, struct nouveau_bo, bo);
+}
+
+static inline struct nouveau_bo *
+nouveau_gem_object(struct drm_gem_object *gem)
+{
+	return gem ? gem->driver_private : NULL;
+}
+
+/* TODO: submit equivalent to TTM generic API upstream? */
+static inline void __iomem *
+nvbo_kmap_obj_iovirtual(struct nouveau_bo *nvbo)
+{
+	bool is_iomem;
+	void __iomem *ioptr = (void __force __iomem *)ttm_kmap_obj_virtual(
+						&nvbo->kmap, &is_iomem);
+	WARN_ON_ONCE(ioptr && !is_iomem);
+	return ioptr;
+}
+
+struct mem_block {
+	struct mem_block *next;
+	struct mem_block *prev;
+	uint64_t start;
+	uint64_t size;
+	struct drm_file *file_priv; /* NULL: free, -1: heap, other: real files */
+};
+
+enum nouveau_flags {
+	NV_NFORCE   = 0x10000000,
+	NV_NFORCE2  = 0x20000000
+};
+
+#define NVOBJ_ENGINE_SW		0
+#define NVOBJ_ENGINE_GR		1
+#define NVOBJ_ENGINE_DISPLAY	2
+#define NVOBJ_ENGINE_INT	0xdeadbeef
+
+#define NVOBJ_FLAG_ALLOW_NO_REFS	(1 << 0)
+#define NVOBJ_FLAG_ZERO_ALLOC		(1 << 1)
+#define NVOBJ_FLAG_ZERO_FREE		(1 << 2)
+#define NVOBJ_FLAG_FAKE			(1 << 3)
+struct nouveau_gpuobj {
+	struct list_head list;
+
+	struct nouveau_channel *im_channel;
+	struct mem_block *im_pramin;
+	struct nouveau_bo *im_backing;
+	uint32_t im_backing_start;
+	uint32_t *im_backing_suspend;
+	int im_bound;
+
+	uint32_t flags;
+	int refcount;
+
+	uint32_t engine;
+	uint32_t class;
+
+	void (*dtor)(struct drm_device *, struct nouveau_gpuobj *);
+	void *priv;
+};
+
+struct nouveau_gpuobj_ref {
+	struct list_head list;
+
+	struct nouveau_gpuobj *gpuobj;
+	uint32_t instance;
+
+	struct nouveau_channel *channel;
+	int handle;
+};
+
+struct nouveau_channel {
+	struct drm_device *dev;
+	int id;
+
+	/* owner of this fifo */
+	struct drm_file *file_priv;
+	/* mapping of the fifo itself */
+	struct drm_local_map *map;
+
+	/* mapping of the regs controling the fifo */
+	void __iomem *user;
+	uint32_t user_get;
+	uint32_t user_put;
+
+	/* Fencing */
+	struct {
+		/* lock protects the pending list only */
+		spinlock_t lock;
+		struct list_head pending;
+		uint32_t sequence;
+		uint32_t sequence_ack;
+		uint32_t last_sequence_irq;
+	} fence;
+
+	/* DMA push buffer */
+	struct nouveau_gpuobj_ref *pushbuf;
+	struct nouveau_bo         *pushbuf_bo;
+	uint32_t                   pushbuf_base;
+
+	/* Notifier memory */
+	struct nouveau_bo *notifier_bo;
+	struct mem_block *notifier_heap;
+
+	/* PFIFO context */
+	struct nouveau_gpuobj_ref *ramfc;
+	struct nouveau_gpuobj_ref *cache;
+
+	/* PGRAPH context */
+	/* XXX may be merge 2 pointers as private data ??? */
+	struct nouveau_gpuobj_ref *ramin_grctx;
+	void *pgraph_ctx;
+
+	/* NV50 VM */
+	struct nouveau_gpuobj     *vm_pd;
+	struct nouveau_gpuobj_ref *vm_gart_pt;
+	struct nouveau_gpuobj_ref *vm_vram_pt[NV50_VM_VRAM_NR];
+
+	/* Objects */
+	struct nouveau_gpuobj_ref *ramin; /* Private instmem */
+	struct mem_block          *ramin_heap; /* Private PRAMIN heap */
+	struct nouveau_gpuobj_ref *ramht; /* Hash table */
+	struct list_head           ramht_refs; /* Objects referenced by RAMHT */
+
+	/* GPU object info for stuff used in-kernel (mm_enabled) */
+	uint32_t m2mf_ntfy;
+	uint32_t vram_handle;
+	uint32_t gart_handle;
+	bool accel_done;
+
+	/* Push buffer state (only for drm's channel on !mm_enabled) */
+	struct {
+		int max;
+		int free;
+		int cur;
+		int put;
+		/* access via pushbuf_bo */
+
+		int ib_base;
+		int ib_max;
+		int ib_free;
+		int ib_put;
+	} dma;
+
+	uint32_t sw_subchannel[8];
+
+	struct {
+		struct nouveau_gpuobj *vblsem;
+		uint32_t vblsem_offset;
+		uint32_t vblsem_rval;
+		struct list_head vbl_wait;
+	} nvsw;
+
+	struct {
+		bool active;
+		char name[32];
+		struct drm_info_list info;
+	} debugfs;
+};
+
+struct nouveau_instmem_engine {
+	void	*priv;
+
+	int	(*init)(struct drm_device *dev);
+	void	(*takedown)(struct drm_device *dev);
+	int	(*suspend)(struct drm_device *dev);
+	void	(*resume)(struct drm_device *dev);
+
+	int	(*populate)(struct drm_device *, struct nouveau_gpuobj *,
+			    uint32_t *size);
+	void	(*clear)(struct drm_device *, struct nouveau_gpuobj *);
+	int	(*bind)(struct drm_device *, struct nouveau_gpuobj *);
+	int	(*unbind)(struct drm_device *, struct nouveau_gpuobj *);
+	void	(*prepare_access)(struct drm_device *, bool write);
+	void	(*finish_access)(struct drm_device *);
+};
+
+struct nouveau_mc_engine {
+	int  (*init)(struct drm_device *dev);
+	void (*takedown)(struct drm_device *dev);
+};
+
+struct nouveau_timer_engine {
+	int      (*init)(struct drm_device *dev);
+	void     (*takedown)(struct drm_device *dev);
+	uint64_t (*read)(struct drm_device *dev);
+};
+
+struct nouveau_fb_engine {
+	int num_tiles;
+
+	int  (*init)(struct drm_device *dev);
+	void (*takedown)(struct drm_device *dev);
+
+	void (*set_region_tiling)(struct drm_device *dev, int i, uint32_t addr,
+				 uint32_t size, uint32_t pitch);
+};
+
+struct nouveau_fifo_engine {
+	void *priv;
+
+	int  channels;
+
+	int  (*init)(struct drm_device *);
+	void (*takedown)(struct drm_device *);
+
+	void (*disable)(struct drm_device *);
+	void (*enable)(struct drm_device *);
+	bool (*reassign)(struct drm_device *, bool enable);
+	bool (*cache_flush)(struct drm_device *dev);
+	bool (*cache_pull)(struct drm_device *dev, bool enable);
+
+	int  (*channel_id)(struct drm_device *);
+
+	int  (*create_context)(struct nouveau_channel *);
+	void (*destroy_context)(struct nouveau_channel *);
+	int  (*load_context)(struct nouveau_channel *);
+	int  (*unload_context)(struct drm_device *);
+};
+
+struct nouveau_pgraph_object_method {
+	int id;
+	int (*exec)(struct nouveau_channel *chan, int grclass, int mthd,
+		      uint32_t data);
+};
+
+struct nouveau_pgraph_object_class {
+	int id;
+	bool software;
+	struct nouveau_pgraph_object_method *methods;
+};
+
+struct nouveau_pgraph_engine {
+	struct nouveau_pgraph_object_class *grclass;
+	bool accel_blocked;
+	void *ctxprog;
+	void *ctxvals;
+	int grctx_size;
+
+	int  (*init)(struct drm_device *);
+	void (*takedown)(struct drm_device *);
+
+	void (*fifo_access)(struct drm_device *, bool);
+
+	struct nouveau_channel *(*channel)(struct drm_device *);
+	int  (*create_context)(struct nouveau_channel *);
+	void (*destroy_context)(struct nouveau_channel *);
+	int  (*load_context)(struct nouveau_channel *);
+	int  (*unload_context)(struct drm_device *);
+
+	void (*set_region_tiling)(struct drm_device *dev, int i, uint32_t addr,
+				  uint32_t size, uint32_t pitch);
+};
+
+struct nouveau_engine {
+	struct nouveau_instmem_engine instmem;
+	struct nouveau_mc_engine      mc;
+	struct nouveau_timer_engine   timer;
+	struct nouveau_fb_engine      fb;
+	struct nouveau_pgraph_engine  graph;
+	struct nouveau_fifo_engine    fifo;
+};
+
+struct nouveau_pll_vals {
+	union {
+		struct {
+#ifdef __BIG_ENDIAN
+			uint8_t N1, M1, N2, M2;
+#else
+			uint8_t M1, N1, M2, N2;
+#endif
+		};
+		struct {
+			uint16_t NM1, NM2;
+		} __attribute__((packed));
+	};
+	int log2P;
+
+	int refclk;
+};
+
+enum nv04_fp_display_regs {
+	FP_DISPLAY_END,
+	FP_TOTAL,
+	FP_CRTC,
+	FP_SYNC_START,
+	FP_SYNC_END,
+	FP_VALID_START,
+	FP_VALID_END
+};
+
+struct nv04_crtc_reg {
+	unsigned char MiscOutReg;     /* */
+	uint8_t CRTC[0x9f];
+	uint8_t CR58[0x10];
+	uint8_t Sequencer[5];
+	uint8_t Graphics[9];
+	uint8_t Attribute[21];
+	unsigned char DAC[768];       /* Internal Colorlookuptable */
+
+	/* PCRTC regs */
+	uint32_t fb_start;
+	uint32_t crtc_cfg;
+	uint32_t cursor_cfg;
+	uint32_t gpio_ext;
+	uint32_t crtc_830;
+	uint32_t crtc_834;
+	uint32_t crtc_850;
+	uint32_t crtc_eng_ctrl;
+
+	/* PRAMDAC regs */
+	uint32_t nv10_cursync;
+	struct nouveau_pll_vals pllvals;
+	uint32_t ramdac_gen_ctrl;
+	uint32_t ramdac_630;
+	uint32_t ramdac_634;
+	uint32_t tv_setup;
+	uint32_t tv_vtotal;
+	uint32_t tv_vskew;
+	uint32_t tv_vsync_delay;
+	uint32_t tv_htotal;
+	uint32_t tv_hskew;
+	uint32_t tv_hsync_delay;
+	uint32_t tv_hsync_delay2;
+	uint32_t fp_horiz_regs[7];
+	uint32_t fp_vert_regs[7];
+	uint32_t dither;
+	uint32_t fp_control;
+	uint32_t dither_regs[6];
+	uint32_t fp_debug_0;
+	uint32_t fp_debug_1;
+	uint32_t fp_debug_2;
+	uint32_t fp_margin_color;
+	uint32_t ramdac_8c0;
+	uint32_t ramdac_a20;
+	uint32_t ramdac_a24;
+	uint32_t ramdac_a34;
+	uint32_t ctv_regs[38];
+};
+
+struct nv04_output_reg {
+	uint32_t output;
+	int head;
+};
+
+struct nv04_mode_state {
+	uint32_t bpp;
+	uint32_t width;
+	uint32_t height;
+	uint32_t interlace;
+	uint32_t repaint0;
+	uint32_t repaint1;
+	uint32_t screen;
+	uint32_t scale;
+	uint32_t dither;
+	uint32_t extra;
+	uint32_t fifo;
+	uint32_t pixel;
+	uint32_t horiz;
+	int arbitration0;
+	int arbitration1;
+	uint32_t pll;
+	uint32_t pllB;
+	uint32_t vpll;
+	uint32_t vpll2;
+	uint32_t vpllB;
+	uint32_t vpll2B;
+	uint32_t pllsel;
+	uint32_t sel_clk;
+	uint32_t general;
+	uint32_t crtcOwner;
+	uint32_t head;
+	uint32_t head2;
+	uint32_t cursorConfig;
+	uint32_t cursor0;
+	uint32_t cursor1;
+	uint32_t cursor2;
+	uint32_t timingH;
+	uint32_t timingV;
+	uint32_t displayV;
+	uint32_t crtcSync;
+
+	struct nv04_crtc_reg crtc_reg[2];
+};
+
+enum nouveau_card_type {
+	NV_04      = 0x00,
+	NV_10      = 0x10,
+	NV_20      = 0x20,
+	NV_30      = 0x30,
+	NV_40      = 0x40,
+	NV_50      = 0x50,
+};
+
+struct drm_nouveau_private {
+	struct drm_device *dev;
+	enum {
+		NOUVEAU_CARD_INIT_DOWN,
+		NOUVEAU_CARD_INIT_DONE,
+		NOUVEAU_CARD_INIT_FAILED
+	} init_state;
+
+	/* the card type, takes NV_* as values */
+	enum nouveau_card_type card_type;
+	/* exact chipset, derived from NV_PMC_BOOT_0 */
+	int chipset;
+	int flags;
+
+	void __iomem *mmio;
+	void __iomem *ramin;
+	uint32_t ramin_size;
+
+	struct nouveau_bo *vga_ram;
+
+	struct workqueue_struct *wq;
+	struct work_struct irq_work;
+
+	struct list_head vbl_waiting;
+
+	struct {
+		struct ttm_global_reference mem_global_ref;
+		struct ttm_bo_global_ref bo_global_ref;
+		struct ttm_bo_device bdev;
+		spinlock_t bo_list_lock;
+		struct list_head bo_list;
+		atomic_t validate_sequence;
+	} ttm;
+
+	struct fb_info *fbdev_info;
+
+	int fifo_alloc_count;
+	struct nouveau_channel *fifos[NOUVEAU_MAX_CHANNEL_NR];
+
+	struct nouveau_engine engine;
+	struct nouveau_channel *channel;
+
+	/* For PFIFO and PGRAPH. */
+	spinlock_t context_switch_lock;
+
+	/* RAMIN configuration, RAMFC, RAMHT and RAMRO offsets */
+	struct nouveau_gpuobj *ramht;
+	uint32_t ramin_rsvd_vram;
+	uint32_t ramht_offset;
+	uint32_t ramht_size;
+	uint32_t ramht_bits;
+	uint32_t ramfc_offset;
+	uint32_t ramfc_size;
+	uint32_t ramro_offset;
+	uint32_t ramro_size;
+
+	/* base physical addresses */
+	uint64_t fb_phys;
+	uint64_t fb_available_size;
+	uint64_t fb_mappable_pages;
+	uint64_t fb_aper_free;
+
+	struct {
+		enum {
+			NOUVEAU_GART_NONE = 0,
+			NOUVEAU_GART_AGP,
+			NOUVEAU_GART_SGDMA
+		} type;
+		uint64_t aper_base;
+		uint64_t aper_size;
+		uint64_t aper_free;
+
+		struct nouveau_gpuobj *sg_ctxdma;
+		struct page *sg_dummy_page;
+		dma_addr_t sg_dummy_bus;
+
+		/* nottm hack */
+		struct drm_ttm_backend *sg_be;
+		unsigned long sg_handle;
+	} gart_info;
+
+	/* nv10-nv40 tiling regions */
+	struct {
+		struct nouveau_tile_reg reg[NOUVEAU_MAX_TILE_NR];
+		spinlock_t lock;
+	} tile;
+
+	/* G8x/G9x virtual address space */
+	uint64_t vm_gart_base;
+	uint64_t vm_gart_size;
+	uint64_t vm_vram_base;
+	uint64_t vm_vram_size;
+	uint64_t vm_end;
+	struct nouveau_gpuobj *vm_vram_pt[NV50_VM_VRAM_NR];
+	int vm_vram_pt_nr;
+	uint64_t vram_sys_base;
+
+	/* the mtrr covering the FB */
+	int fb_mtrr;
+
+	struct mem_block *ramin_heap;
+
+	/* context table pointed to be NV_PGRAPH_CHANNEL_CTX_TABLE (0x400780) */
+	uint32_t ctx_table_size;
+	struct nouveau_gpuobj_ref *ctx_table;
+
+	struct list_head gpuobj_list;
+
+	struct nvbios vbios;
+
+	struct nv04_mode_state mode_reg;
+	struct nv04_mode_state saved_reg;
+	uint32_t saved_vga_font[4][16384];
+	uint32_t crtc_owner;
+	uint32_t dac_users[4];
+
+	struct nouveau_suspend_resume {
+		uint32_t fifo_mode;
+		uint32_t graph_ctx_control;
+		uint32_t graph_state;
+		uint32_t *ramin_copy;
+		uint64_t ramin_size;
+	} susres;
+
+	struct backlight_device *backlight;
+
+	struct nouveau_channel *evo;
+
+	struct {
+		struct dentry *channel_root;
+	} debugfs;
+};
+
+static inline struct drm_nouveau_private *
+nouveau_bdev(struct ttm_bo_device *bd)
+{
+	return container_of(bd, struct drm_nouveau_private, ttm.bdev);
+}
+
+static inline int
+nouveau_bo_ref(struct nouveau_bo *ref, struct nouveau_bo **pnvbo)
+{
+	struct nouveau_bo *prev;
+
+	if (!pnvbo)
+		return -EINVAL;
+	prev = *pnvbo;
+
+	*pnvbo = ref ? nouveau_bo(ttm_bo_reference(&ref->bo)) : NULL;
+	if (prev) {
+		struct ttm_buffer_object *bo = &prev->bo;
+
+		ttm_bo_unref(&bo);
+	}
+
+	return 0;
+}
+
+#define NOUVEAU_CHECK_INITIALISED_WITH_RETURN do {            \
+	struct drm_nouveau_private *nv = dev->dev_private;    \
+	if (nv->init_state != NOUVEAU_CARD_INIT_DONE) {       \
+		NV_ERROR(dev, "called without init\n");       \
+		return -EINVAL;                               \
+	}                                                     \
+} while (0)
+
+#define NOUVEAU_GET_USER_CHANNEL_WITH_RETURN(id, cl, ch) do {    \
+	struct drm_nouveau_private *nv = dev->dev_private;       \
+	if (!nouveau_channel_owner(dev, (cl), (id))) {           \
+		NV_ERROR(dev, "pid %d doesn't own channel %d\n", \
+			 DRM_CURRENTPID, (id));                  \
+		return -EPERM;                                   \
+	}                                                        \
+	(ch) = nv->fifos[(id)];                                  \
+} while (0)
+
+/* nouveau_drv.c */
+extern int nouveau_noagp;
+extern int nouveau_duallink;
+extern int nouveau_uscript_lvds;
+extern int nouveau_uscript_tmds;
+extern int nouveau_vram_pushbuf;
+extern int nouveau_vram_notify;
+extern int nouveau_fbpercrtc;
+extern char *nouveau_tv_norm;
+extern int nouveau_reg_debug;
+extern char *nouveau_vbios;
+extern int nouveau_ctxfw;
+extern int nouveau_ignorelid;
+extern int nouveau_nofbaccel;
+extern int nouveau_noaccel;
+
+extern int nouveau_pci_suspend(struct pci_dev *pdev, pm_message_t pm_state);
+extern int nouveau_pci_resume(struct pci_dev *pdev);
+
+/* nouveau_state.c */
+extern void nouveau_preclose(struct drm_device *dev, struct drm_file *);
+extern int  nouveau_load(struct drm_device *, unsigned long flags);
+extern int  nouveau_firstopen(struct drm_device *);
+extern void nouveau_lastclose(struct drm_device *);
+extern int  nouveau_unload(struct drm_device *);
+extern int  nouveau_ioctl_getparam(struct drm_device *, void *data,
+				   struct drm_file *);
+extern int  nouveau_ioctl_setparam(struct drm_device *, void *data,
+				   struct drm_file *);
+extern bool nouveau_wait_until(struct drm_device *, uint64_t timeout,
+			       uint32_t reg, uint32_t mask, uint32_t val);
+extern bool nouveau_wait_for_idle(struct drm_device *);
+extern int  nouveau_card_init(struct drm_device *);
+
+/* nouveau_mem.c */
+extern int  nouveau_mem_init_heap(struct mem_block **, uint64_t start,
+				 uint64_t size);
+extern struct mem_block *nouveau_mem_alloc_block(struct mem_block *,
+						 uint64_t size, int align2,
+						 struct drm_file *, int tail);
+extern void nouveau_mem_takedown(struct mem_block **heap);
+extern void nouveau_mem_free_block(struct mem_block *);
+extern uint64_t nouveau_mem_fb_amount(struct drm_device *);
+extern void nouveau_mem_release(struct drm_file *, struct mem_block *heap);
+extern int  nouveau_mem_init(struct drm_device *);
+extern int  nouveau_mem_init_agp(struct drm_device *);
+extern void nouveau_mem_close(struct drm_device *);
+extern struct nouveau_tile_reg *nv10_mem_set_tiling(struct drm_device *dev,
+						    uint32_t addr,
+						    uint32_t size,
+						    uint32_t pitch);
+extern void nv10_mem_expire_tiling(struct drm_device *dev,
+				   struct nouveau_tile_reg *tile,
+				   struct nouveau_fence *fence);
+extern int  nv50_mem_vm_bind_linear(struct drm_device *, uint64_t virt,
+				    uint32_t size, uint32_t flags,
+				    uint64_t phys);
+extern void nv50_mem_vm_unbind(struct drm_device *, uint64_t virt,
+			       uint32_t size);
+
+/* nouveau_notifier.c */
+extern int  nouveau_notifier_init_channel(struct nouveau_channel *);
+extern void nouveau_notifier_takedown_channel(struct nouveau_channel *);
+extern int  nouveau_notifier_alloc(struct nouveau_channel *, uint32_t handle,
+				   int cout, uint32_t *offset);
+extern int  nouveau_notifier_offset(struct nouveau_gpuobj *, uint32_t *);
+extern int  nouveau_ioctl_notifier_alloc(struct drm_device *, void *data,
+					 struct drm_file *);
+extern int  nouveau_ioctl_notifier_free(struct drm_device *, void *data,
+					struct drm_file *);
+
+/* nouveau_channel.c */
+extern struct drm_ioctl_desc nouveau_ioctls[];
+extern int nouveau_max_ioctl;
+extern void nouveau_channel_cleanup(struct drm_device *, struct drm_file *);
+extern int  nouveau_channel_owner(struct drm_device *, struct drm_file *,
+				  int channel);
+extern int  nouveau_channel_alloc(struct drm_device *dev,
+				  struct nouveau_channel **chan,
+				  struct drm_file *file_priv,
+				  uint32_t fb_ctxdma, uint32_t tt_ctxdma);
+extern void nouveau_channel_free(struct nouveau_channel *);
+
+/* nouveau_object.c */
+extern int  nouveau_gpuobj_early_init(struct drm_device *);
+extern int  nouveau_gpuobj_init(struct drm_device *);
+extern void nouveau_gpuobj_takedown(struct drm_device *);
+extern void nouveau_gpuobj_late_takedown(struct drm_device *);
+extern int  nouveau_gpuobj_suspend(struct drm_device *dev);
+extern void nouveau_gpuobj_suspend_cleanup(struct drm_device *dev);
+extern void nouveau_gpuobj_resume(struct drm_device *dev);
+extern int nouveau_gpuobj_channel_init(struct nouveau_channel *,
+				       uint32_t vram_h, uint32_t tt_h);
+extern void nouveau_gpuobj_channel_takedown(struct nouveau_channel *);
+extern int nouveau_gpuobj_new(struct drm_device *, struct nouveau_channel *,
+			      uint32_t size, int align, uint32_t flags,
+			      struct nouveau_gpuobj **);
+extern int nouveau_gpuobj_del(struct drm_device *, struct nouveau_gpuobj **);
+extern int nouveau_gpuobj_ref_add(struct drm_device *, struct nouveau_channel *,
+				  uint32_t handle, struct nouveau_gpuobj *,
+				  struct nouveau_gpuobj_ref **);
+extern int nouveau_gpuobj_ref_del(struct drm_device *,
+				  struct nouveau_gpuobj_ref **);
+extern int nouveau_gpuobj_ref_find(struct nouveau_channel *, uint32_t handle,
+				   struct nouveau_gpuobj_ref **ref_ret);
+extern int nouveau_gpuobj_new_ref(struct drm_device *,
+				  struct nouveau_channel *alloc_chan,
+				  struct nouveau_channel *ref_chan,
+				  uint32_t handle, uint32_t size, int align,
+				  uint32_t flags, struct nouveau_gpuobj_ref **);
+extern int nouveau_gpuobj_new_fake(struct drm_device *,
+				   uint32_t p_offset, uint32_t b_offset,
+				   uint32_t size, uint32_t flags,
+				   struct nouveau_gpuobj **,
+				   struct nouveau_gpuobj_ref**);
+extern int nouveau_gpuobj_dma_new(struct nouveau_channel *, int class,
+				  uint64_t offset, uint64_t size, int access,
+				  int target, struct nouveau_gpuobj **);
+extern int nouveau_gpuobj_gart_dma_new(struct nouveau_channel *,
+				       uint64_t offset, uint64_t size,
+				       int access, struct nouveau_gpuobj **,
+				       uint32_t *o_ret);
+extern int nouveau_gpuobj_gr_new(struct nouveau_channel *, int class,
+				 struct nouveau_gpuobj **);
+extern int nouveau_gpuobj_sw_new(struct nouveau_channel *, int class,
+				 struct nouveau_gpuobj **);
+extern int nouveau_ioctl_grobj_alloc(struct drm_device *, void *data,
+				     struct drm_file *);
+extern int nouveau_ioctl_gpuobj_free(struct drm_device *, void *data,
+				     struct drm_file *);
+
+/* nouveau_irq.c */
+extern irqreturn_t nouveau_irq_handler(DRM_IRQ_ARGS);
+extern void        nouveau_irq_preinstall(struct drm_device *);
+extern int         nouveau_irq_postinstall(struct drm_device *);
+extern void        nouveau_irq_uninstall(struct drm_device *);
+
+/* nouveau_sgdma.c */
+extern int nouveau_sgdma_init(struct drm_device *);
+extern void nouveau_sgdma_takedown(struct drm_device *);
+extern int nouveau_sgdma_get_page(struct drm_device *, uint32_t offset,
+				  uint32_t *page);
+extern struct ttm_backend *nouveau_sgdma_init_ttm(struct drm_device *);
+
+/* nouveau_debugfs.c */
+#if defined(CONFIG_DRM_NOUVEAU_DEBUG)
+extern int  nouveau_debugfs_init(struct drm_minor *);
+extern void nouveau_debugfs_takedown(struct drm_minor *);
+extern int  nouveau_debugfs_channel_init(struct nouveau_channel *);
+extern void nouveau_debugfs_channel_fini(struct nouveau_channel *);
+#else
+static inline int
+nouveau_debugfs_init(struct drm_minor *minor)
+{
+	return 0;
+}
+
+static inline void nouveau_debugfs_takedown(struct drm_minor *minor)
+{
+}
+
+static inline int
+nouveau_debugfs_channel_init(struct nouveau_channel *chan)
+{
+	return 0;
+}
+
+static inline void
+nouveau_debugfs_channel_fini(struct nouveau_channel *chan)
+{
+}
+#endif
+
+/* nouveau_dma.c */
+extern void nouveau_dma_pre_init(struct nouveau_channel *);
+extern int  nouveau_dma_init(struct nouveau_channel *);
+extern int  nouveau_dma_wait(struct nouveau_channel *, int slots, int size);
+
+/* nouveau_acpi.c */
+#if defined(CONFIG_ACPI)
+void nouveau_register_dsm_handler(void);
+void nouveau_unregister_dsm_handler(void);
+#else
+static inline void nouveau_register_dsm_handler(void) {}
+static inline void nouveau_unregister_dsm_handler(void) {}
+#endif
+
+/* nouveau_backlight.c */
+#ifdef CONFIG_DRM_NOUVEAU_BACKLIGHT
+extern int nouveau_backlight_init(struct drm_device *);
+extern void nouveau_backlight_exit(struct drm_device *);
+#else
+static inline int nouveau_backlight_init(struct drm_device *dev)
+{
+	return 0;
+}
+
+static inline void nouveau_backlight_exit(struct drm_device *dev) { }
+#endif
+
+/* nouveau_bios.c */
+extern int nouveau_bios_init(struct drm_device *);
+extern void nouveau_bios_takedown(struct drm_device *dev);
+extern int nouveau_run_vbios_init(struct drm_device *);
+extern void nouveau_bios_run_init_table(struct drm_device *, uint16_t table,
+					struct dcb_entry *);
+extern struct dcb_gpio_entry *nouveau_bios_gpio_entry(struct drm_device *,
+						      enum dcb_gpio_tag);
+extern struct dcb_connector_table_entry *
+nouveau_bios_connector_entry(struct drm_device *, int index);
+extern int get_pll_limits(struct drm_device *, uint32_t limit_match,
+			  struct pll_lims *);
+extern int nouveau_bios_run_display_table(struct drm_device *,
+					  struct dcb_entry *,
+					  uint32_t script, int pxclk);
+extern void *nouveau_bios_dp_table(struct drm_device *, struct dcb_entry *,
+				   int *length);
+extern bool nouveau_bios_fp_mode(struct drm_device *, struct drm_display_mode *);
+extern uint8_t *nouveau_bios_embedded_edid(struct drm_device *);
+extern int nouveau_bios_parse_lvds_table(struct drm_device *, int pxclk,
+					 bool *dl, bool *if_is_24bit);
+extern int run_tmds_table(struct drm_device *, struct dcb_entry *,
+			  int head, int pxclk);
+extern int call_lvds_script(struct drm_device *, struct dcb_entry *, int head,
+			    enum LVDS_script, int pxclk);
+
+/* nouveau_ttm.c */
+int nouveau_ttm_global_init(struct drm_nouveau_private *);
+void nouveau_ttm_global_release(struct drm_nouveau_private *);
+int nouveau_ttm_mmap(struct file *, struct vm_area_struct *);
+
+/* nouveau_dp.c */
+int nouveau_dp_auxch(struct nouveau_i2c_chan *auxch, int cmd, int addr,
+		     uint8_t *data, int data_nr);
+bool nouveau_dp_detect(struct drm_encoder *);
+bool nouveau_dp_link_train(struct drm_encoder *);
+
+/* nv04_fb.c */
+extern int  nv04_fb_init(struct drm_device *);
+extern void nv04_fb_takedown(struct drm_device *);
+
+/* nv10_fb.c */
+extern int  nv10_fb_init(struct drm_device *);
+extern void nv10_fb_takedown(struct drm_device *);
+extern void nv10_fb_set_region_tiling(struct drm_device *, int, uint32_t,
+				      uint32_t, uint32_t);
+
+/* nv40_fb.c */
+extern int  nv40_fb_init(struct drm_device *);
+extern void nv40_fb_takedown(struct drm_device *);
+extern void nv40_fb_set_region_tiling(struct drm_device *, int, uint32_t,
+				      uint32_t, uint32_t);
+
+/* nv04_fifo.c */
+extern int  nv04_fifo_init(struct drm_device *);
+extern void nv04_fifo_disable(struct drm_device *);
+extern void nv04_fifo_enable(struct drm_device *);
+extern bool nv04_fifo_reassign(struct drm_device *, bool);
+extern bool nv04_fifo_cache_flush(struct drm_device *);
+extern bool nv04_fifo_cache_pull(struct drm_device *, bool);
+extern int  nv04_fifo_channel_id(struct drm_device *);
+extern int  nv04_fifo_create_context(struct nouveau_channel *);
+extern void nv04_fifo_destroy_context(struct nouveau_channel *);
+extern int  nv04_fifo_load_context(struct nouveau_channel *);
+extern int  nv04_fifo_unload_context(struct drm_device *);
+
+/* nv10_fifo.c */
+extern int  nv10_fifo_init(struct drm_device *);
+extern int  nv10_fifo_channel_id(struct drm_device *);
+extern int  nv10_fifo_create_context(struct nouveau_channel *);
+extern void nv10_fifo_destroy_context(struct nouveau_channel *);
+extern int  nv10_fifo_load_context(struct nouveau_channel *);
+extern int  nv10_fifo_unload_context(struct drm_device *);
+
+/* nv40_fifo.c */
+extern int  nv40_fifo_init(struct drm_device *);
+extern int  nv40_fifo_create_context(struct nouveau_channel *);
+extern void nv40_fifo_destroy_context(struct nouveau_channel *);
+extern int  nv40_fifo_load_context(struct nouveau_channel *);
+extern int  nv40_fifo_unload_context(struct drm_device *);
+
+/* nv50_fifo.c */
+extern int  nv50_fifo_init(struct drm_device *);
+extern void nv50_fifo_takedown(struct drm_device *);
+extern int  nv50_fifo_channel_id(struct drm_device *);
+extern int  nv50_fifo_create_context(struct nouveau_channel *);
+extern void nv50_fifo_destroy_context(struct nouveau_channel *);
+extern int  nv50_fifo_load_context(struct nouveau_channel *);
+extern int  nv50_fifo_unload_context(struct drm_device *);
+
+/* nv04_graph.c */
+extern struct nouveau_pgraph_object_class nv04_graph_grclass[];
+extern int  nv04_graph_init(struct drm_device *);
+extern void nv04_graph_takedown(struct drm_device *);
+extern void nv04_graph_fifo_access(struct drm_device *, bool);
+extern struct nouveau_channel *nv04_graph_channel(struct drm_device *);
+extern int  nv04_graph_create_context(struct nouveau_channel *);
+extern void nv04_graph_destroy_context(struct nouveau_channel *);
+extern int  nv04_graph_load_context(struct nouveau_channel *);
+extern int  nv04_graph_unload_context(struct drm_device *);
+extern void nv04_graph_context_switch(struct drm_device *);
+
+/* nv10_graph.c */
+extern struct nouveau_pgraph_object_class nv10_graph_grclass[];
+extern int  nv10_graph_init(struct drm_device *);
+extern void nv10_graph_takedown(struct drm_device *);
+extern struct nouveau_channel *nv10_graph_channel(struct drm_device *);
+extern int  nv10_graph_create_context(struct nouveau_channel *);
+extern void nv10_graph_destroy_context(struct nouveau_channel *);
+extern int  nv10_graph_load_context(struct nouveau_channel *);
+extern int  nv10_graph_unload_context(struct drm_device *);
+extern void nv10_graph_context_switch(struct drm_device *);
+extern void nv10_graph_set_region_tiling(struct drm_device *, int, uint32_t,
+					 uint32_t, uint32_t);
+
+/* nv20_graph.c */
+extern struct nouveau_pgraph_object_class nv20_graph_grclass[];
+extern struct nouveau_pgraph_object_class nv30_graph_grclass[];
+extern int  nv20_graph_create_context(struct nouveau_channel *);
+extern void nv20_graph_destroy_context(struct nouveau_channel *);
+extern int  nv20_graph_load_context(struct nouveau_channel *);
+extern int  nv20_graph_unload_context(struct drm_device *);
+extern int  nv20_graph_init(struct drm_device *);
+extern void nv20_graph_takedown(struct drm_device *);
+extern int  nv30_graph_init(struct drm_device *);
+extern void nv20_graph_set_region_tiling(struct drm_device *, int, uint32_t,
+					 uint32_t, uint32_t);
+
+/* nv40_graph.c */
+extern struct nouveau_pgraph_object_class nv40_graph_grclass[];
+extern int  nv40_graph_init(struct drm_device *);
+extern void nv40_graph_takedown(struct drm_device *);
+extern struct nouveau_channel *nv40_graph_channel(struct drm_device *);
+extern int  nv40_graph_create_context(struct nouveau_channel *);
+extern void nv40_graph_destroy_context(struct nouveau_channel *);
+extern int  nv40_graph_load_context(struct nouveau_channel *);
+extern int  nv40_graph_unload_context(struct drm_device *);
+extern void nv40_grctx_init(struct nouveau_grctx *);
+extern void nv40_graph_set_region_tiling(struct drm_device *, int, uint32_t,
+					 uint32_t, uint32_t);
+
+/* nv50_graph.c */
+extern struct nouveau_pgraph_object_class nv50_graph_grclass[];
+extern int  nv50_graph_init(struct drm_device *);
+extern void nv50_graph_takedown(struct drm_device *);
+extern void nv50_graph_fifo_access(struct drm_device *, bool);
+extern struct nouveau_channel *nv50_graph_channel(struct drm_device *);
+extern int  nv50_graph_create_context(struct nouveau_channel *);
+extern void nv50_graph_destroy_context(struct nouveau_channel *);
+extern int  nv50_graph_load_context(struct nouveau_channel *);
+extern int  nv50_graph_unload_context(struct drm_device *);
+extern void nv50_graph_context_switch(struct drm_device *);
+extern int  nv50_grctx_init(struct nouveau_grctx *);
+
+/* nouveau_grctx.c */
+extern int  nouveau_grctx_prog_load(struct drm_device *);
+extern void nouveau_grctx_vals_load(struct drm_device *,
+				    struct nouveau_gpuobj *);
+extern void nouveau_grctx_fini(struct drm_device *);
+
+/* nv04_instmem.c */
+extern int  nv04_instmem_init(struct drm_device *);
+extern void nv04_instmem_takedown(struct drm_device *);
+extern int  nv04_instmem_suspend(struct drm_device *);
+extern void nv04_instmem_resume(struct drm_device *);
+extern int  nv04_instmem_populate(struct drm_device *, struct nouveau_gpuobj *,
+				  uint32_t *size);
+extern void nv04_instmem_clear(struct drm_device *, struct nouveau_gpuobj *);
+extern int  nv04_instmem_bind(struct drm_device *, struct nouveau_gpuobj *);
+extern int  nv04_instmem_unbind(struct drm_device *, struct nouveau_gpuobj *);
+extern void nv04_instmem_prepare_access(struct drm_device *, bool write);
+extern void nv04_instmem_finish_access(struct drm_device *);
+
+/* nv50_instmem.c */
+extern int  nv50_instmem_init(struct drm_device *);
+extern void nv50_instmem_takedown(struct drm_device *);
+extern int  nv50_instmem_suspend(struct drm_device *);
+extern void nv50_instmem_resume(struct drm_device *);
+extern int  nv50_instmem_populate(struct drm_device *, struct nouveau_gpuobj *,
+				  uint32_t *size);
+extern void nv50_instmem_clear(struct drm_device *, struct nouveau_gpuobj *);
+extern int  nv50_instmem_bind(struct drm_device *, struct nouveau_gpuobj *);
+extern int  nv50_instmem_unbind(struct drm_device *, struct nouveau_gpuobj *);
+extern void nv50_instmem_prepare_access(struct drm_device *, bool write);
+extern void nv50_instmem_finish_access(struct drm_device *);
+
+/* nv04_mc.c */
+extern int  nv04_mc_init(struct drm_device *);
+extern void nv04_mc_takedown(struct drm_device *);
+
+/* nv40_mc.c */
+extern int  nv40_mc_init(struct drm_device *);
+extern void nv40_mc_takedown(struct drm_device *);
+
+/* nv50_mc.c */
+extern int  nv50_mc_init(struct drm_device *);
+extern void nv50_mc_takedown(struct drm_device *);
+
+/* nv04_timer.c */
+extern int  nv04_timer_init(struct drm_device *);
+extern uint64_t nv04_timer_read(struct drm_device *);
+extern void nv04_timer_takedown(struct drm_device *);
+
+extern long nouveau_compat_ioctl(struct file *file, unsigned int cmd,
+				 unsigned long arg);
+
+/* nv04_dac.c */
+extern int nv04_dac_create(struct drm_device *dev, struct dcb_entry *entry);
+extern uint32_t nv17_dac_sample_load(struct drm_encoder *encoder);
+extern int nv04_dac_output_offset(struct drm_encoder *encoder);
+extern void nv04_dac_update_dacclk(struct drm_encoder *encoder, bool enable);
+
+/* nv04_dfp.c */
+extern int nv04_dfp_create(struct drm_device *dev, struct dcb_entry *entry);
+extern int nv04_dfp_get_bound_head(struct drm_device *dev, struct dcb_entry *dcbent);
+extern void nv04_dfp_bind_head(struct drm_device *dev, struct dcb_entry *dcbent,
+			       int head, bool dl);
+extern void nv04_dfp_disable(struct drm_device *dev, int head);
+extern void nv04_dfp_update_fp_control(struct drm_encoder *encoder, int mode);
+
+/* nv04_tv.c */
+extern int nv04_tv_identify(struct drm_device *dev, int i2c_index);
+extern int nv04_tv_create(struct drm_device *dev, struct dcb_entry *entry);
+
+/* nv17_tv.c */
+extern int nv17_tv_create(struct drm_device *dev, struct dcb_entry *entry);
+
+/* nv04_display.c */
+extern int nv04_display_create(struct drm_device *);
+extern void nv04_display_destroy(struct drm_device *);
+extern void nv04_display_restore(struct drm_device *);
+
+/* nv04_crtc.c */
+extern int nv04_crtc_create(struct drm_device *, int index);
+
+/* nouveau_bo.c */
+extern struct ttm_bo_driver nouveau_bo_driver;
+extern int nouveau_bo_new(struct drm_device *, struct nouveau_channel *,
+			  int size, int align, uint32_t flags,
+			  uint32_t tile_mode, uint32_t tile_flags,
+			  bool no_vm, bool mappable, struct nouveau_bo **);
+extern int nouveau_bo_pin(struct nouveau_bo *, uint32_t flags);
+extern int nouveau_bo_unpin(struct nouveau_bo *);
+extern int nouveau_bo_map(struct nouveau_bo *);
+extern void nouveau_bo_unmap(struct nouveau_bo *);
+extern void nouveau_bo_placement_set(struct nouveau_bo *, uint32_t memtype);
+extern u16 nouveau_bo_rd16(struct nouveau_bo *nvbo, unsigned index);
+extern void nouveau_bo_wr16(struct nouveau_bo *nvbo, unsigned index, u16 val);
+extern u32 nouveau_bo_rd32(struct nouveau_bo *nvbo, unsigned index);
+extern void nouveau_bo_wr32(struct nouveau_bo *nvbo, unsigned index, u32 val);
+
+/* nouveau_fence.c */
+struct nouveau_fence;
+extern int nouveau_fence_init(struct nouveau_channel *);
+extern void nouveau_fence_fini(struct nouveau_channel *);
+extern void nouveau_fence_update(struct nouveau_channel *);
+extern int nouveau_fence_new(struct nouveau_channel *, struct nouveau_fence **,
+			     bool emit);
+extern int nouveau_fence_emit(struct nouveau_fence *);
+struct nouveau_channel *nouveau_fence_channel(struct nouveau_fence *);
+extern bool nouveau_fence_signalled(void *obj, void *arg);
+extern int nouveau_fence_wait(void *obj, void *arg, bool lazy, bool intr);
+extern int nouveau_fence_flush(void *obj, void *arg);
+extern void nouveau_fence_unref(void **obj);
+extern void *nouveau_fence_ref(void *obj);
+extern void nouveau_fence_handler(struct drm_device *dev, int channel);
+
+/* nouveau_gem.c */
+extern int nouveau_gem_new(struct drm_device *, struct nouveau_channel *,
+			   int size, int align, uint32_t flags,
+			   uint32_t tile_mode, uint32_t tile_flags,
+			   bool no_vm, bool mappable, struct nouveau_bo **);
+extern int nouveau_gem_object_new(struct drm_gem_object *);
+extern void nouveau_gem_object_del(struct drm_gem_object *);
+extern int nouveau_gem_ioctl_new(struct drm_device *, void *,
+				 struct drm_file *);
+extern int nouveau_gem_ioctl_pushbuf(struct drm_device *, void *,
+				     struct drm_file *);
+extern int nouveau_gem_ioctl_cpu_prep(struct drm_device *, void *,
+				      struct drm_file *);
+extern int nouveau_gem_ioctl_cpu_fini(struct drm_device *, void *,
+				      struct drm_file *);
+extern int nouveau_gem_ioctl_info(struct drm_device *, void *,
+				  struct drm_file *);
+
+/* nv17_gpio.c */
+int nv17_gpio_get(struct drm_device *dev, enum dcb_gpio_tag tag);
+int nv17_gpio_set(struct drm_device *dev, enum dcb_gpio_tag tag, int state);
+
+#ifndef ioread32_native
+#ifdef __BIG_ENDIAN
+#define ioread16_native ioread16be
+#define iowrite16_native iowrite16be
+#define ioread32_native  ioread32be
+#define iowrite32_native iowrite32be
+#else /* def __BIG_ENDIAN */
+#define ioread16_native ioread16
+#define iowrite16_native iowrite16
+#define ioread32_native  ioread32
+#define iowrite32_native iowrite32
+#endif /* def __BIG_ENDIAN else */
+#endif /* !ioread32_native */
+
+/* channel control reg access */
+static inline u32 nvchan_rd32(struct nouveau_channel *chan, unsigned reg)
+{
+	return ioread32_native(chan->user + reg);
+}
+
+static inline void nvchan_wr32(struct nouveau_channel *chan,
+							unsigned reg, u32 val)
+{
+	iowrite32_native(val, chan->user + reg);
+}
+
+/* register access */
+static inline u32 nv_rd32(struct drm_device *dev, unsigned reg)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	return ioread32_native(dev_priv->mmio + reg);
+}
+
+static inline void nv_wr32(struct drm_device *dev, unsigned reg, u32 val)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	iowrite32_native(val, dev_priv->mmio + reg);
+}
+
+static inline u8 nv_rd08(struct drm_device *dev, unsigned reg)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	return ioread8(dev_priv->mmio + reg);
+}
+
+static inline void nv_wr08(struct drm_device *dev, unsigned reg, u8 val)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	iowrite8(val, dev_priv->mmio + reg);
+}
+
+#define nv_wait(reg, mask, val) \
+	nouveau_wait_until(dev, 2000000000ULL, (reg), (mask), (val))
+
+/* PRAMIN access */
+static inline u32 nv_ri32(struct drm_device *dev, unsigned offset)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	return ioread32_native(dev_priv->ramin + offset);
+}
+
+static inline void nv_wi32(struct drm_device *dev, unsigned offset, u32 val)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	iowrite32_native(val, dev_priv->ramin + offset);
+}
+
+/* object access */
+static inline u32 nv_ro32(struct drm_device *dev, struct nouveau_gpuobj *obj,
+				unsigned index)
+{
+	return nv_ri32(dev, obj->im_pramin->start + index * 4);
+}
+
+static inline void nv_wo32(struct drm_device *dev, struct nouveau_gpuobj *obj,
+				unsigned index, u32 val)
+{
+	nv_wi32(dev, obj->im_pramin->start + index * 4, val);
+}
+
+/*
+ * Logging
+ * Argument d is (struct drm_device *).
+ */
+#define NV_PRINTK(level, d, fmt, arg...) \
+	printk(level "[" DRM_NAME "] " DRIVER_NAME " %s: " fmt, \
+					pci_name(d->pdev), ##arg)
+#ifndef NV_DEBUG_NOTRACE
+#define NV_DEBUG(d, fmt, arg...) do {                                          \
+	if (drm_debug & DRM_UT_DRIVER) {                                       \
+		NV_PRINTK(KERN_DEBUG, d, "%s:%d - " fmt, __func__,             \
+			  __LINE__, ##arg);                                    \
+	}                                                                      \
+} while (0)
+#define NV_DEBUG_KMS(d, fmt, arg...) do {                                      \
+	if (drm_debug & DRM_UT_KMS) {                                          \
+		NV_PRINTK(KERN_DEBUG, d, "%s:%d - " fmt, __func__,             \
+			  __LINE__, ##arg);                                    \
+	}                                                                      \
+} while (0)
+#else
+#define NV_DEBUG(d, fmt, arg...) do {                                          \
+	if (drm_debug & DRM_UT_DRIVER)                                         \
+		NV_PRINTK(KERN_DEBUG, d, fmt, ##arg);                          \
+} while (0)
+#define NV_DEBUG_KMS(d, fmt, arg...) do {                                      \
+	if (drm_debug & DRM_UT_KMS)                                            \
+		NV_PRINTK(KERN_DEBUG, d, fmt, ##arg);                          \
+} while (0)
+#endif
+#define NV_ERROR(d, fmt, arg...) NV_PRINTK(KERN_ERR, d, fmt, ##arg)
+#define NV_INFO(d, fmt, arg...) NV_PRINTK(KERN_INFO, d, fmt, ##arg)
+#define NV_TRACEWARN(d, fmt, arg...) NV_PRINTK(KERN_NOTICE, d, fmt, ##arg)
+#define NV_TRACE(d, fmt, arg...) NV_PRINTK(KERN_INFO, d, fmt, ##arg)
+#define NV_WARN(d, fmt, arg...) NV_PRINTK(KERN_WARNING, d, fmt, ##arg)
+
+/* nouveau_reg_debug bitmask */
+enum {
+	NOUVEAU_REG_DEBUG_MC             = 0x1,
+	NOUVEAU_REG_DEBUG_VIDEO          = 0x2,
+	NOUVEAU_REG_DEBUG_FB             = 0x4,
+	NOUVEAU_REG_DEBUG_EXTDEV         = 0x8,
+	NOUVEAU_REG_DEBUG_CRTC           = 0x10,
+	NOUVEAU_REG_DEBUG_RAMDAC         = 0x20,
+	NOUVEAU_REG_DEBUG_VGACRTC        = 0x40,
+	NOUVEAU_REG_DEBUG_RMVIO          = 0x80,
+	NOUVEAU_REG_DEBUG_VGAATTR        = 0x100,
+	NOUVEAU_REG_DEBUG_EVO            = 0x200,
+};
+
+#define NV_REG_DEBUG(type, dev, fmt, arg...) do { \
+	if (nouveau_reg_debug & NOUVEAU_REG_DEBUG_##type) \
+		NV_PRINTK(KERN_DEBUG, dev, "%s: " fmt, __func__, ##arg); \
+} while (0)
+
+static inline bool
+nv_two_heads(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	const int impl = dev->pci_device & 0x0ff0;
+
+	if (dev_priv->card_type >= NV_10 && impl != 0x0100 &&
+	    impl != 0x0150 && impl != 0x01a0 && impl != 0x0200)
+		return true;
+
+	return false;
+}
+
+static inline bool
+nv_gf4_disp_arch(struct drm_device *dev)
+{
+	return nv_two_heads(dev) && (dev->pci_device & 0x0ff0) != 0x0110;
+}
+
+static inline bool
+nv_two_reg_pll(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	const int impl = dev->pci_device & 0x0ff0;
+
+	if (impl == 0x0310 || impl == 0x0340 || dev_priv->card_type >= NV_40)
+		return true;
+	return false;
+}
+
+#define NV_SW                                                        0x0000506e
+#define NV_SW_DMA_SEMAPHORE                                          0x00000060
+#define NV_SW_SEMAPHORE_OFFSET                                       0x00000064
+#define NV_SW_SEMAPHORE_ACQUIRE                                      0x00000068
+#define NV_SW_SEMAPHORE_RELEASE                                      0x0000006c
+#define NV_SW_DMA_VBLSEM                                             0x0000018c
+#define NV_SW_VBLSEM_OFFSET                                          0x00000400
+#define NV_SW_VBLSEM_RELEASE_VALUE                                   0x00000404
+#define NV_SW_VBLSEM_RELEASE                                         0x00000408
+
+#endif /* __NOUVEAU_DRV_H__ */
diff --git a/drivers/gpu/drm/nouveau/nouveau_encoder.h b/drivers/gpu/drm/nouveau/nouveau_encoder.h
new file mode 100644
index 000000000000..bc4a24029ed1
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_encoder.h
@@ -0,0 +1,91 @@
+/*
+ * Copyright (C) 2008 Maarten Maathuis.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __NOUVEAU_ENCODER_H__
+#define __NOUVEAU_ENCODER_H__
+
+#include "drm_encoder_slave.h"
+#include "nouveau_drv.h"
+
+#define NV_DPMS_CLEARED 0x80
+
+struct nouveau_encoder {
+	struct drm_encoder_slave base;
+
+	struct dcb_entry *dcb;
+	int or;
+
+	struct drm_display_mode mode;
+	int last_dpms;
+
+	struct nv04_output_reg restore;
+
+	void (*disconnect)(struct nouveau_encoder *encoder);
+
+	union {
+		struct {
+			int dpcd_version;
+			int link_nr;
+			int link_bw;
+		} dp;
+	};
+};
+
+static inline struct nouveau_encoder *nouveau_encoder(struct drm_encoder *enc)
+{
+	struct drm_encoder_slave *slave = to_encoder_slave(enc);
+
+	return container_of(slave, struct nouveau_encoder, base);
+}
+
+static inline struct drm_encoder *to_drm_encoder(struct nouveau_encoder *enc)
+{
+	return &enc->base.base;
+}
+
+struct nouveau_connector *
+nouveau_encoder_connector_get(struct nouveau_encoder *encoder);
+int nv50_sor_create(struct drm_device *dev, struct dcb_entry *entry);
+int nv50_dac_create(struct drm_device *dev, struct dcb_entry *entry);
+
+struct bit_displayport_encoder_table {
+	uint32_t match;
+	uint8_t  record_nr;
+	uint8_t  unknown;
+	uint16_t script0;
+	uint16_t script1;
+	uint16_t unknown_table;
+} __attribute__ ((packed));
+
+struct bit_displayport_encoder_table_entry {
+	uint8_t vs_level;
+	uint8_t pre_level;
+	uint8_t reg0;
+	uint8_t reg1;
+	uint8_t reg2;
+} __attribute__ ((packed));
+
+#endif /* __NOUVEAU_ENCODER_H__ */
diff --git a/drivers/gpu/drm/nouveau/nouveau_fb.h b/drivers/gpu/drm/nouveau/nouveau_fb.h
new file mode 100644
index 000000000000..4a3f31aa1949
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_fb.h
@@ -0,0 +1,47 @@
+/*
+ * Copyright (C) 2008 Maarten Maathuis.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __NOUVEAU_FB_H__
+#define __NOUVEAU_FB_H__
+
+struct nouveau_framebuffer {
+	struct drm_framebuffer base;
+	struct nouveau_bo *nvbo;
+};
+
+static inline struct nouveau_framebuffer *
+nouveau_framebuffer(struct drm_framebuffer *fb)
+{
+	return container_of(fb, struct nouveau_framebuffer, base);
+}
+
+extern const struct drm_mode_config_funcs nouveau_mode_config_funcs;
+
+struct drm_framebuffer *
+nouveau_framebuffer_create(struct drm_device *, struct nouveau_bo *,
+			   struct drm_mode_fb_cmd *);
+
+#endif /* __NOUVEAU_FB_H__ */
diff --git a/drivers/gpu/drm/nouveau/nouveau_fbcon.c b/drivers/gpu/drm/nouveau/nouveau_fbcon.c
new file mode 100644
index 000000000000..68cedd9194fe
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_fbcon.c
@@ -0,0 +1,423 @@
+/*
+ * Copyright © 2007 David Airlie
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ *     David Airlie
+ */
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/errno.h>
+#include <linux/string.h>
+#include <linux/mm.h>
+#include <linux/tty.h>
+#include <linux/slab.h>
+#include <linux/sysrq.h>
+#include <linux/delay.h>
+#include <linux/fb.h>
+#include <linux/init.h>
+#include <linux/screen_info.h>
+#include <linux/vga_switcheroo.h>
+
+#include "drmP.h"
+#include "drm.h"
+#include "drm_crtc.h"
+#include "drm_crtc_helper.h"
+#include "drm_fb_helper.h"
+#include "nouveau_drv.h"
+#include "nouveau_drm.h"
+#include "nouveau_crtc.h"
+#include "nouveau_fb.h"
+#include "nouveau_fbcon.h"
+#include "nouveau_dma.h"
+
+static int
+nouveau_fbcon_sync(struct fb_info *info)
+{
+	struct nouveau_fbcon_par *par = info->par;
+	struct drm_device *dev = par->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_channel *chan = dev_priv->channel;
+	int ret, i;
+
+	if (!chan || !chan->accel_done ||
+	    info->state != FBINFO_STATE_RUNNING ||
+	    info->flags & FBINFO_HWACCEL_DISABLED)
+		return 0;
+
+	if (RING_SPACE(chan, 4)) {
+		nouveau_fbcon_gpu_lockup(info);
+		return 0;
+	}
+
+	BEGIN_RING(chan, 0, 0x0104, 1);
+	OUT_RING(chan, 0);
+	BEGIN_RING(chan, 0, 0x0100, 1);
+	OUT_RING(chan, 0);
+	nouveau_bo_wr32(chan->notifier_bo, chan->m2mf_ntfy + 3, 0xffffffff);
+	FIRE_RING(chan);
+
+	ret = -EBUSY;
+	for (i = 0; i < 100000; i++) {
+		if (!nouveau_bo_rd32(chan->notifier_bo, chan->m2mf_ntfy + 3)) {
+			ret = 0;
+			break;
+		}
+		DRM_UDELAY(1);
+	}
+
+	if (ret) {
+		nouveau_fbcon_gpu_lockup(info);
+		return 0;
+	}
+
+	chan->accel_done = false;
+	return 0;
+}
+
+static struct fb_ops nouveau_fbcon_ops = {
+	.owner = THIS_MODULE,
+	.fb_check_var = drm_fb_helper_check_var,
+	.fb_set_par = drm_fb_helper_set_par,
+	.fb_setcolreg = drm_fb_helper_setcolreg,
+	.fb_fillrect = cfb_fillrect,
+	.fb_copyarea = cfb_copyarea,
+	.fb_imageblit = cfb_imageblit,
+	.fb_sync = nouveau_fbcon_sync,
+	.fb_pan_display = drm_fb_helper_pan_display,
+	.fb_blank = drm_fb_helper_blank,
+	.fb_setcmap = drm_fb_helper_setcmap,
+};
+
+static struct fb_ops nv04_fbcon_ops = {
+	.owner = THIS_MODULE,
+	.fb_check_var = drm_fb_helper_check_var,
+	.fb_set_par = drm_fb_helper_set_par,
+	.fb_setcolreg = drm_fb_helper_setcolreg,
+	.fb_fillrect = nv04_fbcon_fillrect,
+	.fb_copyarea = nv04_fbcon_copyarea,
+	.fb_imageblit = nv04_fbcon_imageblit,
+	.fb_sync = nouveau_fbcon_sync,
+	.fb_pan_display = drm_fb_helper_pan_display,
+	.fb_blank = drm_fb_helper_blank,
+	.fb_setcmap = drm_fb_helper_setcmap,
+};
+
+static struct fb_ops nv50_fbcon_ops = {
+	.owner = THIS_MODULE,
+	.fb_check_var = drm_fb_helper_check_var,
+	.fb_set_par = drm_fb_helper_set_par,
+	.fb_setcolreg = drm_fb_helper_setcolreg,
+	.fb_fillrect = nv50_fbcon_fillrect,
+	.fb_copyarea = nv50_fbcon_copyarea,
+	.fb_imageblit = nv50_fbcon_imageblit,
+	.fb_sync = nouveau_fbcon_sync,
+	.fb_pan_display = drm_fb_helper_pan_display,
+	.fb_blank = drm_fb_helper_blank,
+	.fb_setcmap = drm_fb_helper_setcmap,
+};
+
+static void nouveau_fbcon_gamma_set(struct drm_crtc *crtc, u16 red, u16 green,
+				    u16 blue, int regno)
+{
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+
+	nv_crtc->lut.r[regno] = red;
+	nv_crtc->lut.g[regno] = green;
+	nv_crtc->lut.b[regno] = blue;
+}
+
+static void nouveau_fbcon_gamma_get(struct drm_crtc *crtc, u16 *red, u16 *green,
+				    u16 *blue, int regno)
+{
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+
+	*red = nv_crtc->lut.r[regno];
+	*green = nv_crtc->lut.g[regno];
+	*blue = nv_crtc->lut.b[regno];
+}
+
+static struct drm_fb_helper_funcs nouveau_fbcon_helper_funcs = {
+	.gamma_set = nouveau_fbcon_gamma_set,
+	.gamma_get = nouveau_fbcon_gamma_get
+};
+
+#if defined(__i386__) || defined(__x86_64__)
+static bool
+nouveau_fbcon_has_vesafb_or_efifb(struct drm_device *dev)
+{
+	struct pci_dev *pdev = dev->pdev;
+	int ramin;
+
+	if (screen_info.orig_video_isVGA != VIDEO_TYPE_VLFB &&
+	    screen_info.orig_video_isVGA != VIDEO_TYPE_EFI)
+		return false;
+
+	if (screen_info.lfb_base < pci_resource_start(pdev, 1))
+		goto not_fb;
+
+	if (screen_info.lfb_base + screen_info.lfb_size >=
+	    pci_resource_start(pdev, 1) + pci_resource_len(pdev, 1))
+		goto not_fb;
+
+	return true;
+not_fb:
+	ramin = 2;
+	if (pci_resource_len(pdev, ramin) == 0) {
+		ramin = 3;
+		if (pci_resource_len(pdev, ramin) == 0)
+			return false;
+	}
+
+	if (screen_info.lfb_base < pci_resource_start(pdev, ramin))
+		return false;
+
+	if (screen_info.lfb_base + screen_info.lfb_size >=
+	    pci_resource_start(pdev, ramin) + pci_resource_len(pdev, ramin))
+		return false;
+
+	return true;
+}
+#endif
+
+void
+nouveau_fbcon_zfill(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct fb_info *info = dev_priv->fbdev_info;
+	struct fb_fillrect rect;
+
+	/* Clear the entire fbcon.  The drm will program every connector
+	 * with it's preferred mode.  If the sizes differ, one display will
+	 * quite likely have garbage around the console.
+	 */
+	rect.dx = rect.dy = 0;
+	rect.width = info->var.xres_virtual;
+	rect.height = info->var.yres_virtual;
+	rect.color = 0;
+	rect.rop = ROP_COPY;
+	info->fbops->fb_fillrect(info, &rect);
+}
+
+static int
+nouveau_fbcon_create(struct drm_device *dev, uint32_t fb_width,
+		     uint32_t fb_height, uint32_t surface_width,
+		     uint32_t surface_height, uint32_t surface_depth,
+		     uint32_t surface_bpp, struct drm_framebuffer **pfb)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct fb_info *info;
+	struct nouveau_fbcon_par *par;
+	struct drm_framebuffer *fb;
+	struct nouveau_framebuffer *nouveau_fb;
+	struct nouveau_bo *nvbo;
+	struct drm_mode_fb_cmd mode_cmd;
+	struct device *device = &dev->pdev->dev;
+	int size, ret;
+
+	mode_cmd.width = surface_width;
+	mode_cmd.height = surface_height;
+
+	mode_cmd.bpp = surface_bpp;
+	mode_cmd.pitch = mode_cmd.width * (mode_cmd.bpp >> 3);
+	mode_cmd.pitch = roundup(mode_cmd.pitch, 256);
+	mode_cmd.depth = surface_depth;
+
+	size = mode_cmd.pitch * mode_cmd.height;
+	size = roundup(size, PAGE_SIZE);
+
+	ret = nouveau_gem_new(dev, dev_priv->channel, size, 0, TTM_PL_FLAG_VRAM,
+			      0, 0x0000, false, true, &nvbo);
+	if (ret) {
+		NV_ERROR(dev, "failed to allocate framebuffer\n");
+		goto out;
+	}
+
+	ret = nouveau_bo_pin(nvbo, TTM_PL_FLAG_VRAM);
+	if (ret) {
+		NV_ERROR(dev, "failed to pin fb: %d\n", ret);
+		nouveau_bo_ref(NULL, &nvbo);
+		goto out;
+	}
+
+	ret = nouveau_bo_map(nvbo);
+	if (ret) {
+		NV_ERROR(dev, "failed to map fb: %d\n", ret);
+		nouveau_bo_unpin(nvbo);
+		nouveau_bo_ref(NULL, &nvbo);
+		goto out;
+	}
+
+	mutex_lock(&dev->struct_mutex);
+
+	fb = nouveau_framebuffer_create(dev, nvbo, &mode_cmd);
+	if (!fb) {
+		ret = -ENOMEM;
+		NV_ERROR(dev, "failed to allocate fb.\n");
+		goto out_unref;
+	}
+
+	list_add(&fb->filp_head, &dev->mode_config.fb_kernel_list);
+
+	nouveau_fb = nouveau_framebuffer(fb);
+	*pfb = fb;
+
+	info = framebuffer_alloc(sizeof(struct nouveau_fbcon_par), device);
+	if (!info) {
+		ret = -ENOMEM;
+		goto out_unref;
+	}
+
+	par = info->par;
+	par->helper.funcs = &nouveau_fbcon_helper_funcs;
+	par->helper.dev = dev;
+	ret = drm_fb_helper_init_crtc_count(&par->helper, 2, 4);
+	if (ret)
+		goto out_unref;
+	dev_priv->fbdev_info = info;
+
+	strcpy(info->fix.id, "nouveaufb");
+	if (nouveau_nofbaccel)
+		info->flags = FBINFO_DEFAULT | FBINFO_HWACCEL_DISABLED;
+	else
+		info->flags = FBINFO_DEFAULT | FBINFO_HWACCEL_COPYAREA |
+			      FBINFO_HWACCEL_FILLRECT |
+			      FBINFO_HWACCEL_IMAGEBLIT;
+	info->fbops = &nouveau_fbcon_ops;
+	info->fix.smem_start = dev->mode_config.fb_base + nvbo->bo.offset -
+			       dev_priv->vm_vram_base;
+	info->fix.smem_len = size;
+
+	info->screen_base = nvbo_kmap_obj_iovirtual(nouveau_fb->nvbo);
+	info->screen_size = size;
+
+	drm_fb_helper_fill_fix(info, fb->pitch, fb->depth);
+	drm_fb_helper_fill_var(info, fb, fb_width, fb_height);
+
+	/* FIXME: we really shouldn't expose mmio space at all */
+	info->fix.mmio_start = pci_resource_start(dev->pdev, 1);
+	info->fix.mmio_len = pci_resource_len(dev->pdev, 1);
+
+	/* Set aperture base/size for vesafb takeover */
+#if defined(__i386__) || defined(__x86_64__)
+	if (nouveau_fbcon_has_vesafb_or_efifb(dev)) {
+		/* Some NVIDIA VBIOS' are stupid and decide to put the
+		 * framebuffer in the middle of the PRAMIN BAR for
+		 * whatever reason.  We need to know the exact lfb_base
+		 * to get vesafb kicked off, and the only reliable way
+		 * we have left is to find out lfb_base the same way
+		 * vesafb did.
+		 */
+		info->aperture_base = screen_info.lfb_base;
+		info->aperture_size = screen_info.lfb_size;
+		if (screen_info.orig_video_isVGA == VIDEO_TYPE_VLFB)
+			info->aperture_size *= 65536;
+	} else
+#endif
+	{
+		info->aperture_base = info->fix.mmio_start;
+		info->aperture_size = info->fix.mmio_len;
+	}
+
+	info->pixmap.size = 64*1024;
+	info->pixmap.buf_align = 8;
+	info->pixmap.access_align = 32;
+	info->pixmap.flags = FB_PIXMAP_SYSTEM;
+	info->pixmap.scan_align = 1;
+
+	fb->fbdev = info;
+
+	par->nouveau_fb = nouveau_fb;
+	par->dev = dev;
+
+	if (dev_priv->channel && !nouveau_nofbaccel) {
+		switch (dev_priv->card_type) {
+		case NV_50:
+			nv50_fbcon_accel_init(info);
+			info->fbops = &nv50_fbcon_ops;
+			break;
+		default:
+			nv04_fbcon_accel_init(info);
+			info->fbops = &nv04_fbcon_ops;
+			break;
+		};
+	}
+
+	nouveau_fbcon_zfill(dev);
+
+	/* To allow resizeing without swapping buffers */
+	NV_INFO(dev, "allocated %dx%d fb: 0x%lx, bo %p\n",
+						nouveau_fb->base.width,
+						nouveau_fb->base.height,
+						nvbo->bo.offset, nvbo);
+
+	mutex_unlock(&dev->struct_mutex);
+	vga_switcheroo_client_fb_set(dev->pdev, info);
+	return 0;
+
+out_unref:
+	mutex_unlock(&dev->struct_mutex);
+out:
+	return ret;
+}
+
+int
+nouveau_fbcon_probe(struct drm_device *dev)
+{
+	NV_DEBUG_KMS(dev, "\n");
+
+	return drm_fb_helper_single_fb_probe(dev, 32, nouveau_fbcon_create);
+}
+
+int
+nouveau_fbcon_remove(struct drm_device *dev, struct drm_framebuffer *fb)
+{
+	struct nouveau_framebuffer *nouveau_fb = nouveau_framebuffer(fb);
+	struct fb_info *info;
+
+	if (!fb)
+		return -EINVAL;
+
+	info = fb->fbdev;
+	if (info) {
+		struct nouveau_fbcon_par *par = info->par;
+
+		unregister_framebuffer(info);
+		nouveau_bo_unmap(nouveau_fb->nvbo);
+		drm_gem_object_unreference_unlocked(nouveau_fb->nvbo->gem);
+		nouveau_fb->nvbo = NULL;
+		if (par)
+			drm_fb_helper_free(&par->helper);
+		framebuffer_release(info);
+	}
+
+	return 0;
+}
+
+void nouveau_fbcon_gpu_lockup(struct fb_info *info)
+{
+	struct nouveau_fbcon_par *par = info->par;
+	struct drm_device *dev = par->dev;
+
+	NV_ERROR(dev, "GPU lockup - switching to software fbcon\n");
+	info->flags |= FBINFO_HWACCEL_DISABLED;
+}
diff --git a/drivers/gpu/drm/nouveau/nouveau_fbcon.h b/drivers/gpu/drm/nouveau/nouveau_fbcon.h
new file mode 100644
index 000000000000..f9c34e1a8c11
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_fbcon.h
@@ -0,0 +1,54 @@
+/*
+ * Copyright (C) 2008 Maarten Maathuis.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __NOUVEAU_FBCON_H__
+#define __NOUVEAU_FBCON_H__
+
+#include "drm_fb_helper.h"
+
+struct nouveau_fbcon_par {
+	struct drm_fb_helper helper;
+	struct drm_device *dev;
+	struct nouveau_framebuffer *nouveau_fb;
+};
+
+int nouveau_fbcon_probe(struct drm_device *dev);
+int nouveau_fbcon_remove(struct drm_device *dev, struct drm_framebuffer *fb);
+void nouveau_fbcon_restore(void);
+void nouveau_fbcon_zfill(struct drm_device *dev);
+
+void nv04_fbcon_copyarea(struct fb_info *info, const struct fb_copyarea *region);
+void nv04_fbcon_fillrect(struct fb_info *info, const struct fb_fillrect *rect);
+void nv04_fbcon_imageblit(struct fb_info *info, const struct fb_image *image);
+int nv04_fbcon_accel_init(struct fb_info *info);
+void nv50_fbcon_fillrect(struct fb_info *info, const struct fb_fillrect *rect);
+void nv50_fbcon_copyarea(struct fb_info *info, const struct fb_copyarea *region);
+void nv50_fbcon_imageblit(struct fb_info *info, const struct fb_image *image);
+int nv50_fbcon_accel_init(struct fb_info *info);
+
+void nouveau_fbcon_gpu_lockup(struct fb_info *info);
+#endif /* __NV50_FBCON_H__ */
+
diff --git a/drivers/gpu/drm/nouveau/nouveau_fence.c b/drivers/gpu/drm/nouveau/nouveau_fence.c
new file mode 100644
index 000000000000..faddf53ff9ed
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_fence.c
@@ -0,0 +1,262 @@
+/*
+ * Copyright (C) 2007 Ben Skeggs.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "drmP.h"
+#include "drm.h"
+
+#include "nouveau_drv.h"
+#include "nouveau_dma.h"
+
+#define USE_REFCNT (dev_priv->card_type >= NV_10)
+
+struct nouveau_fence {
+	struct nouveau_channel *channel;
+	struct kref refcount;
+	struct list_head entry;
+
+	uint32_t sequence;
+	bool signalled;
+};
+
+static inline struct nouveau_fence *
+nouveau_fence(void *sync_obj)
+{
+	return (struct nouveau_fence *)sync_obj;
+}
+
+static void
+nouveau_fence_del(struct kref *ref)
+{
+	struct nouveau_fence *fence =
+		container_of(ref, struct nouveau_fence, refcount);
+
+	kfree(fence);
+}
+
+void
+nouveau_fence_update(struct nouveau_channel *chan)
+{
+	struct drm_nouveau_private *dev_priv = chan->dev->dev_private;
+	struct list_head *entry, *tmp;
+	struct nouveau_fence *fence;
+	uint32_t sequence;
+
+	if (USE_REFCNT)
+		sequence = nvchan_rd32(chan, 0x48);
+	else
+		sequence = chan->fence.last_sequence_irq;
+
+	if (chan->fence.sequence_ack == sequence)
+		return;
+	chan->fence.sequence_ack = sequence;
+
+	list_for_each_safe(entry, tmp, &chan->fence.pending) {
+		fence = list_entry(entry, struct nouveau_fence, entry);
+
+		sequence = fence->sequence;
+		fence->signalled = true;
+		list_del(&fence->entry);
+		kref_put(&fence->refcount, nouveau_fence_del);
+
+		if (sequence == chan->fence.sequence_ack)
+			break;
+	}
+}
+
+int
+nouveau_fence_new(struct nouveau_channel *chan, struct nouveau_fence **pfence,
+		  bool emit)
+{
+	struct nouveau_fence *fence;
+	int ret = 0;
+
+	fence = kzalloc(sizeof(*fence), GFP_KERNEL);
+	if (!fence)
+		return -ENOMEM;
+	kref_init(&fence->refcount);
+	fence->channel = chan;
+
+	if (emit)
+		ret = nouveau_fence_emit(fence);
+
+	if (ret)
+		nouveau_fence_unref((void *)&fence);
+	*pfence = fence;
+	return ret;
+}
+
+struct nouveau_channel *
+nouveau_fence_channel(struct nouveau_fence *fence)
+{
+	return fence ? fence->channel : NULL;
+}
+
+int
+nouveau_fence_emit(struct nouveau_fence *fence)
+{
+	struct drm_nouveau_private *dev_priv = fence->channel->dev->dev_private;
+	struct nouveau_channel *chan = fence->channel;
+	unsigned long flags;
+	int ret;
+
+	ret = RING_SPACE(chan, 2);
+	if (ret)
+		return ret;
+
+	if (unlikely(chan->fence.sequence == chan->fence.sequence_ack - 1)) {
+		spin_lock_irqsave(&chan->fence.lock, flags);
+		nouveau_fence_update(chan);
+		spin_unlock_irqrestore(&chan->fence.lock, flags);
+
+		BUG_ON(chan->fence.sequence ==
+		       chan->fence.sequence_ack - 1);
+	}
+
+	fence->sequence = ++chan->fence.sequence;
+
+	kref_get(&fence->refcount);
+	spin_lock_irqsave(&chan->fence.lock, flags);
+	list_add_tail(&fence->entry, &chan->fence.pending);
+	spin_unlock_irqrestore(&chan->fence.lock, flags);
+
+	BEGIN_RING(chan, NvSubSw, USE_REFCNT ? 0x0050 : 0x0150, 1);
+	OUT_RING(chan, fence->sequence);
+	FIRE_RING(chan);
+
+	return 0;
+}
+
+void
+nouveau_fence_unref(void **sync_obj)
+{
+	struct nouveau_fence *fence = nouveau_fence(*sync_obj);
+
+	if (fence)
+		kref_put(&fence->refcount, nouveau_fence_del);
+	*sync_obj = NULL;
+}
+
+void *
+nouveau_fence_ref(void *sync_obj)
+{
+	struct nouveau_fence *fence = nouveau_fence(sync_obj);
+
+	kref_get(&fence->refcount);
+	return sync_obj;
+}
+
+bool
+nouveau_fence_signalled(void *sync_obj, void *sync_arg)
+{
+	struct nouveau_fence *fence = nouveau_fence(sync_obj);
+	struct nouveau_channel *chan = fence->channel;
+	unsigned long flags;
+
+	if (fence->signalled)
+		return true;
+
+	spin_lock_irqsave(&chan->fence.lock, flags);
+	nouveau_fence_update(chan);
+	spin_unlock_irqrestore(&chan->fence.lock, flags);
+	return fence->signalled;
+}
+
+int
+nouveau_fence_wait(void *sync_obj, void *sync_arg, bool lazy, bool intr)
+{
+	unsigned long timeout = jiffies + (3 * DRM_HZ);
+	int ret = 0;
+
+	__set_current_state(intr ? TASK_INTERRUPTIBLE : TASK_UNINTERRUPTIBLE);
+
+	while (1) {
+		if (nouveau_fence_signalled(sync_obj, sync_arg))
+			break;
+
+		if (time_after_eq(jiffies, timeout)) {
+			ret = -EBUSY;
+			break;
+		}
+
+		if (lazy)
+			schedule_timeout(1);
+
+		if (intr && signal_pending(current)) {
+			ret = -ERESTARTSYS;
+			break;
+		}
+	}
+
+	__set_current_state(TASK_RUNNING);
+
+	return ret;
+}
+
+int
+nouveau_fence_flush(void *sync_obj, void *sync_arg)
+{
+	return 0;
+}
+
+void
+nouveau_fence_handler(struct drm_device *dev, int channel)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_channel *chan = NULL;
+
+	if (channel >= 0 && channel < dev_priv->engine.fifo.channels)
+		chan = dev_priv->fifos[channel];
+
+	if (chan) {
+		spin_lock_irq(&chan->fence.lock);
+		nouveau_fence_update(chan);
+		spin_unlock_irq(&chan->fence.lock);
+	}
+}
+
+int
+nouveau_fence_init(struct nouveau_channel *chan)
+{
+	INIT_LIST_HEAD(&chan->fence.pending);
+	spin_lock_init(&chan->fence.lock);
+	return 0;
+}
+
+void
+nouveau_fence_fini(struct nouveau_channel *chan)
+{
+	struct list_head *entry, *tmp;
+	struct nouveau_fence *fence;
+
+	list_for_each_safe(entry, tmp, &chan->fence.pending) {
+		fence = list_entry(entry, struct nouveau_fence, entry);
+
+		fence->signalled = true;
+		list_del(&fence->entry);
+		kref_put(&fence->refcount, nouveau_fence_del);
+	}
+}
+
diff --git a/drivers/gpu/drm/nouveau/nouveau_gem.c b/drivers/gpu/drm/nouveau/nouveau_gem.c
new file mode 100644
index 000000000000..0d22f66f1c79
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_gem.c
@@ -0,0 +1,842 @@
+/*
+ * Copyright (C) 2008 Ben Skeggs.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include "drmP.h"
+#include "drm.h"
+
+#include "nouveau_drv.h"
+#include "nouveau_drm.h"
+#include "nouveau_dma.h"
+
+#define nouveau_gem_pushbuf_sync(chan) 0
+
+int
+nouveau_gem_object_new(struct drm_gem_object *gem)
+{
+	return 0;
+}
+
+void
+nouveau_gem_object_del(struct drm_gem_object *gem)
+{
+	struct nouveau_bo *nvbo = gem->driver_private;
+	struct ttm_buffer_object *bo = &nvbo->bo;
+
+	if (!nvbo)
+		return;
+	nvbo->gem = NULL;
+
+	if (unlikely(nvbo->cpu_filp))
+		ttm_bo_synccpu_write_release(bo);
+
+	if (unlikely(nvbo->pin_refcnt)) {
+		nvbo->pin_refcnt = 1;
+		nouveau_bo_unpin(nvbo);
+	}
+
+	ttm_bo_unref(&bo);
+}
+
+int
+nouveau_gem_new(struct drm_device *dev, struct nouveau_channel *chan,
+		int size, int align, uint32_t flags, uint32_t tile_mode,
+		uint32_t tile_flags, bool no_vm, bool mappable,
+		struct nouveau_bo **pnvbo)
+{
+	struct nouveau_bo *nvbo;
+	int ret;
+
+	ret = nouveau_bo_new(dev, chan, size, align, flags, tile_mode,
+			     tile_flags, no_vm, mappable, pnvbo);
+	if (ret)
+		return ret;
+	nvbo = *pnvbo;
+
+	nvbo->gem = drm_gem_object_alloc(dev, nvbo->bo.mem.size);
+	if (!nvbo->gem) {
+		nouveau_bo_ref(NULL, pnvbo);
+		return -ENOMEM;
+	}
+
+	nvbo->bo.persistant_swap_storage = nvbo->gem->filp;
+	nvbo->gem->driver_private = nvbo;
+	return 0;
+}
+
+static int
+nouveau_gem_info(struct drm_gem_object *gem, struct drm_nouveau_gem_info *rep)
+{
+	struct nouveau_bo *nvbo = nouveau_gem_object(gem);
+
+	if (nvbo->bo.mem.mem_type == TTM_PL_TT)
+		rep->domain = NOUVEAU_GEM_DOMAIN_GART;
+	else
+		rep->domain = NOUVEAU_GEM_DOMAIN_VRAM;
+
+	rep->size = nvbo->bo.mem.num_pages << PAGE_SHIFT;
+	rep->offset = nvbo->bo.offset;
+	rep->map_handle = nvbo->mappable ? nvbo->bo.addr_space_offset : 0;
+	rep->tile_mode = nvbo->tile_mode;
+	rep->tile_flags = nvbo->tile_flags;
+	return 0;
+}
+
+static bool
+nouveau_gem_tile_flags_valid(struct drm_device *dev, uint32_t tile_flags) {
+	switch (tile_flags) {
+	case 0x0000:
+	case 0x1800:
+	case 0x2800:
+	case 0x4800:
+	case 0x7000:
+	case 0x7400:
+	case 0x7a00:
+	case 0xe000:
+		break;
+	default:
+		NV_ERROR(dev, "bad page flags: 0x%08x\n", tile_flags);
+		return false;
+	}
+
+	return true;
+}
+
+int
+nouveau_gem_ioctl_new(struct drm_device *dev, void *data,
+		      struct drm_file *file_priv)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct drm_nouveau_gem_new *req = data;
+	struct nouveau_bo *nvbo = NULL;
+	struct nouveau_channel *chan = NULL;
+	uint32_t flags = 0;
+	int ret = 0;
+
+	NOUVEAU_CHECK_INITIALISED_WITH_RETURN;
+
+	if (unlikely(dev_priv->ttm.bdev.dev_mapping == NULL))
+		dev_priv->ttm.bdev.dev_mapping = dev_priv->dev->dev_mapping;
+
+	if (req->channel_hint) {
+		NOUVEAU_GET_USER_CHANNEL_WITH_RETURN(req->channel_hint,
+						     file_priv, chan);
+	}
+
+	if (req->info.domain & NOUVEAU_GEM_DOMAIN_VRAM)
+		flags |= TTM_PL_FLAG_VRAM;
+	if (req->info.domain & NOUVEAU_GEM_DOMAIN_GART)
+		flags |= TTM_PL_FLAG_TT;
+	if (!flags || req->info.domain & NOUVEAU_GEM_DOMAIN_CPU)
+		flags |= TTM_PL_FLAG_SYSTEM;
+
+	if (!nouveau_gem_tile_flags_valid(dev, req->info.tile_flags))
+		return -EINVAL;
+
+	ret = nouveau_gem_new(dev, chan, req->info.size, req->align, flags,
+			      req->info.tile_mode, req->info.tile_flags, false,
+			      (req->info.domain & NOUVEAU_GEM_DOMAIN_MAPPABLE),
+			      &nvbo);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gem_info(nvbo->gem, &req->info);
+	if (ret)
+		goto out;
+
+	ret = drm_gem_handle_create(file_priv, nvbo->gem, &req->info.handle);
+out:
+	drm_gem_object_handle_unreference_unlocked(nvbo->gem);
+
+	if (ret)
+		drm_gem_object_unreference_unlocked(nvbo->gem);
+	return ret;
+}
+
+static int
+nouveau_gem_set_domain(struct drm_gem_object *gem, uint32_t read_domains,
+		       uint32_t write_domains, uint32_t valid_domains)
+{
+	struct nouveau_bo *nvbo = gem->driver_private;
+	struct ttm_buffer_object *bo = &nvbo->bo;
+	uint64_t flags;
+
+	if (!valid_domains || (!read_domains && !write_domains))
+		return -EINVAL;
+
+	if (write_domains) {
+		if ((valid_domains & NOUVEAU_GEM_DOMAIN_VRAM) &&
+		    (write_domains & NOUVEAU_GEM_DOMAIN_VRAM))
+			flags = TTM_PL_FLAG_VRAM;
+		else
+		if ((valid_domains & NOUVEAU_GEM_DOMAIN_GART) &&
+		    (write_domains & NOUVEAU_GEM_DOMAIN_GART))
+			flags = TTM_PL_FLAG_TT;
+		else
+			return -EINVAL;
+	} else {
+		if ((valid_domains & NOUVEAU_GEM_DOMAIN_VRAM) &&
+		    (read_domains & NOUVEAU_GEM_DOMAIN_VRAM) &&
+		    bo->mem.mem_type == TTM_PL_VRAM)
+			flags = TTM_PL_FLAG_VRAM;
+		else
+		if ((valid_domains & NOUVEAU_GEM_DOMAIN_GART) &&
+		    (read_domains & NOUVEAU_GEM_DOMAIN_GART) &&
+		    bo->mem.mem_type == TTM_PL_TT)
+			flags = TTM_PL_FLAG_TT;
+		else
+		if ((valid_domains & NOUVEAU_GEM_DOMAIN_VRAM) &&
+		    (read_domains & NOUVEAU_GEM_DOMAIN_VRAM))
+			flags = TTM_PL_FLAG_VRAM;
+		else
+			flags = TTM_PL_FLAG_TT;
+	}
+
+	nouveau_bo_placement_set(nvbo, flags);
+	return 0;
+}
+
+struct validate_op {
+	struct list_head vram_list;
+	struct list_head gart_list;
+	struct list_head both_list;
+};
+
+static void
+validate_fini_list(struct list_head *list, struct nouveau_fence *fence)
+{
+	struct list_head *entry, *tmp;
+	struct nouveau_bo *nvbo;
+
+	list_for_each_safe(entry, tmp, list) {
+		nvbo = list_entry(entry, struct nouveau_bo, entry);
+		if (likely(fence)) {
+			struct nouveau_fence *prev_fence;
+
+			spin_lock(&nvbo->bo.lock);
+			prev_fence = nvbo->bo.sync_obj;
+			nvbo->bo.sync_obj = nouveau_fence_ref(fence);
+			spin_unlock(&nvbo->bo.lock);
+			nouveau_fence_unref((void *)&prev_fence);
+		}
+
+		if (unlikely(nvbo->validate_mapped)) {
+			ttm_bo_kunmap(&nvbo->kmap);
+			nvbo->validate_mapped = false;
+		}
+
+		list_del(&nvbo->entry);
+		nvbo->reserved_by = NULL;
+		ttm_bo_unreserve(&nvbo->bo);
+		drm_gem_object_unreference(nvbo->gem);
+	}
+}
+
+static void
+validate_fini(struct validate_op *op, struct nouveau_fence* fence)
+{
+	validate_fini_list(&op->vram_list, fence);
+	validate_fini_list(&op->gart_list, fence);
+	validate_fini_list(&op->both_list, fence);
+}
+
+static int
+validate_init(struct nouveau_channel *chan, struct drm_file *file_priv,
+	      struct drm_nouveau_gem_pushbuf_bo *pbbo,
+	      int nr_buffers, struct validate_op *op)
+{
+	struct drm_device *dev = chan->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t sequence;
+	int trycnt = 0;
+	int ret, i;
+
+	sequence = atomic_add_return(1, &dev_priv->ttm.validate_sequence);
+retry:
+	if (++trycnt > 100000) {
+		NV_ERROR(dev, "%s failed and gave up.\n", __func__);
+		return -EINVAL;
+	}
+
+	for (i = 0; i < nr_buffers; i++) {
+		struct drm_nouveau_gem_pushbuf_bo *b = &pbbo[i];
+		struct drm_gem_object *gem;
+		struct nouveau_bo *nvbo;
+
+		gem = drm_gem_object_lookup(dev, file_priv, b->handle);
+		if (!gem) {
+			NV_ERROR(dev, "Unknown handle 0x%08x\n", b->handle);
+			validate_fini(op, NULL);
+			return -EINVAL;
+		}
+		nvbo = gem->driver_private;
+
+		if (nvbo->reserved_by && nvbo->reserved_by == file_priv) {
+			NV_ERROR(dev, "multiple instances of buffer %d on "
+				      "validation list\n", b->handle);
+			validate_fini(op, NULL);
+			return -EINVAL;
+		}
+
+		ret = ttm_bo_reserve(&nvbo->bo, false, false, true, sequence);
+		if (ret) {
+			validate_fini(op, NULL);
+			if (ret == -EAGAIN)
+				ret = ttm_bo_wait_unreserved(&nvbo->bo, false);
+			drm_gem_object_unreference(gem);
+			if (ret) {
+				NV_ERROR(dev, "fail reserve\n");
+				return ret;
+			}
+			goto retry;
+		}
+
+		b->user_priv = (uint64_t)(unsigned long)nvbo;
+		nvbo->reserved_by = file_priv;
+		nvbo->pbbo_index = i;
+		if ((b->valid_domains & NOUVEAU_GEM_DOMAIN_VRAM) &&
+		    (b->valid_domains & NOUVEAU_GEM_DOMAIN_GART))
+			list_add_tail(&nvbo->entry, &op->both_list);
+		else
+		if (b->valid_domains & NOUVEAU_GEM_DOMAIN_VRAM)
+			list_add_tail(&nvbo->entry, &op->vram_list);
+		else
+		if (b->valid_domains & NOUVEAU_GEM_DOMAIN_GART)
+			list_add_tail(&nvbo->entry, &op->gart_list);
+		else {
+			NV_ERROR(dev, "invalid valid domains: 0x%08x\n",
+				 b->valid_domains);
+			list_add_tail(&nvbo->entry, &op->both_list);
+			validate_fini(op, NULL);
+			return -EINVAL;
+		}
+
+		if (unlikely(atomic_read(&nvbo->bo.cpu_writers) > 0)) {
+			validate_fini(op, NULL);
+
+			if (nvbo->cpu_filp == file_priv) {
+				NV_ERROR(dev, "bo %p mapped by process trying "
+					      "to validate it!\n", nvbo);
+				return -EINVAL;
+			}
+
+			ret = ttm_bo_wait_cpu(&nvbo->bo, false);
+			if (ret) {
+				NV_ERROR(dev, "fail wait_cpu\n");
+				return ret;
+			}
+			goto retry;
+		}
+	}
+
+	return 0;
+}
+
+static int
+validate_list(struct nouveau_channel *chan, struct list_head *list,
+	      struct drm_nouveau_gem_pushbuf_bo *pbbo, uint64_t user_pbbo_ptr)
+{
+	struct drm_nouveau_gem_pushbuf_bo __user *upbbo =
+				(void __force __user *)(uintptr_t)user_pbbo_ptr;
+	struct drm_device *dev = chan->dev;
+	struct nouveau_bo *nvbo;
+	int ret, relocs = 0;
+
+	list_for_each_entry(nvbo, list, entry) {
+		struct drm_nouveau_gem_pushbuf_bo *b = &pbbo[nvbo->pbbo_index];
+		struct nouveau_fence *prev_fence = nvbo->bo.sync_obj;
+
+		if (prev_fence && nouveau_fence_channel(prev_fence) != chan) {
+			spin_lock(&nvbo->bo.lock);
+			ret = ttm_bo_wait(&nvbo->bo, false, false, false);
+			spin_unlock(&nvbo->bo.lock);
+			if (unlikely(ret)) {
+				NV_ERROR(dev, "fail wait other chan\n");
+				return ret;
+			}
+		}
+
+		ret = nouveau_gem_set_domain(nvbo->gem, b->read_domains,
+					     b->write_domains,
+					     b->valid_domains);
+		if (unlikely(ret)) {
+			NV_ERROR(dev, "fail set_domain\n");
+			return ret;
+		}
+
+		nvbo->channel = chan;
+		ret = ttm_bo_validate(&nvbo->bo, &nvbo->placement,
+				      false, false);
+		nvbo->channel = NULL;
+		if (unlikely(ret)) {
+			NV_ERROR(dev, "fail ttm_validate\n");
+			return ret;
+		}
+
+		if (nvbo->bo.offset == b->presumed.offset &&
+		    ((nvbo->bo.mem.mem_type == TTM_PL_VRAM &&
+		      b->presumed.domain & NOUVEAU_GEM_DOMAIN_VRAM) ||
+		     (nvbo->bo.mem.mem_type == TTM_PL_TT &&
+		      b->presumed.domain & NOUVEAU_GEM_DOMAIN_GART)))
+			continue;
+
+		if (nvbo->bo.mem.mem_type == TTM_PL_TT)
+			b->presumed.domain = NOUVEAU_GEM_DOMAIN_GART;
+		else
+			b->presumed.domain = NOUVEAU_GEM_DOMAIN_VRAM;
+		b->presumed.offset = nvbo->bo.offset;
+		b->presumed.valid = 0;
+		relocs++;
+
+		if (DRM_COPY_TO_USER(&upbbo[nvbo->pbbo_index].presumed,
+				     &b->presumed, sizeof(b->presumed)))
+			return -EFAULT;
+	}
+
+	return relocs;
+}
+
+static int
+nouveau_gem_pushbuf_validate(struct nouveau_channel *chan,
+			     struct drm_file *file_priv,
+			     struct drm_nouveau_gem_pushbuf_bo *pbbo,
+			     uint64_t user_buffers, int nr_buffers,
+			     struct validate_op *op, int *apply_relocs)
+{
+	struct drm_device *dev = chan->dev;
+	int ret, relocs = 0;
+
+	INIT_LIST_HEAD(&op->vram_list);
+	INIT_LIST_HEAD(&op->gart_list);
+	INIT_LIST_HEAD(&op->both_list);
+
+	if (nr_buffers == 0)
+		return 0;
+
+	ret = validate_init(chan, file_priv, pbbo, nr_buffers, op);
+	if (unlikely(ret)) {
+		NV_ERROR(dev, "validate_init\n");
+		return ret;
+	}
+
+	ret = validate_list(chan, &op->vram_list, pbbo, user_buffers);
+	if (unlikely(ret < 0)) {
+		NV_ERROR(dev, "validate vram_list\n");
+		validate_fini(op, NULL);
+		return ret;
+	}
+	relocs += ret;
+
+	ret = validate_list(chan, &op->gart_list, pbbo, user_buffers);
+	if (unlikely(ret < 0)) {
+		NV_ERROR(dev, "validate gart_list\n");
+		validate_fini(op, NULL);
+		return ret;
+	}
+	relocs += ret;
+
+	ret = validate_list(chan, &op->both_list, pbbo, user_buffers);
+	if (unlikely(ret < 0)) {
+		NV_ERROR(dev, "validate both_list\n");
+		validate_fini(op, NULL);
+		return ret;
+	}
+	relocs += ret;
+
+	*apply_relocs = relocs;
+	return 0;
+}
+
+static inline void *
+u_memcpya(uint64_t user, unsigned nmemb, unsigned size)
+{
+	void *mem;
+	void __user *userptr = (void __force __user *)(uintptr_t)user;
+
+	mem = kmalloc(nmemb * size, GFP_KERNEL);
+	if (!mem)
+		return ERR_PTR(-ENOMEM);
+
+	if (DRM_COPY_FROM_USER(mem, userptr, nmemb * size)) {
+		kfree(mem);
+		return ERR_PTR(-EFAULT);
+	}
+
+	return mem;
+}
+
+static int
+nouveau_gem_pushbuf_reloc_apply(struct drm_device *dev,
+				struct drm_nouveau_gem_pushbuf *req,
+				struct drm_nouveau_gem_pushbuf_bo *bo)
+{
+	struct drm_nouveau_gem_pushbuf_reloc *reloc = NULL;
+	int ret = 0;
+	unsigned i;
+
+	reloc = u_memcpya(req->relocs, req->nr_relocs, sizeof(*reloc));
+	if (IS_ERR(reloc))
+		return PTR_ERR(reloc);
+
+	for (i = 0; i < req->nr_relocs; i++) {
+		struct drm_nouveau_gem_pushbuf_reloc *r = &reloc[i];
+		struct drm_nouveau_gem_pushbuf_bo *b;
+		struct nouveau_bo *nvbo;
+		uint32_t data;
+
+		if (unlikely(r->bo_index > req->nr_buffers)) {
+			NV_ERROR(dev, "reloc bo index invalid\n");
+			ret = -EINVAL;
+			break;
+		}
+
+		b = &bo[r->bo_index];
+		if (b->presumed.valid)
+			continue;
+
+		if (unlikely(r->reloc_bo_index > req->nr_buffers)) {
+			NV_ERROR(dev, "reloc container bo index invalid\n");
+			ret = -EINVAL;
+			break;
+		}
+		nvbo = (void *)(unsigned long)bo[r->reloc_bo_index].user_priv;
+
+		if (unlikely(r->reloc_bo_offset + 4 >
+			     nvbo->bo.mem.num_pages << PAGE_SHIFT)) {
+			NV_ERROR(dev, "reloc outside of bo\n");
+			ret = -EINVAL;
+			break;
+		}
+
+		if (!nvbo->kmap.virtual) {
+			ret = ttm_bo_kmap(&nvbo->bo, 0, nvbo->bo.mem.num_pages,
+					  &nvbo->kmap);
+			if (ret) {
+				NV_ERROR(dev, "failed kmap for reloc\n");
+				break;
+			}
+			nvbo->validate_mapped = true;
+		}
+
+		if (r->flags & NOUVEAU_GEM_RELOC_LOW)
+			data = b->presumed.offset + r->data;
+		else
+		if (r->flags & NOUVEAU_GEM_RELOC_HIGH)
+			data = (b->presumed.offset + r->data) >> 32;
+		else
+			data = r->data;
+
+		if (r->flags & NOUVEAU_GEM_RELOC_OR) {
+			if (b->presumed.domain == NOUVEAU_GEM_DOMAIN_GART)
+				data |= r->tor;
+			else
+				data |= r->vor;
+		}
+
+		spin_lock(&nvbo->bo.lock);
+		ret = ttm_bo_wait(&nvbo->bo, false, false, false);
+		spin_unlock(&nvbo->bo.lock);
+		if (ret) {
+			NV_ERROR(dev, "reloc wait_idle failed: %d\n", ret);
+			break;
+		}
+
+		nouveau_bo_wr32(nvbo, r->reloc_bo_offset >> 2, data);
+	}
+
+	kfree(reloc);
+	return ret;
+}
+
+int
+nouveau_gem_ioctl_pushbuf(struct drm_device *dev, void *data,
+			  struct drm_file *file_priv)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct drm_nouveau_gem_pushbuf *req = data;
+	struct drm_nouveau_gem_pushbuf_push *push;
+	struct drm_nouveau_gem_pushbuf_bo *bo;
+	struct nouveau_channel *chan;
+	struct validate_op op;
+	struct nouveau_fence *fence = 0;
+	int i, j, ret = 0, do_reloc = 0;
+
+	NOUVEAU_CHECK_INITIALISED_WITH_RETURN;
+	NOUVEAU_GET_USER_CHANNEL_WITH_RETURN(req->channel, file_priv, chan);
+
+	req->vram_available = dev_priv->fb_aper_free;
+	req->gart_available = dev_priv->gart_info.aper_free;
+	if (unlikely(req->nr_push == 0))
+		goto out_next;
+
+	if (unlikely(req->nr_push > NOUVEAU_GEM_MAX_PUSH)) {
+		NV_ERROR(dev, "pushbuf push count exceeds limit: %d max %d\n",
+			 req->nr_push, NOUVEAU_GEM_MAX_PUSH);
+		return -EINVAL;
+	}
+
+	if (unlikely(req->nr_buffers > NOUVEAU_GEM_MAX_BUFFERS)) {
+		NV_ERROR(dev, "pushbuf bo count exceeds limit: %d max %d\n",
+			 req->nr_buffers, NOUVEAU_GEM_MAX_BUFFERS);
+		return -EINVAL;
+	}
+
+	if (unlikely(req->nr_relocs > NOUVEAU_GEM_MAX_RELOCS)) {
+		NV_ERROR(dev, "pushbuf reloc count exceeds limit: %d max %d\n",
+			 req->nr_relocs, NOUVEAU_GEM_MAX_RELOCS);
+		return -EINVAL;
+	}
+
+	push = u_memcpya(req->push, req->nr_push, sizeof(*push));
+	if (IS_ERR(push))
+		return PTR_ERR(push);
+
+	bo = u_memcpya(req->buffers, req->nr_buffers, sizeof(*bo));
+	if (IS_ERR(bo)) {
+		kfree(push);
+		return PTR_ERR(bo);
+	}
+
+	mutex_lock(&dev->struct_mutex);
+
+	/* Validate buffer list */
+	ret = nouveau_gem_pushbuf_validate(chan, file_priv, bo, req->buffers,
+					   req->nr_buffers, &op, &do_reloc);
+	if (ret) {
+		NV_ERROR(dev, "validate: %d\n", ret);
+		goto out;
+	}
+
+	/* Apply any relocations that are required */
+	if (do_reloc) {
+		ret = nouveau_gem_pushbuf_reloc_apply(dev, req, bo);
+		if (ret) {
+			NV_ERROR(dev, "reloc apply: %d\n", ret);
+			goto out;
+		}
+	}
+
+	if (chan->dma.ib_max) {
+		ret = nouveau_dma_wait(chan, req->nr_push + 1, 6);
+		if (ret) {
+			NV_INFO(dev, "nv50cal_space: %d\n", ret);
+			goto out;
+		}
+
+		for (i = 0; i < req->nr_push; i++) {
+			struct nouveau_bo *nvbo = (void *)(unsigned long)
+				bo[push[i].bo_index].user_priv;
+
+			nv50_dma_push(chan, nvbo, push[i].offset,
+				      push[i].length);
+		}
+	} else
+	if (dev_priv->card_type >= NV_20) {
+		ret = RING_SPACE(chan, req->nr_push * 2);
+		if (ret) {
+			NV_ERROR(dev, "cal_space: %d\n", ret);
+			goto out;
+		}
+
+		for (i = 0; i < req->nr_push; i++) {
+			struct nouveau_bo *nvbo = (void *)(unsigned long)
+				bo[push[i].bo_index].user_priv;
+			struct drm_mm_node *mem = nvbo->bo.mem.mm_node;
+
+			OUT_RING(chan, ((mem->start << PAGE_SHIFT) +
+					push[i].offset) | 2);
+			OUT_RING(chan, 0);
+		}
+	} else {
+		ret = RING_SPACE(chan, req->nr_push * (2 + NOUVEAU_DMA_SKIPS));
+		if (ret) {
+			NV_ERROR(dev, "jmp_space: %d\n", ret);
+			goto out;
+		}
+
+		for (i = 0; i < req->nr_push; i++) {
+			struct nouveau_bo *nvbo = (void *)(unsigned long)
+				bo[push[i].bo_index].user_priv;
+			struct drm_mm_node *mem = nvbo->bo.mem.mm_node;
+			uint32_t cmd;
+
+			cmd = chan->pushbuf_base + ((chan->dma.cur + 2) << 2);
+			cmd |= 0x20000000;
+			if (unlikely(cmd != req->suffix0)) {
+				if (!nvbo->kmap.virtual) {
+					ret = ttm_bo_kmap(&nvbo->bo, 0,
+							  nvbo->bo.mem.
+							  num_pages,
+							  &nvbo->kmap);
+					if (ret) {
+						WIND_RING(chan);
+						goto out;
+					}
+					nvbo->validate_mapped = true;
+				}
+
+				nouveau_bo_wr32(nvbo, (push[i].offset +
+						push[i].length - 8) / 4, cmd);
+			}
+
+			OUT_RING(chan, ((mem->start << PAGE_SHIFT) +
+					push[i].offset) | 0x20000000);
+			OUT_RING(chan, 0);
+			for (j = 0; j < NOUVEAU_DMA_SKIPS; j++)
+				OUT_RING(chan, 0);
+		}
+	}
+
+	ret = nouveau_fence_new(chan, &fence, true);
+	if (ret) {
+		NV_ERROR(dev, "error fencing pushbuf: %d\n", ret);
+		WIND_RING(chan);
+		goto out;
+	}
+
+out:
+	validate_fini(&op, fence);
+	nouveau_fence_unref((void**)&fence);
+	mutex_unlock(&dev->struct_mutex);
+	kfree(bo);
+	kfree(push);
+
+out_next:
+	if (chan->dma.ib_max) {
+		req->suffix0 = 0x00000000;
+		req->suffix1 = 0x00000000;
+	} else
+	if (dev_priv->card_type >= NV_20) {
+		req->suffix0 = 0x00020000;
+		req->suffix1 = 0x00000000;
+	} else {
+		req->suffix0 = 0x20000000 |
+			      (chan->pushbuf_base + ((chan->dma.cur + 2) << 2));
+		req->suffix1 = 0x00000000;
+	}
+
+	return ret;
+}
+
+static inline uint32_t
+domain_to_ttm(struct nouveau_bo *nvbo, uint32_t domain)
+{
+	uint32_t flags = 0;
+
+	if (domain & NOUVEAU_GEM_DOMAIN_VRAM)
+		flags |= TTM_PL_FLAG_VRAM;
+	if (domain & NOUVEAU_GEM_DOMAIN_GART)
+		flags |= TTM_PL_FLAG_TT;
+
+	return flags;
+}
+
+int
+nouveau_gem_ioctl_cpu_prep(struct drm_device *dev, void *data,
+			   struct drm_file *file_priv)
+{
+	struct drm_nouveau_gem_cpu_prep *req = data;
+	struct drm_gem_object *gem;
+	struct nouveau_bo *nvbo;
+	bool no_wait = !!(req->flags & NOUVEAU_GEM_CPU_PREP_NOWAIT);
+	int ret = -EINVAL;
+
+	NOUVEAU_CHECK_INITIALISED_WITH_RETURN;
+
+	gem = drm_gem_object_lookup(dev, file_priv, req->handle);
+	if (!gem)
+		return ret;
+	nvbo = nouveau_gem_object(gem);
+
+	if (nvbo->cpu_filp) {
+		if (nvbo->cpu_filp == file_priv)
+			goto out;
+
+		ret = ttm_bo_wait_cpu(&nvbo->bo, no_wait);
+		if (ret)
+			goto out;
+	}
+
+	if (req->flags & NOUVEAU_GEM_CPU_PREP_NOBLOCK) {
+		spin_lock(&nvbo->bo.lock);
+		ret = ttm_bo_wait(&nvbo->bo, false, false, no_wait);
+		spin_unlock(&nvbo->bo.lock);
+	} else {
+		ret = ttm_bo_synccpu_write_grab(&nvbo->bo, no_wait);
+		if (ret == 0)
+			nvbo->cpu_filp = file_priv;
+	}
+
+out:
+	drm_gem_object_unreference_unlocked(gem);
+	return ret;
+}
+
+int
+nouveau_gem_ioctl_cpu_fini(struct drm_device *dev, void *data,
+			   struct drm_file *file_priv)
+{
+	struct drm_nouveau_gem_cpu_prep *req = data;
+	struct drm_gem_object *gem;
+	struct nouveau_bo *nvbo;
+	int ret = -EINVAL;
+
+	NOUVEAU_CHECK_INITIALISED_WITH_RETURN;
+
+	gem = drm_gem_object_lookup(dev, file_priv, req->handle);
+	if (!gem)
+		return ret;
+	nvbo = nouveau_gem_object(gem);
+
+	if (nvbo->cpu_filp != file_priv)
+		goto out;
+	nvbo->cpu_filp = NULL;
+
+	ttm_bo_synccpu_write_release(&nvbo->bo);
+	ret = 0;
+
+out:
+	drm_gem_object_unreference_unlocked(gem);
+	return ret;
+}
+
+int
+nouveau_gem_ioctl_info(struct drm_device *dev, void *data,
+		       struct drm_file *file_priv)
+{
+	struct drm_nouveau_gem_info *req = data;
+	struct drm_gem_object *gem;
+	int ret;
+
+	NOUVEAU_CHECK_INITIALISED_WITH_RETURN;
+
+	gem = drm_gem_object_lookup(dev, file_priv, req->handle);
+	if (!gem)
+		return -EINVAL;
+
+	ret = nouveau_gem_info(gem, req);
+	drm_gem_object_unreference_unlocked(gem);
+	return ret;
+}
+
diff --git a/drivers/gpu/drm/nouveau/nouveau_grctx.c b/drivers/gpu/drm/nouveau/nouveau_grctx.c
new file mode 100644
index 000000000000..c7ebec696747
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_grctx.c
@@ -0,0 +1,161 @@
+/*
+ * Copyright 2009 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include <linux/firmware.h>
+
+#include "drmP.h"
+#include "nouveau_drv.h"
+
+struct nouveau_ctxprog {
+	uint32_t signature;
+	uint8_t  version;
+	uint16_t length;
+	uint32_t data[];
+} __attribute__ ((packed));
+
+struct nouveau_ctxvals {
+	uint32_t signature;
+	uint8_t  version;
+	uint32_t length;
+	struct {
+		uint32_t offset;
+		uint32_t value;
+	} data[];
+} __attribute__ ((packed));
+
+int
+nouveau_grctx_prog_load(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
+	const int chipset = dev_priv->chipset;
+	const struct firmware *fw;
+	const struct nouveau_ctxprog *cp;
+	const struct nouveau_ctxvals *cv;
+	char name[32];
+	int ret, i;
+
+	if (pgraph->accel_blocked)
+		return -ENODEV;
+
+	if (!pgraph->ctxprog) {
+		sprintf(name, "nouveau/nv%02x.ctxprog", chipset);
+		ret = request_firmware(&fw, name, &dev->pdev->dev);
+		if (ret) {
+			NV_ERROR(dev, "No ctxprog for NV%02x\n", chipset);
+			return ret;
+		}
+
+		pgraph->ctxprog = kmalloc(fw->size, GFP_KERNEL);
+		if (!pgraph->ctxprog) {
+			NV_ERROR(dev, "OOM copying ctxprog\n");
+			release_firmware(fw);
+			return -ENOMEM;
+		}
+		memcpy(pgraph->ctxprog, fw->data, fw->size);
+
+		cp = pgraph->ctxprog;
+		if (le32_to_cpu(cp->signature) != 0x5043564e ||
+		    cp->version != 0 ||
+		    le16_to_cpu(cp->length) != ((fw->size - 7) / 4)) {
+			NV_ERROR(dev, "ctxprog invalid\n");
+			release_firmware(fw);
+			nouveau_grctx_fini(dev);
+			return -EINVAL;
+		}
+		release_firmware(fw);
+	}
+
+	if (!pgraph->ctxvals) {
+		sprintf(name, "nouveau/nv%02x.ctxvals", chipset);
+		ret = request_firmware(&fw, name, &dev->pdev->dev);
+		if (ret) {
+			NV_ERROR(dev, "No ctxvals for NV%02x\n", chipset);
+			nouveau_grctx_fini(dev);
+			return ret;
+		}
+
+		pgraph->ctxvals = kmalloc(fw->size, GFP_KERNEL);
+		if (!pgraph->ctxvals) {
+			NV_ERROR(dev, "OOM copying ctxvals\n");
+			release_firmware(fw);
+			nouveau_grctx_fini(dev);
+			return -ENOMEM;
+		}
+		memcpy(pgraph->ctxvals, fw->data, fw->size);
+
+		cv = (void *)pgraph->ctxvals;
+		if (le32_to_cpu(cv->signature) != 0x5643564e ||
+		    cv->version != 0 ||
+		    le32_to_cpu(cv->length) != ((fw->size - 9) / 8)) {
+			NV_ERROR(dev, "ctxvals invalid\n");
+			release_firmware(fw);
+			nouveau_grctx_fini(dev);
+			return -EINVAL;
+		}
+		release_firmware(fw);
+	}
+
+	cp = pgraph->ctxprog;
+
+	nv_wr32(dev, NV40_PGRAPH_CTXCTL_UCODE_INDEX, 0);
+	for (i = 0; i < le16_to_cpu(cp->length); i++)
+		nv_wr32(dev, NV40_PGRAPH_CTXCTL_UCODE_DATA,
+			le32_to_cpu(cp->data[i]));
+
+	return 0;
+}
+
+void
+nouveau_grctx_fini(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
+
+	if (pgraph->ctxprog) {
+		kfree(pgraph->ctxprog);
+		pgraph->ctxprog = NULL;
+	}
+
+	if (pgraph->ctxvals) {
+		kfree(pgraph->ctxprog);
+		pgraph->ctxvals = NULL;
+	}
+}
+
+void
+nouveau_grctx_vals_load(struct drm_device *dev, struct nouveau_gpuobj *ctx)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
+	struct nouveau_ctxvals *cv = pgraph->ctxvals;
+	int i;
+
+	if (!cv)
+		return;
+
+	for (i = 0; i < le32_to_cpu(cv->length); i++)
+		nv_wo32(dev, ctx, le32_to_cpu(cv->data[i].offset),
+			le32_to_cpu(cv->data[i].value));
+}
diff --git a/drivers/gpu/drm/nouveau/nouveau_grctx.h b/drivers/gpu/drm/nouveau/nouveau_grctx.h
new file mode 100644
index 000000000000..5d39c4ce8006
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_grctx.h
@@ -0,0 +1,133 @@
+#ifndef __NOUVEAU_GRCTX_H__
+#define __NOUVEAU_GRCTX_H__
+
+struct nouveau_grctx {
+	struct drm_device *dev;
+
+	enum {
+		NOUVEAU_GRCTX_PROG,
+		NOUVEAU_GRCTX_VALS
+	} mode;
+	void *data;
+
+	uint32_t ctxprog_max;
+	uint32_t ctxprog_len;
+	uint32_t ctxprog_reg;
+	int      ctxprog_label[32];
+	uint32_t ctxvals_pos;
+	uint32_t ctxvals_base;
+};
+
+#ifdef CP_CTX
+static inline void
+cp_out(struct nouveau_grctx *ctx, uint32_t inst)
+{
+	uint32_t *ctxprog = ctx->data;
+
+	if (ctx->mode != NOUVEAU_GRCTX_PROG)
+		return;
+
+	BUG_ON(ctx->ctxprog_len == ctx->ctxprog_max);
+	ctxprog[ctx->ctxprog_len++] = inst;
+}
+
+static inline void
+cp_lsr(struct nouveau_grctx *ctx, uint32_t val)
+{
+	cp_out(ctx, CP_LOAD_SR | val);
+}
+
+static inline void
+cp_ctx(struct nouveau_grctx *ctx, uint32_t reg, uint32_t length)
+{
+	ctx->ctxprog_reg = (reg - 0x00400000) >> 2;
+
+	ctx->ctxvals_base = ctx->ctxvals_pos;
+	ctx->ctxvals_pos = ctx->ctxvals_base + length;
+
+	if (length > (CP_CTX_COUNT >> CP_CTX_COUNT_SHIFT)) {
+		cp_lsr(ctx, length);
+		length = 0;
+	}
+
+	cp_out(ctx, CP_CTX | (length << CP_CTX_COUNT_SHIFT) | ctx->ctxprog_reg);
+}
+
+static inline void
+cp_name(struct nouveau_grctx *ctx, int name)
+{
+	uint32_t *ctxprog = ctx->data;
+	int i;
+
+	if (ctx->mode != NOUVEAU_GRCTX_PROG)
+		return;
+
+	ctx->ctxprog_label[name] = ctx->ctxprog_len;
+	for (i = 0; i < ctx->ctxprog_len; i++) {
+		if ((ctxprog[i] & 0xfff00000) != 0xff400000)
+			continue;
+		if ((ctxprog[i] & CP_BRA_IP) != ((name) << CP_BRA_IP_SHIFT))
+			continue;
+		ctxprog[i] = (ctxprog[i] & 0x00ff00ff) |
+			     (ctx->ctxprog_len << CP_BRA_IP_SHIFT);
+	}
+}
+
+static inline void
+_cp_bra(struct nouveau_grctx *ctx, u32 mod, int flag, int state, int name)
+{
+	int ip = 0;
+
+	if (mod != 2) {
+		ip = ctx->ctxprog_label[name] << CP_BRA_IP_SHIFT;
+		if (ip == 0)
+			ip = 0xff000000 | (name << CP_BRA_IP_SHIFT);
+	}
+
+	cp_out(ctx, CP_BRA | (mod << 18) | ip | flag |
+		    (state ? 0 : CP_BRA_IF_CLEAR));
+}
+#define cp_bra(c,f,s,n) _cp_bra((c), 0, CP_FLAG_##f, CP_FLAG_##f##_##s, n)
+#ifdef CP_BRA_MOD
+#define cp_cal(c,f,s,n) _cp_bra((c), 1, CP_FLAG_##f, CP_FLAG_##f##_##s, n)
+#define cp_ret(c,f,s) _cp_bra((c), 2, CP_FLAG_##f, CP_FLAG_##f##_##s, 0)
+#endif
+
+static inline void
+_cp_wait(struct nouveau_grctx *ctx, int flag, int state)
+{
+	cp_out(ctx, CP_WAIT | flag | (state ? CP_WAIT_SET : 0));
+}
+#define cp_wait(c,f,s) _cp_wait((c), CP_FLAG_##f, CP_FLAG_##f##_##s)
+
+static inline void
+_cp_set(struct nouveau_grctx *ctx, int flag, int state)
+{
+	cp_out(ctx, CP_SET | flag | (state ? CP_SET_1 : 0));
+}
+#define cp_set(c,f,s) _cp_set((c), CP_FLAG_##f, CP_FLAG_##f##_##s)
+
+static inline void
+cp_pos(struct nouveau_grctx *ctx, int offset)
+{
+	ctx->ctxvals_pos = offset;
+	ctx->ctxvals_base = ctx->ctxvals_pos;
+
+	cp_lsr(ctx, ctx->ctxvals_pos);
+	cp_out(ctx, CP_SET_CONTEXT_POINTER);
+}
+
+static inline void
+gr_def(struct nouveau_grctx *ctx, uint32_t reg, uint32_t val)
+{
+	if (ctx->mode != NOUVEAU_GRCTX_VALS)
+		return;
+
+	reg = (reg - 0x00400000) / 4;
+	reg = (reg - ctx->ctxprog_reg) + ctx->ctxvals_base;
+
+	nv_wo32(ctx->dev, ctx->data, reg, val);
+}
+#endif
+
+#endif
diff --git a/drivers/gpu/drm/nouveau/nouveau_hw.c b/drivers/gpu/drm/nouveau/nouveau_hw.c
new file mode 100644
index 000000000000..7855b35effc3
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_hw.c
@@ -0,0 +1,1080 @@
+/*
+ * Copyright 2006 Dave Airlie
+ * Copyright 2007 Maarten Maathuis
+ * Copyright 2007-2009 Stuart Bennett
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF
+ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include "drmP.h"
+#include "nouveau_drv.h"
+#include "nouveau_hw.h"
+
+#define CHIPSET_NFORCE 0x01a0
+#define CHIPSET_NFORCE2 0x01f0
+
+/*
+ * misc hw access wrappers/control functions
+ */
+
+void
+NVWriteVgaSeq(struct drm_device *dev, int head, uint8_t index, uint8_t value)
+{
+	NVWritePRMVIO(dev, head, NV_PRMVIO_SRX, index);
+	NVWritePRMVIO(dev, head, NV_PRMVIO_SR, value);
+}
+
+uint8_t
+NVReadVgaSeq(struct drm_device *dev, int head, uint8_t index)
+{
+	NVWritePRMVIO(dev, head, NV_PRMVIO_SRX, index);
+	return NVReadPRMVIO(dev, head, NV_PRMVIO_SR);
+}
+
+void
+NVWriteVgaGr(struct drm_device *dev, int head, uint8_t index, uint8_t value)
+{
+	NVWritePRMVIO(dev, head, NV_PRMVIO_GRX, index);
+	NVWritePRMVIO(dev, head, NV_PRMVIO_GX, value);
+}
+
+uint8_t
+NVReadVgaGr(struct drm_device *dev, int head, uint8_t index)
+{
+	NVWritePRMVIO(dev, head, NV_PRMVIO_GRX, index);
+	return NVReadPRMVIO(dev, head, NV_PRMVIO_GX);
+}
+
+/* CR44 takes values 0 (head A), 3 (head B) and 4 (heads tied)
+ * it affects only the 8 bit vga io regs, which we access using mmio at
+ * 0xc{0,2}3c*, 0x60{1,3}3*, and 0x68{1,3}3d*
+ * in general, the set value of cr44 does not matter: reg access works as
+ * expected and values can be set for the appropriate head by using a 0x2000
+ * offset as required
+ * however:
+ * a) pre nv40, the head B range of PRMVIO regs at 0xc23c* was not exposed and
+ *    cr44 must be set to 0 or 3 for accessing values on the correct head
+ *    through the common 0xc03c* addresses
+ * b) in tied mode (4) head B is programmed to the values set on head A, and
+ *    access using the head B addresses can have strange results, ergo we leave
+ *    tied mode in init once we know to what cr44 should be restored on exit
+ *
+ * the owner parameter is slightly abused:
+ * 0 and 1 are treated as head values and so the set value is (owner * 3)
+ * other values are treated as literal values to set
+ */
+void
+NVSetOwner(struct drm_device *dev, int owner)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	if (owner == 1)
+		owner *= 3;
+
+	if (dev_priv->chipset == 0x11) {
+		/* This might seem stupid, but the blob does it and
+		 * omitting it often locks the system up.
+		 */
+		NVReadVgaCrtc(dev, 0, NV_CIO_SR_LOCK_INDEX);
+		NVReadVgaCrtc(dev, 1, NV_CIO_SR_LOCK_INDEX);
+	}
+
+	/* CR44 is always changed on CRTC0 */
+	NVWriteVgaCrtc(dev, 0, NV_CIO_CRE_44, owner);
+
+	if (dev_priv->chipset == 0x11) {	/* set me harder */
+		NVWriteVgaCrtc(dev, 0, NV_CIO_CRE_2E, owner);
+		NVWriteVgaCrtc(dev, 0, NV_CIO_CRE_2E, owner);
+	}
+}
+
+void
+NVBlankScreen(struct drm_device *dev, int head, bool blank)
+{
+	unsigned char seq1;
+
+	if (nv_two_heads(dev))
+		NVSetOwner(dev, head);
+
+	seq1 = NVReadVgaSeq(dev, head, NV_VIO_SR_CLOCK_INDEX);
+
+	NVVgaSeqReset(dev, head, true);
+	if (blank)
+		NVWriteVgaSeq(dev, head, NV_VIO_SR_CLOCK_INDEX, seq1 | 0x20);
+	else
+		NVWriteVgaSeq(dev, head, NV_VIO_SR_CLOCK_INDEX, seq1 & ~0x20);
+	NVVgaSeqReset(dev, head, false);
+}
+
+/*
+ * PLL setting
+ */
+
+static int
+powerctrl_1_shift(int chip_version, int reg)
+{
+	int shift = -4;
+
+	if (chip_version < 0x17 || chip_version == 0x1a || chip_version == 0x20)
+		return shift;
+
+	switch (reg) {
+	case NV_RAMDAC_VPLL2:
+		shift += 4;
+	case NV_PRAMDAC_VPLL_COEFF:
+		shift += 4;
+	case NV_PRAMDAC_MPLL_COEFF:
+		shift += 4;
+	case NV_PRAMDAC_NVPLL_COEFF:
+		shift += 4;
+	}
+
+	/*
+	 * the shift for vpll regs is only used for nv3x chips with a single
+	 * stage pll
+	 */
+	if (shift > 4 && (chip_version < 0x32 || chip_version == 0x35 ||
+			  chip_version == 0x36 || chip_version >= 0x40))
+		shift = -4;
+
+	return shift;
+}
+
+static void
+setPLL_single(struct drm_device *dev, uint32_t reg, struct nouveau_pll_vals *pv)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	int chip_version = dev_priv->vbios.chip_version;
+	uint32_t oldpll = NVReadRAMDAC(dev, 0, reg);
+	int oldN = (oldpll >> 8) & 0xff, oldM = oldpll & 0xff;
+	uint32_t pll = (oldpll & 0xfff80000) | pv->log2P << 16 | pv->NM1;
+	uint32_t saved_powerctrl_1 = 0;
+	int shift_powerctrl_1 = powerctrl_1_shift(chip_version, reg);
+
+	if (oldpll == pll)
+		return;	/* already set */
+
+	if (shift_powerctrl_1 >= 0) {
+		saved_powerctrl_1 = nvReadMC(dev, NV_PBUS_POWERCTRL_1);
+		nvWriteMC(dev, NV_PBUS_POWERCTRL_1,
+			(saved_powerctrl_1 & ~(0xf << shift_powerctrl_1)) |
+			1 << shift_powerctrl_1);
+	}
+
+	if (oldM && pv->M1 && (oldN / oldM < pv->N1 / pv->M1))
+		/* upclock -- write new post divider first */
+		NVWriteRAMDAC(dev, 0, reg, pv->log2P << 16 | (oldpll & 0xffff));
+	else
+		/* downclock -- write new NM first */
+		NVWriteRAMDAC(dev, 0, reg, (oldpll & 0xffff0000) | pv->NM1);
+
+	if (chip_version < 0x17 && chip_version != 0x11)
+		/* wait a bit on older chips */
+		msleep(64);
+	NVReadRAMDAC(dev, 0, reg);
+
+	/* then write the other half as well */
+	NVWriteRAMDAC(dev, 0, reg, pll);
+
+	if (shift_powerctrl_1 >= 0)
+		nvWriteMC(dev, NV_PBUS_POWERCTRL_1, saved_powerctrl_1);
+}
+
+static uint32_t
+new_ramdac580(uint32_t reg1, bool ss, uint32_t ramdac580)
+{
+	bool head_a = (reg1 == NV_PRAMDAC_VPLL_COEFF);
+
+	if (ss)	/* single stage pll mode */
+		ramdac580 |= head_a ? NV_RAMDAC_580_VPLL1_ACTIVE :
+				      NV_RAMDAC_580_VPLL2_ACTIVE;
+	else
+		ramdac580 &= head_a ? ~NV_RAMDAC_580_VPLL1_ACTIVE :
+				      ~NV_RAMDAC_580_VPLL2_ACTIVE;
+
+	return ramdac580;
+}
+
+static void
+setPLL_double_highregs(struct drm_device *dev, uint32_t reg1,
+		       struct nouveau_pll_vals *pv)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	int chip_version = dev_priv->vbios.chip_version;
+	bool nv3035 = chip_version == 0x30 || chip_version == 0x35;
+	uint32_t reg2 = reg1 + ((reg1 == NV_RAMDAC_VPLL2) ? 0x5c : 0x70);
+	uint32_t oldpll1 = NVReadRAMDAC(dev, 0, reg1);
+	uint32_t oldpll2 = !nv3035 ? NVReadRAMDAC(dev, 0, reg2) : 0;
+	uint32_t pll1 = (oldpll1 & 0xfff80000) | pv->log2P << 16 | pv->NM1;
+	uint32_t pll2 = (oldpll2 & 0x7fff0000) | 1 << 31 | pv->NM2;
+	uint32_t oldramdac580 = 0, ramdac580 = 0;
+	bool single_stage = !pv->NM2 || pv->N2 == pv->M2;	/* nv41+ only */
+	uint32_t saved_powerctrl_1 = 0, savedc040 = 0;
+	int shift_powerctrl_1 = powerctrl_1_shift(chip_version, reg1);
+
+	/* model specific additions to generic pll1 and pll2 set up above */
+	if (nv3035) {
+		pll1 = (pll1 & 0xfcc7ffff) | (pv->N2 & 0x18) << 21 |
+		       (pv->N2 & 0x7) << 19 | 8 << 4 | (pv->M2 & 7) << 4;
+		pll2 = 0;
+	}
+	if (chip_version > 0x40 && reg1 >= NV_PRAMDAC_VPLL_COEFF) { /* !nv40 */
+		oldramdac580 = NVReadRAMDAC(dev, 0, NV_PRAMDAC_580);
+		ramdac580 = new_ramdac580(reg1, single_stage, oldramdac580);
+		if (oldramdac580 != ramdac580)
+			oldpll1 = ~0;	/* force mismatch */
+		if (single_stage)
+			/* magic value used by nvidia in single stage mode */
+			pll2 |= 0x011f;
+	}
+	if (chip_version > 0x70)
+		/* magic bits set by the blob (but not the bios) on g71-73 */
+		pll1 = (pll1 & 0x7fffffff) | (single_stage ? 0x4 : 0xc) << 28;
+
+	if (oldpll1 == pll1 && oldpll2 == pll2)
+		return;	/* already set */
+
+	if (shift_powerctrl_1 >= 0) {
+		saved_powerctrl_1 = nvReadMC(dev, NV_PBUS_POWERCTRL_1);
+		nvWriteMC(dev, NV_PBUS_POWERCTRL_1,
+			(saved_powerctrl_1 & ~(0xf << shift_powerctrl_1)) |
+			1 << shift_powerctrl_1);
+	}
+
+	if (chip_version >= 0x40) {
+		int shift_c040 = 14;
+
+		switch (reg1) {
+		case NV_PRAMDAC_MPLL_COEFF:
+			shift_c040 += 2;
+		case NV_PRAMDAC_NVPLL_COEFF:
+			shift_c040 += 2;
+		case NV_RAMDAC_VPLL2:
+			shift_c040 += 2;
+		case NV_PRAMDAC_VPLL_COEFF:
+			shift_c040 += 2;
+		}
+
+		savedc040 = nvReadMC(dev, 0xc040);
+		if (shift_c040 != 14)
+			nvWriteMC(dev, 0xc040, savedc040 & ~(3 << shift_c040));
+	}
+
+	if (oldramdac580 != ramdac580)
+		NVWriteRAMDAC(dev, 0, NV_PRAMDAC_580, ramdac580);
+
+	if (!nv3035)
+		NVWriteRAMDAC(dev, 0, reg2, pll2);
+	NVWriteRAMDAC(dev, 0, reg1, pll1);
+
+	if (shift_powerctrl_1 >= 0)
+		nvWriteMC(dev, NV_PBUS_POWERCTRL_1, saved_powerctrl_1);
+	if (chip_version >= 0x40)
+		nvWriteMC(dev, 0xc040, savedc040);
+}
+
+static void
+setPLL_double_lowregs(struct drm_device *dev, uint32_t NMNMreg,
+		      struct nouveau_pll_vals *pv)
+{
+	/* When setting PLLs, there is a merry game of disabling and enabling
+	 * various bits of hardware during the process. This function is a
+	 * synthesis of six nv4x traces, nearly each card doing a subtly
+	 * different thing. With luck all the necessary bits for each card are
+	 * combined herein. Without luck it deviates from each card's formula
+	 * so as to not work on any :)
+	 */
+
+	uint32_t Preg = NMNMreg - 4;
+	bool mpll = Preg == 0x4020;
+	uint32_t oldPval = nvReadMC(dev, Preg);
+	uint32_t NMNM = pv->NM2 << 16 | pv->NM1;
+	uint32_t Pval = (oldPval & (mpll ? ~(0x11 << 16) : ~(1 << 16))) |
+			0xc << 28 | pv->log2P << 16;
+	uint32_t saved4600 = 0;
+	/* some cards have different maskc040s */
+	uint32_t maskc040 = ~(3 << 14), savedc040;
+	bool single_stage = !pv->NM2 || pv->N2 == pv->M2;
+
+	if (nvReadMC(dev, NMNMreg) == NMNM && (oldPval & 0xc0070000) == Pval)
+		return;
+
+	if (Preg == 0x4000)
+		maskc040 = ~0x333;
+	if (Preg == 0x4058)
+		maskc040 = ~(0xc << 24);
+
+	if (mpll) {
+		struct pll_lims pll_lim;
+		uint8_t Pval2;
+
+		if (get_pll_limits(dev, Preg, &pll_lim))
+			return;
+
+		Pval2 = pv->log2P + pll_lim.log2p_bias;
+		if (Pval2 > pll_lim.max_log2p)
+			Pval2 = pll_lim.max_log2p;
+		Pval |= 1 << 28 | Pval2 << 20;
+
+		saved4600 = nvReadMC(dev, 0x4600);
+		nvWriteMC(dev, 0x4600, saved4600 | 8 << 28);
+	}
+	if (single_stage)
+		Pval |= mpll ? 1 << 12 : 1 << 8;
+
+	nvWriteMC(dev, Preg, oldPval | 1 << 28);
+	nvWriteMC(dev, Preg, Pval & ~(4 << 28));
+	if (mpll) {
+		Pval |= 8 << 20;
+		nvWriteMC(dev, 0x4020, Pval & ~(0xc << 28));
+		nvWriteMC(dev, 0x4038, Pval & ~(0xc << 28));
+	}
+
+	savedc040 = nvReadMC(dev, 0xc040);
+	nvWriteMC(dev, 0xc040, savedc040 & maskc040);
+
+	nvWriteMC(dev, NMNMreg, NMNM);
+	if (NMNMreg == 0x4024)
+		nvWriteMC(dev, 0x403c, NMNM);
+
+	nvWriteMC(dev, Preg, Pval);
+	if (mpll) {
+		Pval &= ~(8 << 20);
+		nvWriteMC(dev, 0x4020, Pval);
+		nvWriteMC(dev, 0x4038, Pval);
+		nvWriteMC(dev, 0x4600, saved4600);
+	}
+
+	nvWriteMC(dev, 0xc040, savedc040);
+
+	if (mpll) {
+		nvWriteMC(dev, 0x4020, Pval & ~(1 << 28));
+		nvWriteMC(dev, 0x4038, Pval & ~(1 << 28));
+	}
+}
+
+void
+nouveau_hw_setpll(struct drm_device *dev, uint32_t reg1,
+		  struct nouveau_pll_vals *pv)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	int cv = dev_priv->vbios.chip_version;
+
+	if (cv == 0x30 || cv == 0x31 || cv == 0x35 || cv == 0x36 ||
+	    cv >= 0x40) {
+		if (reg1 > 0x405c)
+			setPLL_double_highregs(dev, reg1, pv);
+		else
+			setPLL_double_lowregs(dev, reg1, pv);
+	} else
+		setPLL_single(dev, reg1, pv);
+}
+
+/*
+ * PLL getting
+ */
+
+static void
+nouveau_hw_decode_pll(struct drm_device *dev, uint32_t reg1, uint32_t pll1,
+		      uint32_t pll2, struct nouveau_pll_vals *pllvals)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	/* to force parsing as single stage (i.e. nv40 vplls) pass pll2 as 0 */
+
+	/* log2P is & 0x7 as never more than 7, and nv30/35 only uses 3 bits */
+	pllvals->log2P = (pll1 >> 16) & 0x7;
+	pllvals->N2 = pllvals->M2 = 1;
+
+	if (reg1 <= 0x405c) {
+		pllvals->NM1 = pll2 & 0xffff;
+		/* single stage NVPLL and VPLLs use 1 << 8, MPLL uses 1 << 12 */
+		if (!(pll1 & 0x1100))
+			pllvals->NM2 = pll2 >> 16;
+	} else {
+		pllvals->NM1 = pll1 & 0xffff;
+		if (nv_two_reg_pll(dev) && pll2 & NV31_RAMDAC_ENABLE_VCO2)
+			pllvals->NM2 = pll2 & 0xffff;
+		else if (dev_priv->chipset == 0x30 || dev_priv->chipset == 0x35) {
+			pllvals->M1 &= 0xf; /* only 4 bits */
+			if (pll1 & NV30_RAMDAC_ENABLE_VCO2) {
+				pllvals->M2 = (pll1 >> 4) & 0x7;
+				pllvals->N2 = ((pll1 >> 21) & 0x18) |
+					      ((pll1 >> 19) & 0x7);
+			}
+		}
+	}
+}
+
+int
+nouveau_hw_get_pllvals(struct drm_device *dev, enum pll_types plltype,
+		       struct nouveau_pll_vals *pllvals)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	const uint32_t nv04_regs[MAX_PLL_TYPES] = { NV_PRAMDAC_NVPLL_COEFF,
+						    NV_PRAMDAC_MPLL_COEFF,
+						    NV_PRAMDAC_VPLL_COEFF,
+						    NV_RAMDAC_VPLL2 };
+	const uint32_t nv40_regs[MAX_PLL_TYPES] = { 0x4000,
+						    0x4020,
+						    NV_PRAMDAC_VPLL_COEFF,
+						    NV_RAMDAC_VPLL2 };
+	uint32_t reg1, pll1, pll2 = 0;
+	struct pll_lims pll_lim;
+	int ret;
+
+	if (dev_priv->card_type < NV_40)
+		reg1 = nv04_regs[plltype];
+	else
+		reg1 = nv40_regs[plltype];
+
+	pll1 = nvReadMC(dev, reg1);
+
+	if (reg1 <= 0x405c)
+		pll2 = nvReadMC(dev, reg1 + 4);
+	else if (nv_two_reg_pll(dev)) {
+		uint32_t reg2 = reg1 + (reg1 == NV_RAMDAC_VPLL2 ? 0x5c : 0x70);
+
+		pll2 = nvReadMC(dev, reg2);
+	}
+
+	if (dev_priv->card_type == 0x40 && reg1 >= NV_PRAMDAC_VPLL_COEFF) {
+		uint32_t ramdac580 = NVReadRAMDAC(dev, 0, NV_PRAMDAC_580);
+
+		/* check whether vpll has been forced into single stage mode */
+		if (reg1 == NV_PRAMDAC_VPLL_COEFF) {
+			if (ramdac580 & NV_RAMDAC_580_VPLL1_ACTIVE)
+				pll2 = 0;
+		} else
+			if (ramdac580 & NV_RAMDAC_580_VPLL2_ACTIVE)
+				pll2 = 0;
+	}
+
+	nouveau_hw_decode_pll(dev, reg1, pll1, pll2, pllvals);
+
+	ret = get_pll_limits(dev, plltype, &pll_lim);
+	if (ret)
+		return ret;
+
+	pllvals->refclk = pll_lim.refclk;
+
+	return 0;
+}
+
+int
+nouveau_hw_pllvals_to_clk(struct nouveau_pll_vals *pv)
+{
+	/* Avoid divide by zero if called at an inappropriate time */
+	if (!pv->M1 || !pv->M2)
+		return 0;
+
+	return pv->N1 * pv->N2 * pv->refclk / (pv->M1 * pv->M2) >> pv->log2P;
+}
+
+int
+nouveau_hw_get_clock(struct drm_device *dev, enum pll_types plltype)
+{
+	struct nouveau_pll_vals pllvals;
+
+	if (plltype == MPLL && (dev->pci_device & 0x0ff0) == CHIPSET_NFORCE) {
+		uint32_t mpllP;
+
+		pci_read_config_dword(pci_get_bus_and_slot(0, 3), 0x6c, &mpllP);
+		if (!mpllP)
+			mpllP = 4;
+
+		return 400000 / mpllP;
+	} else
+	if (plltype == MPLL && (dev->pci_device & 0xff0) == CHIPSET_NFORCE2) {
+		uint32_t clock;
+
+		pci_read_config_dword(pci_get_bus_and_slot(0, 5), 0x4c, &clock);
+		return clock;
+	}
+
+	nouveau_hw_get_pllvals(dev, plltype, &pllvals);
+
+	return nouveau_hw_pllvals_to_clk(&pllvals);
+}
+
+static void
+nouveau_hw_fix_bad_vpll(struct drm_device *dev, int head)
+{
+	/* the vpll on an unused head can come up with a random value, way
+	 * beyond the pll limits.  for some reason this causes the chip to
+	 * lock up when reading the dac palette regs, so set a valid pll here
+	 * when such a condition detected.  only seen on nv11 to date
+	 */
+
+	struct pll_lims pll_lim;
+	struct nouveau_pll_vals pv;
+	uint32_t pllreg = head ? NV_RAMDAC_VPLL2 : NV_PRAMDAC_VPLL_COEFF;
+
+	if (get_pll_limits(dev, head ? VPLL2 : VPLL1, &pll_lim))
+		return;
+	nouveau_hw_get_pllvals(dev, head ? VPLL2 : VPLL1, &pv);
+
+	if (pv.M1 >= pll_lim.vco1.min_m && pv.M1 <= pll_lim.vco1.max_m &&
+	    pv.N1 >= pll_lim.vco1.min_n && pv.N1 <= pll_lim.vco1.max_n &&
+	    pv.log2P <= pll_lim.max_log2p)
+		return;
+
+	NV_WARN(dev, "VPLL %d outwith limits, attempting to fix\n", head + 1);
+
+	/* set lowest clock within static limits */
+	pv.M1 = pll_lim.vco1.max_m;
+	pv.N1 = pll_lim.vco1.min_n;
+	pv.log2P = pll_lim.max_usable_log2p;
+	nouveau_hw_setpll(dev, pllreg, &pv);
+}
+
+/*
+ * vga font save/restore
+ */
+
+static void nouveau_vga_font_io(struct drm_device *dev,
+				void __iomem *iovram,
+				bool save, unsigned plane)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	unsigned i;
+
+	NVWriteVgaSeq(dev, 0, NV_VIO_SR_PLANE_MASK_INDEX, 1 << plane);
+	NVWriteVgaGr(dev, 0, NV_VIO_GX_READ_MAP_INDEX, plane);
+	for (i = 0; i < 16384; i++) {
+		if (save) {
+			dev_priv->saved_vga_font[plane][i] =
+					ioread32_native(iovram + i * 4);
+		} else {
+			iowrite32_native(dev_priv->saved_vga_font[plane][i],
+							iovram + i * 4);
+		}
+	}
+}
+
+void
+nouveau_hw_save_vga_fonts(struct drm_device *dev, bool save)
+{
+	uint8_t misc, gr4, gr5, gr6, seq2, seq4;
+	bool graphicsmode;
+	unsigned plane;
+	void __iomem *iovram;
+
+	if (nv_two_heads(dev))
+		NVSetOwner(dev, 0);
+
+	NVSetEnablePalette(dev, 0, true);
+	graphicsmode = NVReadVgaAttr(dev, 0, NV_CIO_AR_MODE_INDEX) & 1;
+	NVSetEnablePalette(dev, 0, false);
+
+	if (graphicsmode) /* graphics mode => framebuffer => no need to save */
+		return;
+
+	NV_INFO(dev, "%sing VGA fonts\n", save ? "Sav" : "Restor");
+
+	/* map first 64KiB of VRAM, holds VGA fonts etc */
+	iovram = ioremap(pci_resource_start(dev->pdev, 1), 65536);
+	if (!iovram) {
+		NV_ERROR(dev, "Failed to map VRAM, "
+					"cannot save/restore VGA fonts.\n");
+		return;
+	}
+
+	if (nv_two_heads(dev))
+		NVBlankScreen(dev, 1, true);
+	NVBlankScreen(dev, 0, true);
+
+	/* save control regs */
+	misc = NVReadPRMVIO(dev, 0, NV_PRMVIO_MISC__READ);
+	seq2 = NVReadVgaSeq(dev, 0, NV_VIO_SR_PLANE_MASK_INDEX);
+	seq4 = NVReadVgaSeq(dev, 0, NV_VIO_SR_MEM_MODE_INDEX);
+	gr4 = NVReadVgaGr(dev, 0, NV_VIO_GX_READ_MAP_INDEX);
+	gr5 = NVReadVgaGr(dev, 0, NV_VIO_GX_MODE_INDEX);
+	gr6 = NVReadVgaGr(dev, 0, NV_VIO_GX_MISC_INDEX);
+
+	NVWritePRMVIO(dev, 0, NV_PRMVIO_MISC__WRITE, 0x67);
+	NVWriteVgaSeq(dev, 0, NV_VIO_SR_MEM_MODE_INDEX, 0x6);
+	NVWriteVgaGr(dev, 0, NV_VIO_GX_MODE_INDEX, 0x0);
+	NVWriteVgaGr(dev, 0, NV_VIO_GX_MISC_INDEX, 0x5);
+
+	/* store font in planes 0..3 */
+	for (plane = 0; plane < 4; plane++)
+		nouveau_vga_font_io(dev, iovram, save, plane);
+
+	/* restore control regs */
+	NVWritePRMVIO(dev, 0, NV_PRMVIO_MISC__WRITE, misc);
+	NVWriteVgaGr(dev, 0, NV_VIO_GX_READ_MAP_INDEX, gr4);
+	NVWriteVgaGr(dev, 0, NV_VIO_GX_MODE_INDEX, gr5);
+	NVWriteVgaGr(dev, 0, NV_VIO_GX_MISC_INDEX, gr6);
+	NVWriteVgaSeq(dev, 0, NV_VIO_SR_PLANE_MASK_INDEX, seq2);
+	NVWriteVgaSeq(dev, 0, NV_VIO_SR_MEM_MODE_INDEX, seq4);
+
+	if (nv_two_heads(dev))
+		NVBlankScreen(dev, 1, false);
+	NVBlankScreen(dev, 0, false);
+
+	iounmap(iovram);
+}
+
+/*
+ * mode state save/load
+ */
+
+static void
+rd_cio_state(struct drm_device *dev, int head,
+	     struct nv04_crtc_reg *crtcstate, int index)
+{
+	crtcstate->CRTC[index] = NVReadVgaCrtc(dev, head, index);
+}
+
+static void
+wr_cio_state(struct drm_device *dev, int head,
+	     struct nv04_crtc_reg *crtcstate, int index)
+{
+	NVWriteVgaCrtc(dev, head, index, crtcstate->CRTC[index]);
+}
+
+static void
+nv_save_state_ramdac(struct drm_device *dev, int head,
+		     struct nv04_mode_state *state)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nv04_crtc_reg *regp = &state->crtc_reg[head];
+	int i;
+
+	if (dev_priv->card_type >= NV_10)
+		regp->nv10_cursync = NVReadRAMDAC(dev, head, NV_RAMDAC_NV10_CURSYNC);
+
+	nouveau_hw_get_pllvals(dev, head ? VPLL2 : VPLL1, &regp->pllvals);
+	state->pllsel = NVReadRAMDAC(dev, 0, NV_PRAMDAC_PLL_COEFF_SELECT);
+	if (nv_two_heads(dev))
+		state->sel_clk = NVReadRAMDAC(dev, 0, NV_PRAMDAC_SEL_CLK);
+	if (dev_priv->chipset == 0x11)
+		regp->dither = NVReadRAMDAC(dev, head, NV_RAMDAC_DITHER_NV11);
+
+	regp->ramdac_gen_ctrl = NVReadRAMDAC(dev, head, NV_PRAMDAC_GENERAL_CONTROL);
+
+	if (nv_gf4_disp_arch(dev))
+		regp->ramdac_630 = NVReadRAMDAC(dev, head, NV_PRAMDAC_630);
+	if (dev_priv->chipset >= 0x30)
+		regp->ramdac_634 = NVReadRAMDAC(dev, head, NV_PRAMDAC_634);
+
+	regp->tv_setup = NVReadRAMDAC(dev, head, NV_PRAMDAC_TV_SETUP);
+	regp->tv_vtotal = NVReadRAMDAC(dev, head, NV_PRAMDAC_TV_VTOTAL);
+	regp->tv_vskew = NVReadRAMDAC(dev, head, NV_PRAMDAC_TV_VSKEW);
+	regp->tv_vsync_delay = NVReadRAMDAC(dev, head, NV_PRAMDAC_TV_VSYNC_DELAY);
+	regp->tv_htotal = NVReadRAMDAC(dev, head, NV_PRAMDAC_TV_HTOTAL);
+	regp->tv_hskew = NVReadRAMDAC(dev, head, NV_PRAMDAC_TV_HSKEW);
+	regp->tv_hsync_delay = NVReadRAMDAC(dev, head, NV_PRAMDAC_TV_HSYNC_DELAY);
+	regp->tv_hsync_delay2 = NVReadRAMDAC(dev, head, NV_PRAMDAC_TV_HSYNC_DELAY2);
+
+	for (i = 0; i < 7; i++) {
+		uint32_t ramdac_reg = NV_PRAMDAC_FP_VDISPLAY_END + (i * 4);
+		regp->fp_vert_regs[i] = NVReadRAMDAC(dev, head, ramdac_reg);
+		regp->fp_horiz_regs[i] = NVReadRAMDAC(dev, head, ramdac_reg + 0x20);
+	}
+
+	if (nv_gf4_disp_arch(dev)) {
+		regp->dither = NVReadRAMDAC(dev, head, NV_RAMDAC_FP_DITHER);
+		for (i = 0; i < 3; i++) {
+			regp->dither_regs[i] = NVReadRAMDAC(dev, head, NV_PRAMDAC_850 + i * 4);
+			regp->dither_regs[i + 3] = NVReadRAMDAC(dev, head, NV_PRAMDAC_85C + i * 4);
+		}
+	}
+
+	regp->fp_control = NVReadRAMDAC(dev, head, NV_PRAMDAC_FP_TG_CONTROL);
+	regp->fp_debug_0 = NVReadRAMDAC(dev, head, NV_PRAMDAC_FP_DEBUG_0);
+	if (!nv_gf4_disp_arch(dev) && head == 0) {
+		/* early chips don't allow access to PRAMDAC_TMDS_* without
+		 * the head A FPCLK on (nv11 even locks up) */
+		NVWriteRAMDAC(dev, 0, NV_PRAMDAC_FP_DEBUG_0, regp->fp_debug_0 &
+			      ~NV_PRAMDAC_FP_DEBUG_0_PWRDOWN_FPCLK);
+	}
+	regp->fp_debug_1 = NVReadRAMDAC(dev, head, NV_PRAMDAC_FP_DEBUG_1);
+	regp->fp_debug_2 = NVReadRAMDAC(dev, head, NV_PRAMDAC_FP_DEBUG_2);
+
+	regp->fp_margin_color = NVReadRAMDAC(dev, head, NV_PRAMDAC_FP_MARGIN_COLOR);
+
+	if (nv_gf4_disp_arch(dev))
+		regp->ramdac_8c0 = NVReadRAMDAC(dev, head, NV_PRAMDAC_8C0);
+
+	if (dev_priv->card_type == NV_40) {
+		regp->ramdac_a20 = NVReadRAMDAC(dev, head, NV_PRAMDAC_A20);
+		regp->ramdac_a24 = NVReadRAMDAC(dev, head, NV_PRAMDAC_A24);
+		regp->ramdac_a34 = NVReadRAMDAC(dev, head, NV_PRAMDAC_A34);
+
+		for (i = 0; i < 38; i++)
+			regp->ctv_regs[i] = NVReadRAMDAC(dev, head,
+							 NV_PRAMDAC_CTV + 4*i);
+	}
+}
+
+static void
+nv_load_state_ramdac(struct drm_device *dev, int head,
+		     struct nv04_mode_state *state)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nv04_crtc_reg *regp = &state->crtc_reg[head];
+	uint32_t pllreg = head ? NV_RAMDAC_VPLL2 : NV_PRAMDAC_VPLL_COEFF;
+	int i;
+
+	if (dev_priv->card_type >= NV_10)
+		NVWriteRAMDAC(dev, head, NV_RAMDAC_NV10_CURSYNC, regp->nv10_cursync);
+
+	nouveau_hw_setpll(dev, pllreg, &regp->pllvals);
+	NVWriteRAMDAC(dev, 0, NV_PRAMDAC_PLL_COEFF_SELECT, state->pllsel);
+	if (nv_two_heads(dev))
+		NVWriteRAMDAC(dev, 0, NV_PRAMDAC_SEL_CLK, state->sel_clk);
+	if (dev_priv->chipset == 0x11)
+		NVWriteRAMDAC(dev, head, NV_RAMDAC_DITHER_NV11, regp->dither);
+
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_GENERAL_CONTROL, regp->ramdac_gen_ctrl);
+
+	if (nv_gf4_disp_arch(dev))
+		NVWriteRAMDAC(dev, head, NV_PRAMDAC_630, regp->ramdac_630);
+	if (dev_priv->chipset >= 0x30)
+		NVWriteRAMDAC(dev, head, NV_PRAMDAC_634, regp->ramdac_634);
+
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_TV_SETUP, regp->tv_setup);
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_TV_VTOTAL, regp->tv_vtotal);
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_TV_VSKEW, regp->tv_vskew);
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_TV_VSYNC_DELAY, regp->tv_vsync_delay);
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_TV_HTOTAL, regp->tv_htotal);
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_TV_HSKEW, regp->tv_hskew);
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_TV_HSYNC_DELAY, regp->tv_hsync_delay);
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_TV_HSYNC_DELAY2, regp->tv_hsync_delay2);
+
+	for (i = 0; i < 7; i++) {
+		uint32_t ramdac_reg = NV_PRAMDAC_FP_VDISPLAY_END + (i * 4);
+
+		NVWriteRAMDAC(dev, head, ramdac_reg, regp->fp_vert_regs[i]);
+		NVWriteRAMDAC(dev, head, ramdac_reg + 0x20, regp->fp_horiz_regs[i]);
+	}
+
+	if (nv_gf4_disp_arch(dev)) {
+		NVWriteRAMDAC(dev, head, NV_RAMDAC_FP_DITHER, regp->dither);
+		for (i = 0; i < 3; i++) {
+			NVWriteRAMDAC(dev, head, NV_PRAMDAC_850 + i * 4, regp->dither_regs[i]);
+			NVWriteRAMDAC(dev, head, NV_PRAMDAC_85C + i * 4, regp->dither_regs[i + 3]);
+		}
+	}
+
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_FP_TG_CONTROL, regp->fp_control);
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_FP_DEBUG_0, regp->fp_debug_0);
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_FP_DEBUG_1, regp->fp_debug_1);
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_FP_DEBUG_2, regp->fp_debug_2);
+
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_FP_MARGIN_COLOR, regp->fp_margin_color);
+
+	if (nv_gf4_disp_arch(dev))
+		NVWriteRAMDAC(dev, head, NV_PRAMDAC_8C0, regp->ramdac_8c0);
+
+	if (dev_priv->card_type == NV_40) {
+		NVWriteRAMDAC(dev, head, NV_PRAMDAC_A20, regp->ramdac_a20);
+		NVWriteRAMDAC(dev, head, NV_PRAMDAC_A24, regp->ramdac_a24);
+		NVWriteRAMDAC(dev, head, NV_PRAMDAC_A34, regp->ramdac_a34);
+
+		for (i = 0; i < 38; i++)
+			NVWriteRAMDAC(dev, head,
+				      NV_PRAMDAC_CTV + 4*i, regp->ctv_regs[i]);
+	}
+}
+
+static void
+nv_save_state_vga(struct drm_device *dev, int head,
+		  struct nv04_mode_state *state)
+{
+	struct nv04_crtc_reg *regp = &state->crtc_reg[head];
+	int i;
+
+	regp->MiscOutReg = NVReadPRMVIO(dev, head, NV_PRMVIO_MISC__READ);
+
+	for (i = 0; i < 25; i++)
+		rd_cio_state(dev, head, regp, i);
+
+	NVSetEnablePalette(dev, head, true);
+	for (i = 0; i < 21; i++)
+		regp->Attribute[i] = NVReadVgaAttr(dev, head, i);
+	NVSetEnablePalette(dev, head, false);
+
+	for (i = 0; i < 9; i++)
+		regp->Graphics[i] = NVReadVgaGr(dev, head, i);
+
+	for (i = 0; i < 5; i++)
+		regp->Sequencer[i] = NVReadVgaSeq(dev, head, i);
+}
+
+static void
+nv_load_state_vga(struct drm_device *dev, int head,
+		  struct nv04_mode_state *state)
+{
+	struct nv04_crtc_reg *regp = &state->crtc_reg[head];
+	int i;
+
+	NVWritePRMVIO(dev, head, NV_PRMVIO_MISC__WRITE, regp->MiscOutReg);
+
+	for (i = 0; i < 5; i++)
+		NVWriteVgaSeq(dev, head, i, regp->Sequencer[i]);
+
+	nv_lock_vga_crtc_base(dev, head, false);
+	for (i = 0; i < 25; i++)
+		wr_cio_state(dev, head, regp, i);
+	nv_lock_vga_crtc_base(dev, head, true);
+
+	for (i = 0; i < 9; i++)
+		NVWriteVgaGr(dev, head, i, regp->Graphics[i]);
+
+	NVSetEnablePalette(dev, head, true);
+	for (i = 0; i < 21; i++)
+		NVWriteVgaAttr(dev, head, i, regp->Attribute[i]);
+	NVSetEnablePalette(dev, head, false);
+}
+
+static void
+nv_save_state_ext(struct drm_device *dev, int head,
+		  struct nv04_mode_state *state)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nv04_crtc_reg *regp = &state->crtc_reg[head];
+	int i;
+
+	rd_cio_state(dev, head, regp, NV_CIO_CRE_LCD__INDEX);
+	rd_cio_state(dev, head, regp, NV_CIO_CRE_RPC0_INDEX);
+	rd_cio_state(dev, head, regp, NV_CIO_CRE_RPC1_INDEX);
+	rd_cio_state(dev, head, regp, NV_CIO_CRE_LSR_INDEX);
+	rd_cio_state(dev, head, regp, NV_CIO_CRE_PIXEL_INDEX);
+	rd_cio_state(dev, head, regp, NV_CIO_CRE_HEB__INDEX);
+	rd_cio_state(dev, head, regp, NV_CIO_CRE_ENH_INDEX);
+
+	rd_cio_state(dev, head, regp, NV_CIO_CRE_FF_INDEX);
+	rd_cio_state(dev, head, regp, NV_CIO_CRE_FFLWM__INDEX);
+	rd_cio_state(dev, head, regp, NV_CIO_CRE_21);
+	if (dev_priv->card_type >= NV_30)
+		rd_cio_state(dev, head, regp, NV_CIO_CRE_47);
+	rd_cio_state(dev, head, regp, NV_CIO_CRE_49);
+	rd_cio_state(dev, head, regp, NV_CIO_CRE_HCUR_ADDR0_INDEX);
+	rd_cio_state(dev, head, regp, NV_CIO_CRE_HCUR_ADDR1_INDEX);
+	rd_cio_state(dev, head, regp, NV_CIO_CRE_HCUR_ADDR2_INDEX);
+	rd_cio_state(dev, head, regp, NV_CIO_CRE_ILACE__INDEX);
+
+	if (dev_priv->card_type >= NV_10) {
+		regp->crtc_830 = NVReadCRTC(dev, head, NV_PCRTC_830);
+		regp->crtc_834 = NVReadCRTC(dev, head, NV_PCRTC_834);
+
+		if (dev_priv->card_type >= NV_30)
+			regp->gpio_ext = NVReadCRTC(dev, head, NV_PCRTC_GPIO_EXT);
+
+		if (dev_priv->card_type == NV_40)
+			regp->crtc_850 = NVReadCRTC(dev, head, NV_PCRTC_850);
+
+		if (nv_two_heads(dev))
+			regp->crtc_eng_ctrl = NVReadCRTC(dev, head, NV_PCRTC_ENGINE_CTRL);
+		regp->cursor_cfg = NVReadCRTC(dev, head, NV_PCRTC_CURSOR_CONFIG);
+	}
+
+	regp->crtc_cfg = NVReadCRTC(dev, head, NV_PCRTC_CONFIG);
+
+	rd_cio_state(dev, head, regp, NV_CIO_CRE_SCRATCH3__INDEX);
+	rd_cio_state(dev, head, regp, NV_CIO_CRE_SCRATCH4__INDEX);
+	if (dev_priv->card_type >= NV_10) {
+		rd_cio_state(dev, head, regp, NV_CIO_CRE_EBR_INDEX);
+		rd_cio_state(dev, head, regp, NV_CIO_CRE_CSB);
+		rd_cio_state(dev, head, regp, NV_CIO_CRE_4B);
+		rd_cio_state(dev, head, regp, NV_CIO_CRE_TVOUT_LATENCY);
+	}
+	/* NV11 and NV20 don't have this, they stop at 0x52. */
+	if (nv_gf4_disp_arch(dev)) {
+		rd_cio_state(dev, head, regp, NV_CIO_CRE_53);
+		rd_cio_state(dev, head, regp, NV_CIO_CRE_54);
+
+		for (i = 0; i < 0x10; i++)
+			regp->CR58[i] = NVReadVgaCrtc5758(dev, head, i);
+		rd_cio_state(dev, head, regp, NV_CIO_CRE_59);
+		rd_cio_state(dev, head, regp, NV_CIO_CRE_5B);
+
+		rd_cio_state(dev, head, regp, NV_CIO_CRE_85);
+		rd_cio_state(dev, head, regp, NV_CIO_CRE_86);
+	}
+
+	regp->fb_start = NVReadCRTC(dev, head, NV_PCRTC_START);
+}
+
+static void
+nv_load_state_ext(struct drm_device *dev, int head,
+		  struct nv04_mode_state *state)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nv04_crtc_reg *regp = &state->crtc_reg[head];
+	uint32_t reg900;
+	int i;
+
+	if (dev_priv->card_type >= NV_10) {
+		if (nv_two_heads(dev))
+			/* setting ENGINE_CTRL (EC) *must* come before
+			 * CIO_CRE_LCD, as writing CRE_LCD sets bits 16 & 17 in
+			 * EC that should not be overwritten by writing stale EC
+			 */
+			NVWriteCRTC(dev, head, NV_PCRTC_ENGINE_CTRL, regp->crtc_eng_ctrl);
+
+		nvWriteVIDEO(dev, NV_PVIDEO_STOP, 1);
+		nvWriteVIDEO(dev, NV_PVIDEO_INTR_EN, 0);
+		nvWriteVIDEO(dev, NV_PVIDEO_OFFSET_BUFF(0), 0);
+		nvWriteVIDEO(dev, NV_PVIDEO_OFFSET_BUFF(1), 0);
+		nvWriteVIDEO(dev, NV_PVIDEO_LIMIT(0), dev_priv->fb_available_size - 1);
+		nvWriteVIDEO(dev, NV_PVIDEO_LIMIT(1), dev_priv->fb_available_size - 1);
+		nvWriteVIDEO(dev, NV_PVIDEO_UVPLANE_LIMIT(0), dev_priv->fb_available_size - 1);
+		nvWriteVIDEO(dev, NV_PVIDEO_UVPLANE_LIMIT(1), dev_priv->fb_available_size - 1);
+		nvWriteMC(dev, NV_PBUS_POWERCTRL_2, 0);
+
+		NVWriteCRTC(dev, head, NV_PCRTC_CURSOR_CONFIG, regp->cursor_cfg);
+		NVWriteCRTC(dev, head, NV_PCRTC_830, regp->crtc_830);
+		NVWriteCRTC(dev, head, NV_PCRTC_834, regp->crtc_834);
+
+		if (dev_priv->card_type >= NV_30)
+			NVWriteCRTC(dev, head, NV_PCRTC_GPIO_EXT, regp->gpio_ext);
+
+		if (dev_priv->card_type == NV_40) {
+			NVWriteCRTC(dev, head, NV_PCRTC_850, regp->crtc_850);
+
+			reg900 = NVReadRAMDAC(dev, head, NV_PRAMDAC_900);
+			if (regp->crtc_cfg == NV_PCRTC_CONFIG_START_ADDRESS_HSYNC)
+				NVWriteRAMDAC(dev, head, NV_PRAMDAC_900, reg900 | 0x10000);
+			else
+				NVWriteRAMDAC(dev, head, NV_PRAMDAC_900, reg900 & ~0x10000);
+		}
+	}
+
+	NVWriteCRTC(dev, head, NV_PCRTC_CONFIG, regp->crtc_cfg);
+
+	wr_cio_state(dev, head, regp, NV_CIO_CRE_RPC0_INDEX);
+	wr_cio_state(dev, head, regp, NV_CIO_CRE_RPC1_INDEX);
+	wr_cio_state(dev, head, regp, NV_CIO_CRE_LSR_INDEX);
+	wr_cio_state(dev, head, regp, NV_CIO_CRE_PIXEL_INDEX);
+	wr_cio_state(dev, head, regp, NV_CIO_CRE_LCD__INDEX);
+	wr_cio_state(dev, head, regp, NV_CIO_CRE_HEB__INDEX);
+	wr_cio_state(dev, head, regp, NV_CIO_CRE_ENH_INDEX);
+	wr_cio_state(dev, head, regp, NV_CIO_CRE_FF_INDEX);
+	wr_cio_state(dev, head, regp, NV_CIO_CRE_FFLWM__INDEX);
+	if (dev_priv->card_type >= NV_30)
+		wr_cio_state(dev, head, regp, NV_CIO_CRE_47);
+
+	wr_cio_state(dev, head, regp, NV_CIO_CRE_49);
+	wr_cio_state(dev, head, regp, NV_CIO_CRE_HCUR_ADDR0_INDEX);
+	wr_cio_state(dev, head, regp, NV_CIO_CRE_HCUR_ADDR1_INDEX);
+	wr_cio_state(dev, head, regp, NV_CIO_CRE_HCUR_ADDR2_INDEX);
+	if (dev_priv->card_type == NV_40)
+		nv_fix_nv40_hw_cursor(dev, head);
+	wr_cio_state(dev, head, regp, NV_CIO_CRE_ILACE__INDEX);
+
+	wr_cio_state(dev, head, regp, NV_CIO_CRE_SCRATCH3__INDEX);
+	wr_cio_state(dev, head, regp, NV_CIO_CRE_SCRATCH4__INDEX);
+	if (dev_priv->card_type >= NV_10) {
+		wr_cio_state(dev, head, regp, NV_CIO_CRE_EBR_INDEX);
+		wr_cio_state(dev, head, regp, NV_CIO_CRE_CSB);
+		wr_cio_state(dev, head, regp, NV_CIO_CRE_4B);
+		wr_cio_state(dev, head, regp, NV_CIO_CRE_TVOUT_LATENCY);
+	}
+	/* NV11 and NV20 stop at 0x52. */
+	if (nv_gf4_disp_arch(dev)) {
+		if (dev_priv->card_type == NV_10) {
+			/* Not waiting for vertical retrace before modifying
+			   CRE_53/CRE_54 causes lockups. */
+			nouveau_wait_until(dev, 650000000, NV_PRMCIO_INP0__COLOR, 0x8, 0x8);
+			nouveau_wait_until(dev, 650000000, NV_PRMCIO_INP0__COLOR, 0x8, 0x0);
+		}
+
+		wr_cio_state(dev, head, regp, NV_CIO_CRE_53);
+		wr_cio_state(dev, head, regp, NV_CIO_CRE_54);
+
+		for (i = 0; i < 0x10; i++)
+			NVWriteVgaCrtc5758(dev, head, i, regp->CR58[i]);
+		wr_cio_state(dev, head, regp, NV_CIO_CRE_59);
+		wr_cio_state(dev, head, regp, NV_CIO_CRE_5B);
+
+		wr_cio_state(dev, head, regp, NV_CIO_CRE_85);
+		wr_cio_state(dev, head, regp, NV_CIO_CRE_86);
+	}
+
+	NVWriteCRTC(dev, head, NV_PCRTC_START, regp->fb_start);
+
+	/* Setting 1 on this value gives you interrupts for every vblank period. */
+	NVWriteCRTC(dev, head, NV_PCRTC_INTR_EN_0, 0);
+	NVWriteCRTC(dev, head, NV_PCRTC_INTR_0, NV_PCRTC_INTR_0_VBLANK);
+}
+
+static void
+nv_save_state_palette(struct drm_device *dev, int head,
+		      struct nv04_mode_state *state)
+{
+	int head_offset = head * NV_PRMDIO_SIZE, i;
+
+	nv_wr08(dev, NV_PRMDIO_PIXEL_MASK + head_offset,
+				NV_PRMDIO_PIXEL_MASK_MASK);
+	nv_wr08(dev, NV_PRMDIO_READ_MODE_ADDRESS + head_offset, 0x0);
+
+	for (i = 0; i < 768; i++) {
+		state->crtc_reg[head].DAC[i] = nv_rd08(dev,
+				NV_PRMDIO_PALETTE_DATA + head_offset);
+	}
+
+	NVSetEnablePalette(dev, head, false);
+}
+
+void
+nouveau_hw_load_state_palette(struct drm_device *dev, int head,
+			      struct nv04_mode_state *state)
+{
+	int head_offset = head * NV_PRMDIO_SIZE, i;
+
+	nv_wr08(dev, NV_PRMDIO_PIXEL_MASK + head_offset,
+				NV_PRMDIO_PIXEL_MASK_MASK);
+	nv_wr08(dev, NV_PRMDIO_WRITE_MODE_ADDRESS + head_offset, 0x0);
+
+	for (i = 0; i < 768; i++) {
+		nv_wr08(dev, NV_PRMDIO_PALETTE_DATA + head_offset,
+				state->crtc_reg[head].DAC[i]);
+	}
+
+	NVSetEnablePalette(dev, head, false);
+}
+
+void nouveau_hw_save_state(struct drm_device *dev, int head,
+			   struct nv04_mode_state *state)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	if (dev_priv->chipset == 0x11)
+		/* NB: no attempt is made to restore the bad pll later on */
+		nouveau_hw_fix_bad_vpll(dev, head);
+	nv_save_state_ramdac(dev, head, state);
+	nv_save_state_vga(dev, head, state);
+	nv_save_state_palette(dev, head, state);
+	nv_save_state_ext(dev, head, state);
+}
+
+void nouveau_hw_load_state(struct drm_device *dev, int head,
+			   struct nv04_mode_state *state)
+{
+	NVVgaProtect(dev, head, true);
+	nv_load_state_ramdac(dev, head, state);
+	nv_load_state_ext(dev, head, state);
+	nouveau_hw_load_state_palette(dev, head, state);
+	nv_load_state_vga(dev, head, state);
+	NVVgaProtect(dev, head, false);
+}
diff --git a/drivers/gpu/drm/nouveau/nouveau_hw.h b/drivers/gpu/drm/nouveau/nouveau_hw.h
new file mode 100644
index 000000000000..869130f83602
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_hw.h
@@ -0,0 +1,455 @@
+/*
+ * Copyright 2008 Stuart Bennett
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF
+ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef __NOUVEAU_HW_H__
+#define __NOUVEAU_HW_H__
+
+#include "drmP.h"
+#include "nouveau_drv.h"
+
+#define MASK(field) ( \
+	(0xffffffff >> (31 - ((1 ? field) - (0 ? field)))) << (0 ? field))
+
+#define XLATE(src, srclowbit, outfield) ( \
+	(((src) >> (srclowbit)) << (0 ? outfield)) & MASK(outfield))
+
+void NVWriteVgaSeq(struct drm_device *, int head, uint8_t index, uint8_t value);
+uint8_t NVReadVgaSeq(struct drm_device *, int head, uint8_t index);
+void NVWriteVgaGr(struct drm_device *, int head, uint8_t index, uint8_t value);
+uint8_t NVReadVgaGr(struct drm_device *, int head, uint8_t index);
+void NVSetOwner(struct drm_device *, int owner);
+void NVBlankScreen(struct drm_device *, int head, bool blank);
+void nouveau_hw_setpll(struct drm_device *, uint32_t reg1,
+		       struct nouveau_pll_vals *pv);
+int nouveau_hw_get_pllvals(struct drm_device *, enum pll_types plltype,
+			   struct nouveau_pll_vals *pllvals);
+int nouveau_hw_pllvals_to_clk(struct nouveau_pll_vals *pllvals);
+int nouveau_hw_get_clock(struct drm_device *, enum pll_types plltype);
+void nouveau_hw_save_vga_fonts(struct drm_device *, bool save);
+void nouveau_hw_save_state(struct drm_device *, int head,
+			   struct nv04_mode_state *state);
+void nouveau_hw_load_state(struct drm_device *, int head,
+			   struct nv04_mode_state *state);
+void nouveau_hw_load_state_palette(struct drm_device *, int head,
+				   struct nv04_mode_state *state);
+
+/* nouveau_calc.c */
+extern void nouveau_calc_arb(struct drm_device *, int vclk, int bpp,
+			     int *burst, int *lwm);
+extern int nouveau_calc_pll_mnp(struct drm_device *, struct pll_lims *pll_lim,
+				int clk, struct nouveau_pll_vals *pv);
+
+static inline uint32_t
+nvReadMC(struct drm_device *dev, uint32_t reg)
+{
+	uint32_t val = nv_rd32(dev, reg);
+	NV_REG_DEBUG(MC, dev, "reg %08x val %08x\n", reg, val);
+	return val;
+}
+
+static inline void
+nvWriteMC(struct drm_device *dev, uint32_t reg, uint32_t val)
+{
+	NV_REG_DEBUG(MC, dev, "reg %08x val %08x\n", reg, val);
+	nv_wr32(dev, reg, val);
+}
+
+static inline uint32_t
+nvReadVIDEO(struct drm_device *dev, uint32_t reg)
+{
+	uint32_t val = nv_rd32(dev, reg);
+	NV_REG_DEBUG(VIDEO, dev, "reg %08x val %08x\n", reg, val);
+	return val;
+}
+
+static inline void
+nvWriteVIDEO(struct drm_device *dev, uint32_t reg, uint32_t val)
+{
+	NV_REG_DEBUG(VIDEO, dev, "reg %08x val %08x\n", reg, val);
+	nv_wr32(dev, reg, val);
+}
+
+static inline uint32_t
+nvReadFB(struct drm_device *dev, uint32_t reg)
+{
+	uint32_t val = nv_rd32(dev, reg);
+	NV_REG_DEBUG(FB, dev, "reg %08x val %08x\n", reg, val);
+	return val;
+}
+
+static inline void
+nvWriteFB(struct drm_device *dev, uint32_t reg, uint32_t val)
+{
+	NV_REG_DEBUG(FB, dev, "reg %08x val %08x\n", reg, val);
+	nv_wr32(dev, reg, val);
+}
+
+static inline uint32_t
+nvReadEXTDEV(struct drm_device *dev, uint32_t reg)
+{
+	uint32_t val = nv_rd32(dev, reg);
+	NV_REG_DEBUG(EXTDEV, dev, "reg %08x val %08x\n", reg, val);
+	return val;
+}
+
+static inline void
+nvWriteEXTDEV(struct drm_device *dev, uint32_t reg, uint32_t val)
+{
+	NV_REG_DEBUG(EXTDEV, dev, "reg %08x val %08x\n", reg, val);
+	nv_wr32(dev, reg, val);
+}
+
+static inline uint32_t NVReadCRTC(struct drm_device *dev,
+					int head, uint32_t reg)
+{
+	uint32_t val;
+	if (head)
+		reg += NV_PCRTC0_SIZE;
+	val = nv_rd32(dev, reg);
+	NV_REG_DEBUG(CRTC, dev, "head %d reg %08x val %08x\n", head, reg, val);
+	return val;
+}
+
+static inline void NVWriteCRTC(struct drm_device *dev,
+					int head, uint32_t reg, uint32_t val)
+{
+	if (head)
+		reg += NV_PCRTC0_SIZE;
+	NV_REG_DEBUG(CRTC, dev, "head %d reg %08x val %08x\n", head, reg, val);
+	nv_wr32(dev, reg, val);
+}
+
+static inline uint32_t NVReadRAMDAC(struct drm_device *dev,
+					int head, uint32_t reg)
+{
+	uint32_t val;
+	if (head)
+		reg += NV_PRAMDAC0_SIZE;
+	val = nv_rd32(dev, reg);
+	NV_REG_DEBUG(RAMDAC, dev, "head %d reg %08x val %08x\n",
+							head, reg, val);
+	return val;
+}
+
+static inline void NVWriteRAMDAC(struct drm_device *dev,
+					int head, uint32_t reg, uint32_t val)
+{
+	if (head)
+		reg += NV_PRAMDAC0_SIZE;
+	NV_REG_DEBUG(RAMDAC, dev, "head %d reg %08x val %08x\n",
+							head, reg, val);
+	nv_wr32(dev, reg, val);
+}
+
+static inline uint8_t nv_read_tmds(struct drm_device *dev,
+					int or, int dl, uint8_t address)
+{
+	int ramdac = (or & OUTPUT_C) >> 2;
+
+	NVWriteRAMDAC(dev, ramdac, NV_PRAMDAC_FP_TMDS_CONTROL + dl * 8,
+	NV_PRAMDAC_FP_TMDS_CONTROL_WRITE_DISABLE | address);
+	return NVReadRAMDAC(dev, ramdac, NV_PRAMDAC_FP_TMDS_DATA + dl * 8);
+}
+
+static inline void nv_write_tmds(struct drm_device *dev,
+					int or, int dl, uint8_t address,
+					uint8_t data)
+{
+	int ramdac = (or & OUTPUT_C) >> 2;
+
+	NVWriteRAMDAC(dev, ramdac, NV_PRAMDAC_FP_TMDS_DATA + dl * 8, data);
+	NVWriteRAMDAC(dev, ramdac, NV_PRAMDAC_FP_TMDS_CONTROL + dl * 8, address);
+}
+
+static inline void NVWriteVgaCrtc(struct drm_device *dev,
+					int head, uint8_t index, uint8_t value)
+{
+	NV_REG_DEBUG(VGACRTC, dev, "head %d index 0x%02x data 0x%02x\n",
+							head, index, value);
+	nv_wr08(dev, NV_PRMCIO_CRX__COLOR + head * NV_PRMCIO_SIZE, index);
+	nv_wr08(dev, NV_PRMCIO_CR__COLOR + head * NV_PRMCIO_SIZE, value);
+}
+
+static inline uint8_t NVReadVgaCrtc(struct drm_device *dev,
+					int head, uint8_t index)
+{
+	uint8_t val;
+	nv_wr08(dev, NV_PRMCIO_CRX__COLOR + head * NV_PRMCIO_SIZE, index);
+	val = nv_rd08(dev, NV_PRMCIO_CR__COLOR + head * NV_PRMCIO_SIZE);
+	NV_REG_DEBUG(VGACRTC, dev, "head %d index 0x%02x data 0x%02x\n",
+							head, index, val);
+	return val;
+}
+
+/* CR57 and CR58 are a fun pair of regs. CR57 provides an index (0-0xf) for CR58
+ * I suspect they in fact do nothing, but are merely a way to carry useful
+ * per-head variables around
+ *
+ * Known uses:
+ * CR57		CR58
+ * 0x00		index to the appropriate dcb entry (or 7f for inactive)
+ * 0x02		dcb entry's "or" value (or 00 for inactive)
+ * 0x03		bit0 set for dual link (LVDS, possibly elsewhere too)
+ * 0x08 or 0x09	pxclk in MHz
+ * 0x0f		laptop panel info -	low nibble for PEXTDEV_BOOT_0 strap
+ * 					high nibble for xlat strap value
+ */
+
+static inline void
+NVWriteVgaCrtc5758(struct drm_device *dev, int head, uint8_t index, uint8_t value)
+{
+	NVWriteVgaCrtc(dev, head, NV_CIO_CRE_57, index);
+	NVWriteVgaCrtc(dev, head, NV_CIO_CRE_58, value);
+}
+
+static inline uint8_t NVReadVgaCrtc5758(struct drm_device *dev, int head, uint8_t index)
+{
+	NVWriteVgaCrtc(dev, head, NV_CIO_CRE_57, index);
+	return NVReadVgaCrtc(dev, head, NV_CIO_CRE_58);
+}
+
+static inline uint8_t NVReadPRMVIO(struct drm_device *dev,
+					int head, uint32_t reg)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint8_t val;
+
+	/* Only NV4x have two pvio ranges; other twoHeads cards MUST call
+	 * NVSetOwner for the relevant head to be programmed */
+	if (head && dev_priv->card_type == NV_40)
+		reg += NV_PRMVIO_SIZE;
+
+	val = nv_rd08(dev, reg);
+	NV_REG_DEBUG(RMVIO, dev, "head %d reg %08x val %02x\n", head, reg, val);
+	return val;
+}
+
+static inline void NVWritePRMVIO(struct drm_device *dev,
+					int head, uint32_t reg, uint8_t value)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	/* Only NV4x have two pvio ranges; other twoHeads cards MUST call
+	 * NVSetOwner for the relevant head to be programmed */
+	if (head && dev_priv->card_type == NV_40)
+		reg += NV_PRMVIO_SIZE;
+
+	NV_REG_DEBUG(RMVIO, dev, "head %d reg %08x val %02x\n",
+						head, reg, value);
+	nv_wr08(dev, reg, value);
+}
+
+static inline void NVSetEnablePalette(struct drm_device *dev, int head, bool enable)
+{
+	nv_rd08(dev, NV_PRMCIO_INP0__COLOR + head * NV_PRMCIO_SIZE);
+	nv_wr08(dev, NV_PRMCIO_ARX + head * NV_PRMCIO_SIZE, enable ? 0 : 0x20);
+}
+
+static inline bool NVGetEnablePalette(struct drm_device *dev, int head)
+{
+	nv_rd08(dev, NV_PRMCIO_INP0__COLOR + head * NV_PRMCIO_SIZE);
+	return !(nv_rd08(dev, NV_PRMCIO_ARX + head * NV_PRMCIO_SIZE) & 0x20);
+}
+
+static inline void NVWriteVgaAttr(struct drm_device *dev,
+					int head, uint8_t index, uint8_t value)
+{
+	if (NVGetEnablePalette(dev, head))
+		index &= ~0x20;
+	else
+		index |= 0x20;
+
+	nv_rd08(dev, NV_PRMCIO_INP0__COLOR + head * NV_PRMCIO_SIZE);
+	NV_REG_DEBUG(VGAATTR, dev, "head %d index 0x%02x data 0x%02x\n",
+							head, index, value);
+	nv_wr08(dev, NV_PRMCIO_ARX + head * NV_PRMCIO_SIZE, index);
+	nv_wr08(dev, NV_PRMCIO_AR__WRITE + head * NV_PRMCIO_SIZE, value);
+}
+
+static inline uint8_t NVReadVgaAttr(struct drm_device *dev,
+					int head, uint8_t index)
+{
+	uint8_t val;
+	if (NVGetEnablePalette(dev, head))
+		index &= ~0x20;
+	else
+		index |= 0x20;
+
+	nv_rd08(dev, NV_PRMCIO_INP0__COLOR + head * NV_PRMCIO_SIZE);
+	nv_wr08(dev, NV_PRMCIO_ARX + head * NV_PRMCIO_SIZE, index);
+	val = nv_rd08(dev, NV_PRMCIO_AR__READ + head * NV_PRMCIO_SIZE);
+	NV_REG_DEBUG(VGAATTR, dev, "head %d index 0x%02x data 0x%02x\n",
+							head, index, val);
+	return val;
+}
+
+static inline void NVVgaSeqReset(struct drm_device *dev, int head, bool start)
+{
+	NVWriteVgaSeq(dev, head, NV_VIO_SR_RESET_INDEX, start ? 0x1 : 0x3);
+}
+
+static inline void NVVgaProtect(struct drm_device *dev, int head, bool protect)
+{
+	uint8_t seq1 = NVReadVgaSeq(dev, head, NV_VIO_SR_CLOCK_INDEX);
+
+	if (protect) {
+		NVVgaSeqReset(dev, head, true);
+		NVWriteVgaSeq(dev, head, NV_VIO_SR_CLOCK_INDEX, seq1 | 0x20);
+	} else {
+		/* Reenable sequencer, then turn on screen */
+		NVWriteVgaSeq(dev, head, NV_VIO_SR_CLOCK_INDEX, seq1 & ~0x20);   /* reenable display */
+		NVVgaSeqReset(dev, head, false);
+	}
+	NVSetEnablePalette(dev, head, protect);
+}
+
+static inline bool
+nv_heads_tied(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	if (dev_priv->chipset == 0x11)
+		return !!(nvReadMC(dev, NV_PBUS_DEBUG_1) & (1 << 28));
+
+	return NVReadVgaCrtc(dev, 0, NV_CIO_CRE_44) & 0x4;
+}
+
+/* makes cr0-7 on the specified head read-only */
+static inline bool
+nv_lock_vga_crtc_base(struct drm_device *dev, int head, bool lock)
+{
+	uint8_t cr11 = NVReadVgaCrtc(dev, head, NV_CIO_CR_VRE_INDEX);
+	bool waslocked = cr11 & 0x80;
+
+	if (lock)
+		cr11 |= 0x80;
+	else
+		cr11 &= ~0x80;
+	NVWriteVgaCrtc(dev, head, NV_CIO_CR_VRE_INDEX, cr11);
+
+	return waslocked;
+}
+
+static inline void
+nv_lock_vga_crtc_shadow(struct drm_device *dev, int head, int lock)
+{
+	/* shadow lock: connects 0x60?3d? regs to "real" 0x3d? regs
+	 * bit7: unlocks HDT, HBS, HBE, HRS, HRE, HEB
+	 * bit6: seems to have some effect on CR09 (double scan, VBS_9)
+	 * bit5: unlocks HDE
+	 * bit4: unlocks VDE
+	 * bit3: unlocks VDT, OVL, VRS, ?VRE?, VBS, VBE, LSR, EBR
+	 * bit2: same as bit 1 of 0x60?804
+	 * bit0: same as bit 0 of 0x60?804
+	 */
+
+	uint8_t cr21 = lock;
+
+	if (lock < 0)
+		/* 0xfa is generic "unlock all" mask */
+		cr21 = NVReadVgaCrtc(dev, head, NV_CIO_CRE_21) | 0xfa;
+
+	NVWriteVgaCrtc(dev, head, NV_CIO_CRE_21, cr21);
+}
+
+/* renders the extended crtc regs (cr19+) on all crtcs impervious:
+ * immutable and unreadable
+ */
+static inline bool
+NVLockVgaCrtcs(struct drm_device *dev, bool lock)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	bool waslocked = !NVReadVgaCrtc(dev, 0, NV_CIO_SR_LOCK_INDEX);
+
+	NVWriteVgaCrtc(dev, 0, NV_CIO_SR_LOCK_INDEX,
+		       lock ? NV_CIO_SR_LOCK_VALUE : NV_CIO_SR_UNLOCK_RW_VALUE);
+	/* NV11 has independently lockable extended crtcs, except when tied */
+	if (dev_priv->chipset == 0x11 && !nv_heads_tied(dev))
+		NVWriteVgaCrtc(dev, 1, NV_CIO_SR_LOCK_INDEX,
+			       lock ? NV_CIO_SR_LOCK_VALUE :
+				      NV_CIO_SR_UNLOCK_RW_VALUE);
+
+	return waslocked;
+}
+
+/* nv04 cursor max dimensions of 32x32 (A1R5G5B5) */
+#define NV04_CURSOR_SIZE 32
+/* limit nv10 cursors to 64x64 (ARGB8) (we could go to 64x255) */
+#define NV10_CURSOR_SIZE 64
+
+static inline int nv_cursor_width(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	return dev_priv->card_type >= NV_10 ? NV10_CURSOR_SIZE : NV04_CURSOR_SIZE;
+}
+
+static inline void
+nv_fix_nv40_hw_cursor(struct drm_device *dev, int head)
+{
+	/* on some nv40 (such as the "true" (in the NV_PFB_BOOT_0 sense) nv40,
+	 * the gf6800gt) a hardware bug requires a write to PRAMDAC_CURSOR_POS
+	 * for changes to the CRTC CURCTL regs to take effect, whether changing
+	 * the pixmap location, or just showing/hiding the cursor
+	 */
+	uint32_t curpos = NVReadRAMDAC(dev, head, NV_PRAMDAC_CU_START_POS);
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_CU_START_POS, curpos);
+}
+
+static inline void
+nv_show_cursor(struct drm_device *dev, int head, bool show)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint8_t *curctl1 =
+		&dev_priv->mode_reg.crtc_reg[head].CRTC[NV_CIO_CRE_HCUR_ADDR1_INDEX];
+
+	if (show)
+		*curctl1 |= MASK(NV_CIO_CRE_HCUR_ADDR1_ENABLE);
+	else
+		*curctl1 &= ~MASK(NV_CIO_CRE_HCUR_ADDR1_ENABLE);
+	NVWriteVgaCrtc(dev, head, NV_CIO_CRE_HCUR_ADDR1_INDEX, *curctl1);
+
+	if (dev_priv->card_type == NV_40)
+		nv_fix_nv40_hw_cursor(dev, head);
+}
+
+static inline uint32_t
+nv_pitch_align(struct drm_device *dev, uint32_t width, int bpp)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	int mask;
+
+	if (bpp == 15)
+		bpp = 16;
+	if (bpp == 24)
+		bpp = 8;
+
+	/* Alignment requirements taken from the Haiku driver */
+	if (dev_priv->card_type == NV_04)
+		mask = 128 / bpp - 1;
+	else
+		mask = 512 / bpp - 1;
+
+	return (width + mask) & ~mask;
+}
+
+#endif	/* __NOUVEAU_HW_H__ */
diff --git a/drivers/gpu/drm/nouveau/nouveau_i2c.c b/drivers/gpu/drm/nouveau/nouveau_i2c.c
new file mode 100644
index 000000000000..88583e7bf651
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_i2c.c
@@ -0,0 +1,269 @@
+/*
+ * Copyright 2009 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+#include "drmP.h"
+#include "nouveau_drv.h"
+#include "nouveau_i2c.h"
+#include "nouveau_hw.h"
+
+static void
+nv04_i2c_setscl(void *data, int state)
+{
+	struct nouveau_i2c_chan *i2c = data;
+	struct drm_device *dev = i2c->dev;
+	uint8_t val;
+
+	val = (NVReadVgaCrtc(dev, 0, i2c->wr) & 0xd0) | (state ? 0x20 : 0);
+	NVWriteVgaCrtc(dev, 0, i2c->wr, val | 0x01);
+}
+
+static void
+nv04_i2c_setsda(void *data, int state)
+{
+	struct nouveau_i2c_chan *i2c = data;
+	struct drm_device *dev = i2c->dev;
+	uint8_t val;
+
+	val = (NVReadVgaCrtc(dev, 0, i2c->wr) & 0xe0) | (state ? 0x10 : 0);
+	NVWriteVgaCrtc(dev, 0, i2c->wr, val | 0x01);
+}
+
+static int
+nv04_i2c_getscl(void *data)
+{
+	struct nouveau_i2c_chan *i2c = data;
+	struct drm_device *dev = i2c->dev;
+
+	return !!(NVReadVgaCrtc(dev, 0, i2c->rd) & 4);
+}
+
+static int
+nv04_i2c_getsda(void *data)
+{
+	struct nouveau_i2c_chan *i2c = data;
+	struct drm_device *dev = i2c->dev;
+
+	return !!(NVReadVgaCrtc(dev, 0, i2c->rd) & 8);
+}
+
+static void
+nv4e_i2c_setscl(void *data, int state)
+{
+	struct nouveau_i2c_chan *i2c = data;
+	struct drm_device *dev = i2c->dev;
+	uint8_t val;
+
+	val = (nv_rd32(dev, i2c->wr) & 0xd0) | (state ? 0x20 : 0);
+	nv_wr32(dev, i2c->wr, val | 0x01);
+}
+
+static void
+nv4e_i2c_setsda(void *data, int state)
+{
+	struct nouveau_i2c_chan *i2c = data;
+	struct drm_device *dev = i2c->dev;
+	uint8_t val;
+
+	val = (nv_rd32(dev, i2c->wr) & 0xe0) | (state ? 0x10 : 0);
+	nv_wr32(dev, i2c->wr, val | 0x01);
+}
+
+static int
+nv4e_i2c_getscl(void *data)
+{
+	struct nouveau_i2c_chan *i2c = data;
+	struct drm_device *dev = i2c->dev;
+
+	return !!((nv_rd32(dev, i2c->rd) >> 16) & 4);
+}
+
+static int
+nv4e_i2c_getsda(void *data)
+{
+	struct nouveau_i2c_chan *i2c = data;
+	struct drm_device *dev = i2c->dev;
+
+	return !!((nv_rd32(dev, i2c->rd) >> 16) & 8);
+}
+
+static int
+nv50_i2c_getscl(void *data)
+{
+	struct nouveau_i2c_chan *i2c = data;
+	struct drm_device *dev = i2c->dev;
+
+	return !!(nv_rd32(dev, i2c->rd) & 1);
+}
+
+
+static int
+nv50_i2c_getsda(void *data)
+{
+	struct nouveau_i2c_chan *i2c = data;
+	struct drm_device *dev = i2c->dev;
+
+	return !!(nv_rd32(dev, i2c->rd) & 2);
+}
+
+static void
+nv50_i2c_setscl(void *data, int state)
+{
+	struct nouveau_i2c_chan *i2c = data;
+	struct drm_device *dev = i2c->dev;
+
+	nv_wr32(dev, i2c->wr, 4 | (i2c->data ? 2 : 0) | (state ? 1 : 0));
+}
+
+static void
+nv50_i2c_setsda(void *data, int state)
+{
+	struct nouveau_i2c_chan *i2c = data;
+	struct drm_device *dev = i2c->dev;
+
+	nv_wr32(dev, i2c->wr,
+			(nv_rd32(dev, i2c->rd) & 1) | 4 | (state ? 2 : 0));
+	i2c->data = state;
+}
+
+static const uint32_t nv50_i2c_port[] = {
+	0x00e138, 0x00e150, 0x00e168, 0x00e180,
+	0x00e254, 0x00e274, 0x00e764, 0x00e780,
+	0x00e79c, 0x00e7b8
+};
+#define NV50_I2C_PORTS ARRAY_SIZE(nv50_i2c_port)
+
+int
+nouveau_i2c_init(struct drm_device *dev, struct dcb_i2c_entry *entry, int index)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_i2c_chan *i2c;
+	int ret;
+
+	if (entry->chan)
+		return -EEXIST;
+
+	if (dev_priv->card_type == NV_50 && entry->read >= NV50_I2C_PORTS) {
+		NV_ERROR(dev, "unknown i2c port %d\n", entry->read);
+		return -EINVAL;
+	}
+
+	i2c = kzalloc(sizeof(*i2c), GFP_KERNEL);
+	if (i2c == NULL)
+		return -ENOMEM;
+
+	switch (entry->port_type) {
+	case 0:
+		i2c->algo.bit.setsda = nv04_i2c_setsda;
+		i2c->algo.bit.setscl = nv04_i2c_setscl;
+		i2c->algo.bit.getsda = nv04_i2c_getsda;
+		i2c->algo.bit.getscl = nv04_i2c_getscl;
+		i2c->rd = entry->read;
+		i2c->wr = entry->write;
+		break;
+	case 4:
+		i2c->algo.bit.setsda = nv4e_i2c_setsda;
+		i2c->algo.bit.setscl = nv4e_i2c_setscl;
+		i2c->algo.bit.getsda = nv4e_i2c_getsda;
+		i2c->algo.bit.getscl = nv4e_i2c_getscl;
+		i2c->rd = 0x600800 + entry->read;
+		i2c->wr = 0x600800 + entry->write;
+		break;
+	case 5:
+		i2c->algo.bit.setsda = nv50_i2c_setsda;
+		i2c->algo.bit.setscl = nv50_i2c_setscl;
+		i2c->algo.bit.getsda = nv50_i2c_getsda;
+		i2c->algo.bit.getscl = nv50_i2c_getscl;
+		i2c->rd = nv50_i2c_port[entry->read];
+		i2c->wr = i2c->rd;
+		break;
+	case 6:
+		i2c->rd = entry->read;
+		i2c->wr = entry->write;
+		break;
+	default:
+		NV_ERROR(dev, "DCB I2C port type %d unknown\n",
+			 entry->port_type);
+		kfree(i2c);
+		return -EINVAL;
+	}
+
+	snprintf(i2c->adapter.name, sizeof(i2c->adapter.name),
+		 "nouveau-%s-%d", pci_name(dev->pdev), index);
+	i2c->adapter.owner = THIS_MODULE;
+	i2c->adapter.dev.parent = &dev->pdev->dev;
+	i2c->dev = dev;
+	i2c_set_adapdata(&i2c->adapter, i2c);
+
+	if (entry->port_type < 6) {
+		i2c->adapter.algo_data = &i2c->algo.bit;
+		i2c->algo.bit.udelay = 40;
+		i2c->algo.bit.timeout = usecs_to_jiffies(5000);
+		i2c->algo.bit.data = i2c;
+		ret = i2c_bit_add_bus(&i2c->adapter);
+	} else {
+		i2c->adapter.algo_data = &i2c->algo.dp;
+		i2c->algo.dp.running = false;
+		i2c->algo.dp.address = 0;
+		i2c->algo.dp.aux_ch = nouveau_dp_i2c_aux_ch;
+		ret = i2c_dp_aux_add_bus(&i2c->adapter);
+	}
+
+	if (ret) {
+		NV_ERROR(dev, "Failed to register i2c %d\n", index);
+		kfree(i2c);
+		return ret;
+	}
+
+	entry->chan = i2c;
+	return 0;
+}
+
+void
+nouveau_i2c_fini(struct drm_device *dev, struct dcb_i2c_entry *entry)
+{
+	if (!entry->chan)
+		return;
+
+	i2c_del_adapter(&entry->chan->adapter);
+	kfree(entry->chan);
+	entry->chan = NULL;
+}
+
+struct nouveau_i2c_chan *
+nouveau_i2c_find(struct drm_device *dev, int index)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nvbios *bios = &dev_priv->vbios;
+
+	if (index >= DCB_MAX_NUM_I2C_ENTRIES)
+		return NULL;
+
+	if (!bios->dcb.i2c[index].chan) {
+		if (nouveau_i2c_init(dev, &bios->dcb.i2c[index], index))
+			return NULL;
+	}
+
+	return bios->dcb.i2c[index].chan;
+}
+
diff --git a/drivers/gpu/drm/nouveau/nouveau_i2c.h b/drivers/gpu/drm/nouveau/nouveau_i2c.h
new file mode 100644
index 000000000000..c8eaf7a9fcbb
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_i2c.h
@@ -0,0 +1,52 @@
+/*
+ * Copyright 2009 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef __NOUVEAU_I2C_H__
+#define __NOUVEAU_I2C_H__
+
+#include <linux/i2c.h>
+#include <linux/i2c-id.h>
+#include <linux/i2c-algo-bit.h>
+#include "drm_dp_helper.h"
+
+struct dcb_i2c_entry;
+
+struct nouveau_i2c_chan {
+	struct i2c_adapter adapter;
+	struct drm_device *dev;
+	union {
+		struct i2c_algo_bit_data bit;
+		struct i2c_algo_dp_aux_data dp;
+	} algo;
+	unsigned rd;
+	unsigned wr;
+	unsigned data;
+};
+
+int nouveau_i2c_init(struct drm_device *, struct dcb_i2c_entry *, int index);
+void nouveau_i2c_fini(struct drm_device *, struct dcb_i2c_entry *);
+struct nouveau_i2c_chan *nouveau_i2c_find(struct drm_device *, int index);
+
+int nouveau_dp_i2c_aux_ch(struct i2c_adapter *, int mode, uint8_t write_byte,
+			  uint8_t *read_byte);
+
+#endif /* __NOUVEAU_I2C_H__ */
diff --git a/drivers/gpu/drm/nouveau/nouveau_ioc32.c b/drivers/gpu/drm/nouveau/nouveau_ioc32.c
new file mode 100644
index 000000000000..475ba810bba3
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_ioc32.c
@@ -0,0 +1,70 @@
+/**
+ * \file mga_ioc32.c
+ *
+ * 32-bit ioctl compatibility routines for the MGA DRM.
+ *
+ * \author Dave Airlie <airlied@linux.ie> with code from patches by Egbert Eich
+ *
+ *
+ * Copyright (C) Paul Mackerras 2005
+ * Copyright (C) Egbert Eich 2003,2004
+ * Copyright (C) Dave Airlie 2005
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHOR BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include <linux/compat.h>
+
+#include "drmP.h"
+#include "drm.h"
+
+#include "nouveau_drv.h"
+
+/**
+ * Called whenever a 32-bit process running under a 64-bit kernel
+ * performs an ioctl on /dev/dri/card<n>.
+ *
+ * \param filp file pointer.
+ * \param cmd command.
+ * \param arg user argument.
+ * \return zero on success or negative number on failure.
+ */
+long nouveau_compat_ioctl(struct file *filp, unsigned int cmd,
+			 unsigned long arg)
+{
+	unsigned int nr = DRM_IOCTL_NR(cmd);
+	drm_ioctl_compat_t *fn = NULL;
+	int ret;
+
+	if (nr < DRM_COMMAND_BASE)
+		return drm_compat_ioctl(filp, cmd, arg);
+
+#if 0
+	if (nr < DRM_COMMAND_BASE + DRM_ARRAY_SIZE(mga_compat_ioctls))
+		fn = nouveau_compat_ioctls[nr - DRM_COMMAND_BASE];
+#endif
+	if (fn != NULL)
+		ret = (*fn)(filp, cmd, arg);
+	else
+		ret = drm_ioctl(filp, cmd, arg);
+
+	return ret;
+}
diff --git a/drivers/gpu/drm/nouveau/nouveau_irq.c b/drivers/gpu/drm/nouveau/nouveau_irq.c
new file mode 100644
index 000000000000..95220ddebb45
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_irq.c
@@ -0,0 +1,742 @@
+/*
+ * Copyright (C) 2006 Ben Skeggs.
+ *
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+/*
+ * Authors:
+ *   Ben Skeggs <darktama@iinet.net.au>
+ */
+
+#include "drmP.h"
+#include "drm.h"
+#include "nouveau_drm.h"
+#include "nouveau_drv.h"
+#include "nouveau_reg.h"
+#include <linux/ratelimit.h>
+
+/* needed for hotplug irq */
+#include "nouveau_connector.h"
+#include "nv50_display.h"
+
+void
+nouveau_irq_preinstall(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	/* Master disable */
+	nv_wr32(dev, NV03_PMC_INTR_EN_0, 0);
+
+	if (dev_priv->card_type == NV_50) {
+		INIT_WORK(&dev_priv->irq_work, nv50_display_irq_handler_bh);
+		INIT_LIST_HEAD(&dev_priv->vbl_waiting);
+	}
+}
+
+int
+nouveau_irq_postinstall(struct drm_device *dev)
+{
+	/* Master enable */
+	nv_wr32(dev, NV03_PMC_INTR_EN_0, NV_PMC_INTR_EN_0_MASTER_ENABLE);
+	return 0;
+}
+
+void
+nouveau_irq_uninstall(struct drm_device *dev)
+{
+	/* Master disable */
+	nv_wr32(dev, NV03_PMC_INTR_EN_0, 0);
+}
+
+static int
+nouveau_call_method(struct nouveau_channel *chan, int class, int mthd, int data)
+{
+	struct drm_nouveau_private *dev_priv = chan->dev->dev_private;
+	struct nouveau_pgraph_object_method *grm;
+	struct nouveau_pgraph_object_class *grc;
+
+	grc = dev_priv->engine.graph.grclass;
+	while (grc->id) {
+		if (grc->id == class)
+			break;
+		grc++;
+	}
+
+	if (grc->id != class || !grc->methods)
+		return -ENOENT;
+
+	grm = grc->methods;
+	while (grm->id) {
+		if (grm->id == mthd)
+			return grm->exec(chan, class, mthd, data);
+		grm++;
+	}
+
+	return -ENOENT;
+}
+
+static bool
+nouveau_fifo_swmthd(struct nouveau_channel *chan, uint32_t addr, uint32_t data)
+{
+	struct drm_device *dev = chan->dev;
+	const int subc = (addr >> 13) & 0x7;
+	const int mthd = addr & 0x1ffc;
+
+	if (mthd == 0x0000) {
+		struct nouveau_gpuobj_ref *ref = NULL;
+
+		if (nouveau_gpuobj_ref_find(chan, data, &ref))
+			return false;
+
+		if (ref->gpuobj->engine != NVOBJ_ENGINE_SW)
+			return false;
+
+		chan->sw_subchannel[subc] = ref->gpuobj->class;
+		nv_wr32(dev, NV04_PFIFO_CACHE1_ENGINE, nv_rd32(dev,
+			NV04_PFIFO_CACHE1_ENGINE) & ~(0xf << subc * 4));
+		return true;
+	}
+
+	/* hw object */
+	if (nv_rd32(dev, NV04_PFIFO_CACHE1_ENGINE) & (1 << (subc*4)))
+		return false;
+
+	if (nouveau_call_method(chan, chan->sw_subchannel[subc], mthd, data))
+		return false;
+
+	return true;
+}
+
+static void
+nouveau_fifo_irq_handler(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_engine *engine = &dev_priv->engine;
+	uint32_t status, reassign;
+	int cnt = 0;
+
+	reassign = nv_rd32(dev, NV03_PFIFO_CACHES) & 1;
+	while ((status = nv_rd32(dev, NV03_PFIFO_INTR_0)) && (cnt++ < 100)) {
+		struct nouveau_channel *chan = NULL;
+		uint32_t chid, get;
+
+		nv_wr32(dev, NV03_PFIFO_CACHES, 0);
+
+		chid = engine->fifo.channel_id(dev);
+		if (chid >= 0 && chid < engine->fifo.channels)
+			chan = dev_priv->fifos[chid];
+		get  = nv_rd32(dev, NV03_PFIFO_CACHE1_GET);
+
+		if (status & NV_PFIFO_INTR_CACHE_ERROR) {
+			uint32_t mthd, data;
+			int ptr;
+
+			/* NV_PFIFO_CACHE1_GET actually goes to 0xffc before
+			 * wrapping on my G80 chips, but CACHE1 isn't big
+			 * enough for this much data.. Tests show that it
+			 * wraps around to the start at GET=0x800.. No clue
+			 * as to why..
+			 */
+			ptr = (get & 0x7ff) >> 2;
+
+			if (dev_priv->card_type < NV_40) {
+				mthd = nv_rd32(dev,
+					NV04_PFIFO_CACHE1_METHOD(ptr));
+				data = nv_rd32(dev,
+					NV04_PFIFO_CACHE1_DATA(ptr));
+			} else {
+				mthd = nv_rd32(dev,
+					NV40_PFIFO_CACHE1_METHOD(ptr));
+				data = nv_rd32(dev,
+					NV40_PFIFO_CACHE1_DATA(ptr));
+			}
+
+			if (!chan || !nouveau_fifo_swmthd(chan, mthd, data)) {
+				NV_INFO(dev, "PFIFO_CACHE_ERROR - Ch %d/%d "
+					     "Mthd 0x%04x Data 0x%08x\n",
+					chid, (mthd >> 13) & 7, mthd & 0x1ffc,
+					data);
+			}
+
+			nv_wr32(dev, NV04_PFIFO_CACHE1_DMA_PUSH, 0);
+			nv_wr32(dev, NV03_PFIFO_INTR_0,
+						NV_PFIFO_INTR_CACHE_ERROR);
+
+			nv_wr32(dev, NV03_PFIFO_CACHE1_PUSH0,
+				nv_rd32(dev, NV03_PFIFO_CACHE1_PUSH0) & ~1);
+			nv_wr32(dev, NV03_PFIFO_CACHE1_GET, get + 4);
+			nv_wr32(dev, NV03_PFIFO_CACHE1_PUSH0,
+				nv_rd32(dev, NV03_PFIFO_CACHE1_PUSH0) | 1);
+			nv_wr32(dev, NV04_PFIFO_CACHE1_HASH, 0);
+
+			nv_wr32(dev, NV04_PFIFO_CACHE1_DMA_PUSH,
+				nv_rd32(dev, NV04_PFIFO_CACHE1_DMA_PUSH) | 1);
+			nv_wr32(dev, NV04_PFIFO_CACHE1_PULL0, 1);
+
+			status &= ~NV_PFIFO_INTR_CACHE_ERROR;
+		}
+
+		if (status & NV_PFIFO_INTR_DMA_PUSHER) {
+			NV_INFO(dev, "PFIFO_DMA_PUSHER - Ch %d\n", chid);
+
+			status &= ~NV_PFIFO_INTR_DMA_PUSHER;
+			nv_wr32(dev, NV03_PFIFO_INTR_0,
+						NV_PFIFO_INTR_DMA_PUSHER);
+
+			nv_wr32(dev, NV04_PFIFO_CACHE1_DMA_STATE, 0x00000000);
+			if (nv_rd32(dev, NV04_PFIFO_CACHE1_DMA_PUT) != get)
+				nv_wr32(dev, NV04_PFIFO_CACHE1_DMA_GET,
+								get + 4);
+		}
+
+		if (status & NV_PFIFO_INTR_SEMAPHORE) {
+			uint32_t sem;
+
+			status &= ~NV_PFIFO_INTR_SEMAPHORE;
+			nv_wr32(dev, NV03_PFIFO_INTR_0,
+				NV_PFIFO_INTR_SEMAPHORE);
+
+			sem = nv_rd32(dev, NV10_PFIFO_CACHE1_SEMAPHORE);
+			nv_wr32(dev, NV10_PFIFO_CACHE1_SEMAPHORE, sem | 0x1);
+
+			nv_wr32(dev, NV03_PFIFO_CACHE1_GET, get + 4);
+			nv_wr32(dev, NV04_PFIFO_CACHE1_PULL0, 1);
+		}
+
+		if (status) {
+			NV_INFO(dev, "PFIFO_INTR 0x%08x - Ch %d\n",
+				status, chid);
+			nv_wr32(dev, NV03_PFIFO_INTR_0, status);
+			status = 0;
+		}
+
+		nv_wr32(dev, NV03_PFIFO_CACHES, reassign);
+	}
+
+	if (status) {
+		NV_INFO(dev, "PFIFO still angry after %d spins, halt\n", cnt);
+		nv_wr32(dev, 0x2140, 0);
+		nv_wr32(dev, 0x140, 0);
+	}
+
+	nv_wr32(dev, NV03_PMC_INTR_0, NV_PMC_INTR_0_PFIFO_PENDING);
+}
+
+struct nouveau_bitfield_names {
+	uint32_t mask;
+	const char *name;
+};
+
+static struct nouveau_bitfield_names nstatus_names[] =
+{
+	{ NV04_PGRAPH_NSTATUS_STATE_IN_USE,       "STATE_IN_USE" },
+	{ NV04_PGRAPH_NSTATUS_INVALID_STATE,      "INVALID_STATE" },
+	{ NV04_PGRAPH_NSTATUS_BAD_ARGUMENT,       "BAD_ARGUMENT" },
+	{ NV04_PGRAPH_NSTATUS_PROTECTION_FAULT,   "PROTECTION_FAULT" }
+};
+
+static struct nouveau_bitfield_names nstatus_names_nv10[] =
+{
+	{ NV10_PGRAPH_NSTATUS_STATE_IN_USE,       "STATE_IN_USE" },
+	{ NV10_PGRAPH_NSTATUS_INVALID_STATE,      "INVALID_STATE" },
+	{ NV10_PGRAPH_NSTATUS_BAD_ARGUMENT,       "BAD_ARGUMENT" },
+	{ NV10_PGRAPH_NSTATUS_PROTECTION_FAULT,   "PROTECTION_FAULT" }
+};
+
+static struct nouveau_bitfield_names nsource_names[] =
+{
+	{ NV03_PGRAPH_NSOURCE_NOTIFICATION,       "NOTIFICATION" },
+	{ NV03_PGRAPH_NSOURCE_DATA_ERROR,         "DATA_ERROR" },
+	{ NV03_PGRAPH_NSOURCE_PROTECTION_ERROR,   "PROTECTION_ERROR" },
+	{ NV03_PGRAPH_NSOURCE_RANGE_EXCEPTION,    "RANGE_EXCEPTION" },
+	{ NV03_PGRAPH_NSOURCE_LIMIT_COLOR,        "LIMIT_COLOR" },
+	{ NV03_PGRAPH_NSOURCE_LIMIT_ZETA,         "LIMIT_ZETA" },
+	{ NV03_PGRAPH_NSOURCE_ILLEGAL_MTHD,       "ILLEGAL_MTHD" },
+	{ NV03_PGRAPH_NSOURCE_DMA_R_PROTECTION,   "DMA_R_PROTECTION" },
+	{ NV03_PGRAPH_NSOURCE_DMA_W_PROTECTION,   "DMA_W_PROTECTION" },
+	{ NV03_PGRAPH_NSOURCE_FORMAT_EXCEPTION,   "FORMAT_EXCEPTION" },
+	{ NV03_PGRAPH_NSOURCE_PATCH_EXCEPTION,    "PATCH_EXCEPTION" },
+	{ NV03_PGRAPH_NSOURCE_STATE_INVALID,      "STATE_INVALID" },
+	{ NV03_PGRAPH_NSOURCE_DOUBLE_NOTIFY,      "DOUBLE_NOTIFY" },
+	{ NV03_PGRAPH_NSOURCE_NOTIFY_IN_USE,      "NOTIFY_IN_USE" },
+	{ NV03_PGRAPH_NSOURCE_METHOD_CNT,         "METHOD_CNT" },
+	{ NV03_PGRAPH_NSOURCE_BFR_NOTIFICATION,   "BFR_NOTIFICATION" },
+	{ NV03_PGRAPH_NSOURCE_DMA_VTX_PROTECTION, "DMA_VTX_PROTECTION" },
+	{ NV03_PGRAPH_NSOURCE_DMA_WIDTH_A,        "DMA_WIDTH_A" },
+	{ NV03_PGRAPH_NSOURCE_DMA_WIDTH_B,        "DMA_WIDTH_B" },
+};
+
+static void
+nouveau_print_bitfield_names_(uint32_t value,
+				const struct nouveau_bitfield_names *namelist,
+				const int namelist_len)
+{
+	/*
+	 * Caller must have already printed the KERN_* log level for us.
+	 * Also the caller is responsible for adding the newline.
+	 */
+	int i;
+	for (i = 0; i < namelist_len; ++i) {
+		uint32_t mask = namelist[i].mask;
+		if (value & mask) {
+			printk(" %s", namelist[i].name);
+			value &= ~mask;
+		}
+	}
+	if (value)
+		printk(" (unknown bits 0x%08x)", value);
+}
+#define nouveau_print_bitfield_names(val, namelist) \
+	nouveau_print_bitfield_names_((val), (namelist), ARRAY_SIZE(namelist))
+
+
+static int
+nouveau_graph_chid_from_grctx(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t inst;
+	int i;
+
+	if (dev_priv->card_type < NV_40)
+		return dev_priv->engine.fifo.channels;
+	else
+	if (dev_priv->card_type < NV_50) {
+		inst = (nv_rd32(dev, 0x40032c) & 0xfffff) << 4;
+
+		for (i = 0; i < dev_priv->engine.fifo.channels; i++) {
+			struct nouveau_channel *chan = dev_priv->fifos[i];
+
+			if (!chan || !chan->ramin_grctx)
+				continue;
+
+			if (inst == chan->ramin_grctx->instance)
+				break;
+		}
+	} else {
+		inst = (nv_rd32(dev, 0x40032c) & 0xfffff) << 12;
+
+		for (i = 0; i < dev_priv->engine.fifo.channels; i++) {
+			struct nouveau_channel *chan = dev_priv->fifos[i];
+
+			if (!chan || !chan->ramin)
+				continue;
+
+			if (inst == chan->ramin->instance)
+				break;
+		}
+	}
+
+
+	return i;
+}
+
+static int
+nouveau_graph_trapped_channel(struct drm_device *dev, int *channel_ret)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_engine *engine = &dev_priv->engine;
+	int channel;
+
+	if (dev_priv->card_type < NV_10)
+		channel = (nv_rd32(dev, NV04_PGRAPH_TRAPPED_ADDR) >> 24) & 0xf;
+	else
+	if (dev_priv->card_type < NV_40)
+		channel = (nv_rd32(dev, NV04_PGRAPH_TRAPPED_ADDR) >> 20) & 0x1f;
+	else
+		channel = nouveau_graph_chid_from_grctx(dev);
+
+	if (channel >= engine->fifo.channels || !dev_priv->fifos[channel]) {
+		NV_ERROR(dev, "AIII, invalid/inactive channel id %d\n", channel);
+		return -EINVAL;
+	}
+
+	*channel_ret = channel;
+	return 0;
+}
+
+struct nouveau_pgraph_trap {
+	int channel;
+	int class;
+	int subc, mthd, size;
+	uint32_t data, data2;
+	uint32_t nsource, nstatus;
+};
+
+static void
+nouveau_graph_trap_info(struct drm_device *dev,
+			struct nouveau_pgraph_trap *trap)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t address;
+
+	trap->nsource = trap->nstatus = 0;
+	if (dev_priv->card_type < NV_50) {
+		trap->nsource = nv_rd32(dev, NV03_PGRAPH_NSOURCE);
+		trap->nstatus = nv_rd32(dev, NV03_PGRAPH_NSTATUS);
+	}
+
+	if (nouveau_graph_trapped_channel(dev, &trap->channel))
+		trap->channel = -1;
+	address = nv_rd32(dev, NV04_PGRAPH_TRAPPED_ADDR);
+
+	trap->mthd = address & 0x1FFC;
+	trap->data = nv_rd32(dev, NV04_PGRAPH_TRAPPED_DATA);
+	if (dev_priv->card_type < NV_10) {
+		trap->subc  = (address >> 13) & 0x7;
+	} else {
+		trap->subc  = (address >> 16) & 0x7;
+		trap->data2 = nv_rd32(dev, NV10_PGRAPH_TRAPPED_DATA_HIGH);
+	}
+
+	if (dev_priv->card_type < NV_10)
+		trap->class = nv_rd32(dev, 0x400180 + trap->subc*4) & 0xFF;
+	else if (dev_priv->card_type < NV_40)
+		trap->class = nv_rd32(dev, 0x400160 + trap->subc*4) & 0xFFF;
+	else if (dev_priv->card_type < NV_50)
+		trap->class = nv_rd32(dev, 0x400160 + trap->subc*4) & 0xFFFF;
+	else
+		trap->class = nv_rd32(dev, 0x400814);
+}
+
+static void
+nouveau_graph_dump_trap_info(struct drm_device *dev, const char *id,
+			     struct nouveau_pgraph_trap *trap)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t nsource = trap->nsource, nstatus = trap->nstatus;
+
+	NV_INFO(dev, "%s - nSource:", id);
+	nouveau_print_bitfield_names(nsource, nsource_names);
+	printk(", nStatus:");
+	if (dev_priv->card_type < NV_10)
+		nouveau_print_bitfield_names(nstatus, nstatus_names);
+	else
+		nouveau_print_bitfield_names(nstatus, nstatus_names_nv10);
+	printk("\n");
+
+	NV_INFO(dev, "%s - Ch %d/%d Class 0x%04x Mthd 0x%04x "
+					"Data 0x%08x:0x%08x\n",
+					id, trap->channel, trap->subc,
+					trap->class, trap->mthd,
+					trap->data2, trap->data);
+}
+
+static int
+nouveau_pgraph_intr_swmthd(struct drm_device *dev,
+			   struct nouveau_pgraph_trap *trap)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	if (trap->channel < 0 ||
+	    trap->channel >= dev_priv->engine.fifo.channels ||
+	    !dev_priv->fifos[trap->channel])
+		return -ENODEV;
+
+	return nouveau_call_method(dev_priv->fifos[trap->channel],
+				   trap->class, trap->mthd, trap->data);
+}
+
+static inline void
+nouveau_pgraph_intr_notify(struct drm_device *dev, uint32_t nsource)
+{
+	struct nouveau_pgraph_trap trap;
+	int unhandled = 0;
+
+	nouveau_graph_trap_info(dev, &trap);
+
+	if (nsource & NV03_PGRAPH_NSOURCE_ILLEGAL_MTHD) {
+		if (nouveau_pgraph_intr_swmthd(dev, &trap))
+			unhandled = 1;
+	} else {
+		unhandled = 1;
+	}
+
+	if (unhandled)
+		nouveau_graph_dump_trap_info(dev, "PGRAPH_NOTIFY", &trap);
+}
+
+static DEFINE_RATELIMIT_STATE(nouveau_ratelimit_state, 3 * HZ, 20);
+
+static int nouveau_ratelimit(void)
+{
+	return __ratelimit(&nouveau_ratelimit_state);
+}
+
+
+static inline void
+nouveau_pgraph_intr_error(struct drm_device *dev, uint32_t nsource)
+{
+	struct nouveau_pgraph_trap trap;
+	int unhandled = 0;
+
+	nouveau_graph_trap_info(dev, &trap);
+	trap.nsource = nsource;
+
+	if (nsource & NV03_PGRAPH_NSOURCE_ILLEGAL_MTHD) {
+		if (nouveau_pgraph_intr_swmthd(dev, &trap))
+			unhandled = 1;
+	} else if (nsource & NV03_PGRAPH_NSOURCE_DMA_VTX_PROTECTION) {
+		uint32_t v = nv_rd32(dev, 0x402000);
+		nv_wr32(dev, 0x402000, v);
+
+		/* dump the error anyway for now: it's useful for
+		   Gallium development */
+		unhandled = 1;
+	} else {
+		unhandled = 1;
+	}
+
+	if (unhandled && nouveau_ratelimit())
+		nouveau_graph_dump_trap_info(dev, "PGRAPH_ERROR", &trap);
+}
+
+static inline void
+nouveau_pgraph_intr_context_switch(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_engine *engine = &dev_priv->engine;
+	uint32_t chid;
+
+	chid = engine->fifo.channel_id(dev);
+	NV_DEBUG(dev, "PGRAPH context switch interrupt channel %x\n", chid);
+
+	switch (dev_priv->card_type) {
+	case NV_04:
+		nv04_graph_context_switch(dev);
+		break;
+	case NV_10:
+		nv10_graph_context_switch(dev);
+		break;
+	default:
+		NV_ERROR(dev, "Context switch not implemented\n");
+		break;
+	}
+}
+
+static void
+nouveau_pgraph_irq_handler(struct drm_device *dev)
+{
+	uint32_t status;
+
+	while ((status = nv_rd32(dev, NV03_PGRAPH_INTR))) {
+		uint32_t nsource = nv_rd32(dev, NV03_PGRAPH_NSOURCE);
+
+		if (status & NV_PGRAPH_INTR_NOTIFY) {
+			nouveau_pgraph_intr_notify(dev, nsource);
+
+			status &= ~NV_PGRAPH_INTR_NOTIFY;
+			nv_wr32(dev, NV03_PGRAPH_INTR, NV_PGRAPH_INTR_NOTIFY);
+		}
+
+		if (status & NV_PGRAPH_INTR_ERROR) {
+			nouveau_pgraph_intr_error(dev, nsource);
+
+			status &= ~NV_PGRAPH_INTR_ERROR;
+			nv_wr32(dev, NV03_PGRAPH_INTR, NV_PGRAPH_INTR_ERROR);
+		}
+
+		if (status & NV_PGRAPH_INTR_CONTEXT_SWITCH) {
+			nouveau_pgraph_intr_context_switch(dev);
+
+			status &= ~NV_PGRAPH_INTR_CONTEXT_SWITCH;
+			nv_wr32(dev, NV03_PGRAPH_INTR,
+				 NV_PGRAPH_INTR_CONTEXT_SWITCH);
+		}
+
+		if (status) {
+			NV_INFO(dev, "Unhandled PGRAPH_INTR - 0x%08x\n", status);
+			nv_wr32(dev, NV03_PGRAPH_INTR, status);
+		}
+
+		if ((nv_rd32(dev, NV04_PGRAPH_FIFO) & (1 << 0)) == 0)
+			nv_wr32(dev, NV04_PGRAPH_FIFO, 1);
+	}
+
+	nv_wr32(dev, NV03_PMC_INTR_0, NV_PMC_INTR_0_PGRAPH_PENDING);
+}
+
+static void
+nv50_pgraph_irq_handler(struct drm_device *dev)
+{
+	uint32_t status;
+
+	while ((status = nv_rd32(dev, NV03_PGRAPH_INTR))) {
+		uint32_t nsource = nv_rd32(dev, NV03_PGRAPH_NSOURCE);
+
+		if (status & 0x00000001) {
+			nouveau_pgraph_intr_notify(dev, nsource);
+			status &= ~0x00000001;
+			nv_wr32(dev, NV03_PGRAPH_INTR, 0x00000001);
+		}
+
+		if (status & 0x00000010) {
+			nouveau_pgraph_intr_error(dev, nsource |
+					NV03_PGRAPH_NSOURCE_ILLEGAL_MTHD);
+
+			status &= ~0x00000010;
+			nv_wr32(dev, NV03_PGRAPH_INTR, 0x00000010);
+		}
+
+		if (status & 0x00001000) {
+			nv_wr32(dev, 0x400500, 0x00000000);
+			nv_wr32(dev, NV03_PGRAPH_INTR,
+				NV_PGRAPH_INTR_CONTEXT_SWITCH);
+			nv_wr32(dev, NV40_PGRAPH_INTR_EN, nv_rd32(dev,
+				NV40_PGRAPH_INTR_EN) &
+				~NV_PGRAPH_INTR_CONTEXT_SWITCH);
+			nv_wr32(dev, 0x400500, 0x00010001);
+
+			nv50_graph_context_switch(dev);
+
+			status &= ~NV_PGRAPH_INTR_CONTEXT_SWITCH;
+		}
+
+		if (status & 0x00100000) {
+			nouveau_pgraph_intr_error(dev, nsource |
+					NV03_PGRAPH_NSOURCE_DATA_ERROR);
+
+			status &= ~0x00100000;
+			nv_wr32(dev, NV03_PGRAPH_INTR, 0x00100000);
+		}
+
+		if (status & 0x00200000) {
+			int r;
+
+			nouveau_pgraph_intr_error(dev, nsource |
+					NV03_PGRAPH_NSOURCE_PROTECTION_ERROR);
+
+			NV_ERROR(dev, "magic set 1:\n");
+			for (r = 0x408900; r <= 0x408910; r += 4)
+				NV_ERROR(dev, "\t0x%08x: 0x%08x\n", r,
+					nv_rd32(dev, r));
+			nv_wr32(dev, 0x408900,
+				nv_rd32(dev, 0x408904) | 0xc0000000);
+			for (r = 0x408e08; r <= 0x408e24; r += 4)
+				NV_ERROR(dev, "\t0x%08x: 0x%08x\n", r,
+							nv_rd32(dev, r));
+			nv_wr32(dev, 0x408e08,
+				nv_rd32(dev, 0x408e08) | 0xc0000000);
+
+			NV_ERROR(dev, "magic set 2:\n");
+			for (r = 0x409900; r <= 0x409910; r += 4)
+				NV_ERROR(dev, "\t0x%08x: 0x%08x\n", r,
+					nv_rd32(dev, r));
+			nv_wr32(dev, 0x409900,
+				nv_rd32(dev, 0x409904) | 0xc0000000);
+			for (r = 0x409e08; r <= 0x409e24; r += 4)
+				NV_ERROR(dev, "\t0x%08x: 0x%08x\n", r,
+					nv_rd32(dev, r));
+			nv_wr32(dev, 0x409e08,
+				nv_rd32(dev, 0x409e08) | 0xc0000000);
+
+			status &= ~0x00200000;
+			nv_wr32(dev, NV03_PGRAPH_NSOURCE, nsource);
+			nv_wr32(dev, NV03_PGRAPH_INTR, 0x00200000);
+		}
+
+		if (status) {
+			NV_INFO(dev, "Unhandled PGRAPH_INTR - 0x%08x\n",
+				status);
+			nv_wr32(dev, NV03_PGRAPH_INTR, status);
+		}
+
+		{
+			const int isb = (1 << 16) | (1 << 0);
+
+			if ((nv_rd32(dev, 0x400500) & isb) != isb)
+				nv_wr32(dev, 0x400500,
+					nv_rd32(dev, 0x400500) | isb);
+		}
+	}
+
+	nv_wr32(dev, NV03_PMC_INTR_0, NV_PMC_INTR_0_PGRAPH_PENDING);
+	nv_wr32(dev, 0x400824, nv_rd32(dev, 0x400824) & ~(1 << 31));
+}
+
+static void
+nouveau_crtc_irq_handler(struct drm_device *dev, int crtc)
+{
+	if (crtc & 1)
+		nv_wr32(dev, NV_CRTC0_INTSTAT, NV_CRTC_INTR_VBLANK);
+
+	if (crtc & 2)
+		nv_wr32(dev, NV_CRTC1_INTSTAT, NV_CRTC_INTR_VBLANK);
+}
+
+irqreturn_t
+nouveau_irq_handler(DRM_IRQ_ARGS)
+{
+	struct drm_device *dev = (struct drm_device *)arg;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t status, fbdev_flags = 0;
+	unsigned long flags;
+
+	status = nv_rd32(dev, NV03_PMC_INTR_0);
+	if (!status)
+		return IRQ_NONE;
+
+	spin_lock_irqsave(&dev_priv->context_switch_lock, flags);
+
+	if (dev_priv->fbdev_info) {
+		fbdev_flags = dev_priv->fbdev_info->flags;
+		dev_priv->fbdev_info->flags |= FBINFO_HWACCEL_DISABLED;
+	}
+
+	if (status & NV_PMC_INTR_0_PFIFO_PENDING) {
+		nouveau_fifo_irq_handler(dev);
+		status &= ~NV_PMC_INTR_0_PFIFO_PENDING;
+	}
+
+	if (status & NV_PMC_INTR_0_PGRAPH_PENDING) {
+		if (dev_priv->card_type >= NV_50)
+			nv50_pgraph_irq_handler(dev);
+		else
+			nouveau_pgraph_irq_handler(dev);
+
+		status &= ~NV_PMC_INTR_0_PGRAPH_PENDING;
+	}
+
+	if (status & NV_PMC_INTR_0_CRTCn_PENDING) {
+		nouveau_crtc_irq_handler(dev, (status>>24)&3);
+		status &= ~NV_PMC_INTR_0_CRTCn_PENDING;
+	}
+
+	if (status & (NV_PMC_INTR_0_NV50_DISPLAY_PENDING |
+		      NV_PMC_INTR_0_NV50_I2C_PENDING)) {
+		nv50_display_irq_handler(dev);
+		status &= ~(NV_PMC_INTR_0_NV50_DISPLAY_PENDING |
+			    NV_PMC_INTR_0_NV50_I2C_PENDING);
+	}
+
+	if (status)
+		NV_ERROR(dev, "Unhandled PMC INTR status bits 0x%08x\n", status);
+
+	if (dev_priv->fbdev_info)
+		dev_priv->fbdev_info->flags = fbdev_flags;
+
+	spin_unlock_irqrestore(&dev_priv->context_switch_lock, flags);
+
+	return IRQ_HANDLED;
+}
diff --git a/drivers/gpu/drm/nouveau/nouveau_mem.c b/drivers/gpu/drm/nouveau/nouveau_mem.c
new file mode 100644
index 000000000000..2dc09dbd817d
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_mem.c
@@ -0,0 +1,699 @@
+/*
+ * Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
+ * Copyright 2005 Stephane Marchesin
+ *
+ * The Weather Channel (TM) funded Tungsten Graphics to develop the
+ * initial release of the Radeon 8500 driver under the XFree86 license.
+ * This notice must be preserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+
+#include "drmP.h"
+#include "drm.h"
+#include "drm_sarea.h"
+#include "nouveau_drv.h"
+
+static struct mem_block *
+split_block(struct mem_block *p, uint64_t start, uint64_t size,
+	    struct drm_file *file_priv)
+{
+	/* Maybe cut off the start of an existing block */
+	if (start > p->start) {
+		struct mem_block *newblock =
+			kmalloc(sizeof(*newblock), GFP_KERNEL);
+		if (!newblock)
+			goto out;
+		newblock->start = start;
+		newblock->size = p->size - (start - p->start);
+		newblock->file_priv = NULL;
+		newblock->next = p->next;
+		newblock->prev = p;
+		p->next->prev = newblock;
+		p->next = newblock;
+		p->size -= newblock->size;
+		p = newblock;
+	}
+
+	/* Maybe cut off the end of an existing block */
+	if (size < p->size) {
+		struct mem_block *newblock =
+			kmalloc(sizeof(*newblock), GFP_KERNEL);
+		if (!newblock)
+			goto out;
+		newblock->start = start + size;
+		newblock->size = p->size - size;
+		newblock->file_priv = NULL;
+		newblock->next = p->next;
+		newblock->prev = p;
+		p->next->prev = newblock;
+		p->next = newblock;
+		p->size = size;
+	}
+
+out:
+	/* Our block is in the middle */
+	p->file_priv = file_priv;
+	return p;
+}
+
+struct mem_block *
+nouveau_mem_alloc_block(struct mem_block *heap, uint64_t size,
+			int align2, struct drm_file *file_priv, int tail)
+{
+	struct mem_block *p;
+	uint64_t mask = (1 << align2) - 1;
+
+	if (!heap)
+		return NULL;
+
+	if (tail) {
+		list_for_each_prev(p, heap) {
+			uint64_t start = ((p->start + p->size) - size) & ~mask;
+
+			if (p->file_priv == NULL && start >= p->start &&
+			    start + size <= p->start + p->size)
+				return split_block(p, start, size, file_priv);
+		}
+	} else {
+		list_for_each(p, heap) {
+			uint64_t start = (p->start + mask) & ~mask;
+
+			if (p->file_priv == NULL &&
+			    start + size <= p->start + p->size)
+				return split_block(p, start, size, file_priv);
+		}
+	}
+
+	return NULL;
+}
+
+void nouveau_mem_free_block(struct mem_block *p)
+{
+	p->file_priv = NULL;
+
+	/* Assumes a single contiguous range.  Needs a special file_priv in
+	 * 'heap' to stop it being subsumed.
+	 */
+	if (p->next->file_priv == NULL) {
+		struct mem_block *q = p->next;
+		p->size += q->size;
+		p->next = q->next;
+		p->next->prev = p;
+		kfree(q);
+	}
+
+	if (p->prev->file_priv == NULL) {
+		struct mem_block *q = p->prev;
+		q->size += p->size;
+		q->next = p->next;
+		q->next->prev = q;
+		kfree(p);
+	}
+}
+
+/* Initialize.  How to check for an uninitialized heap?
+ */
+int nouveau_mem_init_heap(struct mem_block **heap, uint64_t start,
+			  uint64_t size)
+{
+	struct mem_block *blocks = kmalloc(sizeof(*blocks), GFP_KERNEL);
+
+	if (!blocks)
+		return -ENOMEM;
+
+	*heap = kmalloc(sizeof(**heap), GFP_KERNEL);
+	if (!*heap) {
+		kfree(blocks);
+		return -ENOMEM;
+	}
+
+	blocks->start = start;
+	blocks->size = size;
+	blocks->file_priv = NULL;
+	blocks->next = blocks->prev = *heap;
+
+	memset(*heap, 0, sizeof(**heap));
+	(*heap)->file_priv = (struct drm_file *) -1;
+	(*heap)->next = (*heap)->prev = blocks;
+	return 0;
+}
+
+/*
+ * Free all blocks associated with the releasing file_priv
+ */
+void nouveau_mem_release(struct drm_file *file_priv, struct mem_block *heap)
+{
+	struct mem_block *p;
+
+	if (!heap || !heap->next)
+		return;
+
+	list_for_each(p, heap) {
+		if (p->file_priv == file_priv)
+			p->file_priv = NULL;
+	}
+
+	/* Assumes a single contiguous range.  Needs a special file_priv in
+	 * 'heap' to stop it being subsumed.
+	 */
+	list_for_each(p, heap) {
+		while ((p->file_priv == NULL) &&
+					(p->next->file_priv == NULL) &&
+					(p->next != heap)) {
+			struct mem_block *q = p->next;
+			p->size += q->size;
+			p->next = q->next;
+			p->next->prev = p;
+			kfree(q);
+		}
+	}
+}
+
+/*
+ * NV10-NV40 tiling helpers
+ */
+
+static void
+nv10_mem_set_region_tiling(struct drm_device *dev, int i, uint32_t addr,
+			   uint32_t size, uint32_t pitch)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_fifo_engine *pfifo = &dev_priv->engine.fifo;
+	struct nouveau_fb_engine *pfb = &dev_priv->engine.fb;
+	struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
+	struct nouveau_tile_reg *tile = &dev_priv->tile.reg[i];
+
+	tile->addr = addr;
+	tile->size = size;
+	tile->used = !!pitch;
+	nouveau_fence_unref((void **)&tile->fence);
+
+	if (!pfifo->cache_flush(dev))
+		return;
+
+	pfifo->reassign(dev, false);
+	pfifo->cache_flush(dev);
+	pfifo->cache_pull(dev, false);
+
+	nouveau_wait_for_idle(dev);
+
+	pgraph->set_region_tiling(dev, i, addr, size, pitch);
+	pfb->set_region_tiling(dev, i, addr, size, pitch);
+
+	pfifo->cache_pull(dev, true);
+	pfifo->reassign(dev, true);
+}
+
+struct nouveau_tile_reg *
+nv10_mem_set_tiling(struct drm_device *dev, uint32_t addr, uint32_t size,
+		    uint32_t pitch)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_fb_engine *pfb = &dev_priv->engine.fb;
+	struct nouveau_tile_reg *tile = dev_priv->tile.reg, *found = NULL;
+	int i;
+
+	spin_lock(&dev_priv->tile.lock);
+
+	for (i = 0; i < pfb->num_tiles; i++) {
+		if (tile[i].used)
+			/* Tile region in use. */
+			continue;
+
+		if (tile[i].fence &&
+		    !nouveau_fence_signalled(tile[i].fence, NULL))
+			/* Pending tile region. */
+			continue;
+
+		if (max(tile[i].addr, addr) <
+		    min(tile[i].addr + tile[i].size, addr + size))
+			/* Kill an intersecting tile region. */
+			nv10_mem_set_region_tiling(dev, i, 0, 0, 0);
+
+		if (pitch && !found) {
+			/* Free tile region. */
+			nv10_mem_set_region_tiling(dev, i, addr, size, pitch);
+			found = &tile[i];
+		}
+	}
+
+	spin_unlock(&dev_priv->tile.lock);
+
+	return found;
+}
+
+void
+nv10_mem_expire_tiling(struct drm_device *dev, struct nouveau_tile_reg *tile,
+		       struct nouveau_fence *fence)
+{
+	if (fence) {
+		/* Mark it as pending. */
+		tile->fence = fence;
+		nouveau_fence_ref(fence);
+	}
+
+	tile->used = false;
+}
+
+/*
+ * NV50 VM helpers
+ */
+int
+nv50_mem_vm_bind_linear(struct drm_device *dev, uint64_t virt, uint32_t size,
+			uint32_t flags, uint64_t phys)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_gpuobj *pgt;
+	unsigned block;
+	int i;
+
+	virt = ((virt - dev_priv->vm_vram_base) >> 16) << 1;
+	size = (size >> 16) << 1;
+
+	phys |= ((uint64_t)flags << 32);
+	phys |= 1;
+	if (dev_priv->vram_sys_base) {
+		phys += dev_priv->vram_sys_base;
+		phys |= 0x30;
+	}
+
+	dev_priv->engine.instmem.prepare_access(dev, true);
+	while (size) {
+		unsigned offset_h = upper_32_bits(phys);
+		unsigned offset_l = lower_32_bits(phys);
+		unsigned pte, end;
+
+		for (i = 7; i >= 0; i--) {
+			block = 1 << (i + 1);
+			if (size >= block && !(virt & (block - 1)))
+				break;
+		}
+		offset_l |= (i << 7);
+
+		phys += block << 15;
+		size -= block;
+
+		while (block) {
+			pgt = dev_priv->vm_vram_pt[virt >> 14];
+			pte = virt & 0x3ffe;
+
+			end = pte + block;
+			if (end > 16384)
+				end = 16384;
+			block -= (end - pte);
+			virt  += (end - pte);
+
+			while (pte < end) {
+				nv_wo32(dev, pgt, pte++, offset_l);
+				nv_wo32(dev, pgt, pte++, offset_h);
+			}
+		}
+	}
+	dev_priv->engine.instmem.finish_access(dev);
+
+	nv_wr32(dev, 0x100c80, 0x00050001);
+	if (!nv_wait(0x100c80, 0x00000001, 0x00000000)) {
+		NV_ERROR(dev, "timeout: (0x100c80 & 1) == 0 (2)\n");
+		NV_ERROR(dev, "0x100c80 = 0x%08x\n", nv_rd32(dev, 0x100c80));
+		return -EBUSY;
+	}
+
+	nv_wr32(dev, 0x100c80, 0x00000001);
+	if (!nv_wait(0x100c80, 0x00000001, 0x00000000)) {
+		NV_ERROR(dev, "timeout: (0x100c80 & 1) == 0 (2)\n");
+		NV_ERROR(dev, "0x100c80 = 0x%08x\n", nv_rd32(dev, 0x100c80));
+		return -EBUSY;
+	}
+
+	return 0;
+}
+
+void
+nv50_mem_vm_unbind(struct drm_device *dev, uint64_t virt, uint32_t size)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_gpuobj *pgt;
+	unsigned pages, pte, end;
+
+	virt -= dev_priv->vm_vram_base;
+	pages = (size >> 16) << 1;
+
+	dev_priv->engine.instmem.prepare_access(dev, true);
+	while (pages) {
+		pgt = dev_priv->vm_vram_pt[virt >> 29];
+		pte = (virt & 0x1ffe0000ULL) >> 15;
+
+		end = pte + pages;
+		if (end > 16384)
+			end = 16384;
+		pages -= (end - pte);
+		virt  += (end - pte) << 15;
+
+		while (pte < end)
+			nv_wo32(dev, pgt, pte++, 0);
+	}
+	dev_priv->engine.instmem.finish_access(dev);
+
+	nv_wr32(dev, 0x100c80, 0x00050001);
+	if (!nv_wait(0x100c80, 0x00000001, 0x00000000)) {
+		NV_ERROR(dev, "timeout: (0x100c80 & 1) == 0 (2)\n");
+		NV_ERROR(dev, "0x100c80 = 0x%08x\n", nv_rd32(dev, 0x100c80));
+		return;
+	}
+
+	nv_wr32(dev, 0x100c80, 0x00000001);
+	if (!nv_wait(0x100c80, 0x00000001, 0x00000000)) {
+		NV_ERROR(dev, "timeout: (0x100c80 & 1) == 0 (2)\n");
+		NV_ERROR(dev, "0x100c80 = 0x%08x\n", nv_rd32(dev, 0x100c80));
+	}
+}
+
+/*
+ * Cleanup everything
+ */
+void nouveau_mem_takedown(struct mem_block **heap)
+{
+	struct mem_block *p;
+
+	if (!*heap)
+		return;
+
+	for (p = (*heap)->next; p != *heap;) {
+		struct mem_block *q = p;
+		p = p->next;
+		kfree(q);
+	}
+
+	kfree(*heap);
+	*heap = NULL;
+}
+
+void nouveau_mem_close(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	nouveau_bo_unpin(dev_priv->vga_ram);
+	nouveau_bo_ref(NULL, &dev_priv->vga_ram);
+
+	ttm_bo_device_release(&dev_priv->ttm.bdev);
+
+	nouveau_ttm_global_release(dev_priv);
+
+	if (drm_core_has_AGP(dev) && dev->agp &&
+	    drm_core_check_feature(dev, DRIVER_MODESET)) {
+		struct drm_agp_mem *entry, *tempe;
+
+		/* Remove AGP resources, but leave dev->agp
+		   intact until drv_cleanup is called. */
+		list_for_each_entry_safe(entry, tempe, &dev->agp->memory, head) {
+			if (entry->bound)
+				drm_unbind_agp(entry->memory);
+			drm_free_agp(entry->memory, entry->pages);
+			kfree(entry);
+		}
+		INIT_LIST_HEAD(&dev->agp->memory);
+
+		if (dev->agp->acquired)
+			drm_agp_release(dev);
+
+		dev->agp->acquired = 0;
+		dev->agp->enabled = 0;
+	}
+
+	if (dev_priv->fb_mtrr) {
+		drm_mtrr_del(dev_priv->fb_mtrr, drm_get_resource_start(dev, 1),
+			     drm_get_resource_len(dev, 1), DRM_MTRR_WC);
+		dev_priv->fb_mtrr = 0;
+	}
+}
+
+/*XXX won't work on BSD because of pci_read_config_dword */
+static uint32_t
+nouveau_mem_fb_amount_igp(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct pci_dev *bridge;
+	uint32_t mem;
+
+	bridge = pci_get_bus_and_slot(0, PCI_DEVFN(0, 1));
+	if (!bridge) {
+		NV_ERROR(dev, "no bridge device\n");
+		return 0;
+	}
+
+	if (dev_priv->flags&NV_NFORCE) {
+		pci_read_config_dword(bridge, 0x7C, &mem);
+		return (uint64_t)(((mem >> 6) & 31) + 1)*1024*1024;
+	} else
+	if (dev_priv->flags&NV_NFORCE2) {
+		pci_read_config_dword(bridge, 0x84, &mem);
+		return (uint64_t)(((mem >> 4) & 127) + 1)*1024*1024;
+	}
+
+	NV_ERROR(dev, "impossible!\n");
+	return 0;
+}
+
+/* returns the amount of FB ram in bytes */
+uint64_t nouveau_mem_fb_amount(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t boot0;
+
+	switch (dev_priv->card_type) {
+	case NV_04:
+		boot0 = nv_rd32(dev, NV03_BOOT_0);
+		if (boot0 & 0x00000100)
+			return (((boot0 >> 12) & 0xf) * 2 + 2) * 1024 * 1024;
+
+		switch (boot0 & NV03_BOOT_0_RAM_AMOUNT) {
+		case NV04_BOOT_0_RAM_AMOUNT_32MB:
+			return 32 * 1024 * 1024;
+		case NV04_BOOT_0_RAM_AMOUNT_16MB:
+			return 16 * 1024 * 1024;
+		case NV04_BOOT_0_RAM_AMOUNT_8MB:
+			return 8 * 1024 * 1024;
+		case NV04_BOOT_0_RAM_AMOUNT_4MB:
+			return 4 * 1024 * 1024;
+		}
+		break;
+	case NV_10:
+	case NV_20:
+	case NV_30:
+	case NV_40:
+	case NV_50:
+	default:
+		if (dev_priv->flags & (NV_NFORCE | NV_NFORCE2)) {
+			return nouveau_mem_fb_amount_igp(dev);
+		} else {
+			uint64_t mem;
+			mem = (nv_rd32(dev, NV04_FIFO_DATA) &
+					NV10_FIFO_DATA_RAM_AMOUNT_MB_MASK) >>
+					NV10_FIFO_DATA_RAM_AMOUNT_MB_SHIFT;
+			return mem * 1024 * 1024;
+		}
+		break;
+	}
+
+	NV_ERROR(dev,
+		"Unable to detect video ram size. Please report your setup to "
+							DRIVER_EMAIL "\n");
+	return 0;
+}
+
+#if __OS_HAS_AGP
+static void nouveau_mem_reset_agp(struct drm_device *dev)
+{
+	uint32_t saved_pci_nv_1, saved_pci_nv_19, pmc_enable;
+
+	saved_pci_nv_1 = nv_rd32(dev, NV04_PBUS_PCI_NV_1);
+	saved_pci_nv_19 = nv_rd32(dev, NV04_PBUS_PCI_NV_19);
+
+	/* clear busmaster bit */
+	nv_wr32(dev, NV04_PBUS_PCI_NV_1, saved_pci_nv_1 & ~0x4);
+	/* clear SBA and AGP bits */
+	nv_wr32(dev, NV04_PBUS_PCI_NV_19, saved_pci_nv_19 & 0xfffff0ff);
+
+	/* power cycle pgraph, if enabled */
+	pmc_enable = nv_rd32(dev, NV03_PMC_ENABLE);
+	if (pmc_enable & NV_PMC_ENABLE_PGRAPH) {
+		nv_wr32(dev, NV03_PMC_ENABLE,
+				pmc_enable & ~NV_PMC_ENABLE_PGRAPH);
+		nv_wr32(dev, NV03_PMC_ENABLE, nv_rd32(dev, NV03_PMC_ENABLE) |
+				NV_PMC_ENABLE_PGRAPH);
+	}
+
+	/* and restore (gives effect of resetting AGP) */
+	nv_wr32(dev, NV04_PBUS_PCI_NV_19, saved_pci_nv_19);
+	nv_wr32(dev, NV04_PBUS_PCI_NV_1, saved_pci_nv_1);
+}
+#endif
+
+int
+nouveau_mem_init_agp(struct drm_device *dev)
+{
+#if __OS_HAS_AGP
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct drm_agp_info info;
+	struct drm_agp_mode mode;
+	int ret;
+
+	if (nouveau_noagp)
+		return 0;
+
+	nouveau_mem_reset_agp(dev);
+
+	if (!dev->agp->acquired) {
+		ret = drm_agp_acquire(dev);
+		if (ret) {
+			NV_ERROR(dev, "Unable to acquire AGP: %d\n", ret);
+			return ret;
+		}
+	}
+
+	ret = drm_agp_info(dev, &info);
+	if (ret) {
+		NV_ERROR(dev, "Unable to get AGP info: %d\n", ret);
+		return ret;
+	}
+
+	/* see agp.h for the AGPSTAT_* modes available */
+	mode.mode = info.mode;
+	ret = drm_agp_enable(dev, mode);
+	if (ret) {
+		NV_ERROR(dev, "Unable to enable AGP: %d\n", ret);
+		return ret;
+	}
+
+	dev_priv->gart_info.type	= NOUVEAU_GART_AGP;
+	dev_priv->gart_info.aper_base	= info.aperture_base;
+	dev_priv->gart_info.aper_size	= info.aperture_size;
+#endif
+	return 0;
+}
+
+int
+nouveau_mem_init(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct ttm_bo_device *bdev = &dev_priv->ttm.bdev;
+	int ret, dma_bits = 32;
+
+	dev_priv->fb_phys = drm_get_resource_start(dev, 1);
+	dev_priv->gart_info.type = NOUVEAU_GART_NONE;
+
+	if (dev_priv->card_type >= NV_50 &&
+	    pci_dma_supported(dev->pdev, DMA_BIT_MASK(40)))
+		dma_bits = 40;
+
+	ret = pci_set_dma_mask(dev->pdev, DMA_BIT_MASK(dma_bits));
+	if (ret) {
+		NV_ERROR(dev, "Error setting DMA mask: %d\n", ret);
+		return ret;
+	}
+
+	ret = nouveau_ttm_global_init(dev_priv);
+	if (ret)
+		return ret;
+
+	ret = ttm_bo_device_init(&dev_priv->ttm.bdev,
+				 dev_priv->ttm.bo_global_ref.ref.object,
+				 &nouveau_bo_driver, DRM_FILE_PAGE_OFFSET,
+				 dma_bits <= 32 ? true : false);
+	if (ret) {
+		NV_ERROR(dev, "Error initialising bo driver: %d\n", ret);
+		return ret;
+	}
+
+	INIT_LIST_HEAD(&dev_priv->ttm.bo_list);
+	spin_lock_init(&dev_priv->ttm.bo_list_lock);
+	spin_lock_init(&dev_priv->tile.lock);
+
+	dev_priv->fb_available_size = nouveau_mem_fb_amount(dev);
+
+	dev_priv->fb_mappable_pages = dev_priv->fb_available_size;
+	if (dev_priv->fb_mappable_pages > drm_get_resource_len(dev, 1))
+		dev_priv->fb_mappable_pages = drm_get_resource_len(dev, 1);
+	dev_priv->fb_mappable_pages >>= PAGE_SHIFT;
+
+	NV_INFO(dev, "%d MiB VRAM\n", (int)(dev_priv->fb_available_size >> 20));
+
+	/* remove reserved space at end of vram from available amount */
+	dev_priv->fb_available_size -= dev_priv->ramin_rsvd_vram;
+	dev_priv->fb_aper_free = dev_priv->fb_available_size;
+
+	/* mappable vram */
+	ret = ttm_bo_init_mm(bdev, TTM_PL_VRAM,
+			     dev_priv->fb_available_size >> PAGE_SHIFT);
+	if (ret) {
+		NV_ERROR(dev, "Failed VRAM mm init: %d\n", ret);
+		return ret;
+	}
+
+	ret = nouveau_bo_new(dev, NULL, 256*1024, 0, TTM_PL_FLAG_VRAM,
+			     0, 0, true, true, &dev_priv->vga_ram);
+	if (ret == 0)
+		ret = nouveau_bo_pin(dev_priv->vga_ram, TTM_PL_FLAG_VRAM);
+	if (ret) {
+		NV_WARN(dev, "failed to reserve VGA memory\n");
+		nouveau_bo_ref(NULL, &dev_priv->vga_ram);
+	}
+
+	/* GART */
+#if !defined(__powerpc__) && !defined(__ia64__)
+	if (drm_device_is_agp(dev) && dev->agp) {
+		ret = nouveau_mem_init_agp(dev);
+		if (ret)
+			NV_ERROR(dev, "Error initialising AGP: %d\n", ret);
+	}
+#endif
+
+	if (dev_priv->gart_info.type == NOUVEAU_GART_NONE) {
+		ret = nouveau_sgdma_init(dev);
+		if (ret) {
+			NV_ERROR(dev, "Error initialising PCI(E): %d\n", ret);
+			return ret;
+		}
+	}
+
+	NV_INFO(dev, "%d MiB GART (aperture)\n",
+		(int)(dev_priv->gart_info.aper_size >> 20));
+	dev_priv->gart_info.aper_free = dev_priv->gart_info.aper_size;
+
+	ret = ttm_bo_init_mm(bdev, TTM_PL_TT,
+			     dev_priv->gart_info.aper_size >> PAGE_SHIFT);
+	if (ret) {
+		NV_ERROR(dev, "Failed TT mm init: %d\n", ret);
+		return ret;
+	}
+
+	dev_priv->fb_mtrr = drm_mtrr_add(drm_get_resource_start(dev, 1),
+					 drm_get_resource_len(dev, 1),
+					 DRM_MTRR_WC);
+
+	return 0;
+}
+
+
diff --git a/drivers/gpu/drm/nouveau/nouveau_notifier.c b/drivers/gpu/drm/nouveau/nouveau_notifier.c
new file mode 100644
index 000000000000..9537f3e30115
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_notifier.c
@@ -0,0 +1,200 @@
+/*
+ * Copyright (C) 2007 Ben Skeggs.
+ *
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "drmP.h"
+#include "drm.h"
+#include "nouveau_drv.h"
+
+int
+nouveau_notifier_init_channel(struct nouveau_channel *chan)
+{
+	struct drm_device *dev = chan->dev;
+	struct nouveau_bo *ntfy = NULL;
+	uint32_t flags;
+	int ret;
+
+	if (nouveau_vram_notify)
+		flags = TTM_PL_FLAG_VRAM;
+	else
+		flags = TTM_PL_FLAG_TT;
+
+	ret = nouveau_gem_new(dev, NULL, PAGE_SIZE, 0, flags,
+			      0, 0x0000, false, true, &ntfy);
+	if (ret)
+		return ret;
+
+	ret = nouveau_bo_pin(ntfy, flags);
+	if (ret)
+		goto out_err;
+
+	ret = nouveau_bo_map(ntfy);
+	if (ret)
+		goto out_err;
+
+	ret = nouveau_mem_init_heap(&chan->notifier_heap, 0, ntfy->bo.mem.size);
+	if (ret)
+		goto out_err;
+
+	chan->notifier_bo = ntfy;
+out_err:
+	if (ret)
+		drm_gem_object_unreference_unlocked(ntfy->gem);
+
+	return ret;
+}
+
+void
+nouveau_notifier_takedown_channel(struct nouveau_channel *chan)
+{
+	struct drm_device *dev = chan->dev;
+
+	if (!chan->notifier_bo)
+		return;
+
+	nouveau_bo_unmap(chan->notifier_bo);
+	mutex_lock(&dev->struct_mutex);
+	nouveau_bo_unpin(chan->notifier_bo);
+	mutex_unlock(&dev->struct_mutex);
+	drm_gem_object_unreference_unlocked(chan->notifier_bo->gem);
+	nouveau_mem_takedown(&chan->notifier_heap);
+}
+
+static void
+nouveau_notifier_gpuobj_dtor(struct drm_device *dev,
+			     struct nouveau_gpuobj *gpuobj)
+{
+	NV_DEBUG(dev, "\n");
+
+	if (gpuobj->priv)
+		nouveau_mem_free_block(gpuobj->priv);
+}
+
+int
+nouveau_notifier_alloc(struct nouveau_channel *chan, uint32_t handle,
+		       int size, uint32_t *b_offset)
+{
+	struct drm_device *dev = chan->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_gpuobj *nobj = NULL;
+	struct mem_block *mem;
+	uint32_t offset;
+	int target, ret;
+
+	if (!chan->notifier_heap) {
+		NV_ERROR(dev, "Channel %d doesn't have a notifier heap!\n",
+			 chan->id);
+		return -EINVAL;
+	}
+
+	mem = nouveau_mem_alloc_block(chan->notifier_heap, size, 0,
+				      (struct drm_file *)-2, 0);
+	if (!mem) {
+		NV_ERROR(dev, "Channel %d notifier block full\n", chan->id);
+		return -ENOMEM;
+	}
+
+	offset = chan->notifier_bo->bo.mem.mm_node->start << PAGE_SHIFT;
+	if (chan->notifier_bo->bo.mem.mem_type == TTM_PL_VRAM) {
+		target = NV_DMA_TARGET_VIDMEM;
+	} else
+	if (chan->notifier_bo->bo.mem.mem_type == TTM_PL_TT) {
+		if (dev_priv->gart_info.type == NOUVEAU_GART_SGDMA &&
+		    dev_priv->card_type < NV_50) {
+			ret = nouveau_sgdma_get_page(dev, offset, &offset);
+			if (ret)
+				return ret;
+			target = NV_DMA_TARGET_PCI;
+		} else {
+			target = NV_DMA_TARGET_AGP;
+			if (dev_priv->card_type >= NV_50)
+				offset += dev_priv->vm_gart_base;
+		}
+	} else {
+		NV_ERROR(dev, "Bad DMA target, mem_type %d!\n",
+			 chan->notifier_bo->bo.mem.mem_type);
+		return -EINVAL;
+	}
+	offset += mem->start;
+
+	ret = nouveau_gpuobj_dma_new(chan, NV_CLASS_DMA_IN_MEMORY, offset,
+				     mem->size, NV_DMA_ACCESS_RW, target,
+				     &nobj);
+	if (ret) {
+		nouveau_mem_free_block(mem);
+		NV_ERROR(dev, "Error creating notifier ctxdma: %d\n", ret);
+		return ret;
+	}
+	nobj->dtor   = nouveau_notifier_gpuobj_dtor;
+	nobj->priv   = mem;
+
+	ret = nouveau_gpuobj_ref_add(dev, chan, handle, nobj, NULL);
+	if (ret) {
+		nouveau_gpuobj_del(dev, &nobj);
+		nouveau_mem_free_block(mem);
+		NV_ERROR(dev, "Error referencing notifier ctxdma: %d\n", ret);
+		return ret;
+	}
+
+	*b_offset = mem->start;
+	return 0;
+}
+
+int
+nouveau_notifier_offset(struct nouveau_gpuobj *nobj, uint32_t *poffset)
+{
+	if (!nobj || nobj->dtor != nouveau_notifier_gpuobj_dtor)
+		return -EINVAL;
+
+	if (poffset) {
+		struct mem_block *mem = nobj->priv;
+
+		if (*poffset >= mem->size)
+			return false;
+
+		*poffset += mem->start;
+	}
+
+	return 0;
+}
+
+int
+nouveau_ioctl_notifier_alloc(struct drm_device *dev, void *data,
+			     struct drm_file *file_priv)
+{
+	struct drm_nouveau_notifierobj_alloc *na = data;
+	struct nouveau_channel *chan;
+	int ret;
+
+	NOUVEAU_CHECK_INITIALISED_WITH_RETURN;
+	NOUVEAU_GET_USER_CHANNEL_WITH_RETURN(na->channel, file_priv, chan);
+
+	ret = nouveau_notifier_alloc(chan, na->handle, na->size, &na->offset);
+	if (ret)
+		return ret;
+
+	return 0;
+}
diff --git a/drivers/gpu/drm/nouveau/nouveau_object.c b/drivers/gpu/drm/nouveau/nouveau_object.c
new file mode 100644
index 000000000000..e7c100ba63a1
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_object.c
@@ -0,0 +1,1295 @@
+/*
+ * Copyright (C) 2006 Ben Skeggs.
+ *
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+/*
+ * Authors:
+ *   Ben Skeggs <darktama@iinet.net.au>
+ */
+
+#include "drmP.h"
+#include "drm.h"
+#include "nouveau_drv.h"
+#include "nouveau_drm.h"
+
+/* NVidia uses context objects to drive drawing operations.
+
+   Context objects can be selected into 8 subchannels in the FIFO,
+   and then used via DMA command buffers.
+
+   A context object is referenced by a user defined handle (CARD32). The HW
+   looks up graphics objects in a hash table in the instance RAM.
+
+   An entry in the hash table consists of 2 CARD32. The first CARD32 contains
+   the handle, the second one a bitfield, that contains the address of the
+   object in instance RAM.
+
+   The format of the second CARD32 seems to be:
+
+   NV4 to NV30:
+
+   15: 0  instance_addr >> 4
+   17:16  engine (here uses 1 = graphics)
+   28:24  channel id (here uses 0)
+   31	  valid (use 1)
+
+   NV40:
+
+   15: 0  instance_addr >> 4   (maybe 19-0)
+   21:20  engine (here uses 1 = graphics)
+   I'm unsure about the other bits, but using 0 seems to work.
+
+   The key into the hash table depends on the object handle and channel id and
+   is given as:
+*/
+static uint32_t
+nouveau_ramht_hash_handle(struct drm_device *dev, int channel, uint32_t handle)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t hash = 0;
+	int i;
+
+	NV_DEBUG(dev, "ch%d handle=0x%08x\n", channel, handle);
+
+	for (i = 32; i > 0; i -= dev_priv->ramht_bits) {
+		hash ^= (handle & ((1 << dev_priv->ramht_bits) - 1));
+		handle >>= dev_priv->ramht_bits;
+	}
+
+	if (dev_priv->card_type < NV_50)
+		hash ^= channel << (dev_priv->ramht_bits - 4);
+	hash <<= 3;
+
+	NV_DEBUG(dev, "hash=0x%08x\n", hash);
+	return hash;
+}
+
+static int
+nouveau_ramht_entry_valid(struct drm_device *dev, struct nouveau_gpuobj *ramht,
+			  uint32_t offset)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t ctx = nv_ro32(dev, ramht, (offset + 4)/4);
+
+	if (dev_priv->card_type < NV_40)
+		return ((ctx & NV_RAMHT_CONTEXT_VALID) != 0);
+	return (ctx != 0);
+}
+
+static int
+nouveau_ramht_insert(struct drm_device *dev, struct nouveau_gpuobj_ref *ref)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_instmem_engine *instmem = &dev_priv->engine.instmem;
+	struct nouveau_channel *chan = ref->channel;
+	struct nouveau_gpuobj *ramht = chan->ramht ? chan->ramht->gpuobj : NULL;
+	uint32_t ctx, co, ho;
+
+	if (!ramht) {
+		NV_ERROR(dev, "No hash table!\n");
+		return -EINVAL;
+	}
+
+	if (dev_priv->card_type < NV_40) {
+		ctx = NV_RAMHT_CONTEXT_VALID | (ref->instance >> 4) |
+		      (chan->id << NV_RAMHT_CONTEXT_CHANNEL_SHIFT) |
+		      (ref->gpuobj->engine << NV_RAMHT_CONTEXT_ENGINE_SHIFT);
+	} else
+	if (dev_priv->card_type < NV_50) {
+		ctx = (ref->instance >> 4) |
+		      (chan->id << NV40_RAMHT_CONTEXT_CHANNEL_SHIFT) |
+		      (ref->gpuobj->engine << NV40_RAMHT_CONTEXT_ENGINE_SHIFT);
+	} else {
+		if (ref->gpuobj->engine == NVOBJ_ENGINE_DISPLAY) {
+			ctx = (ref->instance << 10) | 2;
+		} else {
+			ctx = (ref->instance >> 4) |
+			      ((ref->gpuobj->engine <<
+				NV40_RAMHT_CONTEXT_ENGINE_SHIFT));
+		}
+	}
+
+	instmem->prepare_access(dev, true);
+	co = ho = nouveau_ramht_hash_handle(dev, chan->id, ref->handle);
+	do {
+		if (!nouveau_ramht_entry_valid(dev, ramht, co)) {
+			NV_DEBUG(dev,
+				 "insert ch%d 0x%08x: h=0x%08x, c=0x%08x\n",
+				 chan->id, co, ref->handle, ctx);
+			nv_wo32(dev, ramht, (co + 0)/4, ref->handle);
+			nv_wo32(dev, ramht, (co + 4)/4, ctx);
+
+			list_add_tail(&ref->list, &chan->ramht_refs);
+			instmem->finish_access(dev);
+			return 0;
+		}
+		NV_DEBUG(dev, "collision ch%d 0x%08x: h=0x%08x\n",
+			 chan->id, co, nv_ro32(dev, ramht, co/4));
+
+		co += 8;
+		if (co >= dev_priv->ramht_size)
+			co = 0;
+	} while (co != ho);
+	instmem->finish_access(dev);
+
+	NV_ERROR(dev, "RAMHT space exhausted. ch=%d\n", chan->id);
+	return -ENOMEM;
+}
+
+static void
+nouveau_ramht_remove(struct drm_device *dev, struct nouveau_gpuobj_ref *ref)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_instmem_engine *instmem = &dev_priv->engine.instmem;
+	struct nouveau_channel *chan = ref->channel;
+	struct nouveau_gpuobj *ramht = chan->ramht ? chan->ramht->gpuobj : NULL;
+	uint32_t co, ho;
+
+	if (!ramht) {
+		NV_ERROR(dev, "No hash table!\n");
+		return;
+	}
+
+	instmem->prepare_access(dev, true);
+	co = ho = nouveau_ramht_hash_handle(dev, chan->id, ref->handle);
+	do {
+		if (nouveau_ramht_entry_valid(dev, ramht, co) &&
+		    (ref->handle == nv_ro32(dev, ramht, (co/4)))) {
+			NV_DEBUG(dev,
+				 "remove ch%d 0x%08x: h=0x%08x, c=0x%08x\n",
+				 chan->id, co, ref->handle,
+				 nv_ro32(dev, ramht, (co + 4)));
+			nv_wo32(dev, ramht, (co + 0)/4, 0x00000000);
+			nv_wo32(dev, ramht, (co + 4)/4, 0x00000000);
+
+			list_del(&ref->list);
+			instmem->finish_access(dev);
+			return;
+		}
+
+		co += 8;
+		if (co >= dev_priv->ramht_size)
+			co = 0;
+	} while (co != ho);
+	list_del(&ref->list);
+	instmem->finish_access(dev);
+
+	NV_ERROR(dev, "RAMHT entry not found. ch=%d, handle=0x%08x\n",
+		 chan->id, ref->handle);
+}
+
+int
+nouveau_gpuobj_new(struct drm_device *dev, struct nouveau_channel *chan,
+		   uint32_t size, int align, uint32_t flags,
+		   struct nouveau_gpuobj **gpuobj_ret)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_engine *engine = &dev_priv->engine;
+	struct nouveau_gpuobj *gpuobj;
+	struct mem_block *pramin = NULL;
+	int ret;
+
+	NV_DEBUG(dev, "ch%d size=%u align=%d flags=0x%08x\n",
+		 chan ? chan->id : -1, size, align, flags);
+
+	if (!dev_priv || !gpuobj_ret || *gpuobj_ret != NULL)
+		return -EINVAL;
+
+	gpuobj = kzalloc(sizeof(*gpuobj), GFP_KERNEL);
+	if (!gpuobj)
+		return -ENOMEM;
+	NV_DEBUG(dev, "gpuobj %p\n", gpuobj);
+	gpuobj->flags = flags;
+	gpuobj->im_channel = chan;
+
+	list_add_tail(&gpuobj->list, &dev_priv->gpuobj_list);
+
+	/* Choose between global instmem heap, and per-channel private
+	 * instmem heap.  On <NV50 allow requests for private instmem
+	 * to be satisfied from global heap if no per-channel area
+	 * available.
+	 */
+	if (chan) {
+		if (chan->ramin_heap) {
+			NV_DEBUG(dev, "private heap\n");
+			pramin = chan->ramin_heap;
+		} else
+		if (dev_priv->card_type < NV_50) {
+			NV_DEBUG(dev, "global heap fallback\n");
+			pramin = dev_priv->ramin_heap;
+		}
+	} else {
+		NV_DEBUG(dev, "global heap\n");
+		pramin = dev_priv->ramin_heap;
+	}
+
+	if (!pramin) {
+		NV_ERROR(dev, "No PRAMIN heap!\n");
+		return -EINVAL;
+	}
+
+	if (!chan) {
+		ret = engine->instmem.populate(dev, gpuobj, &size);
+		if (ret) {
+			nouveau_gpuobj_del(dev, &gpuobj);
+			return ret;
+		}
+	}
+
+	/* Allocate a chunk of the PRAMIN aperture */
+	gpuobj->im_pramin = nouveau_mem_alloc_block(pramin, size,
+						    drm_order(align),
+						    (struct drm_file *)-2, 0);
+	if (!gpuobj->im_pramin) {
+		nouveau_gpuobj_del(dev, &gpuobj);
+		return -ENOMEM;
+	}
+
+	if (!chan) {
+		ret = engine->instmem.bind(dev, gpuobj);
+		if (ret) {
+			nouveau_gpuobj_del(dev, &gpuobj);
+			return ret;
+		}
+	}
+
+	if (gpuobj->flags & NVOBJ_FLAG_ZERO_ALLOC) {
+		int i;
+
+		engine->instmem.prepare_access(dev, true);
+		for (i = 0; i < gpuobj->im_pramin->size; i += 4)
+			nv_wo32(dev, gpuobj, i/4, 0);
+		engine->instmem.finish_access(dev);
+	}
+
+	*gpuobj_ret = gpuobj;
+	return 0;
+}
+
+int
+nouveau_gpuobj_early_init(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	NV_DEBUG(dev, "\n");
+
+	INIT_LIST_HEAD(&dev_priv->gpuobj_list);
+
+	return 0;
+}
+
+int
+nouveau_gpuobj_init(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	int ret;
+
+	NV_DEBUG(dev, "\n");
+
+	if (dev_priv->card_type < NV_50) {
+		ret = nouveau_gpuobj_new_fake(dev,
+			dev_priv->ramht_offset, ~0, dev_priv->ramht_size,
+			NVOBJ_FLAG_ZERO_ALLOC | NVOBJ_FLAG_ALLOW_NO_REFS,
+						&dev_priv->ramht, NULL);
+		if (ret)
+			return ret;
+	}
+
+	return 0;
+}
+
+void
+nouveau_gpuobj_takedown(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	NV_DEBUG(dev, "\n");
+
+	nouveau_gpuobj_del(dev, &dev_priv->ramht);
+}
+
+void
+nouveau_gpuobj_late_takedown(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_gpuobj *gpuobj = NULL;
+	struct list_head *entry, *tmp;
+
+	NV_DEBUG(dev, "\n");
+
+	list_for_each_safe(entry, tmp, &dev_priv->gpuobj_list) {
+		gpuobj = list_entry(entry, struct nouveau_gpuobj, list);
+
+		NV_ERROR(dev, "gpuobj %p still exists at takedown, refs=%d\n",
+			 gpuobj, gpuobj->refcount);
+		gpuobj->refcount = 0;
+		nouveau_gpuobj_del(dev, &gpuobj);
+	}
+}
+
+int
+nouveau_gpuobj_del(struct drm_device *dev, struct nouveau_gpuobj **pgpuobj)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_engine *engine = &dev_priv->engine;
+	struct nouveau_gpuobj *gpuobj;
+	int i;
+
+	NV_DEBUG(dev, "gpuobj %p\n", pgpuobj ? *pgpuobj : NULL);
+
+	if (!dev_priv || !pgpuobj || !(*pgpuobj))
+		return -EINVAL;
+	gpuobj = *pgpuobj;
+
+	if (gpuobj->refcount != 0) {
+		NV_ERROR(dev, "gpuobj refcount is %d\n", gpuobj->refcount);
+		return -EINVAL;
+	}
+
+	if (gpuobj->im_pramin && (gpuobj->flags & NVOBJ_FLAG_ZERO_FREE)) {
+		engine->instmem.prepare_access(dev, true);
+		for (i = 0; i < gpuobj->im_pramin->size; i += 4)
+			nv_wo32(dev, gpuobj, i/4, 0);
+		engine->instmem.finish_access(dev);
+	}
+
+	if (gpuobj->dtor)
+		gpuobj->dtor(dev, gpuobj);
+
+	if (gpuobj->im_backing && !(gpuobj->flags & NVOBJ_FLAG_FAKE))
+		engine->instmem.clear(dev, gpuobj);
+
+	if (gpuobj->im_pramin) {
+		if (gpuobj->flags & NVOBJ_FLAG_FAKE)
+			kfree(gpuobj->im_pramin);
+		else
+			nouveau_mem_free_block(gpuobj->im_pramin);
+	}
+
+	list_del(&gpuobj->list);
+
+	*pgpuobj = NULL;
+	kfree(gpuobj);
+	return 0;
+}
+
+static int
+nouveau_gpuobj_instance_get(struct drm_device *dev,
+			    struct nouveau_channel *chan,
+			    struct nouveau_gpuobj *gpuobj, uint32_t *inst)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_gpuobj *cpramin;
+
+	/* <NV50 use PRAMIN address everywhere */
+	if (dev_priv->card_type < NV_50) {
+		*inst = gpuobj->im_pramin->start;
+		return 0;
+	}
+
+	if (chan && gpuobj->im_channel != chan) {
+		NV_ERROR(dev, "Channel mismatch: obj %d, ref %d\n",
+			 gpuobj->im_channel->id, chan->id);
+		return -EINVAL;
+	}
+
+	/* NV50 channel-local instance */
+	if (chan) {
+		cpramin = chan->ramin->gpuobj;
+		*inst = gpuobj->im_pramin->start - cpramin->im_pramin->start;
+		return 0;
+	}
+
+	/* NV50 global (VRAM) instance */
+	if (!gpuobj->im_channel) {
+		/* ...from global heap */
+		if (!gpuobj->im_backing) {
+			NV_ERROR(dev, "AII, no VRAM backing gpuobj\n");
+			return -EINVAL;
+		}
+		*inst = gpuobj->im_backing_start;
+		return 0;
+	} else {
+		/* ...from local heap */
+		cpramin = gpuobj->im_channel->ramin->gpuobj;
+		*inst = cpramin->im_backing_start +
+			(gpuobj->im_pramin->start - cpramin->im_pramin->start);
+		return 0;
+	}
+
+	return -EINVAL;
+}
+
+int
+nouveau_gpuobj_ref_add(struct drm_device *dev, struct nouveau_channel *chan,
+		       uint32_t handle, struct nouveau_gpuobj *gpuobj,
+		       struct nouveau_gpuobj_ref **ref_ret)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_gpuobj_ref *ref;
+	uint32_t instance;
+	int ret;
+
+	NV_DEBUG(dev, "ch%d h=0x%08x gpuobj=%p\n",
+		 chan ? chan->id : -1, handle, gpuobj);
+
+	if (!dev_priv || !gpuobj || (ref_ret && *ref_ret != NULL))
+		return -EINVAL;
+
+	if (!chan && !ref_ret)
+		return -EINVAL;
+
+	if (gpuobj->engine == NVOBJ_ENGINE_SW && !gpuobj->im_pramin) {
+		/* sw object */
+		instance = 0x40;
+	} else {
+		ret = nouveau_gpuobj_instance_get(dev, chan, gpuobj, &instance);
+		if (ret)
+			return ret;
+	}
+
+	ref = kzalloc(sizeof(*ref), GFP_KERNEL);
+	if (!ref)
+		return -ENOMEM;
+	INIT_LIST_HEAD(&ref->list);
+	ref->gpuobj   = gpuobj;
+	ref->channel  = chan;
+	ref->instance = instance;
+
+	if (!ref_ret) {
+		ref->handle = handle;
+
+		ret = nouveau_ramht_insert(dev, ref);
+		if (ret) {
+			kfree(ref);
+			return ret;
+		}
+	} else {
+		ref->handle = ~0;
+		*ref_ret = ref;
+	}
+
+	ref->gpuobj->refcount++;
+	return 0;
+}
+
+int nouveau_gpuobj_ref_del(struct drm_device *dev, struct nouveau_gpuobj_ref **pref)
+{
+	struct nouveau_gpuobj_ref *ref;
+
+	NV_DEBUG(dev, "ref %p\n", pref ? *pref : NULL);
+
+	if (!dev || !pref || *pref == NULL)
+		return -EINVAL;
+	ref = *pref;
+
+	if (ref->handle != ~0)
+		nouveau_ramht_remove(dev, ref);
+
+	if (ref->gpuobj) {
+		ref->gpuobj->refcount--;
+
+		if (ref->gpuobj->refcount == 0) {
+			if (!(ref->gpuobj->flags & NVOBJ_FLAG_ALLOW_NO_REFS))
+				nouveau_gpuobj_del(dev, &ref->gpuobj);
+		}
+	}
+
+	*pref = NULL;
+	kfree(ref);
+	return 0;
+}
+
+int
+nouveau_gpuobj_new_ref(struct drm_device *dev,
+		       struct nouveau_channel *oc, struct nouveau_channel *rc,
+		       uint32_t handle, uint32_t size, int align,
+		       uint32_t flags, struct nouveau_gpuobj_ref **ref)
+{
+	struct nouveau_gpuobj *gpuobj = NULL;
+	int ret;
+
+	ret = nouveau_gpuobj_new(dev, oc, size, align, flags, &gpuobj);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_ref_add(dev, rc, handle, gpuobj, ref);
+	if (ret) {
+		nouveau_gpuobj_del(dev, &gpuobj);
+		return ret;
+	}
+
+	return 0;
+}
+
+int
+nouveau_gpuobj_ref_find(struct nouveau_channel *chan, uint32_t handle,
+			struct nouveau_gpuobj_ref **ref_ret)
+{
+	struct nouveau_gpuobj_ref *ref;
+	struct list_head *entry, *tmp;
+
+	list_for_each_safe(entry, tmp, &chan->ramht_refs) {
+		ref = list_entry(entry, struct nouveau_gpuobj_ref, list);
+
+		if (ref->handle == handle) {
+			if (ref_ret)
+				*ref_ret = ref;
+			return 0;
+		}
+	}
+
+	return -EINVAL;
+}
+
+int
+nouveau_gpuobj_new_fake(struct drm_device *dev, uint32_t p_offset,
+			uint32_t b_offset, uint32_t size,
+			uint32_t flags, struct nouveau_gpuobj **pgpuobj,
+			struct nouveau_gpuobj_ref **pref)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_gpuobj *gpuobj = NULL;
+	int i;
+
+	NV_DEBUG(dev,
+		 "p_offset=0x%08x b_offset=0x%08x size=0x%08x flags=0x%08x\n",
+		 p_offset, b_offset, size, flags);
+
+	gpuobj = kzalloc(sizeof(*gpuobj), GFP_KERNEL);
+	if (!gpuobj)
+		return -ENOMEM;
+	NV_DEBUG(dev, "gpuobj %p\n", gpuobj);
+	gpuobj->im_channel = NULL;
+	gpuobj->flags      = flags | NVOBJ_FLAG_FAKE;
+
+	list_add_tail(&gpuobj->list, &dev_priv->gpuobj_list);
+
+	if (p_offset != ~0) {
+		gpuobj->im_pramin = kzalloc(sizeof(struct mem_block),
+					    GFP_KERNEL);
+		if (!gpuobj->im_pramin) {
+			nouveau_gpuobj_del(dev, &gpuobj);
+			return -ENOMEM;
+		}
+		gpuobj->im_pramin->start = p_offset;
+		gpuobj->im_pramin->size  = size;
+	}
+
+	if (b_offset != ~0) {
+		gpuobj->im_backing = (struct nouveau_bo *)-1;
+		gpuobj->im_backing_start = b_offset;
+	}
+
+	if (gpuobj->flags & NVOBJ_FLAG_ZERO_ALLOC) {
+		dev_priv->engine.instmem.prepare_access(dev, true);
+		for (i = 0; i < gpuobj->im_pramin->size; i += 4)
+			nv_wo32(dev, gpuobj, i/4, 0);
+		dev_priv->engine.instmem.finish_access(dev);
+	}
+
+	if (pref) {
+		i = nouveau_gpuobj_ref_add(dev, NULL, 0, gpuobj, pref);
+		if (i) {
+			nouveau_gpuobj_del(dev, &gpuobj);
+			return i;
+		}
+	}
+
+	if (pgpuobj)
+		*pgpuobj = gpuobj;
+	return 0;
+}
+
+
+static uint32_t
+nouveau_gpuobj_class_instmem_size(struct drm_device *dev, int class)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	/*XXX: dodgy hack for now */
+	if (dev_priv->card_type >= NV_50)
+		return 24;
+	if (dev_priv->card_type >= NV_40)
+		return 32;
+	return 16;
+}
+
+/*
+   DMA objects are used to reference a piece of memory in the
+   framebuffer, PCI or AGP address space. Each object is 16 bytes big
+   and looks as follows:
+
+   entry[0]
+   11:0  class (seems like I can always use 0 here)
+   12    page table present?
+   13    page entry linear?
+   15:14 access: 0 rw, 1 ro, 2 wo
+   17:16 target: 0 NV memory, 1 NV memory tiled, 2 PCI, 3 AGP
+   31:20 dma adjust (bits 0-11 of the address)
+   entry[1]
+   dma limit (size of transfer)
+   entry[X]
+   1     0 readonly, 1 readwrite
+   31:12 dma frame address of the page (bits 12-31 of the address)
+   entry[N]
+   page table terminator, same value as the first pte, as does nvidia
+   rivatv uses 0xffffffff
+
+   Non linear page tables need a list of frame addresses afterwards,
+   the rivatv project has some info on this.
+
+   The method below creates a DMA object in instance RAM and returns a handle
+   to it that can be used to set up context objects.
+*/
+int
+nouveau_gpuobj_dma_new(struct nouveau_channel *chan, int class,
+		       uint64_t offset, uint64_t size, int access,
+		       int target, struct nouveau_gpuobj **gpuobj)
+{
+	struct drm_device *dev = chan->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_instmem_engine *instmem = &dev_priv->engine.instmem;
+	int ret;
+
+	NV_DEBUG(dev, "ch%d class=0x%04x offset=0x%llx size=0x%llx\n",
+		 chan->id, class, offset, size);
+	NV_DEBUG(dev, "access=%d target=%d\n", access, target);
+
+	switch (target) {
+	case NV_DMA_TARGET_AGP:
+		offset += dev_priv->gart_info.aper_base;
+		break;
+	default:
+		break;
+	}
+
+	ret = nouveau_gpuobj_new(dev, chan,
+				 nouveau_gpuobj_class_instmem_size(dev, class),
+				 16, NVOBJ_FLAG_ZERO_ALLOC |
+				 NVOBJ_FLAG_ZERO_FREE, gpuobj);
+	if (ret) {
+		NV_ERROR(dev, "Error creating gpuobj: %d\n", ret);
+		return ret;
+	}
+
+	instmem->prepare_access(dev, true);
+
+	if (dev_priv->card_type < NV_50) {
+		uint32_t frame, adjust, pte_flags = 0;
+
+		if (access != NV_DMA_ACCESS_RO)
+			pte_flags |= (1<<1);
+		adjust = offset &  0x00000fff;
+		frame  = offset & ~0x00000fff;
+
+		nv_wo32(dev, *gpuobj, 0, ((1<<12) | (1<<13) |
+				(adjust << 20) |
+				 (access << 14) |
+				 (target << 16) |
+				  class));
+		nv_wo32(dev, *gpuobj, 1, size - 1);
+		nv_wo32(dev, *gpuobj, 2, frame | pte_flags);
+		nv_wo32(dev, *gpuobj, 3, frame | pte_flags);
+	} else {
+		uint64_t limit = offset + size - 1;
+		uint32_t flags0, flags5;
+
+		if (target == NV_DMA_TARGET_VIDMEM) {
+			flags0 = 0x00190000;
+			flags5 = 0x00010000;
+		} else {
+			flags0 = 0x7fc00000;
+			flags5 = 0x00080000;
+		}
+
+		nv_wo32(dev, *gpuobj, 0, flags0 | class);
+		nv_wo32(dev, *gpuobj, 1, lower_32_bits(limit));
+		nv_wo32(dev, *gpuobj, 2, lower_32_bits(offset));
+		nv_wo32(dev, *gpuobj, 3, ((upper_32_bits(limit) & 0xff) << 24) |
+					(upper_32_bits(offset) & 0xff));
+		nv_wo32(dev, *gpuobj, 5, flags5);
+	}
+
+	instmem->finish_access(dev);
+
+	(*gpuobj)->engine = NVOBJ_ENGINE_SW;
+	(*gpuobj)->class  = class;
+	return 0;
+}
+
+int
+nouveau_gpuobj_gart_dma_new(struct nouveau_channel *chan,
+			    uint64_t offset, uint64_t size, int access,
+			    struct nouveau_gpuobj **gpuobj,
+			    uint32_t *o_ret)
+{
+	struct drm_device *dev = chan->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	int ret;
+
+	if (dev_priv->gart_info.type == NOUVEAU_GART_AGP ||
+	    (dev_priv->card_type >= NV_50 &&
+	     dev_priv->gart_info.type == NOUVEAU_GART_SGDMA)) {
+		ret = nouveau_gpuobj_dma_new(chan, NV_CLASS_DMA_IN_MEMORY,
+					     offset + dev_priv->vm_gart_base,
+					     size, access, NV_DMA_TARGET_AGP,
+					     gpuobj);
+		if (o_ret)
+			*o_ret = 0;
+	} else
+	if (dev_priv->gart_info.type == NOUVEAU_GART_SGDMA) {
+		*gpuobj = dev_priv->gart_info.sg_ctxdma;
+		if (offset & ~0xffffffffULL) {
+			NV_ERROR(dev, "obj offset exceeds 32-bits\n");
+			return -EINVAL;
+		}
+		if (o_ret)
+			*o_ret = (uint32_t)offset;
+		ret = (*gpuobj != NULL) ? 0 : -EINVAL;
+	} else {
+		NV_ERROR(dev, "Invalid GART type %d\n", dev_priv->gart_info.type);
+		return -EINVAL;
+	}
+
+	return ret;
+}
+
+/* Context objects in the instance RAM have the following structure.
+ * On NV40 they are 32 byte long, on NV30 and smaller 16 bytes.
+
+   NV4 - NV30:
+
+   entry[0]
+   11:0 class
+   12   chroma key enable
+   13   user clip enable
+   14   swizzle enable
+   17:15 patch config:
+       scrcopy_and, rop_and, blend_and, scrcopy, srccopy_pre, blend_pre
+   18   synchronize enable
+   19   endian: 1 big, 0 little
+   21:20 dither mode
+   23    single step enable
+   24    patch status: 0 invalid, 1 valid
+   25    context_surface 0: 1 valid
+   26    context surface 1: 1 valid
+   27    context pattern: 1 valid
+   28    context rop: 1 valid
+   29,30 context beta, beta4
+   entry[1]
+   7:0   mono format
+   15:8  color format
+   31:16 notify instance address
+   entry[2]
+   15:0  dma 0 instance address
+   31:16 dma 1 instance address
+   entry[3]
+   dma method traps
+
+   NV40:
+   No idea what the exact format is. Here's what can be deducted:
+
+   entry[0]:
+   11:0  class  (maybe uses more bits here?)
+   17    user clip enable
+   21:19 patch config
+   25    patch status valid ?
+   entry[1]:
+   15:0  DMA notifier  (maybe 20:0)
+   entry[2]:
+   15:0  DMA 0 instance (maybe 20:0)
+   24    big endian
+   entry[3]:
+   15:0  DMA 1 instance (maybe 20:0)
+   entry[4]:
+   entry[5]:
+   set to 0?
+*/
+int
+nouveau_gpuobj_gr_new(struct nouveau_channel *chan, int class,
+		      struct nouveau_gpuobj **gpuobj)
+{
+	struct drm_device *dev = chan->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	int ret;
+
+	NV_DEBUG(dev, "ch%d class=0x%04x\n", chan->id, class);
+
+	ret = nouveau_gpuobj_new(dev, chan,
+				 nouveau_gpuobj_class_instmem_size(dev, class),
+				 16,
+				 NVOBJ_FLAG_ZERO_ALLOC | NVOBJ_FLAG_ZERO_FREE,
+				 gpuobj);
+	if (ret) {
+		NV_ERROR(dev, "Error creating gpuobj: %d\n", ret);
+		return ret;
+	}
+
+	dev_priv->engine.instmem.prepare_access(dev, true);
+	if (dev_priv->card_type >= NV_50) {
+		nv_wo32(dev, *gpuobj, 0, class);
+		nv_wo32(dev, *gpuobj, 5, 0x00010000);
+	} else {
+		switch (class) {
+		case NV_CLASS_NULL:
+			nv_wo32(dev, *gpuobj, 0, 0x00001030);
+			nv_wo32(dev, *gpuobj, 1, 0xFFFFFFFF);
+			break;
+		default:
+			if (dev_priv->card_type >= NV_40) {
+				nv_wo32(dev, *gpuobj, 0, class);
+#ifdef __BIG_ENDIAN
+				nv_wo32(dev, *gpuobj, 2, 0x01000000);
+#endif
+			} else {
+#ifdef __BIG_ENDIAN
+				nv_wo32(dev, *gpuobj, 0, class | 0x00080000);
+#else
+				nv_wo32(dev, *gpuobj, 0, class);
+#endif
+			}
+		}
+	}
+	dev_priv->engine.instmem.finish_access(dev);
+
+	(*gpuobj)->engine = NVOBJ_ENGINE_GR;
+	(*gpuobj)->class  = class;
+	return 0;
+}
+
+int
+nouveau_gpuobj_sw_new(struct nouveau_channel *chan, int class,
+		      struct nouveau_gpuobj **gpuobj_ret)
+{
+	struct drm_nouveau_private *dev_priv;
+	struct nouveau_gpuobj *gpuobj;
+
+	if (!chan || !gpuobj_ret || *gpuobj_ret != NULL)
+		return -EINVAL;
+	dev_priv = chan->dev->dev_private;
+
+	gpuobj = kzalloc(sizeof(*gpuobj), GFP_KERNEL);
+	if (!gpuobj)
+		return -ENOMEM;
+	gpuobj->engine = NVOBJ_ENGINE_SW;
+	gpuobj->class = class;
+
+	list_add_tail(&gpuobj->list, &dev_priv->gpuobj_list);
+	*gpuobj_ret = gpuobj;
+	return 0;
+}
+
+static int
+nouveau_gpuobj_channel_init_pramin(struct nouveau_channel *chan)
+{
+	struct drm_device *dev = chan->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_gpuobj *pramin = NULL;
+	uint32_t size;
+	uint32_t base;
+	int ret;
+
+	NV_DEBUG(dev, "ch%d\n", chan->id);
+
+	/* Base amount for object storage (4KiB enough?) */
+	size = 0x1000;
+	base = 0;
+
+	/* PGRAPH context */
+
+	if (dev_priv->card_type == NV_50) {
+		/* Various fixed table thingos */
+		size += 0x1400; /* mostly unknown stuff */
+		size += 0x4000; /* vm pd */
+		base  = 0x6000;
+		/* RAMHT, not sure about setting size yet, 32KiB to be safe */
+		size += 0x8000;
+		/* RAMFC */
+		size += 0x1000;
+		/* PGRAPH context */
+		size += 0x70000;
+	}
+
+	NV_DEBUG(dev, "ch%d PRAMIN size: 0x%08x bytes, base alloc=0x%08x\n",
+		 chan->id, size, base);
+	ret = nouveau_gpuobj_new_ref(dev, NULL, NULL, 0, size, 0x1000, 0,
+				     &chan->ramin);
+	if (ret) {
+		NV_ERROR(dev, "Error allocating channel PRAMIN: %d\n", ret);
+		return ret;
+	}
+	pramin = chan->ramin->gpuobj;
+
+	ret = nouveau_mem_init_heap(&chan->ramin_heap,
+				    pramin->im_pramin->start + base, size);
+	if (ret) {
+		NV_ERROR(dev, "Error creating PRAMIN heap: %d\n", ret);
+		nouveau_gpuobj_ref_del(dev, &chan->ramin);
+		return ret;
+	}
+
+	return 0;
+}
+
+int
+nouveau_gpuobj_channel_init(struct nouveau_channel *chan,
+			    uint32_t vram_h, uint32_t tt_h)
+{
+	struct drm_device *dev = chan->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_instmem_engine *instmem = &dev_priv->engine.instmem;
+	struct nouveau_gpuobj *vram = NULL, *tt = NULL;
+	int ret, i;
+
+	INIT_LIST_HEAD(&chan->ramht_refs);
+
+	NV_DEBUG(dev, "ch%d vram=0x%08x tt=0x%08x\n", chan->id, vram_h, tt_h);
+
+	/* Reserve a block of PRAMIN for the channel
+	 *XXX: maybe on <NV50 too at some point
+	 */
+	if (0 || dev_priv->card_type == NV_50) {
+		ret = nouveau_gpuobj_channel_init_pramin(chan);
+		if (ret) {
+			NV_ERROR(dev, "init pramin\n");
+			return ret;
+		}
+	}
+
+	/* NV50 VM
+	 *  - Allocate per-channel page-directory
+	 *  - Map GART and VRAM into the channel's address space at the
+	 *    locations determined during init.
+	 */
+	if (dev_priv->card_type >= NV_50) {
+		uint32_t vm_offset, pde;
+
+		instmem->prepare_access(dev, true);
+
+		vm_offset = (dev_priv->chipset & 0xf0) == 0x50 ? 0x1400 : 0x200;
+		vm_offset += chan->ramin->gpuobj->im_pramin->start;
+
+		ret = nouveau_gpuobj_new_fake(dev, vm_offset, ~0, 0x4000,
+							0, &chan->vm_pd, NULL);
+		if (ret) {
+			instmem->finish_access(dev);
+			return ret;
+		}
+		for (i = 0; i < 0x4000; i += 8) {
+			nv_wo32(dev, chan->vm_pd, (i+0)/4, 0x00000000);
+			nv_wo32(dev, chan->vm_pd, (i+4)/4, 0xdeadcafe);
+		}
+
+		pde = (dev_priv->vm_gart_base / (512*1024*1024)) * 2;
+		ret = nouveau_gpuobj_ref_add(dev, NULL, 0,
+					     dev_priv->gart_info.sg_ctxdma,
+					     &chan->vm_gart_pt);
+		if (ret) {
+			instmem->finish_access(dev);
+			return ret;
+		}
+		nv_wo32(dev, chan->vm_pd, pde++,
+			    chan->vm_gart_pt->instance | 0x03);
+		nv_wo32(dev, chan->vm_pd, pde++, 0x00000000);
+
+		pde = (dev_priv->vm_vram_base / (512*1024*1024)) * 2;
+		for (i = 0; i < dev_priv->vm_vram_pt_nr; i++) {
+			ret = nouveau_gpuobj_ref_add(dev, NULL, 0,
+						     dev_priv->vm_vram_pt[i],
+						     &chan->vm_vram_pt[i]);
+			if (ret) {
+				instmem->finish_access(dev);
+				return ret;
+			}
+
+			nv_wo32(dev, chan->vm_pd, pde++,
+				    chan->vm_vram_pt[i]->instance | 0x61);
+			nv_wo32(dev, chan->vm_pd, pde++, 0x00000000);
+		}
+
+		instmem->finish_access(dev);
+	}
+
+	/* RAMHT */
+	if (dev_priv->card_type < NV_50) {
+		ret = nouveau_gpuobj_ref_add(dev, NULL, 0, dev_priv->ramht,
+					     &chan->ramht);
+		if (ret)
+			return ret;
+	} else {
+		ret = nouveau_gpuobj_new_ref(dev, chan, chan, 0,
+					     0x8000, 16,
+					     NVOBJ_FLAG_ZERO_ALLOC,
+					     &chan->ramht);
+		if (ret)
+			return ret;
+	}
+
+	/* VRAM ctxdma */
+	if (dev_priv->card_type >= NV_50) {
+		ret = nouveau_gpuobj_dma_new(chan, NV_CLASS_DMA_IN_MEMORY,
+					     0, dev_priv->vm_end,
+					     NV_DMA_ACCESS_RW,
+					     NV_DMA_TARGET_AGP, &vram);
+		if (ret) {
+			NV_ERROR(dev, "Error creating VRAM ctxdma: %d\n", ret);
+			return ret;
+		}
+	} else {
+		ret = nouveau_gpuobj_dma_new(chan, NV_CLASS_DMA_IN_MEMORY,
+						0, dev_priv->fb_available_size,
+						NV_DMA_ACCESS_RW,
+						NV_DMA_TARGET_VIDMEM, &vram);
+		if (ret) {
+			NV_ERROR(dev, "Error creating VRAM ctxdma: %d\n", ret);
+			return ret;
+		}
+	}
+
+	ret = nouveau_gpuobj_ref_add(dev, chan, vram_h, vram, NULL);
+	if (ret) {
+		NV_ERROR(dev, "Error referencing VRAM ctxdma: %d\n", ret);
+		return ret;
+	}
+
+	/* TT memory ctxdma */
+	if (dev_priv->card_type >= NV_50) {
+		tt = vram;
+	} else
+	if (dev_priv->gart_info.type != NOUVEAU_GART_NONE) {
+		ret = nouveau_gpuobj_gart_dma_new(chan, 0,
+						  dev_priv->gart_info.aper_size,
+						  NV_DMA_ACCESS_RW, &tt, NULL);
+	} else {
+		NV_ERROR(dev, "Invalid GART type %d\n", dev_priv->gart_info.type);
+		ret = -EINVAL;
+	}
+
+	if (ret) {
+		NV_ERROR(dev, "Error creating TT ctxdma: %d\n", ret);
+		return ret;
+	}
+
+	ret = nouveau_gpuobj_ref_add(dev, chan, tt_h, tt, NULL);
+	if (ret) {
+		NV_ERROR(dev, "Error referencing TT ctxdma: %d\n", ret);
+		return ret;
+	}
+
+	return 0;
+}
+
+void
+nouveau_gpuobj_channel_takedown(struct nouveau_channel *chan)
+{
+	struct drm_nouveau_private *dev_priv = chan->dev->dev_private;
+	struct drm_device *dev = chan->dev;
+	struct list_head *entry, *tmp;
+	struct nouveau_gpuobj_ref *ref;
+	int i;
+
+	NV_DEBUG(dev, "ch%d\n", chan->id);
+
+	if (!chan->ramht_refs.next)
+		return;
+
+	list_for_each_safe(entry, tmp, &chan->ramht_refs) {
+		ref = list_entry(entry, struct nouveau_gpuobj_ref, list);
+
+		nouveau_gpuobj_ref_del(dev, &ref);
+	}
+
+	nouveau_gpuobj_ref_del(dev, &chan->ramht);
+
+	nouveau_gpuobj_del(dev, &chan->vm_pd);
+	nouveau_gpuobj_ref_del(dev, &chan->vm_gart_pt);
+	for (i = 0; i < dev_priv->vm_vram_pt_nr; i++)
+		nouveau_gpuobj_ref_del(dev, &chan->vm_vram_pt[i]);
+
+	if (chan->ramin_heap)
+		nouveau_mem_takedown(&chan->ramin_heap);
+	if (chan->ramin)
+		nouveau_gpuobj_ref_del(dev, &chan->ramin);
+
+}
+
+int
+nouveau_gpuobj_suspend(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_gpuobj *gpuobj;
+	int i;
+
+	if (dev_priv->card_type < NV_50) {
+		dev_priv->susres.ramin_copy = vmalloc(dev_priv->ramin_rsvd_vram);
+		if (!dev_priv->susres.ramin_copy)
+			return -ENOMEM;
+
+		for (i = 0; i < dev_priv->ramin_rsvd_vram; i += 4)
+			dev_priv->susres.ramin_copy[i/4] = nv_ri32(dev, i);
+		return 0;
+	}
+
+	list_for_each_entry(gpuobj, &dev_priv->gpuobj_list, list) {
+		if (!gpuobj->im_backing || (gpuobj->flags & NVOBJ_FLAG_FAKE))
+			continue;
+
+		gpuobj->im_backing_suspend = vmalloc(gpuobj->im_pramin->size);
+		if (!gpuobj->im_backing_suspend) {
+			nouveau_gpuobj_resume(dev);
+			return -ENOMEM;
+		}
+
+		dev_priv->engine.instmem.prepare_access(dev, false);
+		for (i = 0; i < gpuobj->im_pramin->size / 4; i++)
+			gpuobj->im_backing_suspend[i] = nv_ro32(dev, gpuobj, i);
+		dev_priv->engine.instmem.finish_access(dev);
+	}
+
+	return 0;
+}
+
+void
+nouveau_gpuobj_suspend_cleanup(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_gpuobj *gpuobj;
+
+	if (dev_priv->card_type < NV_50) {
+		vfree(dev_priv->susres.ramin_copy);
+		dev_priv->susres.ramin_copy = NULL;
+		return;
+	}
+
+	list_for_each_entry(gpuobj, &dev_priv->gpuobj_list, list) {
+		if (!gpuobj->im_backing_suspend)
+			continue;
+
+		vfree(gpuobj->im_backing_suspend);
+		gpuobj->im_backing_suspend = NULL;
+	}
+}
+
+void
+nouveau_gpuobj_resume(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_gpuobj *gpuobj;
+	int i;
+
+	if (dev_priv->card_type < NV_50) {
+		for (i = 0; i < dev_priv->ramin_rsvd_vram; i += 4)
+			nv_wi32(dev, i, dev_priv->susres.ramin_copy[i/4]);
+		nouveau_gpuobj_suspend_cleanup(dev);
+		return;
+	}
+
+	list_for_each_entry(gpuobj, &dev_priv->gpuobj_list, list) {
+		if (!gpuobj->im_backing_suspend)
+			continue;
+
+		dev_priv->engine.instmem.prepare_access(dev, true);
+		for (i = 0; i < gpuobj->im_pramin->size / 4; i++)
+			nv_wo32(dev, gpuobj, i, gpuobj->im_backing_suspend[i]);
+		dev_priv->engine.instmem.finish_access(dev);
+	}
+
+	nouveau_gpuobj_suspend_cleanup(dev);
+}
+
+int nouveau_ioctl_grobj_alloc(struct drm_device *dev, void *data,
+			      struct drm_file *file_priv)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct drm_nouveau_grobj_alloc *init = data;
+	struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
+	struct nouveau_pgraph_object_class *grc;
+	struct nouveau_gpuobj *gr = NULL;
+	struct nouveau_channel *chan;
+	int ret;
+
+	NOUVEAU_CHECK_INITIALISED_WITH_RETURN;
+	NOUVEAU_GET_USER_CHANNEL_WITH_RETURN(init->channel, file_priv, chan);
+
+	if (init->handle == ~0)
+		return -EINVAL;
+
+	grc = pgraph->grclass;
+	while (grc->id) {
+		if (grc->id == init->class)
+			break;
+		grc++;
+	}
+
+	if (!grc->id) {
+		NV_ERROR(dev, "Illegal object class: 0x%x\n", init->class);
+		return -EPERM;
+	}
+
+	if (nouveau_gpuobj_ref_find(chan, init->handle, NULL) == 0)
+		return -EEXIST;
+
+	if (!grc->software)
+		ret = nouveau_gpuobj_gr_new(chan, grc->id, &gr);
+	else
+		ret = nouveau_gpuobj_sw_new(chan, grc->id, &gr);
+
+	if (ret) {
+		NV_ERROR(dev, "Error creating object: %d (%d/0x%08x)\n",
+			 ret, init->channel, init->handle);
+		return ret;
+	}
+
+	ret = nouveau_gpuobj_ref_add(dev, chan, init->handle, gr, NULL);
+	if (ret) {
+		NV_ERROR(dev, "Error referencing object: %d (%d/0x%08x)\n",
+			 ret, init->channel, init->handle);
+		nouveau_gpuobj_del(dev, &gr);
+		return ret;
+	}
+
+	return 0;
+}
+
+int nouveau_ioctl_gpuobj_free(struct drm_device *dev, void *data,
+			      struct drm_file *file_priv)
+{
+	struct drm_nouveau_gpuobj_free *objfree = data;
+	struct nouveau_gpuobj_ref *ref;
+	struct nouveau_channel *chan;
+	int ret;
+
+	NOUVEAU_CHECK_INITIALISED_WITH_RETURN;
+	NOUVEAU_GET_USER_CHANNEL_WITH_RETURN(objfree->channel, file_priv, chan);
+
+	ret = nouveau_gpuobj_ref_find(chan, objfree->handle, &ref);
+	if (ret)
+		return ret;
+	nouveau_gpuobj_ref_del(dev, &ref);
+
+	return 0;
+}
diff --git a/drivers/gpu/drm/nouveau/nouveau_reg.h b/drivers/gpu/drm/nouveau/nouveau_reg.h
new file mode 100644
index 000000000000..aa9b310e41be
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_reg.h
@@ -0,0 +1,837 @@
+
+
+#define NV03_BOOT_0                                        0x00100000
+#    define NV03_BOOT_0_RAM_AMOUNT                         0x00000003
+#    define NV03_BOOT_0_RAM_AMOUNT_8MB                     0x00000000
+#    define NV03_BOOT_0_RAM_AMOUNT_2MB                     0x00000001
+#    define NV03_BOOT_0_RAM_AMOUNT_4MB                     0x00000002
+#    define NV03_BOOT_0_RAM_AMOUNT_8MB_SDRAM               0x00000003
+#    define NV04_BOOT_0_RAM_AMOUNT_32MB                    0x00000000
+#    define NV04_BOOT_0_RAM_AMOUNT_4MB                     0x00000001
+#    define NV04_BOOT_0_RAM_AMOUNT_8MB                     0x00000002
+#    define NV04_BOOT_0_RAM_AMOUNT_16MB                    0x00000003
+
+#define NV04_FIFO_DATA                                     0x0010020c
+#    define NV10_FIFO_DATA_RAM_AMOUNT_MB_MASK              0xfff00000
+#    define NV10_FIFO_DATA_RAM_AMOUNT_MB_SHIFT             20
+
+#define NV_RAMIN                                           0x00700000
+
+#define NV_RAMHT_HANDLE_OFFSET                             0
+#define NV_RAMHT_CONTEXT_OFFSET                            4
+#    define NV_RAMHT_CONTEXT_VALID                         (1<<31)
+#    define NV_RAMHT_CONTEXT_CHANNEL_SHIFT                 24
+#    define NV_RAMHT_CONTEXT_ENGINE_SHIFT                  16
+#        define NV_RAMHT_CONTEXT_ENGINE_SOFTWARE           0
+#        define NV_RAMHT_CONTEXT_ENGINE_GRAPHICS           1
+#    define NV_RAMHT_CONTEXT_INSTANCE_SHIFT                0
+#    define NV40_RAMHT_CONTEXT_CHANNEL_SHIFT               23
+#    define NV40_RAMHT_CONTEXT_ENGINE_SHIFT                20
+#    define NV40_RAMHT_CONTEXT_INSTANCE_SHIFT              0
+
+/* DMA object defines */
+#define NV_DMA_ACCESS_RW 0
+#define NV_DMA_ACCESS_RO 1
+#define NV_DMA_ACCESS_WO 2
+#define NV_DMA_TARGET_VIDMEM 0
+#define NV_DMA_TARGET_PCI    2
+#define NV_DMA_TARGET_AGP    3
+/* The following is not a real value used by the card, it's changed by
+ * nouveau_object_dma_create */
+#define NV_DMA_TARGET_PCI_NONLINEAR 8
+
+/* Some object classes we care about in the drm */
+#define NV_CLASS_DMA_FROM_MEMORY                           0x00000002
+#define NV_CLASS_DMA_TO_MEMORY                             0x00000003
+#define NV_CLASS_NULL                                      0x00000030
+#define NV_CLASS_DMA_IN_MEMORY                             0x0000003D
+
+#define NV03_USER(i)                             (0x00800000+(i*NV03_USER_SIZE))
+#define NV03_USER__SIZE                                                       16
+#define NV10_USER__SIZE                                                       32
+#define NV03_USER_SIZE                                                0x00010000
+#define NV03_USER_DMA_PUT(i)                     (0x00800040+(i*NV03_USER_SIZE))
+#define NV03_USER_DMA_PUT__SIZE                                               16
+#define NV10_USER_DMA_PUT__SIZE                                               32
+#define NV03_USER_DMA_GET(i)                     (0x00800044+(i*NV03_USER_SIZE))
+#define NV03_USER_DMA_GET__SIZE                                               16
+#define NV10_USER_DMA_GET__SIZE                                               32
+#define NV03_USER_REF_CNT(i)                     (0x00800048+(i*NV03_USER_SIZE))
+#define NV03_USER_REF_CNT__SIZE                                               16
+#define NV10_USER_REF_CNT__SIZE                                               32
+
+#define NV40_USER(i)                             (0x00c00000+(i*NV40_USER_SIZE))
+#define NV40_USER_SIZE                                                0x00001000
+#define NV40_USER_DMA_PUT(i)                     (0x00c00040+(i*NV40_USER_SIZE))
+#define NV40_USER_DMA_PUT__SIZE                                               32
+#define NV40_USER_DMA_GET(i)                     (0x00c00044+(i*NV40_USER_SIZE))
+#define NV40_USER_DMA_GET__SIZE                                               32
+#define NV40_USER_REF_CNT(i)                     (0x00c00048+(i*NV40_USER_SIZE))
+#define NV40_USER_REF_CNT__SIZE                                               32
+
+#define NV50_USER(i)                             (0x00c00000+(i*NV50_USER_SIZE))
+#define NV50_USER_SIZE                                                0x00002000
+#define NV50_USER_DMA_PUT(i)                     (0x00c00040+(i*NV50_USER_SIZE))
+#define NV50_USER_DMA_PUT__SIZE                                              128
+#define NV50_USER_DMA_GET(i)                     (0x00c00044+(i*NV50_USER_SIZE))
+#define NV50_USER_DMA_GET__SIZE                                              128
+#define NV50_USER_REF_CNT(i)                     (0x00c00048+(i*NV50_USER_SIZE))
+#define NV50_USER_REF_CNT__SIZE                                              128
+
+#define NV03_FIFO_SIZE                                     0x8000UL
+
+#define NV03_PMC_BOOT_0                                    0x00000000
+#define NV03_PMC_BOOT_1                                    0x00000004
+#define NV03_PMC_INTR_0                                    0x00000100
+#    define NV_PMC_INTR_0_PFIFO_PENDING                        (1<<8)
+#    define NV_PMC_INTR_0_PGRAPH_PENDING                      (1<<12)
+#    define NV_PMC_INTR_0_NV50_I2C_PENDING                    (1<<21)
+#    define NV_PMC_INTR_0_CRTC0_PENDING                       (1<<24)
+#    define NV_PMC_INTR_0_CRTC1_PENDING                       (1<<25)
+#    define NV_PMC_INTR_0_NV50_DISPLAY_PENDING                (1<<26)
+#    define NV_PMC_INTR_0_CRTCn_PENDING                       (3<<24)
+#define NV03_PMC_INTR_EN_0                                 0x00000140
+#    define NV_PMC_INTR_EN_0_MASTER_ENABLE                     (1<<0)
+#define NV03_PMC_ENABLE                                    0x00000200
+#    define NV_PMC_ENABLE_PFIFO                                (1<<8)
+#    define NV_PMC_ENABLE_PGRAPH                              (1<<12)
+/* Disabling the below bit breaks newer (G7X only?) mobile chipsets,
+ * the card will hang early on in the X init process.
+ */
+#    define NV_PMC_ENABLE_UNK13                               (1<<13)
+#define NV40_PMC_GRAPH_UNITS				   0x00001540
+#define NV40_PMC_BACKLIGHT				   0x000015f0
+#	define NV40_PMC_BACKLIGHT_MASK			   0x001f0000
+#define NV40_PMC_1700                                      0x00001700
+#define NV40_PMC_1704                                      0x00001704
+#define NV40_PMC_1708                                      0x00001708
+#define NV40_PMC_170C                                      0x0000170C
+
+/* probably PMC ? */
+#define NV50_PUNK_BAR0_PRAMIN                              0x00001700
+#define NV50_PUNK_BAR_CFG_BASE                             0x00001704
+#define NV50_PUNK_BAR_CFG_BASE_VALID                          (1<<30)
+#define NV50_PUNK_BAR1_CTXDMA                              0x00001708
+#define NV50_PUNK_BAR1_CTXDMA_VALID                           (1<<31)
+#define NV50_PUNK_BAR3_CTXDMA                              0x0000170C
+#define NV50_PUNK_BAR3_CTXDMA_VALID                           (1<<31)
+#define NV50_PUNK_UNK1710                                  0x00001710
+
+#define NV04_PBUS_PCI_NV_1                                 0x00001804
+#define NV04_PBUS_PCI_NV_19                                0x0000184C
+#define NV04_PBUS_PCI_NV_20				0x00001850
+#	define NV04_PBUS_PCI_NV_20_ROM_SHADOW_DISABLED		(0 << 0)
+#	define NV04_PBUS_PCI_NV_20_ROM_SHADOW_ENABLED		(1 << 0)
+
+#define NV04_PTIMER_INTR_0                                 0x00009100
+#define NV04_PTIMER_INTR_EN_0                              0x00009140
+#define NV04_PTIMER_NUMERATOR                              0x00009200
+#define NV04_PTIMER_DENOMINATOR                            0x00009210
+#define NV04_PTIMER_TIME_0                                 0x00009400
+#define NV04_PTIMER_TIME_1                                 0x00009410
+#define NV04_PTIMER_ALARM_0                                0x00009420
+
+#define NV04_PFB_CFG0                                      0x00100200
+#define NV04_PFB_CFG1                                      0x00100204
+#define NV40_PFB_020C                                      0x0010020C
+#define NV10_PFB_TILE(i)                                   (0x00100240 + (i*16))
+#define NV10_PFB_TILE__SIZE                                8
+#define NV10_PFB_TLIMIT(i)                                 (0x00100244 + (i*16))
+#define NV10_PFB_TSIZE(i)                                  (0x00100248 + (i*16))
+#define NV10_PFB_TSTATUS(i)                                (0x0010024C + (i*16))
+#define NV10_PFB_CLOSE_PAGE2                               0x0010033C
+#define NV40_PFB_TILE(i)                                   (0x00100600 + (i*16))
+#define NV40_PFB_TILE__SIZE_0                              12
+#define NV40_PFB_TILE__SIZE_1                              15
+#define NV40_PFB_TLIMIT(i)                                 (0x00100604 + (i*16))
+#define NV40_PFB_TSIZE(i)                                  (0x00100608 + (i*16))
+#define NV40_PFB_TSTATUS(i)                                (0x0010060C + (i*16))
+#define NV40_PFB_UNK_800					0x00100800
+
+#define NV04_PGRAPH_DEBUG_0                                0x00400080
+#define NV04_PGRAPH_DEBUG_1                                0x00400084
+#define NV04_PGRAPH_DEBUG_2                                0x00400088
+#define NV04_PGRAPH_DEBUG_3                                0x0040008c
+#define NV10_PGRAPH_DEBUG_4                                0x00400090
+#define NV03_PGRAPH_INTR                                   0x00400100
+#define NV03_PGRAPH_NSTATUS                                0x00400104
+#    define NV04_PGRAPH_NSTATUS_STATE_IN_USE                  (1<<11)
+#    define NV04_PGRAPH_NSTATUS_INVALID_STATE                 (1<<12)
+#    define NV04_PGRAPH_NSTATUS_BAD_ARGUMENT                  (1<<13)
+#    define NV04_PGRAPH_NSTATUS_PROTECTION_FAULT              (1<<14)
+#    define NV10_PGRAPH_NSTATUS_STATE_IN_USE                  (1<<23)
+#    define NV10_PGRAPH_NSTATUS_INVALID_STATE                 (1<<24)
+#    define NV10_PGRAPH_NSTATUS_BAD_ARGUMENT                  (1<<25)
+#    define NV10_PGRAPH_NSTATUS_PROTECTION_FAULT              (1<<26)
+#define NV03_PGRAPH_NSOURCE                                0x00400108
+#    define NV03_PGRAPH_NSOURCE_NOTIFICATION                   (1<<0)
+#    define NV03_PGRAPH_NSOURCE_DATA_ERROR                     (1<<1)
+#    define NV03_PGRAPH_NSOURCE_PROTECTION_ERROR               (1<<2)
+#    define NV03_PGRAPH_NSOURCE_RANGE_EXCEPTION                (1<<3)
+#    define NV03_PGRAPH_NSOURCE_LIMIT_COLOR                    (1<<4)
+#    define NV03_PGRAPH_NSOURCE_LIMIT_ZETA                     (1<<5)
+#    define NV03_PGRAPH_NSOURCE_ILLEGAL_MTHD                   (1<<6)
+#    define NV03_PGRAPH_NSOURCE_DMA_R_PROTECTION               (1<<7)
+#    define NV03_PGRAPH_NSOURCE_DMA_W_PROTECTION               (1<<8)
+#    define NV03_PGRAPH_NSOURCE_FORMAT_EXCEPTION               (1<<9)
+#    define NV03_PGRAPH_NSOURCE_PATCH_EXCEPTION               (1<<10)
+#    define NV03_PGRAPH_NSOURCE_STATE_INVALID                 (1<<11)
+#    define NV03_PGRAPH_NSOURCE_DOUBLE_NOTIFY                 (1<<12)
+#    define NV03_PGRAPH_NSOURCE_NOTIFY_IN_USE                 (1<<13)
+#    define NV03_PGRAPH_NSOURCE_METHOD_CNT                    (1<<14)
+#    define NV03_PGRAPH_NSOURCE_BFR_NOTIFICATION              (1<<15)
+#    define NV03_PGRAPH_NSOURCE_DMA_VTX_PROTECTION            (1<<16)
+#    define NV03_PGRAPH_NSOURCE_DMA_WIDTH_A                   (1<<17)
+#    define NV03_PGRAPH_NSOURCE_DMA_WIDTH_B                   (1<<18)
+#define NV03_PGRAPH_INTR_EN                                0x00400140
+#define NV40_PGRAPH_INTR_EN                                0x0040013C
+#    define NV_PGRAPH_INTR_NOTIFY                              (1<<0)
+#    define NV_PGRAPH_INTR_MISSING_HW                          (1<<4)
+#    define NV_PGRAPH_INTR_CONTEXT_SWITCH                     (1<<12)
+#    define NV_PGRAPH_INTR_BUFFER_NOTIFY                      (1<<16)
+#    define NV_PGRAPH_INTR_ERROR                              (1<<20)
+#define NV10_PGRAPH_CTX_CONTROL                            0x00400144
+#define NV10_PGRAPH_CTX_USER                               0x00400148
+#define NV10_PGRAPH_CTX_SWITCH1                            0x0040014C
+#define NV10_PGRAPH_CTX_SWITCH2                            0x00400150
+#define NV10_PGRAPH_CTX_SWITCH3                            0x00400154
+#define NV10_PGRAPH_CTX_SWITCH4                            0x00400158
+#define NV10_PGRAPH_CTX_SWITCH5                            0x0040015C
+#define NV04_PGRAPH_CTX_SWITCH1                            0x00400160
+#define NV10_PGRAPH_CTX_CACHE1                             0x00400160
+#define NV04_PGRAPH_CTX_SWITCH2                            0x00400164
+#define NV04_PGRAPH_CTX_SWITCH3                            0x00400168
+#define NV04_PGRAPH_CTX_SWITCH4                            0x0040016C
+#define NV04_PGRAPH_CTX_CONTROL                            0x00400170
+#define NV04_PGRAPH_CTX_USER                               0x00400174
+#define NV04_PGRAPH_CTX_CACHE1                             0x00400180
+#define NV10_PGRAPH_CTX_CACHE2                             0x00400180
+#define NV03_PGRAPH_CTX_CONTROL                            0x00400190
+#define NV03_PGRAPH_CTX_USER                               0x00400194
+#define NV04_PGRAPH_CTX_CACHE2                             0x004001A0
+#define NV10_PGRAPH_CTX_CACHE3                             0x004001A0
+#define NV04_PGRAPH_CTX_CACHE3                             0x004001C0
+#define NV10_PGRAPH_CTX_CACHE4                             0x004001C0
+#define NV04_PGRAPH_CTX_CACHE4                             0x004001E0
+#define NV10_PGRAPH_CTX_CACHE5                             0x004001E0
+#define NV40_PGRAPH_CTXCTL_0304                            0x00400304
+#define NV40_PGRAPH_CTXCTL_0304_XFER_CTX                   0x00000001
+#define NV40_PGRAPH_CTXCTL_UCODE_STAT                      0x00400308
+#define NV40_PGRAPH_CTXCTL_UCODE_STAT_IP_MASK              0xff000000
+#define NV40_PGRAPH_CTXCTL_UCODE_STAT_IP_SHIFT                     24
+#define NV40_PGRAPH_CTXCTL_UCODE_STAT_OP_MASK              0x00ffffff
+#define NV40_PGRAPH_CTXCTL_0310                            0x00400310
+#define NV40_PGRAPH_CTXCTL_0310_XFER_SAVE                  0x00000020
+#define NV40_PGRAPH_CTXCTL_0310_XFER_LOAD                  0x00000040
+#define NV40_PGRAPH_CTXCTL_030C                            0x0040030c
+#define NV40_PGRAPH_CTXCTL_UCODE_INDEX                     0x00400324
+#define NV40_PGRAPH_CTXCTL_UCODE_DATA                      0x00400328
+#define NV40_PGRAPH_CTXCTL_CUR                             0x0040032c
+#define NV40_PGRAPH_CTXCTL_CUR_LOADED                      0x01000000
+#define NV40_PGRAPH_CTXCTL_CUR_INSTANCE                    0x000FFFFF
+#define NV40_PGRAPH_CTXCTL_NEXT                            0x00400330
+#define NV40_PGRAPH_CTXCTL_NEXT_INSTANCE                   0x000fffff
+#define NV50_PGRAPH_CTXCTL_CUR                             0x0040032c
+#define NV50_PGRAPH_CTXCTL_CUR_LOADED                      0x80000000
+#define NV50_PGRAPH_CTXCTL_CUR_INSTANCE                    0x00ffffff
+#define NV50_PGRAPH_CTXCTL_NEXT                            0x00400330
+#define NV50_PGRAPH_CTXCTL_NEXT_INSTANCE                   0x00ffffff
+#define NV03_PGRAPH_ABS_X_RAM                              0x00400400
+#define NV03_PGRAPH_ABS_Y_RAM                              0x00400480
+#define NV03_PGRAPH_X_MISC                                 0x00400500
+#define NV03_PGRAPH_Y_MISC                                 0x00400504
+#define NV04_PGRAPH_VALID1                                 0x00400508
+#define NV04_PGRAPH_SOURCE_COLOR                           0x0040050C
+#define NV04_PGRAPH_MISC24_0                               0x00400510
+#define NV03_PGRAPH_XY_LOGIC_MISC0                         0x00400514
+#define NV03_PGRAPH_XY_LOGIC_MISC1                         0x00400518
+#define NV03_PGRAPH_XY_LOGIC_MISC2                         0x0040051C
+#define NV03_PGRAPH_XY_LOGIC_MISC3                         0x00400520
+#define NV03_PGRAPH_CLIPX_0                                0x00400524
+#define NV03_PGRAPH_CLIPX_1                                0x00400528
+#define NV03_PGRAPH_CLIPY_0                                0x0040052C
+#define NV03_PGRAPH_CLIPY_1                                0x00400530
+#define NV03_PGRAPH_ABS_ICLIP_XMAX                         0x00400534
+#define NV03_PGRAPH_ABS_ICLIP_YMAX                         0x00400538
+#define NV03_PGRAPH_ABS_UCLIP_XMIN                         0x0040053C
+#define NV03_PGRAPH_ABS_UCLIP_YMIN                         0x00400540
+#define NV03_PGRAPH_ABS_UCLIP_XMAX                         0x00400544
+#define NV03_PGRAPH_ABS_UCLIP_YMAX                         0x00400548
+#define NV03_PGRAPH_ABS_UCLIPA_XMIN                        0x00400560
+#define NV03_PGRAPH_ABS_UCLIPA_YMIN                        0x00400564
+#define NV03_PGRAPH_ABS_UCLIPA_XMAX                        0x00400568
+#define NV03_PGRAPH_ABS_UCLIPA_YMAX                        0x0040056C
+#define NV04_PGRAPH_MISC24_1                               0x00400570
+#define NV04_PGRAPH_MISC24_2                               0x00400574
+#define NV04_PGRAPH_VALID2                                 0x00400578
+#define NV04_PGRAPH_PASSTHRU_0                             0x0040057C
+#define NV04_PGRAPH_PASSTHRU_1                             0x00400580
+#define NV04_PGRAPH_PASSTHRU_2                             0x00400584
+#define NV10_PGRAPH_DIMX_TEXTURE                           0x00400588
+#define NV10_PGRAPH_WDIMX_TEXTURE                          0x0040058C
+#define NV04_PGRAPH_COMBINE_0_ALPHA                        0x00400590
+#define NV04_PGRAPH_COMBINE_0_COLOR                        0x00400594
+#define NV04_PGRAPH_COMBINE_1_ALPHA                        0x00400598
+#define NV04_PGRAPH_COMBINE_1_COLOR                        0x0040059C
+#define NV04_PGRAPH_FORMAT_0                               0x004005A8
+#define NV04_PGRAPH_FORMAT_1                               0x004005AC
+#define NV04_PGRAPH_FILTER_0                               0x004005B0
+#define NV04_PGRAPH_FILTER_1                               0x004005B4
+#define NV03_PGRAPH_MONO_COLOR0                            0x00400600
+#define NV04_PGRAPH_ROP3                                   0x00400604
+#define NV04_PGRAPH_BETA_AND                               0x00400608
+#define NV04_PGRAPH_BETA_PREMULT                           0x0040060C
+#define NV04_PGRAPH_LIMIT_VIOL_PIX                         0x00400610
+#define NV04_PGRAPH_FORMATS                                0x00400618
+#define NV10_PGRAPH_DEBUG_2                                0x00400620
+#define NV04_PGRAPH_BOFFSET0                               0x00400640
+#define NV04_PGRAPH_BOFFSET1                               0x00400644
+#define NV04_PGRAPH_BOFFSET2                               0x00400648
+#define NV04_PGRAPH_BOFFSET3                               0x0040064C
+#define NV04_PGRAPH_BOFFSET4                               0x00400650
+#define NV04_PGRAPH_BOFFSET5                               0x00400654
+#define NV04_PGRAPH_BBASE0                                 0x00400658
+#define NV04_PGRAPH_BBASE1                                 0x0040065C
+#define NV04_PGRAPH_BBASE2                                 0x00400660
+#define NV04_PGRAPH_BBASE3                                 0x00400664
+#define NV04_PGRAPH_BBASE4                                 0x00400668
+#define NV04_PGRAPH_BBASE5                                 0x0040066C
+#define NV04_PGRAPH_BPITCH0                                0x00400670
+#define NV04_PGRAPH_BPITCH1                                0x00400674
+#define NV04_PGRAPH_BPITCH2                                0x00400678
+#define NV04_PGRAPH_BPITCH3                                0x0040067C
+#define NV04_PGRAPH_BPITCH4                                0x00400680
+#define NV04_PGRAPH_BLIMIT0                                0x00400684
+#define NV04_PGRAPH_BLIMIT1                                0x00400688
+#define NV04_PGRAPH_BLIMIT2                                0x0040068C
+#define NV04_PGRAPH_BLIMIT3                                0x00400690
+#define NV04_PGRAPH_BLIMIT4                                0x00400694
+#define NV04_PGRAPH_BLIMIT5                                0x00400698
+#define NV04_PGRAPH_BSWIZZLE2                              0x0040069C
+#define NV04_PGRAPH_BSWIZZLE5                              0x004006A0
+#define NV03_PGRAPH_STATUS                                 0x004006B0
+#define NV04_PGRAPH_STATUS                                 0x00400700
+#define NV04_PGRAPH_TRAPPED_ADDR                           0x00400704
+#define NV04_PGRAPH_TRAPPED_DATA                           0x00400708
+#define NV04_PGRAPH_SURFACE                                0x0040070C
+#define NV10_PGRAPH_TRAPPED_DATA_HIGH                      0x0040070C
+#define NV04_PGRAPH_STATE                                  0x00400710
+#define NV10_PGRAPH_SURFACE                                0x00400710
+#define NV04_PGRAPH_NOTIFY                                 0x00400714
+#define NV10_PGRAPH_STATE                                  0x00400714
+#define NV10_PGRAPH_NOTIFY                                 0x00400718
+
+#define NV04_PGRAPH_FIFO                                   0x00400720
+
+#define NV04_PGRAPH_BPIXEL                                 0x00400724
+#define NV10_PGRAPH_RDI_INDEX                              0x00400750
+#define NV04_PGRAPH_FFINTFC_ST2                            0x00400754
+#define NV10_PGRAPH_RDI_DATA                               0x00400754
+#define NV04_PGRAPH_DMA_PITCH                              0x00400760
+#define NV10_PGRAPH_FFINTFC_ST2                            0x00400764
+#define NV04_PGRAPH_DVD_COLORFMT                           0x00400764
+#define NV04_PGRAPH_SCALED_FORMAT                          0x00400768
+#define NV10_PGRAPH_DMA_PITCH                              0x00400770
+#define NV10_PGRAPH_DVD_COLORFMT                           0x00400774
+#define NV10_PGRAPH_SCALED_FORMAT                          0x00400778
+#define NV20_PGRAPH_CHANNEL_CTX_TABLE                      0x00400780
+#define NV20_PGRAPH_CHANNEL_CTX_POINTER                    0x00400784
+#define NV20_PGRAPH_CHANNEL_CTX_XFER                       0x00400788
+#define NV20_PGRAPH_CHANNEL_CTX_XFER_LOAD                  0x00000001
+#define NV20_PGRAPH_CHANNEL_CTX_XFER_SAVE                  0x00000002
+#define NV04_PGRAPH_PATT_COLOR0                            0x00400800
+#define NV04_PGRAPH_PATT_COLOR1                            0x00400804
+#define NV04_PGRAPH_PATTERN                                0x00400808
+#define NV04_PGRAPH_PATTERN_SHAPE                          0x00400810
+#define NV04_PGRAPH_CHROMA                                 0x00400814
+#define NV04_PGRAPH_CONTROL0                               0x00400818
+#define NV04_PGRAPH_CONTROL1                               0x0040081C
+#define NV04_PGRAPH_CONTROL2                               0x00400820
+#define NV04_PGRAPH_BLEND                                  0x00400824
+#define NV04_PGRAPH_STORED_FMT                             0x00400830
+#define NV04_PGRAPH_PATT_COLORRAM                          0x00400900
+#define NV20_PGRAPH_TILE(i)                                (0x00400900 + (i*16))
+#define NV20_PGRAPH_TLIMIT(i)                              (0x00400904 + (i*16))
+#define NV20_PGRAPH_TSIZE(i)                               (0x00400908 + (i*16))
+#define NV20_PGRAPH_TSTATUS(i)                             (0x0040090C + (i*16))
+#define NV10_PGRAPH_TILE(i)                                (0x00400B00 + (i*16))
+#define NV10_PGRAPH_TLIMIT(i)                              (0x00400B04 + (i*16))
+#define NV10_PGRAPH_TSIZE(i)                               (0x00400B08 + (i*16))
+#define NV10_PGRAPH_TSTATUS(i)                             (0x00400B0C + (i*16))
+#define NV04_PGRAPH_U_RAM                                  0x00400D00
+#define NV47_PGRAPH_TILE(i)                                (0x00400D00 + (i*16))
+#define NV47_PGRAPH_TLIMIT(i)                              (0x00400D04 + (i*16))
+#define NV47_PGRAPH_TSIZE(i)                               (0x00400D08 + (i*16))
+#define NV47_PGRAPH_TSTATUS(i)                             (0x00400D0C + (i*16))
+#define NV04_PGRAPH_V_RAM                                  0x00400D40
+#define NV04_PGRAPH_W_RAM                                  0x00400D80
+#define NV10_PGRAPH_COMBINER0_IN_ALPHA                     0x00400E40
+#define NV10_PGRAPH_COMBINER1_IN_ALPHA                     0x00400E44
+#define NV10_PGRAPH_COMBINER0_IN_RGB                       0x00400E48
+#define NV10_PGRAPH_COMBINER1_IN_RGB                       0x00400E4C
+#define NV10_PGRAPH_COMBINER_COLOR0                        0x00400E50
+#define NV10_PGRAPH_COMBINER_COLOR1                        0x00400E54
+#define NV10_PGRAPH_COMBINER0_OUT_ALPHA                    0x00400E58
+#define NV10_PGRAPH_COMBINER1_OUT_ALPHA                    0x00400E5C
+#define NV10_PGRAPH_COMBINER0_OUT_RGB                      0x00400E60
+#define NV10_PGRAPH_COMBINER1_OUT_RGB                      0x00400E64
+#define NV10_PGRAPH_COMBINER_FINAL0                        0x00400E68
+#define NV10_PGRAPH_COMBINER_FINAL1                        0x00400E6C
+#define NV10_PGRAPH_WINDOWCLIP_HORIZONTAL                  0x00400F00
+#define NV10_PGRAPH_WINDOWCLIP_VERTICAL                    0x00400F20
+#define NV10_PGRAPH_XFMODE0                                0x00400F40
+#define NV10_PGRAPH_XFMODE1                                0x00400F44
+#define NV10_PGRAPH_GLOBALSTATE0                           0x00400F48
+#define NV10_PGRAPH_GLOBALSTATE1                           0x00400F4C
+#define NV10_PGRAPH_PIPE_ADDRESS                           0x00400F50
+#define NV10_PGRAPH_PIPE_DATA                              0x00400F54
+#define NV04_PGRAPH_DMA_START_0                            0x00401000
+#define NV04_PGRAPH_DMA_START_1                            0x00401004
+#define NV04_PGRAPH_DMA_LENGTH                             0x00401008
+#define NV04_PGRAPH_DMA_MISC                               0x0040100C
+#define NV04_PGRAPH_DMA_DATA_0                             0x00401020
+#define NV04_PGRAPH_DMA_DATA_1                             0x00401024
+#define NV04_PGRAPH_DMA_RM                                 0x00401030
+#define NV04_PGRAPH_DMA_A_XLATE_INST                       0x00401040
+#define NV04_PGRAPH_DMA_A_CONTROL                          0x00401044
+#define NV04_PGRAPH_DMA_A_LIMIT                            0x00401048
+#define NV04_PGRAPH_DMA_A_TLB_PTE                          0x0040104C
+#define NV04_PGRAPH_DMA_A_TLB_TAG                          0x00401050
+#define NV04_PGRAPH_DMA_A_ADJ_OFFSET                       0x00401054
+#define NV04_PGRAPH_DMA_A_OFFSET                           0x00401058
+#define NV04_PGRAPH_DMA_A_SIZE                             0x0040105C
+#define NV04_PGRAPH_DMA_A_Y_SIZE                           0x00401060
+#define NV04_PGRAPH_DMA_B_XLATE_INST                       0x00401080
+#define NV04_PGRAPH_DMA_B_CONTROL                          0x00401084
+#define NV04_PGRAPH_DMA_B_LIMIT                            0x00401088
+#define NV04_PGRAPH_DMA_B_TLB_PTE                          0x0040108C
+#define NV04_PGRAPH_DMA_B_TLB_TAG                          0x00401090
+#define NV04_PGRAPH_DMA_B_ADJ_OFFSET                       0x00401094
+#define NV04_PGRAPH_DMA_B_OFFSET                           0x00401098
+#define NV04_PGRAPH_DMA_B_SIZE                             0x0040109C
+#define NV04_PGRAPH_DMA_B_Y_SIZE                           0x004010A0
+#define NV40_PGRAPH_TILE1(i)                               (0x00406900 + (i*16))
+#define NV40_PGRAPH_TLIMIT1(i)                             (0x00406904 + (i*16))
+#define NV40_PGRAPH_TSIZE1(i)                              (0x00406908 + (i*16))
+#define NV40_PGRAPH_TSTATUS1(i)                            (0x0040690C + (i*16))
+
+
+/* It's a guess that this works on NV03. Confirmed on NV04, though */
+#define NV04_PFIFO_DELAY_0                                 0x00002040
+#define NV04_PFIFO_DMA_TIMESLICE                           0x00002044
+#define NV04_PFIFO_NEXT_CHANNEL                            0x00002050
+#define NV03_PFIFO_INTR_0                                  0x00002100
+#define NV03_PFIFO_INTR_EN_0                               0x00002140
+#    define NV_PFIFO_INTR_CACHE_ERROR                          (1<<0)
+#    define NV_PFIFO_INTR_RUNOUT                               (1<<4)
+#    define NV_PFIFO_INTR_RUNOUT_OVERFLOW                      (1<<8)
+#    define NV_PFIFO_INTR_DMA_PUSHER                          (1<<12)
+#    define NV_PFIFO_INTR_DMA_PT                              (1<<16)
+#    define NV_PFIFO_INTR_SEMAPHORE                           (1<<20)
+#    define NV_PFIFO_INTR_ACQUIRE_TIMEOUT                     (1<<24)
+#define NV03_PFIFO_RAMHT                                   0x00002210
+#define NV03_PFIFO_RAMFC                                   0x00002214
+#define NV03_PFIFO_RAMRO                                   0x00002218
+#define NV40_PFIFO_RAMFC                                   0x00002220
+#define NV03_PFIFO_CACHES                                  0x00002500
+#define NV04_PFIFO_MODE                                    0x00002504
+#define NV04_PFIFO_DMA                                     0x00002508
+#define NV04_PFIFO_SIZE                                    0x0000250c
+#define NV50_PFIFO_CTX_TABLE(c)                        (0x2600+(c)*4)
+#define NV50_PFIFO_CTX_TABLE__SIZE                                128
+#define NV50_PFIFO_CTX_TABLE_CHANNEL_ENABLED                  (1<<31)
+#define NV50_PFIFO_CTX_TABLE_UNK30_BAD                        (1<<30)
+#define NV50_PFIFO_CTX_TABLE_INSTANCE_MASK_G80             0x0FFFFFFF
+#define NV50_PFIFO_CTX_TABLE_INSTANCE_MASK_G84             0x00FFFFFF
+#define NV03_PFIFO_CACHE0_PUSH0                            0x00003000
+#define NV03_PFIFO_CACHE0_PULL0                            0x00003040
+#define NV04_PFIFO_CACHE0_PULL0                            0x00003050
+#define NV04_PFIFO_CACHE0_PULL1                            0x00003054
+#define NV03_PFIFO_CACHE1_PUSH0                            0x00003200
+#define NV03_PFIFO_CACHE1_PUSH1                            0x00003204
+#define NV03_PFIFO_CACHE1_PUSH1_DMA                            (1<<8)
+#define NV40_PFIFO_CACHE1_PUSH1_DMA                           (1<<16)
+#define NV03_PFIFO_CACHE1_PUSH1_CHID_MASK                  0x0000000f
+#define NV10_PFIFO_CACHE1_PUSH1_CHID_MASK                  0x0000001f
+#define NV50_PFIFO_CACHE1_PUSH1_CHID_MASK                  0x0000007f
+#define NV03_PFIFO_CACHE1_PUT                              0x00003210
+#define NV04_PFIFO_CACHE1_DMA_PUSH                         0x00003220
+#define NV04_PFIFO_CACHE1_DMA_FETCH                        0x00003224
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_8_BYTES         0x00000000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_16_BYTES        0x00000008
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_24_BYTES        0x00000010
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_32_BYTES        0x00000018
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_40_BYTES        0x00000020
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_48_BYTES        0x00000028
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_56_BYTES        0x00000030
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_64_BYTES        0x00000038
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_72_BYTES        0x00000040
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_80_BYTES        0x00000048
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_88_BYTES        0x00000050
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_96_BYTES        0x00000058
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_104_BYTES       0x00000060
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_112_BYTES       0x00000068
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_120_BYTES       0x00000070
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_128_BYTES       0x00000078
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_136_BYTES       0x00000080
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_144_BYTES       0x00000088
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_152_BYTES       0x00000090
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_160_BYTES       0x00000098
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_168_BYTES       0x000000A0
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_176_BYTES       0x000000A8
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_184_BYTES       0x000000B0
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_192_BYTES       0x000000B8
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_200_BYTES       0x000000C0
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_208_BYTES       0x000000C8
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_216_BYTES       0x000000D0
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_224_BYTES       0x000000D8
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_232_BYTES       0x000000E0
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_240_BYTES       0x000000E8
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_248_BYTES       0x000000F0
+#    define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_256_BYTES       0x000000F8
+#    define NV_PFIFO_CACHE1_DMA_FETCH_SIZE                 0x0000E000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_SIZE_32_BYTES        0x00000000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_SIZE_64_BYTES        0x00002000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_SIZE_96_BYTES        0x00004000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_SIZE_128_BYTES       0x00006000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_SIZE_160_BYTES       0x00008000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_SIZE_192_BYTES       0x0000A000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_SIZE_224_BYTES       0x0000C000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_SIZE_256_BYTES       0x0000E000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS             0x001F0000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_0           0x00000000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_1           0x00010000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_2           0x00020000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_3           0x00030000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_4           0x00040000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_5           0x00050000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_6           0x00060000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_7           0x00070000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_8           0x00080000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_9           0x00090000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_10          0x000A0000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_11          0x000B0000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_12          0x000C0000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_13          0x000D0000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_14          0x000E0000
+#    define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_15          0x000F0000
+#    define NV_PFIFO_CACHE1_ENDIAN                         0x80000000
+#    define NV_PFIFO_CACHE1_LITTLE_ENDIAN                  0x7FFFFFFF
+#    define NV_PFIFO_CACHE1_BIG_ENDIAN                     0x80000000
+#define NV04_PFIFO_CACHE1_DMA_STATE                        0x00003228
+#define NV04_PFIFO_CACHE1_DMA_INSTANCE                     0x0000322c
+#define NV04_PFIFO_CACHE1_DMA_CTL                          0x00003230
+#define NV04_PFIFO_CACHE1_DMA_PUT                          0x00003240
+#define NV04_PFIFO_CACHE1_DMA_GET                          0x00003244
+#define NV10_PFIFO_CACHE1_REF_CNT                          0x00003248
+#define NV10_PFIFO_CACHE1_DMA_SUBROUTINE                   0x0000324C
+#define NV03_PFIFO_CACHE1_PULL0                            0x00003240
+#define NV04_PFIFO_CACHE1_PULL0                            0x00003250
+#define NV03_PFIFO_CACHE1_PULL1                            0x00003250
+#define NV04_PFIFO_CACHE1_PULL1                            0x00003254
+#define NV04_PFIFO_CACHE1_HASH                             0x00003258
+#define NV10_PFIFO_CACHE1_ACQUIRE_TIMEOUT                  0x00003260
+#define NV10_PFIFO_CACHE1_ACQUIRE_TIMESTAMP                0x00003264
+#define NV10_PFIFO_CACHE1_ACQUIRE_VALUE                    0x00003268
+#define NV10_PFIFO_CACHE1_SEMAPHORE                        0x0000326C
+#define NV03_PFIFO_CACHE1_GET                              0x00003270
+#define NV04_PFIFO_CACHE1_ENGINE                           0x00003280
+#define NV04_PFIFO_CACHE1_DMA_DCOUNT                       0x000032A0
+#define NV40_PFIFO_GRCTX_INSTANCE                          0x000032E0
+#define NV40_PFIFO_UNK32E4                                 0x000032E4
+#define NV04_PFIFO_CACHE1_METHOD(i)                (0x00003800+(i*8))
+#define NV04_PFIFO_CACHE1_DATA(i)                  (0x00003804+(i*8))
+#define NV40_PFIFO_CACHE1_METHOD(i)                (0x00090000+(i*8))
+#define NV40_PFIFO_CACHE1_DATA(i)                  (0x00090004+(i*8))
+
+#define NV_CRTC0_INTSTAT                                   0x00600100
+#define NV_CRTC0_INTEN                                     0x00600140
+#define NV_CRTC1_INTSTAT                                   0x00602100
+#define NV_CRTC1_INTEN                                     0x00602140
+#    define NV_CRTC_INTR_VBLANK                                (1<<0)
+
+#define NV04_PRAMIN						0x00700000
+
+/* Fifo commands. These are not regs, neither masks */
+#define NV03_FIFO_CMD_JUMP                                 0x20000000
+#define NV03_FIFO_CMD_JUMP_OFFSET_MASK                     0x1ffffffc
+#define NV03_FIFO_CMD_REWIND                               (NV03_FIFO_CMD_JUMP | (0 & NV03_FIFO_CMD_JUMP_OFFSET_MASK))
+
+/* This is a partial import from rules-ng, a few things may be duplicated.
+ * Eventually we should completely import everything from rules-ng.
+ * For the moment check rules-ng for docs.
+  */
+
+#define NV50_PMC                                            0x00000000
+#define NV50_PMC__LEN                                              0x1
+#define NV50_PMC__ESIZE                                         0x2000
+#    define NV50_PMC_BOOT_0                                 0x00000000
+#        define NV50_PMC_BOOT_0_REVISION                    0x000000ff
+#        define NV50_PMC_BOOT_0_REVISION__SHIFT                      0
+#        define NV50_PMC_BOOT_0_ARCH                        0x0ff00000
+#        define NV50_PMC_BOOT_0_ARCH__SHIFT                         20
+#    define NV50_PMC_INTR_0                                 0x00000100
+#        define NV50_PMC_INTR_0_PFIFO                           (1<<8)
+#        define NV50_PMC_INTR_0_PGRAPH                         (1<<12)
+#        define NV50_PMC_INTR_0_PTIMER                         (1<<20)
+#        define NV50_PMC_INTR_0_HOTPLUG                        (1<<21)
+#        define NV50_PMC_INTR_0_DISPLAY                        (1<<26)
+#    define NV50_PMC_INTR_EN_0                              0x00000140
+#        define NV50_PMC_INTR_EN_0_MASTER                       (1<<0)
+#            define NV50_PMC_INTR_EN_0_MASTER_DISABLED          (0<<0)
+#            define NV50_PMC_INTR_EN_0_MASTER_ENABLED           (1<<0)
+#    define NV50_PMC_ENABLE                                 0x00000200
+#        define NV50_PMC_ENABLE_PFIFO                           (1<<8)
+#        define NV50_PMC_ENABLE_PGRAPH                         (1<<12)
+
+#define NV50_PCONNECTOR                                     0x0000e000
+#define NV50_PCONNECTOR__LEN                                       0x1
+#define NV50_PCONNECTOR__ESIZE                                  0x1000
+#    define NV50_PCONNECTOR_HOTPLUG_INTR                    0x0000e050
+#        define NV50_PCONNECTOR_HOTPLUG_INTR_PLUG_I2C0          (1<<0)
+#        define NV50_PCONNECTOR_HOTPLUG_INTR_PLUG_I2C1          (1<<1)
+#        define NV50_PCONNECTOR_HOTPLUG_INTR_PLUG_I2C2          (1<<2)
+#        define NV50_PCONNECTOR_HOTPLUG_INTR_PLUG_I2C3          (1<<3)
+#        define NV50_PCONNECTOR_HOTPLUG_INTR_UNPLUG_I2C0       (1<<16)
+#        define NV50_PCONNECTOR_HOTPLUG_INTR_UNPLUG_I2C1       (1<<17)
+#        define NV50_PCONNECTOR_HOTPLUG_INTR_UNPLUG_I2C2       (1<<18)
+#        define NV50_PCONNECTOR_HOTPLUG_INTR_UNPLUG_I2C3       (1<<19)
+#    define NV50_PCONNECTOR_HOTPLUG_CTRL                    0x0000e054
+#        define NV50_PCONNECTOR_HOTPLUG_CTRL_PLUG_I2C0          (1<<0)
+#        define NV50_PCONNECTOR_HOTPLUG_CTRL_PLUG_I2C1          (1<<1)
+#        define NV50_PCONNECTOR_HOTPLUG_CTRL_PLUG_I2C2          (1<<2)
+#        define NV50_PCONNECTOR_HOTPLUG_CTRL_PLUG_I2C3          (1<<3)
+#        define NV50_PCONNECTOR_HOTPLUG_CTRL_UNPLUG_I2C0       (1<<16)
+#        define NV50_PCONNECTOR_HOTPLUG_CTRL_UNPLUG_I2C1       (1<<17)
+#        define NV50_PCONNECTOR_HOTPLUG_CTRL_UNPLUG_I2C2       (1<<18)
+#        define NV50_PCONNECTOR_HOTPLUG_CTRL_UNPLUG_I2C3       (1<<19)
+#    define NV50_PCONNECTOR_HOTPLUG_STATE                   0x0000e104
+#        define NV50_PCONNECTOR_HOTPLUG_STATE_PIN_CONNECTED_I2C0 (1<<2)
+#        define NV50_PCONNECTOR_HOTPLUG_STATE_PIN_CONNECTED_I2C1 (1<<6)
+#        define NV50_PCONNECTOR_HOTPLUG_STATE_PIN_CONNECTED_I2C2 (1<<10)
+#        define NV50_PCONNECTOR_HOTPLUG_STATE_PIN_CONNECTED_I2C3 (1<<14)
+#    define NV50_PCONNECTOR_I2C_PORT_0                      0x0000e138
+#    define NV50_PCONNECTOR_I2C_PORT_1                      0x0000e150
+#    define NV50_PCONNECTOR_I2C_PORT_2                      0x0000e168
+#    define NV50_PCONNECTOR_I2C_PORT_3                      0x0000e180
+#    define NV50_PCONNECTOR_I2C_PORT_4                      0x0000e240
+#    define NV50_PCONNECTOR_I2C_PORT_5                      0x0000e258
+
+#define NV50_AUXCH_DATA_OUT(i,n)             ((n) * 4 + (i) * 0x50 + 0x0000e4c0)
+#define NV50_AUXCH_DATA_OUT__SIZE                                             4
+#define NV50_AUXCH_DATA_IN(i,n)              ((n) * 4 + (i) * 0x50 + 0x0000e4d0)
+#define NV50_AUXCH_DATA_IN__SIZE                                              4
+#define NV50_AUXCH_ADDR(i)                             ((i) * 0x50 + 0x0000e4e0)
+#define NV50_AUXCH_CTRL(i)                             ((i) * 0x50 + 0x0000e4e4)
+#define NV50_AUXCH_CTRL_LINKSTAT                                     0x01000000
+#define NV50_AUXCH_CTRL_LINKSTAT_NOT_READY                           0x00000000
+#define NV50_AUXCH_CTRL_LINKSTAT_READY                               0x01000000
+#define NV50_AUXCH_CTRL_LINKEN                                       0x00100000
+#define NV50_AUXCH_CTRL_LINKEN_DISABLED                              0x00000000
+#define NV50_AUXCH_CTRL_LINKEN_ENABLED                               0x00100000
+#define NV50_AUXCH_CTRL_EXEC                                         0x00010000
+#define NV50_AUXCH_CTRL_EXEC_COMPLETE                                0x00000000
+#define NV50_AUXCH_CTRL_EXEC_IN_PROCESS                              0x00010000
+#define NV50_AUXCH_CTRL_CMD                                          0x0000f000
+#define NV50_AUXCH_CTRL_CMD_SHIFT                                            12
+#define NV50_AUXCH_CTRL_LEN                                          0x0000000f
+#define NV50_AUXCH_CTRL_LEN_SHIFT                                             0
+#define NV50_AUXCH_STAT(i)                             ((i) * 0x50 + 0x0000e4e8)
+#define NV50_AUXCH_STAT_STATE                                        0x10000000
+#define NV50_AUXCH_STAT_STATE_NOT_READY                              0x00000000
+#define NV50_AUXCH_STAT_STATE_READY                                  0x10000000
+#define NV50_AUXCH_STAT_REPLY                                        0x000f0000
+#define NV50_AUXCH_STAT_REPLY_AUX                                    0x00030000
+#define NV50_AUXCH_STAT_REPLY_AUX_ACK                                0x00000000
+#define NV50_AUXCH_STAT_REPLY_AUX_NACK                               0x00010000
+#define NV50_AUXCH_STAT_REPLY_AUX_DEFER                              0x00020000
+#define NV50_AUXCH_STAT_REPLY_I2C                                    0x000c0000
+#define NV50_AUXCH_STAT_REPLY_I2C_ACK                                0x00000000
+#define NV50_AUXCH_STAT_REPLY_I2C_NACK                               0x00040000
+#define NV50_AUXCH_STAT_REPLY_I2C_DEFER                              0x00080000
+#define NV50_AUXCH_STAT_COUNT                                        0x0000001f
+
+#define NV50_PBUS                                           0x00088000
+#define NV50_PBUS__LEN                                             0x1
+#define NV50_PBUS__ESIZE                                        0x1000
+#    define NV50_PBUS_PCI_ID                                0x00088000
+#        define NV50_PBUS_PCI_ID_VENDOR_ID                  0x0000ffff
+#        define NV50_PBUS_PCI_ID_VENDOR_ID__SHIFT                    0
+#        define NV50_PBUS_PCI_ID_DEVICE_ID                  0xffff0000
+#        define NV50_PBUS_PCI_ID_DEVICE_ID__SHIFT                   16
+
+#define NV50_PFB                                            0x00100000
+#define NV50_PFB__LEN                                              0x1
+#define NV50_PFB__ESIZE                                         0x1000
+
+#define NV50_PEXTDEV                                        0x00101000
+#define NV50_PEXTDEV__LEN                                          0x1
+#define NV50_PEXTDEV__ESIZE                                     0x1000
+
+#define NV50_PROM                                           0x00300000
+#define NV50_PROM__LEN                                             0x1
+#define NV50_PROM__ESIZE                                       0x10000
+
+#define NV50_PGRAPH                                         0x00400000
+#define NV50_PGRAPH__LEN                                           0x1
+#define NV50_PGRAPH__ESIZE                                     0x10000
+
+#define NV50_PDISPLAY                                                0x00610000
+#define NV50_PDISPLAY_OBJECTS                                        0x00610010
+#define NV50_PDISPLAY_INTR_0                                         0x00610020
+#define NV50_PDISPLAY_INTR_1                                         0x00610024
+#define NV50_PDISPLAY_INTR_1_VBLANK_CRTC                             0x0000000c
+#define NV50_PDISPLAY_INTR_1_VBLANK_CRTC_SHIFT                                2
+#define NV50_PDISPLAY_INTR_1_VBLANK_CRTC_(n)                   (1 << ((n) + 2))
+#define NV50_PDISPLAY_INTR_1_VBLANK_CRTC_0                           0x00000004
+#define NV50_PDISPLAY_INTR_1_VBLANK_CRTC_1                           0x00000008
+#define NV50_PDISPLAY_INTR_1_CLK_UNK10                               0x00000010
+#define NV50_PDISPLAY_INTR_1_CLK_UNK20                               0x00000020
+#define NV50_PDISPLAY_INTR_1_CLK_UNK40                               0x00000040
+#define NV50_PDISPLAY_INTR_EN                                        0x0061002c
+#define NV50_PDISPLAY_INTR_EN_VBLANK_CRTC                            0x0000000c
+#define NV50_PDISPLAY_INTR_EN_VBLANK_CRTC_(n)                   (1 << ((n) + 2))
+#define NV50_PDISPLAY_INTR_EN_VBLANK_CRTC_0                          0x00000004
+#define NV50_PDISPLAY_INTR_EN_VBLANK_CRTC_1                          0x00000008
+#define NV50_PDISPLAY_INTR_EN_CLK_UNK10                              0x00000010
+#define NV50_PDISPLAY_INTR_EN_CLK_UNK20                              0x00000020
+#define NV50_PDISPLAY_INTR_EN_CLK_UNK40                              0x00000040
+#define NV50_PDISPLAY_UNK30_CTRL                                     0x00610030
+#define NV50_PDISPLAY_UNK30_CTRL_UPDATE_VCLK0                        0x00000200
+#define NV50_PDISPLAY_UNK30_CTRL_UPDATE_VCLK1                        0x00000400
+#define NV50_PDISPLAY_UNK30_CTRL_PENDING                             0x80000000
+#define NV50_PDISPLAY_TRAPPED_ADDR                                   0x00610080
+#define NV50_PDISPLAY_TRAPPED_DATA                                   0x00610084
+#define NV50_PDISPLAY_CHANNEL_STAT(i)                  ((i) * 0x10 + 0x00610200)
+#define NV50_PDISPLAY_CHANNEL_STAT_DMA                               0x00000010
+#define NV50_PDISPLAY_CHANNEL_STAT_DMA_DISABLED                      0x00000000
+#define NV50_PDISPLAY_CHANNEL_STAT_DMA_ENABLED                       0x00000010
+#define NV50_PDISPLAY_CHANNEL_DMA_CB(i)                ((i) * 0x10 + 0x00610204)
+#define NV50_PDISPLAY_CHANNEL_DMA_CB_LOCATION                        0x00000002
+#define NV50_PDISPLAY_CHANNEL_DMA_CB_LOCATION_VRAM                   0x00000000
+#define NV50_PDISPLAY_CHANNEL_DMA_CB_LOCATION_SYSTEM                 0x00000002
+#define NV50_PDISPLAY_CHANNEL_DMA_CB_VALID                           0x00000001
+#define NV50_PDISPLAY_CHANNEL_UNK2(i)                  ((i) * 0x10 + 0x00610208)
+#define NV50_PDISPLAY_CHANNEL_UNK3(i)                  ((i) * 0x10 + 0x0061020c)
+
+#define NV50_PDISPLAY_CURSOR                                         0x00610270
+#define NV50_PDISPLAY_CURSOR_CURSOR_CTRL2(i)           ((i) * 0x10 + 0x00610270)
+#define NV50_PDISPLAY_CURSOR_CURSOR_CTRL2_ON                         0x00000001
+#define NV50_PDISPLAY_CURSOR_CURSOR_CTRL2_STATUS                     0x00030000
+#define NV50_PDISPLAY_CURSOR_CURSOR_CTRL2_STATUS_ACTIVE              0x00010000
+
+#define NV50_PDISPLAY_CTRL_STATE                                     0x00610300
+#define NV50_PDISPLAY_CTRL_STATE_PENDING                             0x80000000
+#define NV50_PDISPLAY_CTRL_STATE_METHOD                              0x00001ffc
+#define NV50_PDISPLAY_CTRL_STATE_ENABLE                              0x00000001
+#define NV50_PDISPLAY_CTRL_VAL                                       0x00610304
+#define NV50_PDISPLAY_UNK_380                                        0x00610380
+#define NV50_PDISPLAY_RAM_AMOUNT                                     0x00610384
+#define NV50_PDISPLAY_UNK_388                                        0x00610388
+#define NV50_PDISPLAY_UNK_38C                                        0x0061038c
+
+#define NV50_PDISPLAY_CRTC_P(i, r)        ((i) * 0x540 + NV50_PDISPLAY_CRTC_##r)
+#define NV50_PDISPLAY_CRTC_C(i, r)    (4 + (i) * 0x540 + NV50_PDISPLAY_CRTC_##r)
+#define NV50_PDISPLAY_CRTC_UNK_0A18 /* mthd 0x0900 */                0x00610a18
+#define NV50_PDISPLAY_CRTC_CLUT_MODE                                 0x00610a24
+#define NV50_PDISPLAY_CRTC_INTERLACE                                 0x00610a48
+#define NV50_PDISPLAY_CRTC_SCALE_CTRL                                0x00610a50
+#define NV50_PDISPLAY_CRTC_CURSOR_CTRL                               0x00610a58
+#define NV50_PDISPLAY_CRTC_UNK0A78 /* mthd 0x0904 */                 0x00610a78
+#define NV50_PDISPLAY_CRTC_UNK0AB8                                   0x00610ab8
+#define NV50_PDISPLAY_CRTC_DEPTH                                     0x00610ac8
+#define NV50_PDISPLAY_CRTC_CLOCK                                     0x00610ad0
+#define NV50_PDISPLAY_CRTC_COLOR_CTRL                                0x00610ae0
+#define NV50_PDISPLAY_CRTC_SYNC_START_TO_BLANK_END                   0x00610ae8
+#define NV50_PDISPLAY_CRTC_MODE_UNK1                                 0x00610af0
+#define NV50_PDISPLAY_CRTC_DISPLAY_TOTAL                             0x00610af8
+#define NV50_PDISPLAY_CRTC_SYNC_DURATION                             0x00610b00
+#define NV50_PDISPLAY_CRTC_MODE_UNK2                                 0x00610b08
+#define NV50_PDISPLAY_CRTC_UNK_0B10 /* mthd 0x0828 */                0x00610b10
+#define NV50_PDISPLAY_CRTC_FB_SIZE                                   0x00610b18
+#define NV50_PDISPLAY_CRTC_FB_PITCH                                  0x00610b20
+#define NV50_PDISPLAY_CRTC_FB_PITCH_LINEAR                           0x00100000
+#define NV50_PDISPLAY_CRTC_FB_POS                                    0x00610b28
+#define NV50_PDISPLAY_CRTC_SCALE_CENTER_OFFSET                       0x00610b38
+#define NV50_PDISPLAY_CRTC_REAL_RES                                  0x00610b40
+#define NV50_PDISPLAY_CRTC_SCALE_RES1                                0x00610b48
+#define NV50_PDISPLAY_CRTC_SCALE_RES2                                0x00610b50
+
+#define NV50_PDISPLAY_DAC_MODE_CTRL_P(i)                (0x00610b58 + (i) * 0x8)
+#define NV50_PDISPLAY_DAC_MODE_CTRL_C(i)                (0x00610b5c + (i) * 0x8)
+#define NV50_PDISPLAY_SOR_MODE_CTRL_P(i)                (0x00610b70 + (i) * 0x8)
+#define NV50_PDISPLAY_SOR_MODE_CTRL_C(i)                (0x00610b74 + (i) * 0x8)
+#define NV50_PDISPLAY_DAC_MODE_CTRL2_P(i)               (0x00610bdc + (i) * 0x8)
+#define NV50_PDISPLAY_DAC_MODE_CTRL2_C(i)               (0x00610be0 + (i) * 0x8)
+
+#define NV90_PDISPLAY_SOR_MODE_CTRL_P(i)                (0x00610794 + (i) * 0x8)
+#define NV90_PDISPLAY_SOR_MODE_CTRL_C(i)                (0x00610798 + (i) * 0x8)
+#define NV90_PDISPLAY_DAC_MODE_CTRL_P(i)                (0x00610b58 + (i) * 0x8)
+#define NV90_PDISPLAY_DAC_MODE_CTRL_C(i)                (0x00610b5c + (i) * 0x8)
+#define NV90_PDISPLAY_DAC_MODE_CTRL2_P(i)               (0x00610b80 + (i) * 0x8)
+#define NV90_PDISPLAY_DAC_MODE_CTRL2_C(i)               (0x00610b84 + (i) * 0x8)
+
+#define NV50_PDISPLAY_CRTC_CLK                                       0x00614000
+#define NV50_PDISPLAY_CRTC_CLK_CTRL1(i)                 ((i) * 0x800 + 0x614100)
+#define NV50_PDISPLAY_CRTC_CLK_CTRL1_CONNECTED                       0x00000600
+#define NV50_PDISPLAY_CRTC_CLK_VPLL_A(i)                ((i) * 0x800 + 0x614104)
+#define NV50_PDISPLAY_CRTC_CLK_VPLL_B(i)                ((i) * 0x800 + 0x614108)
+#define NV50_PDISPLAY_CRTC_CLK_CTRL2(i)                 ((i) * 0x800 + 0x614200)
+
+#define NV50_PDISPLAY_DAC_CLK                                        0x00614000
+#define NV50_PDISPLAY_DAC_CLK_CTRL2(i)                  ((i) * 0x800 + 0x614280)
+
+#define NV50_PDISPLAY_SOR_CLK                                        0x00614000
+#define NV50_PDISPLAY_SOR_CLK_CTRL2(i)                  ((i) * 0x800 + 0x614300)
+
+#define NV50_PDISPLAY_VGACRTC(r)                                ((r) + 0x619400)
+
+#define NV50_PDISPLAY_DAC                                            0x0061a000
+#define NV50_PDISPLAY_DAC_DPMS_CTRL(i)                (0x0061a004 + (i) * 0x800)
+#define NV50_PDISPLAY_DAC_DPMS_CTRL_HSYNC_OFF                        0x00000001
+#define NV50_PDISPLAY_DAC_DPMS_CTRL_VSYNC_OFF                        0x00000004
+#define NV50_PDISPLAY_DAC_DPMS_CTRL_BLANKED                          0x00000010
+#define NV50_PDISPLAY_DAC_DPMS_CTRL_OFF                              0x00000040
+#define NV50_PDISPLAY_DAC_DPMS_CTRL_PENDING                          0x80000000
+#define NV50_PDISPLAY_DAC_LOAD_CTRL(i)                (0x0061a00c + (i) * 0x800)
+#define NV50_PDISPLAY_DAC_LOAD_CTRL_ACTIVE                           0x00100000
+#define NV50_PDISPLAY_DAC_LOAD_CTRL_PRESENT                          0x38000000
+#define NV50_PDISPLAY_DAC_LOAD_CTRL_DONE                             0x80000000
+#define NV50_PDISPLAY_DAC_CLK_CTRL1(i)                (0x0061a010 + (i) * 0x800)
+#define NV50_PDISPLAY_DAC_CLK_CTRL1_CONNECTED                        0x00000600
+
+#define NV50_PDISPLAY_SOR                                            0x0061c000
+#define NV50_PDISPLAY_SOR_DPMS_CTRL(i)                (0x0061c004 + (i) * 0x800)
+#define NV50_PDISPLAY_SOR_DPMS_CTRL_PENDING                          0x80000000
+#define NV50_PDISPLAY_SOR_DPMS_CTRL_ON                               0x00000001
+#define NV50_PDISPLAY_SOR_CLK_CTRL1(i)                (0x0061c008 + (i) * 0x800)
+#define NV50_PDISPLAY_SOR_CLK_CTRL1_CONNECTED                        0x00000600
+#define NV50_PDISPLAY_SOR_DPMS_STATE(i)               (0x0061c030 + (i) * 0x800)
+#define NV50_PDISPLAY_SOR_DPMS_STATE_ACTIVE                          0x00030000
+#define NV50_PDISPLAY_SOR_DPMS_STATE_BLANKED                         0x00080000
+#define NV50_PDISPLAY_SOR_DPMS_STATE_WAIT                            0x10000000
+#define NV50_PDISPLAY_SOR_BACKLIGHT                                  0x0061c084
+#define NV50_PDISPLAY_SOR_BACKLIGHT_ENABLE                           0x80000000
+#define NV50_PDISPLAY_SOR_BACKLIGHT_LEVEL                            0x00000fff
+#define NV50_SOR_DP_CTRL(i,l)            (0x0061c10c + (i) * 0x800 + (l) * 0x80)
+#define NV50_SOR_DP_CTRL_ENHANCED_FRAME_ENABLED                      0x00004000
+#define NV50_SOR_DP_CTRL_LANE_MASK                                   0x001f0000
+#define NV50_SOR_DP_CTRL_LANE_0_ENABLED                              0x00010000
+#define NV50_SOR_DP_CTRL_LANE_1_ENABLED                              0x00020000
+#define NV50_SOR_DP_CTRL_LANE_2_ENABLED                              0x00040000
+#define NV50_SOR_DP_CTRL_LANE_3_ENABLED                              0x00080000
+#define NV50_SOR_DP_CTRL_TRAINING_PATTERN                            0x0f000000
+#define NV50_SOR_DP_CTRL_TRAINING_PATTERN_DISABLED                   0x00000000
+#define NV50_SOR_DP_CTRL_TRAINING_PATTERN_1                          0x01000000
+#define NV50_SOR_DP_CTRL_TRAINING_PATTERN_2                          0x02000000
+#define NV50_SOR_DP_UNK118(i,l)          (0x0061c118 + (i) * 0x800 + (l) * 0x80)
+#define NV50_SOR_DP_UNK120(i,l)          (0x0061c120 + (i) * 0x800 + (l) * 0x80)
+#define NV50_SOR_DP_UNK130(i,l)          (0x0061c130 + (i) * 0x800 + (l) * 0x80)
+
+#define NV50_PDISPLAY_USER(i)                        ((i) * 0x1000 + 0x00640000)
+#define NV50_PDISPLAY_USER_PUT(i)                    ((i) * 0x1000 + 0x00640000)
+#define NV50_PDISPLAY_USER_GET(i)                    ((i) * 0x1000 + 0x00640004)
+
+#define NV50_PDISPLAY_CURSOR_USER                                    0x00647000
+#define NV50_PDISPLAY_CURSOR_USER_POS_CTRL(i)        ((i) * 0x1000 + 0x00647080)
+#define NV50_PDISPLAY_CURSOR_USER_POS(i)             ((i) * 0x1000 + 0x00647084)
diff --git a/drivers/gpu/drm/nouveau/nouveau_sgdma.c b/drivers/gpu/drm/nouveau/nouveau_sgdma.c
new file mode 100644
index 000000000000..ed1590577b6c
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_sgdma.c
@@ -0,0 +1,322 @@
+#include "drmP.h"
+#include "nouveau_drv.h"
+#include <linux/pagemap.h>
+
+#define NV_CTXDMA_PAGE_SHIFT 12
+#define NV_CTXDMA_PAGE_SIZE  (1 << NV_CTXDMA_PAGE_SHIFT)
+#define NV_CTXDMA_PAGE_MASK  (NV_CTXDMA_PAGE_SIZE - 1)
+
+struct nouveau_sgdma_be {
+	struct ttm_backend backend;
+	struct drm_device *dev;
+
+	dma_addr_t *pages;
+	unsigned nr_pages;
+
+	unsigned pte_start;
+	bool bound;
+};
+
+static int
+nouveau_sgdma_populate(struct ttm_backend *be, unsigned long num_pages,
+		       struct page **pages, struct page *dummy_read_page)
+{
+	struct nouveau_sgdma_be *nvbe = (struct nouveau_sgdma_be *)be;
+	struct drm_device *dev = nvbe->dev;
+
+	NV_DEBUG(nvbe->dev, "num_pages = %ld\n", num_pages);
+
+	if (nvbe->pages)
+		return -EINVAL;
+
+	nvbe->pages = kmalloc(sizeof(dma_addr_t) * num_pages, GFP_KERNEL);
+	if (!nvbe->pages)
+		return -ENOMEM;
+
+	nvbe->nr_pages = 0;
+	while (num_pages--) {
+		nvbe->pages[nvbe->nr_pages] =
+			pci_map_page(dev->pdev, pages[nvbe->nr_pages], 0,
+				     PAGE_SIZE, PCI_DMA_BIDIRECTIONAL);
+		if (pci_dma_mapping_error(dev->pdev,
+					  nvbe->pages[nvbe->nr_pages])) {
+			be->func->clear(be);
+			return -EFAULT;
+		}
+
+		nvbe->nr_pages++;
+	}
+
+	return 0;
+}
+
+static void
+nouveau_sgdma_clear(struct ttm_backend *be)
+{
+	struct nouveau_sgdma_be *nvbe = (struct nouveau_sgdma_be *)be;
+	struct drm_device *dev;
+
+	if (nvbe && nvbe->pages) {
+		dev = nvbe->dev;
+		NV_DEBUG(dev, "\n");
+
+		if (nvbe->bound)
+			be->func->unbind(be);
+
+		while (nvbe->nr_pages--) {
+			pci_unmap_page(dev->pdev, nvbe->pages[nvbe->nr_pages],
+				       PAGE_SIZE, PCI_DMA_BIDIRECTIONAL);
+		}
+		kfree(nvbe->pages);
+		nvbe->pages = NULL;
+		nvbe->nr_pages = 0;
+	}
+}
+
+static inline unsigned
+nouveau_sgdma_pte(struct drm_device *dev, uint64_t offset)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	unsigned pte = (offset >> NV_CTXDMA_PAGE_SHIFT);
+
+	if (dev_priv->card_type < NV_50)
+		return pte + 2;
+
+	return pte << 1;
+}
+
+static int
+nouveau_sgdma_bind(struct ttm_backend *be, struct ttm_mem_reg *mem)
+{
+	struct nouveau_sgdma_be *nvbe = (struct nouveau_sgdma_be *)be;
+	struct drm_device *dev = nvbe->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_gpuobj *gpuobj = dev_priv->gart_info.sg_ctxdma;
+	unsigned i, j, pte;
+
+	NV_DEBUG(dev, "pg=0x%lx\n", mem->mm_node->start);
+
+	dev_priv->engine.instmem.prepare_access(nvbe->dev, true);
+	pte = nouveau_sgdma_pte(nvbe->dev, mem->mm_node->start << PAGE_SHIFT);
+	nvbe->pte_start = pte;
+	for (i = 0; i < nvbe->nr_pages; i++) {
+		dma_addr_t dma_offset = nvbe->pages[i];
+		uint32_t offset_l = lower_32_bits(dma_offset);
+		uint32_t offset_h = upper_32_bits(dma_offset);
+
+		for (j = 0; j < PAGE_SIZE / NV_CTXDMA_PAGE_SIZE; j++) {
+			if (dev_priv->card_type < NV_50)
+				nv_wo32(dev, gpuobj, pte++, offset_l | 3);
+			else {
+				nv_wo32(dev, gpuobj, pte++, offset_l | 0x21);
+				nv_wo32(dev, gpuobj, pte++, offset_h & 0xff);
+			}
+
+			dma_offset += NV_CTXDMA_PAGE_SIZE;
+		}
+	}
+	dev_priv->engine.instmem.finish_access(nvbe->dev);
+
+	if (dev_priv->card_type == NV_50) {
+		nv_wr32(dev, 0x100c80, 0x00050001);
+		if (!nv_wait(0x100c80, 0x00000001, 0x00000000)) {
+			NV_ERROR(dev, "timeout: (0x100c80 & 1) == 0 (2)\n");
+			NV_ERROR(dev, "0x100c80 = 0x%08x\n",
+						nv_rd32(dev, 0x100c80));
+			return -EBUSY;
+		}
+
+		nv_wr32(dev, 0x100c80, 0x00000001);
+		if (!nv_wait(0x100c80, 0x00000001, 0x00000000)) {
+			NV_ERROR(dev, "timeout: (0x100c80 & 1) == 0 (2)\n");
+			NV_ERROR(dev, "0x100c80 = 0x%08x\n",
+						nv_rd32(dev, 0x100c80));
+			return -EBUSY;
+		}
+	}
+
+	nvbe->bound = true;
+	return 0;
+}
+
+static int
+nouveau_sgdma_unbind(struct ttm_backend *be)
+{
+	struct nouveau_sgdma_be *nvbe = (struct nouveau_sgdma_be *)be;
+	struct drm_device *dev = nvbe->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_gpuobj *gpuobj = dev_priv->gart_info.sg_ctxdma;
+	unsigned i, j, pte;
+
+	NV_DEBUG(dev, "\n");
+
+	if (!nvbe->bound)
+		return 0;
+
+	dev_priv->engine.instmem.prepare_access(nvbe->dev, true);
+	pte = nvbe->pte_start;
+	for (i = 0; i < nvbe->nr_pages; i++) {
+		dma_addr_t dma_offset = dev_priv->gart_info.sg_dummy_bus;
+
+		for (j = 0; j < PAGE_SIZE / NV_CTXDMA_PAGE_SIZE; j++) {
+			if (dev_priv->card_type < NV_50)
+				nv_wo32(dev, gpuobj, pte++, dma_offset | 3);
+			else {
+				nv_wo32(dev, gpuobj, pte++, dma_offset | 0x21);
+				nv_wo32(dev, gpuobj, pte++, 0x00000000);
+			}
+
+			dma_offset += NV_CTXDMA_PAGE_SIZE;
+		}
+	}
+	dev_priv->engine.instmem.finish_access(nvbe->dev);
+
+	nvbe->bound = false;
+	return 0;
+}
+
+static void
+nouveau_sgdma_destroy(struct ttm_backend *be)
+{
+	struct nouveau_sgdma_be *nvbe = (struct nouveau_sgdma_be *)be;
+
+	if (be) {
+		NV_DEBUG(nvbe->dev, "\n");
+
+		if (nvbe) {
+			if (nvbe->pages)
+				be->func->clear(be);
+			kfree(nvbe);
+		}
+	}
+}
+
+static struct ttm_backend_func nouveau_sgdma_backend = {
+	.populate		= nouveau_sgdma_populate,
+	.clear			= nouveau_sgdma_clear,
+	.bind			= nouveau_sgdma_bind,
+	.unbind			= nouveau_sgdma_unbind,
+	.destroy		= nouveau_sgdma_destroy
+};
+
+struct ttm_backend *
+nouveau_sgdma_init_ttm(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_sgdma_be *nvbe;
+
+	if (!dev_priv->gart_info.sg_ctxdma)
+		return NULL;
+
+	nvbe = kzalloc(sizeof(*nvbe), GFP_KERNEL);
+	if (!nvbe)
+		return NULL;
+
+	nvbe->dev = dev;
+
+	nvbe->backend.func	= &nouveau_sgdma_backend;
+
+	return &nvbe->backend;
+}
+
+int
+nouveau_sgdma_init(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_gpuobj *gpuobj = NULL;
+	uint32_t aper_size, obj_size;
+	int i, ret;
+
+	if (dev_priv->card_type < NV_50) {
+		aper_size = (64 * 1024 * 1024);
+		obj_size  = (aper_size >> NV_CTXDMA_PAGE_SHIFT) * 4;
+		obj_size += 8; /* ctxdma header */
+	} else {
+		/* 1 entire VM page table */
+		aper_size = (512 * 1024 * 1024);
+		obj_size  = (aper_size >> NV_CTXDMA_PAGE_SHIFT) * 8;
+	}
+
+	ret = nouveau_gpuobj_new(dev, NULL, obj_size, 16,
+				      NVOBJ_FLAG_ALLOW_NO_REFS |
+				      NVOBJ_FLAG_ZERO_ALLOC |
+				      NVOBJ_FLAG_ZERO_FREE, &gpuobj);
+	if (ret) {
+		NV_ERROR(dev, "Error creating sgdma object: %d\n", ret);
+		return ret;
+	}
+
+	dev_priv->gart_info.sg_dummy_page =
+		alloc_page(GFP_KERNEL|__GFP_DMA32);
+	set_bit(PG_locked, &dev_priv->gart_info.sg_dummy_page->flags);
+	dev_priv->gart_info.sg_dummy_bus =
+		pci_map_page(dev->pdev, dev_priv->gart_info.sg_dummy_page, 0,
+			     PAGE_SIZE, PCI_DMA_BIDIRECTIONAL);
+
+	dev_priv->engine.instmem.prepare_access(dev, true);
+	if (dev_priv->card_type < NV_50) {
+		/* Maybe use NV_DMA_TARGET_AGP for PCIE? NVIDIA do this, and
+		 * confirmed to work on c51.  Perhaps means NV_DMA_TARGET_PCIE
+		 * on those cards? */
+		nv_wo32(dev, gpuobj, 0, NV_CLASS_DMA_IN_MEMORY |
+				       (1 << 12) /* PT present */ |
+				       (0 << 13) /* PT *not* linear */ |
+				       (NV_DMA_ACCESS_RW  << 14) |
+				       (NV_DMA_TARGET_PCI << 16));
+		nv_wo32(dev, gpuobj, 1, aper_size - 1);
+		for (i = 2; i < 2 + (aper_size >> 12); i++) {
+			nv_wo32(dev, gpuobj, i,
+				    dev_priv->gart_info.sg_dummy_bus | 3);
+		}
+	} else {
+		for (i = 0; i < obj_size; i += 8) {
+			nv_wo32(dev, gpuobj, (i+0)/4,
+				    dev_priv->gart_info.sg_dummy_bus | 0x21);
+			nv_wo32(dev, gpuobj, (i+4)/4, 0);
+		}
+	}
+	dev_priv->engine.instmem.finish_access(dev);
+
+	dev_priv->gart_info.type      = NOUVEAU_GART_SGDMA;
+	dev_priv->gart_info.aper_base = 0;
+	dev_priv->gart_info.aper_size = aper_size;
+	dev_priv->gart_info.sg_ctxdma = gpuobj;
+	return 0;
+}
+
+void
+nouveau_sgdma_takedown(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	if (dev_priv->gart_info.sg_dummy_page) {
+		pci_unmap_page(dev->pdev, dev_priv->gart_info.sg_dummy_bus,
+			       NV_CTXDMA_PAGE_SIZE, PCI_DMA_BIDIRECTIONAL);
+		unlock_page(dev_priv->gart_info.sg_dummy_page);
+		__free_page(dev_priv->gart_info.sg_dummy_page);
+		dev_priv->gart_info.sg_dummy_page = NULL;
+		dev_priv->gart_info.sg_dummy_bus = 0;
+	}
+
+	nouveau_gpuobj_del(dev, &dev_priv->gart_info.sg_ctxdma);
+}
+
+int
+nouveau_sgdma_get_page(struct drm_device *dev, uint32_t offset, uint32_t *page)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_gpuobj *gpuobj = dev_priv->gart_info.sg_ctxdma;
+	struct nouveau_instmem_engine *instmem = &dev_priv->engine.instmem;
+	int pte;
+
+	pte = (offset >> NV_CTXDMA_PAGE_SHIFT);
+	if (dev_priv->card_type < NV_50) {
+		instmem->prepare_access(dev, false);
+		*page = nv_ro32(dev, gpuobj, (pte + 2)) & ~NV_CTXDMA_PAGE_MASK;
+		instmem->finish_access(dev);
+		return 0;
+	}
+
+	NV_ERROR(dev, "Unimplemented on NV50\n");
+	return -EINVAL;
+}
diff --git a/drivers/gpu/drm/nouveau/nouveau_state.c b/drivers/gpu/drm/nouveau/nouveau_state.c
new file mode 100644
index 000000000000..eb8f084d5f53
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_state.c
@@ -0,0 +1,913 @@
+/*
+ * Copyright 2005 Stephane Marchesin
+ * Copyright 2008 Stuart Bennett
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include <linux/swab.h>
+#include "drmP.h"
+#include "drm.h"
+#include "drm_sarea.h"
+#include "drm_crtc_helper.h"
+#include <linux/vgaarb.h>
+#include <linux/vga_switcheroo.h>
+
+#include "nouveau_drv.h"
+#include "nouveau_drm.h"
+#include "nv50_display.h"
+
+static int nouveau_stub_init(struct drm_device *dev) { return 0; }
+static void nouveau_stub_takedown(struct drm_device *dev) {}
+
+static int nouveau_init_engine_ptrs(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_engine *engine = &dev_priv->engine;
+
+	switch (dev_priv->chipset & 0xf0) {
+	case 0x00:
+		engine->instmem.init		= nv04_instmem_init;
+		engine->instmem.takedown	= nv04_instmem_takedown;
+		engine->instmem.suspend		= nv04_instmem_suspend;
+		engine->instmem.resume		= nv04_instmem_resume;
+		engine->instmem.populate	= nv04_instmem_populate;
+		engine->instmem.clear		= nv04_instmem_clear;
+		engine->instmem.bind		= nv04_instmem_bind;
+		engine->instmem.unbind		= nv04_instmem_unbind;
+		engine->instmem.prepare_access	= nv04_instmem_prepare_access;
+		engine->instmem.finish_access	= nv04_instmem_finish_access;
+		engine->mc.init			= nv04_mc_init;
+		engine->mc.takedown		= nv04_mc_takedown;
+		engine->timer.init		= nv04_timer_init;
+		engine->timer.read		= nv04_timer_read;
+		engine->timer.takedown		= nv04_timer_takedown;
+		engine->fb.init			= nv04_fb_init;
+		engine->fb.takedown		= nv04_fb_takedown;
+		engine->graph.grclass		= nv04_graph_grclass;
+		engine->graph.init		= nv04_graph_init;
+		engine->graph.takedown		= nv04_graph_takedown;
+		engine->graph.fifo_access	= nv04_graph_fifo_access;
+		engine->graph.channel		= nv04_graph_channel;
+		engine->graph.create_context	= nv04_graph_create_context;
+		engine->graph.destroy_context	= nv04_graph_destroy_context;
+		engine->graph.load_context	= nv04_graph_load_context;
+		engine->graph.unload_context	= nv04_graph_unload_context;
+		engine->fifo.channels		= 16;
+		engine->fifo.init		= nv04_fifo_init;
+		engine->fifo.takedown		= nouveau_stub_takedown;
+		engine->fifo.disable		= nv04_fifo_disable;
+		engine->fifo.enable		= nv04_fifo_enable;
+		engine->fifo.reassign		= nv04_fifo_reassign;
+		engine->fifo.cache_flush	= nv04_fifo_cache_flush;
+		engine->fifo.cache_pull		= nv04_fifo_cache_pull;
+		engine->fifo.channel_id		= nv04_fifo_channel_id;
+		engine->fifo.create_context	= nv04_fifo_create_context;
+		engine->fifo.destroy_context	= nv04_fifo_destroy_context;
+		engine->fifo.load_context	= nv04_fifo_load_context;
+		engine->fifo.unload_context	= nv04_fifo_unload_context;
+		break;
+	case 0x10:
+		engine->instmem.init		= nv04_instmem_init;
+		engine->instmem.takedown	= nv04_instmem_takedown;
+		engine->instmem.suspend		= nv04_instmem_suspend;
+		engine->instmem.resume		= nv04_instmem_resume;
+		engine->instmem.populate	= nv04_instmem_populate;
+		engine->instmem.clear		= nv04_instmem_clear;
+		engine->instmem.bind		= nv04_instmem_bind;
+		engine->instmem.unbind		= nv04_instmem_unbind;
+		engine->instmem.prepare_access	= nv04_instmem_prepare_access;
+		engine->instmem.finish_access	= nv04_instmem_finish_access;
+		engine->mc.init			= nv04_mc_init;
+		engine->mc.takedown		= nv04_mc_takedown;
+		engine->timer.init		= nv04_timer_init;
+		engine->timer.read		= nv04_timer_read;
+		engine->timer.takedown		= nv04_timer_takedown;
+		engine->fb.init			= nv10_fb_init;
+		engine->fb.takedown		= nv10_fb_takedown;
+		engine->fb.set_region_tiling	= nv10_fb_set_region_tiling;
+		engine->graph.grclass		= nv10_graph_grclass;
+		engine->graph.init		= nv10_graph_init;
+		engine->graph.takedown		= nv10_graph_takedown;
+		engine->graph.channel		= nv10_graph_channel;
+		engine->graph.create_context	= nv10_graph_create_context;
+		engine->graph.destroy_context	= nv10_graph_destroy_context;
+		engine->graph.fifo_access	= nv04_graph_fifo_access;
+		engine->graph.load_context	= nv10_graph_load_context;
+		engine->graph.unload_context	= nv10_graph_unload_context;
+		engine->graph.set_region_tiling	= nv10_graph_set_region_tiling;
+		engine->fifo.channels		= 32;
+		engine->fifo.init		= nv10_fifo_init;
+		engine->fifo.takedown		= nouveau_stub_takedown;
+		engine->fifo.disable		= nv04_fifo_disable;
+		engine->fifo.enable		= nv04_fifo_enable;
+		engine->fifo.reassign		= nv04_fifo_reassign;
+		engine->fifo.cache_flush	= nv04_fifo_cache_flush;
+		engine->fifo.cache_pull		= nv04_fifo_cache_pull;
+		engine->fifo.channel_id		= nv10_fifo_channel_id;
+		engine->fifo.create_context	= nv10_fifo_create_context;
+		engine->fifo.destroy_context	= nv10_fifo_destroy_context;
+		engine->fifo.load_context	= nv10_fifo_load_context;
+		engine->fifo.unload_context	= nv10_fifo_unload_context;
+		break;
+	case 0x20:
+		engine->instmem.init		= nv04_instmem_init;
+		engine->instmem.takedown	= nv04_instmem_takedown;
+		engine->instmem.suspend		= nv04_instmem_suspend;
+		engine->instmem.resume		= nv04_instmem_resume;
+		engine->instmem.populate	= nv04_instmem_populate;
+		engine->instmem.clear		= nv04_instmem_clear;
+		engine->instmem.bind		= nv04_instmem_bind;
+		engine->instmem.unbind		= nv04_instmem_unbind;
+		engine->instmem.prepare_access	= nv04_instmem_prepare_access;
+		engine->instmem.finish_access	= nv04_instmem_finish_access;
+		engine->mc.init			= nv04_mc_init;
+		engine->mc.takedown		= nv04_mc_takedown;
+		engine->timer.init		= nv04_timer_init;
+		engine->timer.read		= nv04_timer_read;
+		engine->timer.takedown		= nv04_timer_takedown;
+		engine->fb.init			= nv10_fb_init;
+		engine->fb.takedown		= nv10_fb_takedown;
+		engine->fb.set_region_tiling	= nv10_fb_set_region_tiling;
+		engine->graph.grclass		= nv20_graph_grclass;
+		engine->graph.init		= nv20_graph_init;
+		engine->graph.takedown		= nv20_graph_takedown;
+		engine->graph.channel		= nv10_graph_channel;
+		engine->graph.create_context	= nv20_graph_create_context;
+		engine->graph.destroy_context	= nv20_graph_destroy_context;
+		engine->graph.fifo_access	= nv04_graph_fifo_access;
+		engine->graph.load_context	= nv20_graph_load_context;
+		engine->graph.unload_context	= nv20_graph_unload_context;
+		engine->graph.set_region_tiling	= nv20_graph_set_region_tiling;
+		engine->fifo.channels		= 32;
+		engine->fifo.init		= nv10_fifo_init;
+		engine->fifo.takedown		= nouveau_stub_takedown;
+		engine->fifo.disable		= nv04_fifo_disable;
+		engine->fifo.enable		= nv04_fifo_enable;
+		engine->fifo.reassign		= nv04_fifo_reassign;
+		engine->fifo.cache_flush	= nv04_fifo_cache_flush;
+		engine->fifo.cache_pull		= nv04_fifo_cache_pull;
+		engine->fifo.channel_id		= nv10_fifo_channel_id;
+		engine->fifo.create_context	= nv10_fifo_create_context;
+		engine->fifo.destroy_context	= nv10_fifo_destroy_context;
+		engine->fifo.load_context	= nv10_fifo_load_context;
+		engine->fifo.unload_context	= nv10_fifo_unload_context;
+		break;
+	case 0x30:
+		engine->instmem.init		= nv04_instmem_init;
+		engine->instmem.takedown	= nv04_instmem_takedown;
+		engine->instmem.suspend		= nv04_instmem_suspend;
+		engine->instmem.resume		= nv04_instmem_resume;
+		engine->instmem.populate	= nv04_instmem_populate;
+		engine->instmem.clear		= nv04_instmem_clear;
+		engine->instmem.bind		= nv04_instmem_bind;
+		engine->instmem.unbind		= nv04_instmem_unbind;
+		engine->instmem.prepare_access	= nv04_instmem_prepare_access;
+		engine->instmem.finish_access	= nv04_instmem_finish_access;
+		engine->mc.init			= nv04_mc_init;
+		engine->mc.takedown		= nv04_mc_takedown;
+		engine->timer.init		= nv04_timer_init;
+		engine->timer.read		= nv04_timer_read;
+		engine->timer.takedown		= nv04_timer_takedown;
+		engine->fb.init			= nv10_fb_init;
+		engine->fb.takedown		= nv10_fb_takedown;
+		engine->fb.set_region_tiling	= nv10_fb_set_region_tiling;
+		engine->graph.grclass		= nv30_graph_grclass;
+		engine->graph.init		= nv30_graph_init;
+		engine->graph.takedown		= nv20_graph_takedown;
+		engine->graph.fifo_access	= nv04_graph_fifo_access;
+		engine->graph.channel		= nv10_graph_channel;
+		engine->graph.create_context	= nv20_graph_create_context;
+		engine->graph.destroy_context	= nv20_graph_destroy_context;
+		engine->graph.load_context	= nv20_graph_load_context;
+		engine->graph.unload_context	= nv20_graph_unload_context;
+		engine->graph.set_region_tiling	= nv20_graph_set_region_tiling;
+		engine->fifo.channels		= 32;
+		engine->fifo.init		= nv10_fifo_init;
+		engine->fifo.takedown		= nouveau_stub_takedown;
+		engine->fifo.disable		= nv04_fifo_disable;
+		engine->fifo.enable		= nv04_fifo_enable;
+		engine->fifo.reassign		= nv04_fifo_reassign;
+		engine->fifo.cache_flush	= nv04_fifo_cache_flush;
+		engine->fifo.cache_pull		= nv04_fifo_cache_pull;
+		engine->fifo.channel_id		= nv10_fifo_channel_id;
+		engine->fifo.create_context	= nv10_fifo_create_context;
+		engine->fifo.destroy_context	= nv10_fifo_destroy_context;
+		engine->fifo.load_context	= nv10_fifo_load_context;
+		engine->fifo.unload_context	= nv10_fifo_unload_context;
+		break;
+	case 0x40:
+	case 0x60:
+		engine->instmem.init		= nv04_instmem_init;
+		engine->instmem.takedown	= nv04_instmem_takedown;
+		engine->instmem.suspend		= nv04_instmem_suspend;
+		engine->instmem.resume		= nv04_instmem_resume;
+		engine->instmem.populate	= nv04_instmem_populate;
+		engine->instmem.clear		= nv04_instmem_clear;
+		engine->instmem.bind		= nv04_instmem_bind;
+		engine->instmem.unbind		= nv04_instmem_unbind;
+		engine->instmem.prepare_access	= nv04_instmem_prepare_access;
+		engine->instmem.finish_access	= nv04_instmem_finish_access;
+		engine->mc.init			= nv40_mc_init;
+		engine->mc.takedown		= nv40_mc_takedown;
+		engine->timer.init		= nv04_timer_init;
+		engine->timer.read		= nv04_timer_read;
+		engine->timer.takedown		= nv04_timer_takedown;
+		engine->fb.init			= nv40_fb_init;
+		engine->fb.takedown		= nv40_fb_takedown;
+		engine->fb.set_region_tiling	= nv40_fb_set_region_tiling;
+		engine->graph.grclass		= nv40_graph_grclass;
+		engine->graph.init		= nv40_graph_init;
+		engine->graph.takedown		= nv40_graph_takedown;
+		engine->graph.fifo_access	= nv04_graph_fifo_access;
+		engine->graph.channel		= nv40_graph_channel;
+		engine->graph.create_context	= nv40_graph_create_context;
+		engine->graph.destroy_context	= nv40_graph_destroy_context;
+		engine->graph.load_context	= nv40_graph_load_context;
+		engine->graph.unload_context	= nv40_graph_unload_context;
+		engine->graph.set_region_tiling	= nv40_graph_set_region_tiling;
+		engine->fifo.channels		= 32;
+		engine->fifo.init		= nv40_fifo_init;
+		engine->fifo.takedown		= nouveau_stub_takedown;
+		engine->fifo.disable		= nv04_fifo_disable;
+		engine->fifo.enable		= nv04_fifo_enable;
+		engine->fifo.reassign		= nv04_fifo_reassign;
+		engine->fifo.cache_flush	= nv04_fifo_cache_flush;
+		engine->fifo.cache_pull		= nv04_fifo_cache_pull;
+		engine->fifo.channel_id		= nv10_fifo_channel_id;
+		engine->fifo.create_context	= nv40_fifo_create_context;
+		engine->fifo.destroy_context	= nv40_fifo_destroy_context;
+		engine->fifo.load_context	= nv40_fifo_load_context;
+		engine->fifo.unload_context	= nv40_fifo_unload_context;
+		break;
+	case 0x50:
+	case 0x80: /* gotta love NVIDIA's consistency.. */
+	case 0x90:
+	case 0xA0:
+		engine->instmem.init		= nv50_instmem_init;
+		engine->instmem.takedown	= nv50_instmem_takedown;
+		engine->instmem.suspend		= nv50_instmem_suspend;
+		engine->instmem.resume		= nv50_instmem_resume;
+		engine->instmem.populate	= nv50_instmem_populate;
+		engine->instmem.clear		= nv50_instmem_clear;
+		engine->instmem.bind		= nv50_instmem_bind;
+		engine->instmem.unbind		= nv50_instmem_unbind;
+		engine->instmem.prepare_access	= nv50_instmem_prepare_access;
+		engine->instmem.finish_access	= nv50_instmem_finish_access;
+		engine->mc.init			= nv50_mc_init;
+		engine->mc.takedown		= nv50_mc_takedown;
+		engine->timer.init		= nv04_timer_init;
+		engine->timer.read		= nv04_timer_read;
+		engine->timer.takedown		= nv04_timer_takedown;
+		engine->fb.init			= nouveau_stub_init;
+		engine->fb.takedown		= nouveau_stub_takedown;
+		engine->graph.grclass		= nv50_graph_grclass;
+		engine->graph.init		= nv50_graph_init;
+		engine->graph.takedown		= nv50_graph_takedown;
+		engine->graph.fifo_access	= nv50_graph_fifo_access;
+		engine->graph.channel		= nv50_graph_channel;
+		engine->graph.create_context	= nv50_graph_create_context;
+		engine->graph.destroy_context	= nv50_graph_destroy_context;
+		engine->graph.load_context	= nv50_graph_load_context;
+		engine->graph.unload_context	= nv50_graph_unload_context;
+		engine->fifo.channels		= 128;
+		engine->fifo.init		= nv50_fifo_init;
+		engine->fifo.takedown		= nv50_fifo_takedown;
+		engine->fifo.disable		= nv04_fifo_disable;
+		engine->fifo.enable		= nv04_fifo_enable;
+		engine->fifo.reassign		= nv04_fifo_reassign;
+		engine->fifo.channel_id		= nv50_fifo_channel_id;
+		engine->fifo.create_context	= nv50_fifo_create_context;
+		engine->fifo.destroy_context	= nv50_fifo_destroy_context;
+		engine->fifo.load_context	= nv50_fifo_load_context;
+		engine->fifo.unload_context	= nv50_fifo_unload_context;
+		break;
+	default:
+		NV_ERROR(dev, "NV%02x unsupported\n", dev_priv->chipset);
+		return 1;
+	}
+
+	return 0;
+}
+
+static unsigned int
+nouveau_vga_set_decode(void *priv, bool state)
+{
+	struct drm_device *dev = priv;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	if (dev_priv->chipset >= 0x40)
+		nv_wr32(dev, 0x88054, state);
+	else
+		nv_wr32(dev, 0x1854, state);
+
+	if (state)
+		return VGA_RSRC_LEGACY_IO | VGA_RSRC_LEGACY_MEM |
+		       VGA_RSRC_NORMAL_IO | VGA_RSRC_NORMAL_MEM;
+	else
+		return VGA_RSRC_NORMAL_IO | VGA_RSRC_NORMAL_MEM;
+}
+
+static int
+nouveau_card_init_channel(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_gpuobj *gpuobj;
+	int ret;
+
+	ret = nouveau_channel_alloc(dev, &dev_priv->channel,
+				    (struct drm_file *)-2,
+				    NvDmaFB, NvDmaTT);
+	if (ret)
+		return ret;
+
+	gpuobj = NULL;
+	ret = nouveau_gpuobj_dma_new(dev_priv->channel, NV_CLASS_DMA_IN_MEMORY,
+				     0, nouveau_mem_fb_amount(dev),
+				     NV_DMA_ACCESS_RW, NV_DMA_TARGET_VIDMEM,
+				     &gpuobj);
+	if (ret)
+		goto out_err;
+
+	ret = nouveau_gpuobj_ref_add(dev, dev_priv->channel, NvDmaVRAM,
+				     gpuobj, NULL);
+	if (ret)
+		goto out_err;
+
+	gpuobj = NULL;
+	ret = nouveau_gpuobj_gart_dma_new(dev_priv->channel, 0,
+					  dev_priv->gart_info.aper_size,
+					  NV_DMA_ACCESS_RW, &gpuobj, NULL);
+	if (ret)
+		goto out_err;
+
+	ret = nouveau_gpuobj_ref_add(dev, dev_priv->channel, NvDmaGART,
+				     gpuobj, NULL);
+	if (ret)
+		goto out_err;
+
+	return 0;
+out_err:
+	nouveau_gpuobj_del(dev, &gpuobj);
+	nouveau_channel_free(dev_priv->channel);
+	dev_priv->channel = NULL;
+	return ret;
+}
+
+static void nouveau_switcheroo_set_state(struct pci_dev *pdev,
+					 enum vga_switcheroo_state state)
+{
+	pm_message_t pmm = { .event = PM_EVENT_SUSPEND };
+	if (state == VGA_SWITCHEROO_ON) {
+		printk(KERN_ERR "VGA switcheroo: switched nouveau on\n");
+		nouveau_pci_resume(pdev);
+	} else {
+		printk(KERN_ERR "VGA switcheroo: switched nouveau off\n");
+		nouveau_pci_suspend(pdev, pmm);
+	}
+}
+
+static bool nouveau_switcheroo_can_switch(struct pci_dev *pdev)
+{
+	struct drm_device *dev = pci_get_drvdata(pdev);
+	bool can_switch;
+
+	spin_lock(&dev->count_lock);
+	can_switch = (dev->open_count == 0);
+	spin_unlock(&dev->count_lock);
+	return can_switch;
+}
+
+int
+nouveau_card_init(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_engine *engine;
+	int ret;
+
+	NV_DEBUG(dev, "prev state = %d\n", dev_priv->init_state);
+
+	if (dev_priv->init_state == NOUVEAU_CARD_INIT_DONE)
+		return 0;
+
+	vga_client_register(dev->pdev, dev, NULL, nouveau_vga_set_decode);
+	vga_switcheroo_register_client(dev->pdev, nouveau_switcheroo_set_state,
+				       nouveau_switcheroo_can_switch);
+
+	/* Initialise internal driver API hooks */
+	ret = nouveau_init_engine_ptrs(dev);
+	if (ret)
+		goto out;
+	engine = &dev_priv->engine;
+	dev_priv->init_state = NOUVEAU_CARD_INIT_FAILED;
+	spin_lock_init(&dev_priv->context_switch_lock);
+
+	/* Parse BIOS tables / Run init tables if card not POSTed */
+	if (drm_core_check_feature(dev, DRIVER_MODESET)) {
+		ret = nouveau_bios_init(dev);
+		if (ret)
+			goto out;
+	}
+
+	ret = nouveau_gpuobj_early_init(dev);
+	if (ret)
+		goto out_bios;
+
+	/* Initialise instance memory, must happen before mem_init so we
+	 * know exactly how much VRAM we're able to use for "normal"
+	 * purposes.
+	 */
+	ret = engine->instmem.init(dev);
+	if (ret)
+		goto out_gpuobj_early;
+
+	/* Setup the memory manager */
+	ret = nouveau_mem_init(dev);
+	if (ret)
+		goto out_instmem;
+
+	ret = nouveau_gpuobj_init(dev);
+	if (ret)
+		goto out_mem;
+
+	/* PMC */
+	ret = engine->mc.init(dev);
+	if (ret)
+		goto out_gpuobj;
+
+	/* PTIMER */
+	ret = engine->timer.init(dev);
+	if (ret)
+		goto out_mc;
+
+	/* PFB */
+	ret = engine->fb.init(dev);
+	if (ret)
+		goto out_timer;
+
+	if (nouveau_noaccel)
+		engine->graph.accel_blocked = true;
+	else {
+		/* PGRAPH */
+		ret = engine->graph.init(dev);
+		if (ret)
+			goto out_fb;
+
+		/* PFIFO */
+		ret = engine->fifo.init(dev);
+		if (ret)
+			goto out_graph;
+	}
+
+	/* this call irq_preinstall, register irq handler and
+	 * call irq_postinstall
+	 */
+	ret = drm_irq_install(dev);
+	if (ret)
+		goto out_fifo;
+
+	ret = drm_vblank_init(dev, 0);
+	if (ret)
+		goto out_irq;
+
+	/* what about PVIDEO/PCRTC/PRAMDAC etc? */
+
+	if (!engine->graph.accel_blocked) {
+		ret = nouveau_card_init_channel(dev);
+		if (ret)
+			goto out_irq;
+	}
+
+	if (drm_core_check_feature(dev, DRIVER_MODESET)) {
+		if (dev_priv->card_type >= NV_50)
+			ret = nv50_display_create(dev);
+		else
+			ret = nv04_display_create(dev);
+		if (ret)
+			goto out_irq;
+	}
+
+	ret = nouveau_backlight_init(dev);
+	if (ret)
+		NV_ERROR(dev, "Error %d registering backlight\n", ret);
+
+	dev_priv->init_state = NOUVEAU_CARD_INIT_DONE;
+
+	if (drm_core_check_feature(dev, DRIVER_MODESET))
+		drm_helper_initial_config(dev);
+
+	return 0;
+
+out_irq:
+	drm_irq_uninstall(dev);
+out_fifo:
+	if (!nouveau_noaccel)
+		engine->fifo.takedown(dev);
+out_graph:
+	if (!nouveau_noaccel)
+		engine->graph.takedown(dev);
+out_fb:
+	engine->fb.takedown(dev);
+out_timer:
+	engine->timer.takedown(dev);
+out_mc:
+	engine->mc.takedown(dev);
+out_gpuobj:
+	nouveau_gpuobj_takedown(dev);
+out_mem:
+	nouveau_mem_close(dev);
+out_instmem:
+	engine->instmem.takedown(dev);
+out_gpuobj_early:
+	nouveau_gpuobj_late_takedown(dev);
+out_bios:
+	nouveau_bios_takedown(dev);
+out:
+	vga_client_register(dev->pdev, NULL, NULL, NULL);
+	return ret;
+}
+
+static void nouveau_card_takedown(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_engine *engine = &dev_priv->engine;
+
+	NV_DEBUG(dev, "prev state = %d\n", dev_priv->init_state);
+
+	if (dev_priv->init_state != NOUVEAU_CARD_INIT_DOWN) {
+		nouveau_backlight_exit(dev);
+
+		if (dev_priv->channel) {
+			nouveau_channel_free(dev_priv->channel);
+			dev_priv->channel = NULL;
+		}
+
+		if (!nouveau_noaccel) {
+			engine->fifo.takedown(dev);
+			engine->graph.takedown(dev);
+		}
+		engine->fb.takedown(dev);
+		engine->timer.takedown(dev);
+		engine->mc.takedown(dev);
+
+		mutex_lock(&dev->struct_mutex);
+		ttm_bo_clean_mm(&dev_priv->ttm.bdev, TTM_PL_VRAM);
+		ttm_bo_clean_mm(&dev_priv->ttm.bdev, TTM_PL_TT);
+		mutex_unlock(&dev->struct_mutex);
+		nouveau_sgdma_takedown(dev);
+
+		nouveau_gpuobj_takedown(dev);
+		nouveau_mem_close(dev);
+		engine->instmem.takedown(dev);
+
+		if (drm_core_check_feature(dev, DRIVER_MODESET))
+			drm_irq_uninstall(dev);
+
+		nouveau_gpuobj_late_takedown(dev);
+		nouveau_bios_takedown(dev);
+
+		vga_client_register(dev->pdev, NULL, NULL, NULL);
+
+		dev_priv->init_state = NOUVEAU_CARD_INIT_DOWN;
+	}
+}
+
+/* here a client dies, release the stuff that was allocated for its
+ * file_priv */
+void nouveau_preclose(struct drm_device *dev, struct drm_file *file_priv)
+{
+	nouveau_channel_cleanup(dev, file_priv);
+}
+
+/* first module load, setup the mmio/fb mapping */
+/* KMS: we need mmio at load time, not when the first drm client opens. */
+int nouveau_firstopen(struct drm_device *dev)
+{
+	return 0;
+}
+
+/* if we have an OF card, copy vbios to RAMIN */
+static void nouveau_OF_copy_vbios_to_ramin(struct drm_device *dev)
+{
+#if defined(__powerpc__)
+	int size, i;
+	const uint32_t *bios;
+	struct device_node *dn = pci_device_to_OF_node(dev->pdev);
+	if (!dn) {
+		NV_INFO(dev, "Unable to get the OF node\n");
+		return;
+	}
+
+	bios = of_get_property(dn, "NVDA,BMP", &size);
+	if (bios) {
+		for (i = 0; i < size; i += 4)
+			nv_wi32(dev, i, bios[i/4]);
+		NV_INFO(dev, "OF bios successfully copied (%d bytes)\n", size);
+	} else {
+		NV_INFO(dev, "Unable to get the OF bios\n");
+	}
+#endif
+}
+
+int nouveau_load(struct drm_device *dev, unsigned long flags)
+{
+	struct drm_nouveau_private *dev_priv;
+	uint32_t reg0;
+	resource_size_t mmio_start_offs;
+
+	dev_priv = kzalloc(sizeof(*dev_priv), GFP_KERNEL);
+	if (!dev_priv)
+		return -ENOMEM;
+	dev->dev_private = dev_priv;
+	dev_priv->dev = dev;
+
+	dev_priv->flags = flags & NOUVEAU_FLAGS;
+	dev_priv->init_state = NOUVEAU_CARD_INIT_DOWN;
+
+	NV_DEBUG(dev, "vendor: 0x%X device: 0x%X class: 0x%X\n",
+		 dev->pci_vendor, dev->pci_device, dev->pdev->class);
+
+	dev_priv->wq = create_workqueue("nouveau");
+	if (!dev_priv->wq)
+		return -EINVAL;
+
+	/* resource 0 is mmio regs */
+	/* resource 1 is linear FB */
+	/* resource 2 is RAMIN (mmio regs + 0x1000000) */
+	/* resource 6 is bios */
+
+	/* map the mmio regs */
+	mmio_start_offs = pci_resource_start(dev->pdev, 0);
+	dev_priv->mmio = ioremap(mmio_start_offs, 0x00800000);
+	if (!dev_priv->mmio) {
+		NV_ERROR(dev, "Unable to initialize the mmio mapping. "
+			 "Please report your setup to " DRIVER_EMAIL "\n");
+		return -EINVAL;
+	}
+	NV_DEBUG(dev, "regs mapped ok at 0x%llx\n",
+					(unsigned long long)mmio_start_offs);
+
+#ifdef __BIG_ENDIAN
+	/* Put the card in BE mode if it's not */
+	if (nv_rd32(dev, NV03_PMC_BOOT_1))
+		nv_wr32(dev, NV03_PMC_BOOT_1, 0x00000001);
+
+	DRM_MEMORYBARRIER();
+#endif
+
+	/* Time to determine the card architecture */
+	reg0 = nv_rd32(dev, NV03_PMC_BOOT_0);
+
+	/* We're dealing with >=NV10 */
+	if ((reg0 & 0x0f000000) > 0) {
+		/* Bit 27-20 contain the architecture in hex */
+		dev_priv->chipset = (reg0 & 0xff00000) >> 20;
+	/* NV04 or NV05 */
+	} else if ((reg0 & 0xff00fff0) == 0x20004000) {
+		if (reg0 & 0x00f00000)
+			dev_priv->chipset = 0x05;
+		else
+			dev_priv->chipset = 0x04;
+	} else
+		dev_priv->chipset = 0xff;
+
+	switch (dev_priv->chipset & 0xf0) {
+	case 0x00:
+	case 0x10:
+	case 0x20:
+	case 0x30:
+		dev_priv->card_type = dev_priv->chipset & 0xf0;
+		break;
+	case 0x40:
+	case 0x60:
+		dev_priv->card_type = NV_40;
+		break;
+	case 0x50:
+	case 0x80:
+	case 0x90:
+	case 0xa0:
+		dev_priv->card_type = NV_50;
+		break;
+	default:
+		NV_INFO(dev, "Unsupported chipset 0x%08x\n", reg0);
+		return -EINVAL;
+	}
+
+	NV_INFO(dev, "Detected an NV%2x generation card (0x%08x)\n",
+		dev_priv->card_type, reg0);
+
+	/* map larger RAMIN aperture on NV40 cards */
+	dev_priv->ramin  = NULL;
+	if (dev_priv->card_type >= NV_40) {
+		int ramin_bar = 2;
+		if (pci_resource_len(dev->pdev, ramin_bar) == 0)
+			ramin_bar = 3;
+
+		dev_priv->ramin_size = pci_resource_len(dev->pdev, ramin_bar);
+		dev_priv->ramin = ioremap(
+				pci_resource_start(dev->pdev, ramin_bar),
+				dev_priv->ramin_size);
+		if (!dev_priv->ramin) {
+			NV_ERROR(dev, "Failed to init RAMIN mapping, "
+				      "limited instance memory available\n");
+		}
+	}
+
+	/* On older cards (or if the above failed), create a map covering
+	 * the BAR0 PRAMIN aperture */
+	if (!dev_priv->ramin) {
+		dev_priv->ramin_size = 1 * 1024 * 1024;
+		dev_priv->ramin = ioremap(mmio_start_offs + NV_RAMIN,
+							dev_priv->ramin_size);
+		if (!dev_priv->ramin) {
+			NV_ERROR(dev, "Failed to map BAR0 PRAMIN.\n");
+			return -ENOMEM;
+		}
+	}
+
+	nouveau_OF_copy_vbios_to_ramin(dev);
+
+	/* Special flags */
+	if (dev->pci_device == 0x01a0)
+		dev_priv->flags |= NV_NFORCE;
+	else if (dev->pci_device == 0x01f0)
+		dev_priv->flags |= NV_NFORCE2;
+
+	/* For kernel modesetting, init card now and bring up fbcon */
+	if (drm_core_check_feature(dev, DRIVER_MODESET)) {
+		int ret = nouveau_card_init(dev);
+		if (ret)
+			return ret;
+	}
+
+	return 0;
+}
+
+static void nouveau_close(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	/* In the case of an error dev_priv may not be allocated yet */
+	if (dev_priv)
+		nouveau_card_takedown(dev);
+}
+
+/* KMS: we need mmio at load time, not when the first drm client opens. */
+void nouveau_lastclose(struct drm_device *dev)
+{
+	if (drm_core_check_feature(dev, DRIVER_MODESET))
+		return;
+
+	nouveau_close(dev);
+}
+
+int nouveau_unload(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	if (drm_core_check_feature(dev, DRIVER_MODESET)) {
+		if (dev_priv->card_type >= NV_50)
+			nv50_display_destroy(dev);
+		else
+			nv04_display_destroy(dev);
+		nouveau_close(dev);
+	}
+
+	iounmap(dev_priv->mmio);
+	iounmap(dev_priv->ramin);
+
+	kfree(dev_priv);
+	dev->dev_private = NULL;
+	return 0;
+}
+
+int nouveau_ioctl_getparam(struct drm_device *dev, void *data,
+						struct drm_file *file_priv)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct drm_nouveau_getparam *getparam = data;
+
+	NOUVEAU_CHECK_INITIALISED_WITH_RETURN;
+
+	switch (getparam->param) {
+	case NOUVEAU_GETPARAM_CHIPSET_ID:
+		getparam->value = dev_priv->chipset;
+		break;
+	case NOUVEAU_GETPARAM_PCI_VENDOR:
+		getparam->value = dev->pci_vendor;
+		break;
+	case NOUVEAU_GETPARAM_PCI_DEVICE:
+		getparam->value = dev->pci_device;
+		break;
+	case NOUVEAU_GETPARAM_BUS_TYPE:
+		if (drm_device_is_agp(dev))
+			getparam->value = NV_AGP;
+		else if (drm_device_is_pcie(dev))
+			getparam->value = NV_PCIE;
+		else
+			getparam->value = NV_PCI;
+		break;
+	case NOUVEAU_GETPARAM_FB_PHYSICAL:
+		getparam->value = dev_priv->fb_phys;
+		break;
+	case NOUVEAU_GETPARAM_AGP_PHYSICAL:
+		getparam->value = dev_priv->gart_info.aper_base;
+		break;
+	case NOUVEAU_GETPARAM_PCI_PHYSICAL:
+		if (dev->sg) {
+			getparam->value = (unsigned long)dev->sg->virtual;
+		} else {
+			NV_ERROR(dev, "Requested PCIGART address, "
+					"while no PCIGART was created\n");
+			return -EINVAL;
+		}
+		break;
+	case NOUVEAU_GETPARAM_FB_SIZE:
+		getparam->value = dev_priv->fb_available_size;
+		break;
+	case NOUVEAU_GETPARAM_AGP_SIZE:
+		getparam->value = dev_priv->gart_info.aper_size;
+		break;
+	case NOUVEAU_GETPARAM_VM_VRAM_BASE:
+		getparam->value = dev_priv->vm_vram_base;
+		break;
+	case NOUVEAU_GETPARAM_GRAPH_UNITS:
+		/* NV40 and NV50 versions are quite different, but register
+		 * address is the same. User is supposed to know the card
+		 * family anyway... */
+		if (dev_priv->chipset >= 0x40) {
+			getparam->value = nv_rd32(dev, NV40_PMC_GRAPH_UNITS);
+			break;
+		}
+		/* FALLTHRU */
+	default:
+		NV_ERROR(dev, "unknown parameter %lld\n", getparam->param);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+int
+nouveau_ioctl_setparam(struct drm_device *dev, void *data,
+		       struct drm_file *file_priv)
+{
+	struct drm_nouveau_setparam *setparam = data;
+
+	NOUVEAU_CHECK_INITIALISED_WITH_RETURN;
+
+	switch (setparam->param) {
+	default:
+		NV_ERROR(dev, "unknown parameter %lld\n", setparam->param);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+/* Wait until (value(reg) & mask) == val, up until timeout has hit */
+bool nouveau_wait_until(struct drm_device *dev, uint64_t timeout,
+			uint32_t reg, uint32_t mask, uint32_t val)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_timer_engine *ptimer = &dev_priv->engine.timer;
+	uint64_t start = ptimer->read(dev);
+
+	do {
+		if ((nv_rd32(dev, reg) & mask) == val)
+			return true;
+	} while (ptimer->read(dev) - start < timeout);
+
+	return false;
+}
+
+/* Waits for PGRAPH to go completely idle */
+bool nouveau_wait_for_idle(struct drm_device *dev)
+{
+	if (!nv_wait(NV04_PGRAPH_STATUS, 0xffffffff, 0x00000000)) {
+		NV_ERROR(dev, "PGRAPH idle timed out with status 0x%08x\n",
+			 nv_rd32(dev, NV04_PGRAPH_STATUS));
+		return false;
+	}
+
+	return true;
+}
+
diff --git a/drivers/gpu/drm/nouveau/nouveau_ttm.c b/drivers/gpu/drm/nouveau/nouveau_ttm.c
new file mode 100644
index 000000000000..c385d50f041b
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_ttm.c
@@ -0,0 +1,103 @@
+/*
+ * Copyright (c) 2007-2008 Tungsten Graphics, Inc., Cedar Park, TX., USA,
+ * All Rights Reserved.
+ * Copyright (c) 2009 VMware, Inc., Palo Alto, CA., USA,
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sub license,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "drmP.h"
+
+#include "nouveau_drv.h"
+
+int
+nouveau_ttm_mmap(struct file *filp, struct vm_area_struct *vma)
+{
+	struct drm_file *file_priv = filp->private_data;
+	struct drm_nouveau_private *dev_priv =
+		file_priv->minor->dev->dev_private;
+
+	if (unlikely(vma->vm_pgoff < DRM_FILE_PAGE_OFFSET))
+		return drm_mmap(filp, vma);
+
+	return ttm_bo_mmap(filp, vma, &dev_priv->ttm.bdev);
+}
+
+static int
+nouveau_ttm_mem_global_init(struct ttm_global_reference *ref)
+{
+	return ttm_mem_global_init(ref->object);
+}
+
+static void
+nouveau_ttm_mem_global_release(struct ttm_global_reference *ref)
+{
+	ttm_mem_global_release(ref->object);
+}
+
+int
+nouveau_ttm_global_init(struct drm_nouveau_private *dev_priv)
+{
+	struct ttm_global_reference *global_ref;
+	int ret;
+
+	global_ref = &dev_priv->ttm.mem_global_ref;
+	global_ref->global_type = TTM_GLOBAL_TTM_MEM;
+	global_ref->size = sizeof(struct ttm_mem_global);
+	global_ref->init = &nouveau_ttm_mem_global_init;
+	global_ref->release = &nouveau_ttm_mem_global_release;
+
+	ret = ttm_global_item_ref(global_ref);
+	if (unlikely(ret != 0)) {
+		DRM_ERROR("Failed setting up TTM memory accounting\n");
+		dev_priv->ttm.mem_global_ref.release = NULL;
+		return ret;
+	}
+
+	dev_priv->ttm.bo_global_ref.mem_glob = global_ref->object;
+	global_ref = &dev_priv->ttm.bo_global_ref.ref;
+	global_ref->global_type = TTM_GLOBAL_TTM_BO;
+	global_ref->size = sizeof(struct ttm_bo_global);
+	global_ref->init = &ttm_bo_global_init;
+	global_ref->release = &ttm_bo_global_release;
+
+	ret = ttm_global_item_ref(global_ref);
+	if (unlikely(ret != 0)) {
+		DRM_ERROR("Failed setting up TTM BO subsystem\n");
+		ttm_global_item_unref(&dev_priv->ttm.mem_global_ref);
+		dev_priv->ttm.mem_global_ref.release = NULL;
+		return ret;
+	}
+
+	return 0;
+}
+
+void
+nouveau_ttm_global_release(struct drm_nouveau_private *dev_priv)
+{
+	if (dev_priv->ttm.mem_global_ref.release == NULL)
+		return;
+
+	ttm_global_item_unref(&dev_priv->ttm.bo_global_ref.ref);
+	ttm_global_item_unref(&dev_priv->ttm.mem_global_ref);
+	dev_priv->ttm.mem_global_ref.release = NULL;
+}
+
diff --git a/drivers/gpu/drm/nouveau/nv04_crtc.c b/drivers/gpu/drm/nouveau/nv04_crtc.c
new file mode 100644
index 000000000000..a1d1ebb073d9
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv04_crtc.c
@@ -0,0 +1,1000 @@
+/*
+ * Copyright 1993-2003 NVIDIA, Corporation
+ * Copyright 2006 Dave Airlie
+ * Copyright 2007 Maarten Maathuis
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include "drmP.h"
+#include "drm_crtc_helper.h"
+
+#include "nouveau_drv.h"
+#include "nouveau_encoder.h"
+#include "nouveau_connector.h"
+#include "nouveau_crtc.h"
+#include "nouveau_fb.h"
+#include "nouveau_hw.h"
+#include "nvreg.h"
+
+static int
+nv04_crtc_mode_set_base(struct drm_crtc *crtc, int x, int y,
+			struct drm_framebuffer *old_fb);
+
+static void
+crtc_wr_cio_state(struct drm_crtc *crtc, struct nv04_crtc_reg *crtcstate, int index)
+{
+	NVWriteVgaCrtc(crtc->dev, nouveau_crtc(crtc)->index, index,
+		       crtcstate->CRTC[index]);
+}
+
+static void nv_crtc_set_digital_vibrance(struct drm_crtc *crtc, int level)
+{
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+	struct drm_nouveau_private *dev_priv = crtc->dev->dev_private;
+	struct nv04_crtc_reg *regp = &dev_priv->mode_reg.crtc_reg[nv_crtc->index];
+
+	regp->CRTC[NV_CIO_CRE_CSB] = nv_crtc->saturation = level;
+	if (nv_crtc->saturation && nv_gf4_disp_arch(crtc->dev)) {
+		regp->CRTC[NV_CIO_CRE_CSB] = 0x80;
+		regp->CRTC[NV_CIO_CRE_5B] = nv_crtc->saturation << 2;
+		crtc_wr_cio_state(crtc, regp, NV_CIO_CRE_5B);
+	}
+	crtc_wr_cio_state(crtc, regp, NV_CIO_CRE_CSB);
+}
+
+static void nv_crtc_set_image_sharpening(struct drm_crtc *crtc, int level)
+{
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+	struct drm_nouveau_private *dev_priv = crtc->dev->dev_private;
+	struct nv04_crtc_reg *regp = &dev_priv->mode_reg.crtc_reg[nv_crtc->index];
+
+	nv_crtc->sharpness = level;
+	if (level < 0)	/* blur is in hw range 0x3f -> 0x20 */
+		level += 0x40;
+	regp->ramdac_634 = level;
+	NVWriteRAMDAC(crtc->dev, nv_crtc->index, NV_PRAMDAC_634, regp->ramdac_634);
+}
+
+#define PLLSEL_VPLL1_MASK				\
+	(NV_PRAMDAC_PLL_COEFF_SELECT_SOURCE_PROG_VPLL	\
+	 | NV_PRAMDAC_PLL_COEFF_SELECT_VCLK_RATIO_DB2)
+#define PLLSEL_VPLL2_MASK				\
+	(NV_PRAMDAC_PLL_COEFF_SELECT_PLL_SOURCE_VPLL2		\
+	 | NV_PRAMDAC_PLL_COEFF_SELECT_VCLK2_RATIO_DB2)
+#define PLLSEL_TV_MASK					\
+	(NV_PRAMDAC_PLL_COEFF_SELECT_TV_VSCLK1		\
+	 | NV_PRAMDAC_PLL_COEFF_SELECT_TV_PCLK1		\
+	 | NV_PRAMDAC_PLL_COEFF_SELECT_TV_VSCLK2	\
+	 | NV_PRAMDAC_PLL_COEFF_SELECT_TV_PCLK2)
+
+/* NV4x 0x40.. pll notes:
+ * gpu pll: 0x4000 + 0x4004
+ * ?gpu? pll: 0x4008 + 0x400c
+ * vpll1: 0x4010 + 0x4014
+ * vpll2: 0x4018 + 0x401c
+ * mpll: 0x4020 + 0x4024
+ * mpll: 0x4038 + 0x403c
+ *
+ * the first register of each pair has some unknown details:
+ * bits 0-7: redirected values from elsewhere? (similar to PLL_SETUP_CONTROL?)
+ * bits 20-23: (mpll) something to do with post divider?
+ * bits 28-31: related to single stage mode? (bit 8/12)
+ */
+
+static void nv_crtc_calc_state_ext(struct drm_crtc *crtc, struct drm_display_mode * mode, int dot_clock)
+{
+	struct drm_device *dev = crtc->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+	struct nv04_mode_state *state = &dev_priv->mode_reg;
+	struct nv04_crtc_reg *regp = &state->crtc_reg[nv_crtc->index];
+	struct nouveau_pll_vals *pv = &regp->pllvals;
+	struct pll_lims pll_lim;
+
+	if (get_pll_limits(dev, nv_crtc->index ? VPLL2 : VPLL1, &pll_lim))
+		return;
+
+	/* NM2 == 0 is used to determine single stage mode on two stage plls */
+	pv->NM2 = 0;
+
+	/* for newer nv4x the blob uses only the first stage of the vpll below a
+	 * certain clock.  for a certain nv4b this is 150MHz.  since the max
+	 * output frequency of the first stage for this card is 300MHz, it is
+	 * assumed the threshold is given by vco1 maxfreq/2
+	 */
+	/* for early nv4x, specifically nv40 and *some* nv43 (devids 0 and 6,
+	 * not 8, others unknown), the blob always uses both plls.  no problem
+	 * has yet been observed in allowing the use a single stage pll on all
+	 * nv43 however.  the behaviour of single stage use is untested on nv40
+	 */
+	if (dev_priv->chipset > 0x40 && dot_clock <= (pll_lim.vco1.maxfreq / 2))
+		memset(&pll_lim.vco2, 0, sizeof(pll_lim.vco2));
+
+	if (!nouveau_calc_pll_mnp(dev, &pll_lim, dot_clock, pv))
+		return;
+
+	state->pllsel &= PLLSEL_VPLL1_MASK | PLLSEL_VPLL2_MASK | PLLSEL_TV_MASK;
+
+	/* The blob uses this always, so let's do the same */
+	if (dev_priv->card_type == NV_40)
+		state->pllsel |= NV_PRAMDAC_PLL_COEFF_SELECT_USE_VPLL2_TRUE;
+	/* again nv40 and some nv43 act more like nv3x as described above */
+	if (dev_priv->chipset < 0x41)
+		state->pllsel |= NV_PRAMDAC_PLL_COEFF_SELECT_SOURCE_PROG_MPLL |
+				 NV_PRAMDAC_PLL_COEFF_SELECT_SOURCE_PROG_NVPLL;
+	state->pllsel |= nv_crtc->index ? PLLSEL_VPLL2_MASK : PLLSEL_VPLL1_MASK;
+
+	if (pv->NM2)
+		NV_DEBUG_KMS(dev, "vpll: n1 %d n2 %d m1 %d m2 %d log2p %d\n",
+			 pv->N1, pv->N2, pv->M1, pv->M2, pv->log2P);
+	else
+		NV_DEBUG_KMS(dev, "vpll: n %d m %d log2p %d\n",
+			 pv->N1, pv->M1, pv->log2P);
+
+	nv_crtc->cursor.set_offset(nv_crtc, nv_crtc->cursor.offset);
+}
+
+static void
+nv_crtc_dpms(struct drm_crtc *crtc, int mode)
+{
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+	struct drm_device *dev = crtc->dev;
+	unsigned char seq1 = 0, crtc17 = 0;
+	unsigned char crtc1A;
+
+	NV_DEBUG_KMS(dev, "Setting dpms mode %d on CRTC %d\n", mode,
+							nv_crtc->index);
+
+	if (nv_crtc->last_dpms == mode) /* Don't do unnecesary mode changes. */
+		return;
+
+	nv_crtc->last_dpms = mode;
+
+	if (nv_two_heads(dev))
+		NVSetOwner(dev, nv_crtc->index);
+
+	/* nv4ref indicates these two RPC1 bits inhibit h/v sync */
+	crtc1A = NVReadVgaCrtc(dev, nv_crtc->index,
+					NV_CIO_CRE_RPC1_INDEX) & ~0xC0;
+	switch (mode) {
+	case DRM_MODE_DPMS_STANDBY:
+		/* Screen: Off; HSync: Off, VSync: On -- Not Supported */
+		seq1 = 0x20;
+		crtc17 = 0x80;
+		crtc1A |= 0x80;
+		break;
+	case DRM_MODE_DPMS_SUSPEND:
+		/* Screen: Off; HSync: On, VSync: Off -- Not Supported */
+		seq1 = 0x20;
+		crtc17 = 0x80;
+		crtc1A |= 0x40;
+		break;
+	case DRM_MODE_DPMS_OFF:
+		/* Screen: Off; HSync: Off, VSync: Off */
+		seq1 = 0x20;
+		crtc17 = 0x00;
+		crtc1A |= 0xC0;
+		break;
+	case DRM_MODE_DPMS_ON:
+	default:
+		/* Screen: On; HSync: On, VSync: On */
+		seq1 = 0x00;
+		crtc17 = 0x80;
+		break;
+	}
+
+	NVVgaSeqReset(dev, nv_crtc->index, true);
+	/* Each head has it's own sequencer, so we can turn it off when we want */
+	seq1 |= (NVReadVgaSeq(dev, nv_crtc->index, NV_VIO_SR_CLOCK_INDEX) & ~0x20);
+	NVWriteVgaSeq(dev, nv_crtc->index, NV_VIO_SR_CLOCK_INDEX, seq1);
+	crtc17 |= (NVReadVgaCrtc(dev, nv_crtc->index, NV_CIO_CR_MODE_INDEX) & ~0x80);
+	mdelay(10);
+	NVWriteVgaCrtc(dev, nv_crtc->index, NV_CIO_CR_MODE_INDEX, crtc17);
+	NVVgaSeqReset(dev, nv_crtc->index, false);
+
+	NVWriteVgaCrtc(dev, nv_crtc->index, NV_CIO_CRE_RPC1_INDEX, crtc1A);
+}
+
+static bool
+nv_crtc_mode_fixup(struct drm_crtc *crtc, struct drm_display_mode *mode,
+		   struct drm_display_mode *adjusted_mode)
+{
+	return true;
+}
+
+static void
+nv_crtc_mode_set_vga(struct drm_crtc *crtc, struct drm_display_mode *mode)
+{
+	struct drm_device *dev = crtc->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+	struct nv04_crtc_reg *regp = &dev_priv->mode_reg.crtc_reg[nv_crtc->index];
+	struct drm_framebuffer *fb = crtc->fb;
+
+	/* Calculate our timings */
+	int horizDisplay	= (mode->crtc_hdisplay >> 3) 	- 1;
+	int horizStart		= (mode->crtc_hsync_start >> 3) 	- 1;
+	int horizEnd		= (mode->crtc_hsync_end >> 3) 	- 1;
+	int horizTotal		= (mode->crtc_htotal >> 3)		- 5;
+	int horizBlankStart	= (mode->crtc_hdisplay >> 3)		- 1;
+	int horizBlankEnd	= (mode->crtc_htotal >> 3)		- 1;
+	int vertDisplay		= mode->crtc_vdisplay			- 1;
+	int vertStart		= mode->crtc_vsync_start 		- 1;
+	int vertEnd		= mode->crtc_vsync_end			- 1;
+	int vertTotal		= mode->crtc_vtotal 			- 2;
+	int vertBlankStart	= mode->crtc_vdisplay 			- 1;
+	int vertBlankEnd	= mode->crtc_vtotal			- 1;
+
+	struct drm_encoder *encoder;
+	bool fp_output = false;
+
+	list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) {
+		struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+
+		if (encoder->crtc == crtc &&
+		    (nv_encoder->dcb->type == OUTPUT_LVDS ||
+		     nv_encoder->dcb->type == OUTPUT_TMDS))
+			fp_output = true;
+	}
+
+	if (fp_output) {
+		vertStart = vertTotal - 3;
+		vertEnd = vertTotal - 2;
+		vertBlankStart = vertStart;
+		horizStart = horizTotal - 5;
+		horizEnd = horizTotal - 2;
+		horizBlankEnd = horizTotal + 4;
+#if 0
+		if (dev->overlayAdaptor && dev_priv->card_type >= NV_10)
+			/* This reportedly works around some video overlay bandwidth problems */
+			horizTotal += 2;
+#endif
+	}
+
+	if (mode->flags & DRM_MODE_FLAG_INTERLACE)
+		vertTotal |= 1;
+
+#if 0
+	ErrorF("horizDisplay: 0x%X \n", horizDisplay);
+	ErrorF("horizStart: 0x%X \n", horizStart);
+	ErrorF("horizEnd: 0x%X \n", horizEnd);
+	ErrorF("horizTotal: 0x%X \n", horizTotal);
+	ErrorF("horizBlankStart: 0x%X \n", horizBlankStart);
+	ErrorF("horizBlankEnd: 0x%X \n", horizBlankEnd);
+	ErrorF("vertDisplay: 0x%X \n", vertDisplay);
+	ErrorF("vertStart: 0x%X \n", vertStart);
+	ErrorF("vertEnd: 0x%X \n", vertEnd);
+	ErrorF("vertTotal: 0x%X \n", vertTotal);
+	ErrorF("vertBlankStart: 0x%X \n", vertBlankStart);
+	ErrorF("vertBlankEnd: 0x%X \n", vertBlankEnd);
+#endif
+
+	/*
+	* compute correct Hsync & Vsync polarity
+	*/
+	if ((mode->flags & (DRM_MODE_FLAG_PHSYNC | DRM_MODE_FLAG_NHSYNC))
+		&& (mode->flags & (DRM_MODE_FLAG_PVSYNC | DRM_MODE_FLAG_NVSYNC))) {
+
+		regp->MiscOutReg = 0x23;
+		if (mode->flags & DRM_MODE_FLAG_NHSYNC)
+			regp->MiscOutReg |= 0x40;
+		if (mode->flags & DRM_MODE_FLAG_NVSYNC)
+			regp->MiscOutReg |= 0x80;
+	} else {
+		int vdisplay = mode->vdisplay;
+		if (mode->flags & DRM_MODE_FLAG_DBLSCAN)
+			vdisplay *= 2;
+		if (mode->vscan > 1)
+			vdisplay *= mode->vscan;
+		if (vdisplay < 400)
+			regp->MiscOutReg = 0xA3;	/* +hsync -vsync */
+		else if (vdisplay < 480)
+			regp->MiscOutReg = 0x63;	/* -hsync +vsync */
+		else if (vdisplay < 768)
+			regp->MiscOutReg = 0xE3;	/* -hsync -vsync */
+		else
+			regp->MiscOutReg = 0x23;	/* +hsync +vsync */
+	}
+
+	regp->MiscOutReg |= (mode->clock_index & 0x03) << 2;
+
+	/*
+	 * Time Sequencer
+	 */
+	regp->Sequencer[NV_VIO_SR_RESET_INDEX] = 0x00;
+	/* 0x20 disables the sequencer */
+	if (mode->flags & DRM_MODE_FLAG_CLKDIV2)
+		regp->Sequencer[NV_VIO_SR_CLOCK_INDEX] = 0x29;
+	else
+		regp->Sequencer[NV_VIO_SR_CLOCK_INDEX] = 0x21;
+	regp->Sequencer[NV_VIO_SR_PLANE_MASK_INDEX] = 0x0F;
+	regp->Sequencer[NV_VIO_SR_CHAR_MAP_INDEX] = 0x00;
+	regp->Sequencer[NV_VIO_SR_MEM_MODE_INDEX] = 0x0E;
+
+	/*
+	 * CRTC
+	 */
+	regp->CRTC[NV_CIO_CR_HDT_INDEX] = horizTotal;
+	regp->CRTC[NV_CIO_CR_HDE_INDEX] = horizDisplay;
+	regp->CRTC[NV_CIO_CR_HBS_INDEX] = horizBlankStart;
+	regp->CRTC[NV_CIO_CR_HBE_INDEX] = (1 << 7) |
+					  XLATE(horizBlankEnd, 0, NV_CIO_CR_HBE_4_0);
+	regp->CRTC[NV_CIO_CR_HRS_INDEX] = horizStart;
+	regp->CRTC[NV_CIO_CR_HRE_INDEX] = XLATE(horizBlankEnd, 5, NV_CIO_CR_HRE_HBE_5) |
+					  XLATE(horizEnd, 0, NV_CIO_CR_HRE_4_0);
+	regp->CRTC[NV_CIO_CR_VDT_INDEX] = vertTotal;
+	regp->CRTC[NV_CIO_CR_OVL_INDEX] = XLATE(vertStart, 9, NV_CIO_CR_OVL_VRS_9) |
+					  XLATE(vertDisplay, 9, NV_CIO_CR_OVL_VDE_9) |
+					  XLATE(vertTotal, 9, NV_CIO_CR_OVL_VDT_9) |
+					  (1 << 4) |
+					  XLATE(vertBlankStart, 8, NV_CIO_CR_OVL_VBS_8) |
+					  XLATE(vertStart, 8, NV_CIO_CR_OVL_VRS_8) |
+					  XLATE(vertDisplay, 8, NV_CIO_CR_OVL_VDE_8) |
+					  XLATE(vertTotal, 8, NV_CIO_CR_OVL_VDT_8);
+	regp->CRTC[NV_CIO_CR_RSAL_INDEX] = 0x00;
+	regp->CRTC[NV_CIO_CR_CELL_HT_INDEX] = ((mode->flags & DRM_MODE_FLAG_DBLSCAN) ? MASK(NV_CIO_CR_CELL_HT_SCANDBL) : 0) |
+					      1 << 6 |
+					      XLATE(vertBlankStart, 9, NV_CIO_CR_CELL_HT_VBS_9);
+	regp->CRTC[NV_CIO_CR_CURS_ST_INDEX] = 0x00;
+	regp->CRTC[NV_CIO_CR_CURS_END_INDEX] = 0x00;
+	regp->CRTC[NV_CIO_CR_SA_HI_INDEX] = 0x00;
+	regp->CRTC[NV_CIO_CR_SA_LO_INDEX] = 0x00;
+	regp->CRTC[NV_CIO_CR_TCOFF_HI_INDEX] = 0x00;
+	regp->CRTC[NV_CIO_CR_TCOFF_LO_INDEX] = 0x00;
+	regp->CRTC[NV_CIO_CR_VRS_INDEX] = vertStart;
+	regp->CRTC[NV_CIO_CR_VRE_INDEX] = 1 << 5 | XLATE(vertEnd, 0, NV_CIO_CR_VRE_3_0);
+	regp->CRTC[NV_CIO_CR_VDE_INDEX] = vertDisplay;
+	/* framebuffer can be larger than crtc scanout area. */
+	regp->CRTC[NV_CIO_CR_OFFSET_INDEX] = fb->pitch / 8;
+	regp->CRTC[NV_CIO_CR_ULINE_INDEX] = 0x00;
+	regp->CRTC[NV_CIO_CR_VBS_INDEX] = vertBlankStart;
+	regp->CRTC[NV_CIO_CR_VBE_INDEX] = vertBlankEnd;
+	regp->CRTC[NV_CIO_CR_MODE_INDEX] = 0x43;
+	regp->CRTC[NV_CIO_CR_LCOMP_INDEX] = 0xff;
+
+	/*
+	 * Some extended CRTC registers (they are not saved with the rest of the vga regs).
+	 */
+
+	/* framebuffer can be larger than crtc scanout area. */
+	regp->CRTC[NV_CIO_CRE_RPC0_INDEX] = XLATE(fb->pitch / 8, 8, NV_CIO_CRE_RPC0_OFFSET_10_8);
+	regp->CRTC[NV_CIO_CRE_RPC1_INDEX] = mode->crtc_hdisplay < 1280 ?
+					    MASK(NV_CIO_CRE_RPC1_LARGE) : 0x00;
+	regp->CRTC[NV_CIO_CRE_LSR_INDEX] = XLATE(horizBlankEnd, 6, NV_CIO_CRE_LSR_HBE_6) |
+					   XLATE(vertBlankStart, 10, NV_CIO_CRE_LSR_VBS_10) |
+					   XLATE(vertStart, 10, NV_CIO_CRE_LSR_VRS_10) |
+					   XLATE(vertDisplay, 10, NV_CIO_CRE_LSR_VDE_10) |
+					   XLATE(vertTotal, 10, NV_CIO_CRE_LSR_VDT_10);
+	regp->CRTC[NV_CIO_CRE_HEB__INDEX] = XLATE(horizStart, 8, NV_CIO_CRE_HEB_HRS_8) |
+					    XLATE(horizBlankStart, 8, NV_CIO_CRE_HEB_HBS_8) |
+					    XLATE(horizDisplay, 8, NV_CIO_CRE_HEB_HDE_8) |
+					    XLATE(horizTotal, 8, NV_CIO_CRE_HEB_HDT_8);
+	regp->CRTC[NV_CIO_CRE_EBR_INDEX] = XLATE(vertBlankStart, 11, NV_CIO_CRE_EBR_VBS_11) |
+					   XLATE(vertStart, 11, NV_CIO_CRE_EBR_VRS_11) |
+					   XLATE(vertDisplay, 11, NV_CIO_CRE_EBR_VDE_11) |
+					   XLATE(vertTotal, 11, NV_CIO_CRE_EBR_VDT_11);
+
+	if (mode->flags & DRM_MODE_FLAG_INTERLACE) {
+		horizTotal = (horizTotal >> 1) & ~1;
+		regp->CRTC[NV_CIO_CRE_ILACE__INDEX] = horizTotal;
+		regp->CRTC[NV_CIO_CRE_HEB__INDEX] |= XLATE(horizTotal, 8, NV_CIO_CRE_HEB_ILC_8);
+	} else
+		regp->CRTC[NV_CIO_CRE_ILACE__INDEX] = 0xff;  /* interlace off */
+
+	/*
+	* Graphics Display Controller
+	*/
+	regp->Graphics[NV_VIO_GX_SR_INDEX] = 0x00;
+	regp->Graphics[NV_VIO_GX_SREN_INDEX] = 0x00;
+	regp->Graphics[NV_VIO_GX_CCOMP_INDEX] = 0x00;
+	regp->Graphics[NV_VIO_GX_ROP_INDEX] = 0x00;
+	regp->Graphics[NV_VIO_GX_READ_MAP_INDEX] = 0x00;
+	regp->Graphics[NV_VIO_GX_MODE_INDEX] = 0x40; /* 256 color mode */
+	regp->Graphics[NV_VIO_GX_MISC_INDEX] = 0x05; /* map 64k mem + graphic mode */
+	regp->Graphics[NV_VIO_GX_DONT_CARE_INDEX] = 0x0F;
+	regp->Graphics[NV_VIO_GX_BIT_MASK_INDEX] = 0xFF;
+
+	regp->Attribute[0]  = 0x00; /* standard colormap translation */
+	regp->Attribute[1]  = 0x01;
+	regp->Attribute[2]  = 0x02;
+	regp->Attribute[3]  = 0x03;
+	regp->Attribute[4]  = 0x04;
+	regp->Attribute[5]  = 0x05;
+	regp->Attribute[6]  = 0x06;
+	regp->Attribute[7]  = 0x07;
+	regp->Attribute[8]  = 0x08;
+	regp->Attribute[9]  = 0x09;
+	regp->Attribute[10] = 0x0A;
+	regp->Attribute[11] = 0x0B;
+	regp->Attribute[12] = 0x0C;
+	regp->Attribute[13] = 0x0D;
+	regp->Attribute[14] = 0x0E;
+	regp->Attribute[15] = 0x0F;
+	regp->Attribute[NV_CIO_AR_MODE_INDEX] = 0x01; /* Enable graphic mode */
+	/* Non-vga */
+	regp->Attribute[NV_CIO_AR_OSCAN_INDEX] = 0x00;
+	regp->Attribute[NV_CIO_AR_PLANE_INDEX] = 0x0F; /* enable all color planes */
+	regp->Attribute[NV_CIO_AR_HPP_INDEX] = 0x00;
+	regp->Attribute[NV_CIO_AR_CSEL_INDEX] = 0x00;
+}
+
+/**
+ * Sets up registers for the given mode/adjusted_mode pair.
+ *
+ * The clocks, CRTCs and outputs attached to this CRTC must be off.
+ *
+ * This shouldn't enable any clocks, CRTCs, or outputs, but they should
+ * be easily turned on/off after this.
+ */
+static void
+nv_crtc_mode_set_regs(struct drm_crtc *crtc, struct drm_display_mode * mode)
+{
+	struct drm_device *dev = crtc->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+	struct nv04_crtc_reg *regp = &dev_priv->mode_reg.crtc_reg[nv_crtc->index];
+	struct nv04_crtc_reg *savep = &dev_priv->saved_reg.crtc_reg[nv_crtc->index];
+	struct drm_encoder *encoder;
+	bool lvds_output = false, tmds_output = false, tv_output = false,
+		off_chip_digital = false;
+
+	list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) {
+		struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+		bool digital = false;
+
+		if (encoder->crtc != crtc)
+			continue;
+
+		if (nv_encoder->dcb->type == OUTPUT_LVDS)
+			digital = lvds_output = true;
+		if (nv_encoder->dcb->type == OUTPUT_TV)
+			tv_output = true;
+		if (nv_encoder->dcb->type == OUTPUT_TMDS)
+			digital = tmds_output = true;
+		if (nv_encoder->dcb->location != DCB_LOC_ON_CHIP && digital)
+			off_chip_digital = true;
+	}
+
+	/* Registers not directly related to the (s)vga mode */
+
+	/* What is the meaning of this register? */
+	/* A few popular values are 0x18, 0x1c, 0x38, 0x3c */
+	regp->CRTC[NV_CIO_CRE_ENH_INDEX] = savep->CRTC[NV_CIO_CRE_ENH_INDEX] & ~(1<<5);
+
+	regp->crtc_eng_ctrl = 0;
+	/* Except for rare conditions I2C is enabled on the primary crtc */
+	if (nv_crtc->index == 0)
+		regp->crtc_eng_ctrl |= NV_CRTC_FSEL_I2C;
+#if 0
+	/* Set overlay to desired crtc. */
+	if (dev->overlayAdaptor) {
+		NVPortPrivPtr pPriv = GET_OVERLAY_PRIVATE(dev);
+		if (pPriv->overlayCRTC == nv_crtc->index)
+			regp->crtc_eng_ctrl |= NV_CRTC_FSEL_OVERLAY;
+	}
+#endif
+
+	/* ADDRESS_SPACE_PNVM is the same as setting HCUR_ASI */
+	regp->cursor_cfg = NV_PCRTC_CURSOR_CONFIG_CUR_LINES_64 |
+			     NV_PCRTC_CURSOR_CONFIG_CUR_PIXELS_64 |
+			     NV_PCRTC_CURSOR_CONFIG_ADDRESS_SPACE_PNVM;
+	if (dev_priv->chipset >= 0x11)
+		regp->cursor_cfg |= NV_PCRTC_CURSOR_CONFIG_CUR_BPP_32;
+	if (mode->flags & DRM_MODE_FLAG_DBLSCAN)
+		regp->cursor_cfg |= NV_PCRTC_CURSOR_CONFIG_DOUBLE_SCAN_ENABLE;
+
+	/* Unblock some timings */
+	regp->CRTC[NV_CIO_CRE_53] = 0;
+	regp->CRTC[NV_CIO_CRE_54] = 0;
+
+	/* 0x00 is disabled, 0x11 is lvds, 0x22 crt and 0x88 tmds */
+	if (lvds_output)
+		regp->CRTC[NV_CIO_CRE_SCRATCH3__INDEX] = 0x11;
+	else if (tmds_output)
+		regp->CRTC[NV_CIO_CRE_SCRATCH3__INDEX] = 0x88;
+	else
+		regp->CRTC[NV_CIO_CRE_SCRATCH3__INDEX] = 0x22;
+
+	/* These values seem to vary */
+	/* This register seems to be used by the bios to make certain decisions on some G70 cards? */
+	regp->CRTC[NV_CIO_CRE_SCRATCH4__INDEX] = savep->CRTC[NV_CIO_CRE_SCRATCH4__INDEX];
+
+	nv_crtc_set_digital_vibrance(crtc, nv_crtc->saturation);
+
+	/* probably a scratch reg, but kept for cargo-cult purposes:
+	 * bit0: crtc0?, head A
+	 * bit6: lvds, head A
+	 * bit7: (only in X), head A
+	 */
+	if (nv_crtc->index == 0)
+		regp->CRTC[NV_CIO_CRE_4B] = savep->CRTC[NV_CIO_CRE_4B] | 0x80;
+
+	/* The blob seems to take the current value from crtc 0, add 4 to that
+	 * and reuse the old value for crtc 1 */
+	regp->CRTC[NV_CIO_CRE_TVOUT_LATENCY] = dev_priv->saved_reg.crtc_reg[0].CRTC[NV_CIO_CRE_TVOUT_LATENCY];
+	if (!nv_crtc->index)
+		regp->CRTC[NV_CIO_CRE_TVOUT_LATENCY] += 4;
+
+	/* the blob sometimes sets |= 0x10 (which is the same as setting |=
+	 * 1 << 30 on 0x60.830), for no apparent reason */
+	regp->CRTC[NV_CIO_CRE_59] = off_chip_digital;
+
+	regp->crtc_830 = mode->crtc_vdisplay - 3;
+	regp->crtc_834 = mode->crtc_vdisplay - 1;
+
+	if (dev_priv->card_type == NV_40)
+		/* This is what the blob does */
+		regp->crtc_850 = NVReadCRTC(dev, 0, NV_PCRTC_850);
+
+	if (dev_priv->card_type >= NV_30)
+		regp->gpio_ext = NVReadCRTC(dev, 0, NV_PCRTC_GPIO_EXT);
+
+	regp->crtc_cfg = NV_PCRTC_CONFIG_START_ADDRESS_HSYNC;
+
+	/* Some misc regs */
+	if (dev_priv->card_type == NV_40) {
+		regp->CRTC[NV_CIO_CRE_85] = 0xFF;
+		regp->CRTC[NV_CIO_CRE_86] = 0x1;
+	}
+
+	regp->CRTC[NV_CIO_CRE_PIXEL_INDEX] = (crtc->fb->depth + 1) / 8;
+	/* Enable slaved mode (called MODE_TV in nv4ref.h) */
+	if (lvds_output || tmds_output || tv_output)
+		regp->CRTC[NV_CIO_CRE_PIXEL_INDEX] |= (1 << 7);
+
+	/* Generic PRAMDAC regs */
+
+	if (dev_priv->card_type >= NV_10)
+		/* Only bit that bios and blob set. */
+		regp->nv10_cursync = (1 << 25);
+
+	regp->ramdac_gen_ctrl = NV_PRAMDAC_GENERAL_CONTROL_BPC_8BITS |
+				NV_PRAMDAC_GENERAL_CONTROL_VGA_STATE_SEL |
+				NV_PRAMDAC_GENERAL_CONTROL_PIXMIX_ON;
+	if (crtc->fb->depth == 16)
+		regp->ramdac_gen_ctrl |= NV_PRAMDAC_GENERAL_CONTROL_ALT_MODE_SEL;
+	if (dev_priv->chipset >= 0x11)
+		regp->ramdac_gen_ctrl |= NV_PRAMDAC_GENERAL_CONTROL_PIPE_LONG;
+
+	regp->ramdac_630 = 0; /* turn off green mode (tv test pattern?) */
+	regp->tv_setup = 0;
+
+	nv_crtc_set_image_sharpening(crtc, nv_crtc->sharpness);
+
+	/* Some values the blob sets */
+	regp->ramdac_8c0 = 0x100;
+	regp->ramdac_a20 = 0x0;
+	regp->ramdac_a24 = 0xfffff;
+	regp->ramdac_a34 = 0x1;
+}
+
+/**
+ * Sets up registers for the given mode/adjusted_mode pair.
+ *
+ * The clocks, CRTCs and outputs attached to this CRTC must be off.
+ *
+ * This shouldn't enable any clocks, CRTCs, or outputs, but they should
+ * be easily turned on/off after this.
+ */
+static int
+nv_crtc_mode_set(struct drm_crtc *crtc, struct drm_display_mode *mode,
+		 struct drm_display_mode *adjusted_mode,
+		 int x, int y, struct drm_framebuffer *old_fb)
+{
+	struct drm_device *dev = crtc->dev;
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	NV_DEBUG_KMS(dev, "CTRC mode on CRTC %d:\n", nv_crtc->index);
+	drm_mode_debug_printmodeline(adjusted_mode);
+
+	/* unlock must come after turning off FP_TG_CONTROL in output_prepare */
+	nv_lock_vga_crtc_shadow(dev, nv_crtc->index, -1);
+
+	nv_crtc_mode_set_vga(crtc, adjusted_mode);
+	/* calculated in nv04_dfp_prepare, nv40 needs it written before calculating PLLs */
+	if (dev_priv->card_type == NV_40)
+		NVWriteRAMDAC(dev, 0, NV_PRAMDAC_SEL_CLK, dev_priv->mode_reg.sel_clk);
+	nv_crtc_mode_set_regs(crtc, adjusted_mode);
+	nv_crtc_calc_state_ext(crtc, mode, adjusted_mode->clock);
+	return 0;
+}
+
+static void nv_crtc_save(struct drm_crtc *crtc)
+{
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+	struct drm_nouveau_private *dev_priv = crtc->dev->dev_private;
+	struct nv04_mode_state *state = &dev_priv->mode_reg;
+	struct nv04_crtc_reg *crtc_state = &state->crtc_reg[nv_crtc->index];
+	struct nv04_mode_state *saved = &dev_priv->saved_reg;
+	struct nv04_crtc_reg *crtc_saved = &saved->crtc_reg[nv_crtc->index];
+
+	if (nv_two_heads(crtc->dev))
+		NVSetOwner(crtc->dev, nv_crtc->index);
+
+	nouveau_hw_save_state(crtc->dev, nv_crtc->index, saved);
+
+	/* init some state to saved value */
+	state->sel_clk = saved->sel_clk & ~(0x5 << 16);
+	crtc_state->CRTC[NV_CIO_CRE_LCD__INDEX] = crtc_saved->CRTC[NV_CIO_CRE_LCD__INDEX];
+	state->pllsel = saved->pllsel & ~(PLLSEL_VPLL1_MASK | PLLSEL_VPLL2_MASK | PLLSEL_TV_MASK);
+	crtc_state->gpio_ext = crtc_saved->gpio_ext;
+}
+
+static void nv_crtc_restore(struct drm_crtc *crtc)
+{
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+	struct drm_nouveau_private *dev_priv = crtc->dev->dev_private;
+	int head = nv_crtc->index;
+	uint8_t saved_cr21 = dev_priv->saved_reg.crtc_reg[head].CRTC[NV_CIO_CRE_21];
+
+	if (nv_two_heads(crtc->dev))
+		NVSetOwner(crtc->dev, head);
+
+	nouveau_hw_load_state(crtc->dev, head, &dev_priv->saved_reg);
+	nv_lock_vga_crtc_shadow(crtc->dev, head, saved_cr21);
+
+	nv_crtc->last_dpms = NV_DPMS_CLEARED;
+}
+
+static void nv_crtc_prepare(struct drm_crtc *crtc)
+{
+	struct drm_device *dev = crtc->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+	struct drm_crtc_helper_funcs *funcs = crtc->helper_private;
+
+	if (nv_two_heads(dev))
+		NVSetOwner(dev, nv_crtc->index);
+
+	funcs->dpms(crtc, DRM_MODE_DPMS_OFF);
+
+	NVBlankScreen(dev, nv_crtc->index, true);
+
+	/* Some more preperation. */
+	NVWriteCRTC(dev, nv_crtc->index, NV_PCRTC_CONFIG, NV_PCRTC_CONFIG_START_ADDRESS_NON_VGA);
+	if (dev_priv->card_type == NV_40) {
+		uint32_t reg900 = NVReadRAMDAC(dev, nv_crtc->index, NV_PRAMDAC_900);
+		NVWriteRAMDAC(dev, nv_crtc->index, NV_PRAMDAC_900, reg900 & ~0x10000);
+	}
+}
+
+static void nv_crtc_commit(struct drm_crtc *crtc)
+{
+	struct drm_device *dev = crtc->dev;
+	struct drm_crtc_helper_funcs *funcs = crtc->helper_private;
+	struct drm_nouveau_private *dev_priv = crtc->dev->dev_private;
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+
+	nouveau_hw_load_state(dev, nv_crtc->index, &dev_priv->mode_reg);
+	nv04_crtc_mode_set_base(crtc, crtc->x, crtc->y, NULL);
+
+#ifdef __BIG_ENDIAN
+	/* turn on LFB swapping */
+	{
+		uint8_t tmp = NVReadVgaCrtc(dev, nv_crtc->index, NV_CIO_CRE_RCR);
+		tmp |= MASK(NV_CIO_CRE_RCR_ENDIAN_BIG);
+		NVWriteVgaCrtc(dev, nv_crtc->index, NV_CIO_CRE_RCR, tmp);
+	}
+#endif
+
+	funcs->dpms(crtc, DRM_MODE_DPMS_ON);
+}
+
+static void nv_crtc_destroy(struct drm_crtc *crtc)
+{
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+
+	NV_DEBUG_KMS(crtc->dev, "\n");
+
+	if (!nv_crtc)
+		return;
+
+	drm_crtc_cleanup(crtc);
+
+	nouveau_bo_ref(NULL, &nv_crtc->cursor.nvbo);
+	kfree(nv_crtc);
+}
+
+static void
+nv_crtc_gamma_load(struct drm_crtc *crtc)
+{
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+	struct drm_device *dev = nv_crtc->base.dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct rgb { uint8_t r, g, b; } __attribute__((packed)) *rgbs;
+	int i;
+
+	rgbs = (struct rgb *)dev_priv->mode_reg.crtc_reg[nv_crtc->index].DAC;
+	for (i = 0; i < 256; i++) {
+		rgbs[i].r = nv_crtc->lut.r[i] >> 8;
+		rgbs[i].g = nv_crtc->lut.g[i] >> 8;
+		rgbs[i].b = nv_crtc->lut.b[i] >> 8;
+	}
+
+	nouveau_hw_load_state_palette(dev, nv_crtc->index, &dev_priv->mode_reg);
+}
+
+static void
+nv_crtc_gamma_set(struct drm_crtc *crtc, u16 *r, u16 *g, u16 *b, uint32_t size)
+{
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+	int i;
+
+	if (size != 256)
+		return;
+
+	for (i = 0; i < 256; i++) {
+		nv_crtc->lut.r[i] = r[i];
+		nv_crtc->lut.g[i] = g[i];
+		nv_crtc->lut.b[i] = b[i];
+	}
+
+	/* We need to know the depth before we upload, but it's possible to
+	 * get called before a framebuffer is bound.  If this is the case,
+	 * mark the lut values as dirty by setting depth==0, and it'll be
+	 * uploaded on the first mode_set_base()
+	 */
+	if (!nv_crtc->base.fb) {
+		nv_crtc->lut.depth = 0;
+		return;
+	}
+
+	nv_crtc_gamma_load(crtc);
+}
+
+static int
+nv04_crtc_mode_set_base(struct drm_crtc *crtc, int x, int y,
+			struct drm_framebuffer *old_fb)
+{
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+	struct drm_device *dev = crtc->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nv04_crtc_reg *regp = &dev_priv->mode_reg.crtc_reg[nv_crtc->index];
+	struct drm_framebuffer *drm_fb = nv_crtc->base.fb;
+	struct nouveau_framebuffer *fb = nouveau_framebuffer(drm_fb);
+	int arb_burst, arb_lwm;
+	int ret;
+
+	ret = nouveau_bo_pin(fb->nvbo, TTM_PL_FLAG_VRAM);
+	if (ret)
+		return ret;
+
+	if (old_fb) {
+		struct nouveau_framebuffer *ofb = nouveau_framebuffer(old_fb);
+		nouveau_bo_unpin(ofb->nvbo);
+	}
+
+	nv_crtc->fb.offset = fb->nvbo->bo.offset;
+
+	if (nv_crtc->lut.depth != drm_fb->depth) {
+		nv_crtc->lut.depth = drm_fb->depth;
+		nv_crtc_gamma_load(crtc);
+	}
+
+	/* Update the framebuffer format. */
+	regp->CRTC[NV_CIO_CRE_PIXEL_INDEX] &= ~3;
+	regp->CRTC[NV_CIO_CRE_PIXEL_INDEX] |= (crtc->fb->depth + 1) / 8;
+	regp->ramdac_gen_ctrl &= ~NV_PRAMDAC_GENERAL_CONTROL_ALT_MODE_SEL;
+	if (crtc->fb->depth == 16)
+		regp->ramdac_gen_ctrl |= NV_PRAMDAC_GENERAL_CONTROL_ALT_MODE_SEL;
+	crtc_wr_cio_state(crtc, regp, NV_CIO_CRE_PIXEL_INDEX);
+	NVWriteRAMDAC(dev, nv_crtc->index, NV_PRAMDAC_GENERAL_CONTROL,
+		      regp->ramdac_gen_ctrl);
+
+	regp->CRTC[NV_CIO_CR_OFFSET_INDEX] = drm_fb->pitch >> 3;
+	regp->CRTC[NV_CIO_CRE_RPC0_INDEX] =
+		XLATE(drm_fb->pitch >> 3, 8, NV_CIO_CRE_RPC0_OFFSET_10_8);
+	crtc_wr_cio_state(crtc, regp, NV_CIO_CRE_RPC0_INDEX);
+	crtc_wr_cio_state(crtc, regp, NV_CIO_CR_OFFSET_INDEX);
+
+	/* Update the framebuffer location. */
+	regp->fb_start = nv_crtc->fb.offset & ~3;
+	regp->fb_start += (y * drm_fb->pitch) + (x * drm_fb->bits_per_pixel / 8);
+	NVWriteCRTC(dev, nv_crtc->index, NV_PCRTC_START, regp->fb_start);
+
+	/* Update the arbitration parameters. */
+	nouveau_calc_arb(dev, crtc->mode.clock, drm_fb->bits_per_pixel,
+			 &arb_burst, &arb_lwm);
+
+	regp->CRTC[NV_CIO_CRE_FF_INDEX] = arb_burst;
+	regp->CRTC[NV_CIO_CRE_FFLWM__INDEX] = arb_lwm & 0xff;
+	crtc_wr_cio_state(crtc, regp, NV_CIO_CRE_FF_INDEX);
+	crtc_wr_cio_state(crtc, regp, NV_CIO_CRE_FFLWM__INDEX);
+
+	if (dev_priv->card_type >= NV_30) {
+		regp->CRTC[NV_CIO_CRE_47] = arb_lwm >> 8;
+		crtc_wr_cio_state(crtc, regp, NV_CIO_CRE_47);
+	}
+
+	return 0;
+}
+
+static void nv04_cursor_upload(struct drm_device *dev, struct nouveau_bo *src,
+			       struct nouveau_bo *dst)
+{
+	int width = nv_cursor_width(dev);
+	uint32_t pixel;
+	int i, j;
+
+	for (i = 0; i < width; i++) {
+		for (j = 0; j < width; j++) {
+			pixel = nouveau_bo_rd32(src, i*64 + j);
+
+			nouveau_bo_wr16(dst, i*width + j, (pixel & 0x80000000) >> 16
+				     | (pixel & 0xf80000) >> 9
+				     | (pixel & 0xf800) >> 6
+				     | (pixel & 0xf8) >> 3);
+		}
+	}
+}
+
+static void nv11_cursor_upload(struct drm_device *dev, struct nouveau_bo *src,
+			       struct nouveau_bo *dst)
+{
+	uint32_t pixel;
+	int alpha, i;
+
+	/* nv11+ supports premultiplied (PM), or non-premultiplied (NPM) alpha
+	 * cursors (though NPM in combination with fp dithering may not work on
+	 * nv11, from "nv" driver history)
+	 * NPM mode needs NV_PCRTC_CURSOR_CONFIG_ALPHA_BLEND set and is what the
+	 * blob uses, however we get given PM cursors so we use PM mode
+	 */
+	for (i = 0; i < 64 * 64; i++) {
+		pixel = nouveau_bo_rd32(src, i);
+
+		/* hw gets unhappy if alpha <= rgb values.  for a PM image "less
+		 * than" shouldn't happen; fix "equal to" case by adding one to
+		 * alpha channel (slightly inaccurate, but so is attempting to
+		 * get back to NPM images, due to limits of integer precision)
+		 */
+		alpha = pixel >> 24;
+		if (alpha > 0 && alpha < 255)
+			pixel = (pixel & 0x00ffffff) | ((alpha + 1) << 24);
+
+#ifdef __BIG_ENDIAN
+		{
+			struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+			if (dev_priv->chipset == 0x11) {
+				pixel = ((pixel & 0x000000ff) << 24) |
+					((pixel & 0x0000ff00) << 8) |
+					((pixel & 0x00ff0000) >> 8) |
+					((pixel & 0xff000000) >> 24);
+			}
+		}
+#endif
+
+		nouveau_bo_wr32(dst, i, pixel);
+	}
+}
+
+static int
+nv04_crtc_cursor_set(struct drm_crtc *crtc, struct drm_file *file_priv,
+		     uint32_t buffer_handle, uint32_t width, uint32_t height)
+{
+	struct drm_nouveau_private *dev_priv = crtc->dev->dev_private;
+	struct drm_device *dev = dev_priv->dev;
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+	struct nouveau_bo *cursor = NULL;
+	struct drm_gem_object *gem;
+	int ret = 0;
+
+	if (width != 64 || height != 64)
+		return -EINVAL;
+
+	if (!buffer_handle) {
+		nv_crtc->cursor.hide(nv_crtc, true);
+		return 0;
+	}
+
+	gem = drm_gem_object_lookup(dev, file_priv, buffer_handle);
+	if (!gem)
+		return -EINVAL;
+	cursor = nouveau_gem_object(gem);
+
+	ret = nouveau_bo_map(cursor);
+	if (ret)
+		goto out;
+
+	if (dev_priv->chipset >= 0x11)
+		nv11_cursor_upload(dev, cursor, nv_crtc->cursor.nvbo);
+	else
+		nv04_cursor_upload(dev, cursor, nv_crtc->cursor.nvbo);
+
+	nouveau_bo_unmap(cursor);
+	nv_crtc->cursor.offset = nv_crtc->cursor.nvbo->bo.offset;
+	nv_crtc->cursor.set_offset(nv_crtc, nv_crtc->cursor.offset);
+	nv_crtc->cursor.show(nv_crtc, true);
+out:
+	drm_gem_object_unreference_unlocked(gem);
+	return ret;
+}
+
+static int
+nv04_crtc_cursor_move(struct drm_crtc *crtc, int x, int y)
+{
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+
+	nv_crtc->cursor.set_pos(nv_crtc, x, y);
+	return 0;
+}
+
+static const struct drm_crtc_funcs nv04_crtc_funcs = {
+	.save = nv_crtc_save,
+	.restore = nv_crtc_restore,
+	.cursor_set = nv04_crtc_cursor_set,
+	.cursor_move = nv04_crtc_cursor_move,
+	.gamma_set = nv_crtc_gamma_set,
+	.set_config = drm_crtc_helper_set_config,
+	.destroy = nv_crtc_destroy,
+};
+
+static const struct drm_crtc_helper_funcs nv04_crtc_helper_funcs = {
+	.dpms = nv_crtc_dpms,
+	.prepare = nv_crtc_prepare,
+	.commit = nv_crtc_commit,
+	.mode_fixup = nv_crtc_mode_fixup,
+	.mode_set = nv_crtc_mode_set,
+	.mode_set_base = nv04_crtc_mode_set_base,
+	.load_lut = nv_crtc_gamma_load,
+};
+
+int
+nv04_crtc_create(struct drm_device *dev, int crtc_num)
+{
+	struct nouveau_crtc *nv_crtc;
+	int ret, i;
+
+	nv_crtc = kzalloc(sizeof(*nv_crtc), GFP_KERNEL);
+	if (!nv_crtc)
+		return -ENOMEM;
+
+	for (i = 0; i < 256; i++) {
+		nv_crtc->lut.r[i] = i << 8;
+		nv_crtc->lut.g[i] = i << 8;
+		nv_crtc->lut.b[i] = i << 8;
+	}
+	nv_crtc->lut.depth = 0;
+
+	nv_crtc->index = crtc_num;
+	nv_crtc->last_dpms = NV_DPMS_CLEARED;
+
+	drm_crtc_init(dev, &nv_crtc->base, &nv04_crtc_funcs);
+	drm_crtc_helper_add(&nv_crtc->base, &nv04_crtc_helper_funcs);
+	drm_mode_crtc_set_gamma_size(&nv_crtc->base, 256);
+
+	ret = nouveau_bo_new(dev, NULL, 64*64*4, 0x100, TTM_PL_FLAG_VRAM,
+			     0, 0x0000, false, true, &nv_crtc->cursor.nvbo);
+	if (!ret) {
+		ret = nouveau_bo_pin(nv_crtc->cursor.nvbo, TTM_PL_FLAG_VRAM);
+		if (!ret)
+			ret = nouveau_bo_map(nv_crtc->cursor.nvbo);
+		if (ret)
+			nouveau_bo_ref(NULL, &nv_crtc->cursor.nvbo);
+	}
+
+	nv04_cursor_init(nv_crtc);
+
+	return 0;
+}
+
diff --git a/drivers/gpu/drm/nouveau/nv04_cursor.c b/drivers/gpu/drm/nouveau/nv04_cursor.c
new file mode 100644
index 000000000000..89a91b9d8b25
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv04_cursor.c
@@ -0,0 +1,70 @@
+#include "drmP.h"
+#include "drm_mode.h"
+#include "nouveau_reg.h"
+#include "nouveau_drv.h"
+#include "nouveau_crtc.h"
+#include "nouveau_hw.h"
+
+static void
+nv04_cursor_show(struct nouveau_crtc *nv_crtc, bool update)
+{
+	nv_show_cursor(nv_crtc->base.dev, nv_crtc->index, true);
+}
+
+static void
+nv04_cursor_hide(struct nouveau_crtc *nv_crtc, bool update)
+{
+	nv_show_cursor(nv_crtc->base.dev, nv_crtc->index, false);
+}
+
+static void
+nv04_cursor_set_pos(struct nouveau_crtc *nv_crtc, int x, int y)
+{
+	NVWriteRAMDAC(nv_crtc->base.dev, nv_crtc->index,
+		      NV_PRAMDAC_CU_START_POS,
+		      XLATE(y, 0, NV_PRAMDAC_CU_START_POS_Y) |
+		      XLATE(x, 0, NV_PRAMDAC_CU_START_POS_X));
+}
+
+static void
+crtc_wr_cio_state(struct drm_crtc *crtc, struct nv04_crtc_reg *crtcstate, int index)
+{
+	NVWriteVgaCrtc(crtc->dev, nouveau_crtc(crtc)->index, index,
+		       crtcstate->CRTC[index]);
+}
+
+static void
+nv04_cursor_set_offset(struct nouveau_crtc *nv_crtc, uint32_t offset)
+{
+	struct drm_device *dev = nv_crtc->base.dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nv04_crtc_reg *regp = &dev_priv->mode_reg.crtc_reg[nv_crtc->index];
+	struct drm_crtc *crtc = &nv_crtc->base;
+
+	regp->CRTC[NV_CIO_CRE_HCUR_ADDR0_INDEX] =
+		MASK(NV_CIO_CRE_HCUR_ASI) |
+		XLATE(offset, 17, NV_CIO_CRE_HCUR_ADDR0_ADR);
+	regp->CRTC[NV_CIO_CRE_HCUR_ADDR1_INDEX] =
+		XLATE(offset, 11, NV_CIO_CRE_HCUR_ADDR1_ADR);
+	if (crtc->mode.flags & DRM_MODE_FLAG_DBLSCAN)
+		regp->CRTC[NV_CIO_CRE_HCUR_ADDR1_INDEX] |=
+			MASK(NV_CIO_CRE_HCUR_ADDR1_CUR_DBL);
+	regp->CRTC[NV_CIO_CRE_HCUR_ADDR2_INDEX] = offset >> 24;
+
+	crtc_wr_cio_state(crtc, regp, NV_CIO_CRE_HCUR_ADDR0_INDEX);
+	crtc_wr_cio_state(crtc, regp, NV_CIO_CRE_HCUR_ADDR1_INDEX);
+	crtc_wr_cio_state(crtc, regp, NV_CIO_CRE_HCUR_ADDR2_INDEX);
+	if (dev_priv->card_type == NV_40)
+		nv_fix_nv40_hw_cursor(dev, nv_crtc->index);
+}
+
+int
+nv04_cursor_init(struct nouveau_crtc *crtc)
+{
+	crtc->cursor.set_offset = nv04_cursor_set_offset;
+	crtc->cursor.set_pos = nv04_cursor_set_pos;
+	crtc->cursor.hide = nv04_cursor_hide;
+	crtc->cursor.show = nv04_cursor_show;
+	return 0;
+}
+
diff --git a/drivers/gpu/drm/nouveau/nv04_dac.c b/drivers/gpu/drm/nouveau/nv04_dac.c
new file mode 100644
index 000000000000..1cb19e3acb55
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv04_dac.c
@@ -0,0 +1,531 @@
+/*
+ * Copyright 2003 NVIDIA, Corporation
+ * Copyright 2006 Dave Airlie
+ * Copyright 2007 Maarten Maathuis
+ * Copyright 2007-2009 Stuart Bennett
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include "drmP.h"
+#include "drm_crtc_helper.h"
+
+#include "nouveau_drv.h"
+#include "nouveau_encoder.h"
+#include "nouveau_connector.h"
+#include "nouveau_crtc.h"
+#include "nouveau_hw.h"
+#include "nvreg.h"
+
+int nv04_dac_output_offset(struct drm_encoder *encoder)
+{
+	struct dcb_entry *dcb = nouveau_encoder(encoder)->dcb;
+	int offset = 0;
+
+	if (dcb->or & (8 | OUTPUT_C))
+		offset += 0x68;
+	if (dcb->or & (8 | OUTPUT_B))
+		offset += 0x2000;
+
+	return offset;
+}
+
+/*
+ * arbitrary limit to number of sense oscillations tolerated in one sample
+ * period (observed to be at least 13 in "nvidia")
+ */
+#define MAX_HBLANK_OSC 20
+
+/*
+ * arbitrary limit to number of conflicting sample pairs to tolerate at a
+ * voltage step (observed to be at least 5 in "nvidia")
+ */
+#define MAX_SAMPLE_PAIRS 10
+
+static int sample_load_twice(struct drm_device *dev, bool sense[2])
+{
+	int i;
+
+	for (i = 0; i < 2; i++) {
+		bool sense_a, sense_b, sense_b_prime;
+		int j = 0;
+
+		/*
+		 * wait for bit 0 clear -- out of hblank -- (say reg value 0x4),
+		 * then wait for transition 0x4->0x5->0x4: enter hblank, leave
+		 * hblank again
+		 * use a 10ms timeout (guards against crtc being inactive, in
+		 * which case blank state would never change)
+		 */
+		if (!nouveau_wait_until(dev, 10000000, NV_PRMCIO_INP0__COLOR,
+					0x00000001, 0x00000000))
+			return -EBUSY;
+		if (!nouveau_wait_until(dev, 10000000, NV_PRMCIO_INP0__COLOR,
+					0x00000001, 0x00000001))
+			return -EBUSY;
+		if (!nouveau_wait_until(dev, 10000000, NV_PRMCIO_INP0__COLOR,
+					0x00000001, 0x00000000))
+			return -EBUSY;
+
+		udelay(100);
+		/* when level triggers, sense is _LO_ */
+		sense_a = nv_rd08(dev, NV_PRMCIO_INP0) & 0x10;
+
+		/* take another reading until it agrees with sense_a... */
+		do {
+			udelay(100);
+			sense_b = nv_rd08(dev, NV_PRMCIO_INP0) & 0x10;
+			if (sense_a != sense_b) {
+				sense_b_prime =
+					nv_rd08(dev, NV_PRMCIO_INP0) & 0x10;
+				if (sense_b == sense_b_prime) {
+					/* ... unless two consecutive subsequent
+					 * samples agree; sense_a is replaced */
+					sense_a = sense_b;
+					/* force mis-match so we loop */
+					sense_b = !sense_a;
+				}
+			}
+		} while ((sense_a != sense_b) && ++j < MAX_HBLANK_OSC);
+
+		if (j == MAX_HBLANK_OSC)
+			/* with so much oscillation, default to sense:LO */
+			sense[i] = false;
+		else
+			sense[i] = sense_a;
+	}
+
+	return 0;
+}
+
+static enum drm_connector_status nv04_dac_detect(struct drm_encoder *encoder,
+						 struct drm_connector *connector)
+{
+	struct drm_device *dev = encoder->dev;
+	uint8_t saved_seq1, saved_pi, saved_rpc1, saved_cr_mode;
+	uint8_t saved_palette0[3], saved_palette_mask;
+	uint32_t saved_rtest_ctrl, saved_rgen_ctrl;
+	int i;
+	uint8_t blue;
+	bool sense = true;
+
+	/*
+	 * for this detection to work, there needs to be a mode set up on the
+	 * CRTC.  this is presumed to be the case
+	 */
+
+	if (nv_two_heads(dev))
+		/* only implemented for head A for now */
+		NVSetOwner(dev, 0);
+
+	saved_cr_mode = NVReadVgaCrtc(dev, 0, NV_CIO_CR_MODE_INDEX);
+	NVWriteVgaCrtc(dev, 0, NV_CIO_CR_MODE_INDEX, saved_cr_mode | 0x80);
+
+	saved_seq1 = NVReadVgaSeq(dev, 0, NV_VIO_SR_CLOCK_INDEX);
+	NVWriteVgaSeq(dev, 0, NV_VIO_SR_CLOCK_INDEX, saved_seq1 & ~0x20);
+
+	saved_rtest_ctrl = NVReadRAMDAC(dev, 0, NV_PRAMDAC_TEST_CONTROL);
+	NVWriteRAMDAC(dev, 0, NV_PRAMDAC_TEST_CONTROL,
+		      saved_rtest_ctrl & ~NV_PRAMDAC_TEST_CONTROL_PWRDWN_DAC_OFF);
+
+	msleep(10);
+
+	saved_pi = NVReadVgaCrtc(dev, 0, NV_CIO_CRE_PIXEL_INDEX);
+	NVWriteVgaCrtc(dev, 0, NV_CIO_CRE_PIXEL_INDEX,
+		       saved_pi & ~(0x80 | MASK(NV_CIO_CRE_PIXEL_FORMAT)));
+	saved_rpc1 = NVReadVgaCrtc(dev, 0, NV_CIO_CRE_RPC1_INDEX);
+	NVWriteVgaCrtc(dev, 0, NV_CIO_CRE_RPC1_INDEX, saved_rpc1 & ~0xc0);
+
+	nv_wr08(dev, NV_PRMDIO_READ_MODE_ADDRESS, 0x0);
+	for (i = 0; i < 3; i++)
+		saved_palette0[i] = nv_rd08(dev, NV_PRMDIO_PALETTE_DATA);
+	saved_palette_mask = nv_rd08(dev, NV_PRMDIO_PIXEL_MASK);
+	nv_wr08(dev, NV_PRMDIO_PIXEL_MASK, 0);
+
+	saved_rgen_ctrl = NVReadRAMDAC(dev, 0, NV_PRAMDAC_GENERAL_CONTROL);
+	NVWriteRAMDAC(dev, 0, NV_PRAMDAC_GENERAL_CONTROL,
+		      (saved_rgen_ctrl & ~(NV_PRAMDAC_GENERAL_CONTROL_BPC_8BITS |
+					   NV_PRAMDAC_GENERAL_CONTROL_TERMINATION_75OHM)) |
+		      NV_PRAMDAC_GENERAL_CONTROL_PIXMIX_ON);
+
+	blue = 8;	/* start of test range */
+
+	do {
+		bool sense_pair[2];
+
+		nv_wr08(dev, NV_PRMDIO_WRITE_MODE_ADDRESS, 0);
+		nv_wr08(dev, NV_PRMDIO_PALETTE_DATA, 0);
+		nv_wr08(dev, NV_PRMDIO_PALETTE_DATA, 0);
+		/* testing blue won't find monochrome monitors.  I don't care */
+		nv_wr08(dev, NV_PRMDIO_PALETTE_DATA, blue);
+
+		i = 0;
+		/* take sample pairs until both samples in the pair agree */
+		do {
+			if (sample_load_twice(dev, sense_pair))
+				goto out;
+		} while ((sense_pair[0] != sense_pair[1]) &&
+							++i < MAX_SAMPLE_PAIRS);
+
+		if (i == MAX_SAMPLE_PAIRS)
+			/* too much oscillation defaults to LO */
+			sense = false;
+		else
+			sense = sense_pair[0];
+
+	/*
+	 * if sense goes LO before blue ramps to 0x18, monitor is not connected.
+	 * ergo, if blue gets to 0x18, monitor must be connected
+	 */
+	} while (++blue < 0x18 && sense);
+
+out:
+	nv_wr08(dev, NV_PRMDIO_PIXEL_MASK, saved_palette_mask);
+	NVWriteRAMDAC(dev, 0, NV_PRAMDAC_GENERAL_CONTROL, saved_rgen_ctrl);
+	nv_wr08(dev, NV_PRMDIO_WRITE_MODE_ADDRESS, 0);
+	for (i = 0; i < 3; i++)
+		nv_wr08(dev, NV_PRMDIO_PALETTE_DATA, saved_palette0[i]);
+	NVWriteRAMDAC(dev, 0, NV_PRAMDAC_TEST_CONTROL, saved_rtest_ctrl);
+	NVWriteVgaCrtc(dev, 0, NV_CIO_CRE_PIXEL_INDEX, saved_pi);
+	NVWriteVgaCrtc(dev, 0, NV_CIO_CRE_RPC1_INDEX, saved_rpc1);
+	NVWriteVgaSeq(dev, 0, NV_VIO_SR_CLOCK_INDEX, saved_seq1);
+	NVWriteVgaCrtc(dev, 0, NV_CIO_CR_MODE_INDEX, saved_cr_mode);
+
+	if (blue == 0x18) {
+		NV_INFO(dev, "Load detected on head A\n");
+		return connector_status_connected;
+	}
+
+	return connector_status_disconnected;
+}
+
+uint32_t nv17_dac_sample_load(struct drm_encoder *encoder)
+{
+	struct drm_device *dev = encoder->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct dcb_entry *dcb = nouveau_encoder(encoder)->dcb;
+	uint32_t sample, testval, regoffset = nv04_dac_output_offset(encoder);
+	uint32_t saved_powerctrl_2 = 0, saved_powerctrl_4 = 0, saved_routput,
+		saved_rtest_ctrl, saved_gpio0, saved_gpio1, temp, routput;
+	int head;
+
+#define RGB_TEST_DATA(r, g, b) (r << 0 | g << 10 | b << 20)
+	if (dcb->type == OUTPUT_TV) {
+		testval = RGB_TEST_DATA(0xa0, 0xa0, 0xa0);
+
+		if (dev_priv->vbios.tvdactestval)
+			testval = dev_priv->vbios.tvdactestval;
+	} else {
+		testval = RGB_TEST_DATA(0x140, 0x140, 0x140); /* 0x94050140 */
+
+		if (dev_priv->vbios.dactestval)
+			testval = dev_priv->vbios.dactestval;
+	}
+
+	saved_rtest_ctrl = NVReadRAMDAC(dev, 0, NV_PRAMDAC_TEST_CONTROL + regoffset);
+	NVWriteRAMDAC(dev, 0, NV_PRAMDAC_TEST_CONTROL + regoffset,
+		      saved_rtest_ctrl & ~NV_PRAMDAC_TEST_CONTROL_PWRDWN_DAC_OFF);
+
+	saved_powerctrl_2 = nvReadMC(dev, NV_PBUS_POWERCTRL_2);
+
+	nvWriteMC(dev, NV_PBUS_POWERCTRL_2, saved_powerctrl_2 & 0xd7ffffff);
+	if (regoffset == 0x68) {
+		saved_powerctrl_4 = nvReadMC(dev, NV_PBUS_POWERCTRL_4);
+		nvWriteMC(dev, NV_PBUS_POWERCTRL_4, saved_powerctrl_4 & 0xffffffcf);
+	}
+
+	saved_gpio1 = nv17_gpio_get(dev, DCB_GPIO_TVDAC1);
+	saved_gpio0 = nv17_gpio_get(dev, DCB_GPIO_TVDAC0);
+
+	nv17_gpio_set(dev, DCB_GPIO_TVDAC1, dcb->type == OUTPUT_TV);
+	nv17_gpio_set(dev, DCB_GPIO_TVDAC0, dcb->type == OUTPUT_TV);
+
+	msleep(4);
+
+	saved_routput = NVReadRAMDAC(dev, 0, NV_PRAMDAC_DACCLK + regoffset);
+	head = (saved_routput & 0x100) >> 8;
+#if 0
+	/* if there's a spare crtc, using it will minimise flicker for the case
+	 * where the in-use crtc is in use by an off-chip tmds encoder */
+	if (xf86_config->crtc[head]->enabled && !xf86_config->crtc[head ^ 1]->enabled)
+		head ^= 1;
+#endif
+	/* nv driver and nv31 use 0xfffffeee, nv34 and 6600 use 0xfffffece */
+	routput = (saved_routput & 0xfffffece) | head << 8;
+
+	if (dev_priv->card_type >= NV_40) {
+		if (dcb->type == OUTPUT_TV)
+			routput |= 0x1a << 16;
+		else
+			routput &= ~(0x1a << 16);
+	}
+
+	NVWriteRAMDAC(dev, 0, NV_PRAMDAC_DACCLK + regoffset, routput);
+	msleep(1);
+
+	temp = NVReadRAMDAC(dev, 0, NV_PRAMDAC_DACCLK + regoffset);
+	NVWriteRAMDAC(dev, 0, NV_PRAMDAC_DACCLK + regoffset, temp | 1);
+
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_TESTPOINT_DATA,
+		      NV_PRAMDAC_TESTPOINT_DATA_NOTBLANK | testval);
+	temp = NVReadRAMDAC(dev, head, NV_PRAMDAC_TEST_CONTROL);
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_TEST_CONTROL,
+		      temp | NV_PRAMDAC_TEST_CONTROL_TP_INS_EN_ASSERTED);
+	msleep(5);
+
+	sample = NVReadRAMDAC(dev, 0, NV_PRAMDAC_TEST_CONTROL + regoffset);
+
+	temp = NVReadRAMDAC(dev, head, NV_PRAMDAC_TEST_CONTROL);
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_TEST_CONTROL,
+		      temp & ~NV_PRAMDAC_TEST_CONTROL_TP_INS_EN_ASSERTED);
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_TESTPOINT_DATA, 0);
+
+	/* bios does something more complex for restoring, but I think this is good enough */
+	NVWriteRAMDAC(dev, 0, NV_PRAMDAC_DACCLK + regoffset, saved_routput);
+	NVWriteRAMDAC(dev, 0, NV_PRAMDAC_TEST_CONTROL + regoffset, saved_rtest_ctrl);
+	if (regoffset == 0x68)
+		nvWriteMC(dev, NV_PBUS_POWERCTRL_4, saved_powerctrl_4);
+	nvWriteMC(dev, NV_PBUS_POWERCTRL_2, saved_powerctrl_2);
+
+	nv17_gpio_set(dev, DCB_GPIO_TVDAC1, saved_gpio1);
+	nv17_gpio_set(dev, DCB_GPIO_TVDAC0, saved_gpio0);
+
+	return sample;
+}
+
+static enum drm_connector_status
+nv17_dac_detect(struct drm_encoder *encoder, struct drm_connector *connector)
+{
+	struct drm_device *dev = encoder->dev;
+	struct dcb_entry *dcb = nouveau_encoder(encoder)->dcb;
+	uint32_t sample = nv17_dac_sample_load(encoder);
+
+	if (sample & NV_PRAMDAC_TEST_CONTROL_SENSEB_ALLHI) {
+		NV_INFO(dev, "Load detected on output %c\n",
+			'@' + ffs(dcb->or));
+		return connector_status_connected;
+	} else {
+		return connector_status_disconnected;
+	}
+}
+
+static bool nv04_dac_mode_fixup(struct drm_encoder *encoder,
+				struct drm_display_mode *mode,
+				struct drm_display_mode *adjusted_mode)
+{
+	return true;
+}
+
+static void nv04_dac_prepare(struct drm_encoder *encoder)
+{
+	struct drm_encoder_helper_funcs *helper = encoder->helper_private;
+	struct drm_device *dev = encoder->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	int head = nouveau_crtc(encoder->crtc)->index;
+	struct nv04_crtc_reg *crtcstate = dev_priv->mode_reg.crtc_reg;
+
+	helper->dpms(encoder, DRM_MODE_DPMS_OFF);
+
+	nv04_dfp_disable(dev, head);
+
+	/* Some NV4x have unknown values (0x3f, 0x50, 0x54, 0x6b, 0x79, 0x7f)
+	 * at LCD__INDEX which we don't alter
+	 */
+	if (!(crtcstate[head].CRTC[NV_CIO_CRE_LCD__INDEX] & 0x44))
+		crtcstate[head].CRTC[NV_CIO_CRE_LCD__INDEX] = 0;
+}
+
+
+static void nv04_dac_mode_set(struct drm_encoder *encoder,
+			      struct drm_display_mode *mode,
+			      struct drm_display_mode *adjusted_mode)
+{
+	struct drm_device *dev = encoder->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	int head = nouveau_crtc(encoder->crtc)->index;
+
+	if (nv_gf4_disp_arch(dev)) {
+		struct drm_encoder *rebind;
+		uint32_t dac_offset = nv04_dac_output_offset(encoder);
+		uint32_t otherdac;
+
+		/* bit 16-19 are bits that are set on some G70 cards,
+		 * but don't seem to have much effect */
+		NVWriteRAMDAC(dev, 0, NV_PRAMDAC_DACCLK + dac_offset,
+			      head << 8 | NV_PRAMDAC_DACCLK_SEL_DACCLK);
+		/* force any other vga encoders to bind to the other crtc */
+		list_for_each_entry(rebind, &dev->mode_config.encoder_list, head) {
+			if (rebind == encoder
+			    || nouveau_encoder(rebind)->dcb->type != OUTPUT_ANALOG)
+				continue;
+
+			dac_offset = nv04_dac_output_offset(rebind);
+			otherdac = NVReadRAMDAC(dev, 0, NV_PRAMDAC_DACCLK + dac_offset);
+			NVWriteRAMDAC(dev, 0, NV_PRAMDAC_DACCLK + dac_offset,
+				      (otherdac & ~0x0100) | (head ^ 1) << 8);
+		}
+	}
+
+	/* This could use refinement for flatpanels, but it should work this way */
+	if (dev_priv->chipset < 0x44)
+		NVWriteRAMDAC(dev, 0, NV_PRAMDAC_TEST_CONTROL + nv04_dac_output_offset(encoder), 0xf0000000);
+	else
+		NVWriteRAMDAC(dev, 0, NV_PRAMDAC_TEST_CONTROL + nv04_dac_output_offset(encoder), 0x00100000);
+}
+
+static void nv04_dac_commit(struct drm_encoder *encoder)
+{
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	struct drm_device *dev = encoder->dev;
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(encoder->crtc);
+	struct drm_encoder_helper_funcs *helper = encoder->helper_private;
+
+	helper->dpms(encoder, DRM_MODE_DPMS_ON);
+
+	NV_INFO(dev, "Output %s is running on CRTC %d using output %c\n",
+		drm_get_connector_name(&nouveau_encoder_connector_get(nv_encoder)->base),
+		nv_crtc->index, '@' + ffs(nv_encoder->dcb->or));
+}
+
+void nv04_dac_update_dacclk(struct drm_encoder *encoder, bool enable)
+{
+	struct drm_device *dev = encoder->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct dcb_entry *dcb = nouveau_encoder(encoder)->dcb;
+
+	if (nv_gf4_disp_arch(dev)) {
+		uint32_t *dac_users = &dev_priv->dac_users[ffs(dcb->or) - 1];
+		int dacclk_off = NV_PRAMDAC_DACCLK + nv04_dac_output_offset(encoder);
+		uint32_t dacclk = NVReadRAMDAC(dev, 0, dacclk_off);
+
+		if (enable) {
+			*dac_users |= 1 << dcb->index;
+			NVWriteRAMDAC(dev, 0, dacclk_off, dacclk | NV_PRAMDAC_DACCLK_SEL_DACCLK);
+
+		} else {
+			*dac_users &= ~(1 << dcb->index);
+			if (!*dac_users)
+				NVWriteRAMDAC(dev, 0, dacclk_off,
+					dacclk & ~NV_PRAMDAC_DACCLK_SEL_DACCLK);
+		}
+	}
+}
+
+static void nv04_dac_dpms(struct drm_encoder *encoder, int mode)
+{
+	struct drm_device *dev = encoder->dev;
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+
+	if (nv_encoder->last_dpms == mode)
+		return;
+	nv_encoder->last_dpms = mode;
+
+	NV_INFO(dev, "Setting dpms mode %d on vga encoder (output %d)\n",
+		     mode, nv_encoder->dcb->index);
+
+	nv04_dac_update_dacclk(encoder, mode == DRM_MODE_DPMS_ON);
+}
+
+static void nv04_dac_save(struct drm_encoder *encoder)
+{
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	struct drm_device *dev = encoder->dev;
+
+	if (nv_gf4_disp_arch(dev))
+		nv_encoder->restore.output = NVReadRAMDAC(dev, 0, NV_PRAMDAC_DACCLK +
+							  nv04_dac_output_offset(encoder));
+}
+
+static void nv04_dac_restore(struct drm_encoder *encoder)
+{
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	struct drm_device *dev = encoder->dev;
+
+	if (nv_gf4_disp_arch(dev))
+		NVWriteRAMDAC(dev, 0, NV_PRAMDAC_DACCLK + nv04_dac_output_offset(encoder),
+			      nv_encoder->restore.output);
+
+	nv_encoder->last_dpms = NV_DPMS_CLEARED;
+}
+
+static void nv04_dac_destroy(struct drm_encoder *encoder)
+{
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+
+	NV_DEBUG_KMS(encoder->dev, "\n");
+
+	drm_encoder_cleanup(encoder);
+	kfree(nv_encoder);
+}
+
+static const struct drm_encoder_helper_funcs nv04_dac_helper_funcs = {
+	.dpms = nv04_dac_dpms,
+	.save = nv04_dac_save,
+	.restore = nv04_dac_restore,
+	.mode_fixup = nv04_dac_mode_fixup,
+	.prepare = nv04_dac_prepare,
+	.commit = nv04_dac_commit,
+	.mode_set = nv04_dac_mode_set,
+	.detect = nv04_dac_detect
+};
+
+static const struct drm_encoder_helper_funcs nv17_dac_helper_funcs = {
+	.dpms = nv04_dac_dpms,
+	.save = nv04_dac_save,
+	.restore = nv04_dac_restore,
+	.mode_fixup = nv04_dac_mode_fixup,
+	.prepare = nv04_dac_prepare,
+	.commit = nv04_dac_commit,
+	.mode_set = nv04_dac_mode_set,
+	.detect = nv17_dac_detect
+};
+
+static const struct drm_encoder_funcs nv04_dac_funcs = {
+	.destroy = nv04_dac_destroy,
+};
+
+int nv04_dac_create(struct drm_device *dev, struct dcb_entry *entry)
+{
+	const struct drm_encoder_helper_funcs *helper;
+	struct drm_encoder *encoder;
+	struct nouveau_encoder *nv_encoder = NULL;
+
+	nv_encoder = kzalloc(sizeof(*nv_encoder), GFP_KERNEL);
+	if (!nv_encoder)
+		return -ENOMEM;
+
+	encoder = to_drm_encoder(nv_encoder);
+
+	nv_encoder->dcb = entry;
+	nv_encoder->or = ffs(entry->or) - 1;
+
+	if (nv_gf4_disp_arch(dev))
+		helper = &nv17_dac_helper_funcs;
+	else
+		helper = &nv04_dac_helper_funcs;
+
+	drm_encoder_init(dev, encoder, &nv04_dac_funcs, DRM_MODE_ENCODER_DAC);
+	drm_encoder_helper_add(encoder, helper);
+
+	encoder->possible_crtcs = entry->heads;
+	encoder->possible_clones = 0;
+
+	return 0;
+}
diff --git a/drivers/gpu/drm/nouveau/nv04_dfp.c b/drivers/gpu/drm/nouveau/nv04_dfp.c
new file mode 100644
index 000000000000..41634d4752fe
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv04_dfp.c
@@ -0,0 +1,623 @@
+/*
+ * Copyright 2003 NVIDIA, Corporation
+ * Copyright 2006 Dave Airlie
+ * Copyright 2007 Maarten Maathuis
+ * Copyright 2007-2009 Stuart Bennett
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include "drmP.h"
+#include "drm_crtc_helper.h"
+
+#include "nouveau_drv.h"
+#include "nouveau_encoder.h"
+#include "nouveau_connector.h"
+#include "nouveau_crtc.h"
+#include "nouveau_hw.h"
+#include "nvreg.h"
+
+#define FP_TG_CONTROL_ON  (NV_PRAMDAC_FP_TG_CONTROL_DISPEN_POS |	\
+			   NV_PRAMDAC_FP_TG_CONTROL_HSYNC_POS |		\
+			   NV_PRAMDAC_FP_TG_CONTROL_VSYNC_POS)
+#define FP_TG_CONTROL_OFF (NV_PRAMDAC_FP_TG_CONTROL_DISPEN_DISABLE |	\
+			   NV_PRAMDAC_FP_TG_CONTROL_HSYNC_DISABLE |	\
+			   NV_PRAMDAC_FP_TG_CONTROL_VSYNC_DISABLE)
+
+static inline bool is_fpc_off(uint32_t fpc)
+{
+	return ((fpc & (FP_TG_CONTROL_ON | FP_TG_CONTROL_OFF)) ==
+			FP_TG_CONTROL_OFF);
+}
+
+int nv04_dfp_get_bound_head(struct drm_device *dev, struct dcb_entry *dcbent)
+{
+	/* special case of nv_read_tmds to find crtc associated with an output.
+	 * this does not give a correct answer for off-chip dvi, but there's no
+	 * use for such an answer anyway
+	 */
+	int ramdac = (dcbent->or & OUTPUT_C) >> 2;
+
+	NVWriteRAMDAC(dev, ramdac, NV_PRAMDAC_FP_TMDS_CONTROL,
+	NV_PRAMDAC_FP_TMDS_CONTROL_WRITE_DISABLE | 0x4);
+	return ((NVReadRAMDAC(dev, ramdac, NV_PRAMDAC_FP_TMDS_DATA) & 0x8) >> 3) ^ ramdac;
+}
+
+void nv04_dfp_bind_head(struct drm_device *dev, struct dcb_entry *dcbent,
+			int head, bool dl)
+{
+	/* The BIOS scripts don't do this for us, sadly
+	 * Luckily we do know the values ;-)
+	 *
+	 * head < 0 indicates we wish to force a setting with the overrideval
+	 * (for VT restore etc.)
+	 */
+
+	int ramdac = (dcbent->or & OUTPUT_C) >> 2;
+	uint8_t tmds04 = 0x80;
+
+	if (head != ramdac)
+		tmds04 = 0x88;
+
+	if (dcbent->type == OUTPUT_LVDS)
+		tmds04 |= 0x01;
+
+	nv_write_tmds(dev, dcbent->or, 0, 0x04, tmds04);
+
+	if (dl)	/* dual link */
+		nv_write_tmds(dev, dcbent->or, 1, 0x04, tmds04 ^ 0x08);
+}
+
+void nv04_dfp_disable(struct drm_device *dev, int head)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nv04_crtc_reg *crtcstate = dev_priv->mode_reg.crtc_reg;
+
+	if (NVReadRAMDAC(dev, head, NV_PRAMDAC_FP_TG_CONTROL) &
+	    FP_TG_CONTROL_ON) {
+		/* digital remnants must be cleaned before new crtc
+		 * values programmed.  delay is time for the vga stuff
+		 * to realise it's in control again
+		 */
+		NVWriteRAMDAC(dev, head, NV_PRAMDAC_FP_TG_CONTROL,
+			      FP_TG_CONTROL_OFF);
+		msleep(50);
+	}
+	/* don't inadvertently turn it on when state written later */
+	crtcstate[head].fp_control = FP_TG_CONTROL_OFF;
+}
+
+void nv04_dfp_update_fp_control(struct drm_encoder *encoder, int mode)
+{
+	struct drm_device *dev = encoder->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct drm_crtc *crtc;
+	struct nouveau_crtc *nv_crtc;
+	uint32_t *fpc;
+
+	if (mode == DRM_MODE_DPMS_ON) {
+		nv_crtc = nouveau_crtc(encoder->crtc);
+		fpc = &dev_priv->mode_reg.crtc_reg[nv_crtc->index].fp_control;
+
+		if (is_fpc_off(*fpc)) {
+			/* using saved value is ok, as (is_digital && dpms_on &&
+			 * fp_control==OFF) is (at present) *only* true when
+			 * fpc's most recent change was by below "off" code
+			 */
+			*fpc = nv_crtc->dpms_saved_fp_control;
+		}
+
+		nv_crtc->fp_users |= 1 << nouveau_encoder(encoder)->dcb->index;
+		NVWriteRAMDAC(dev, nv_crtc->index, NV_PRAMDAC_FP_TG_CONTROL, *fpc);
+	} else {
+		list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
+			nv_crtc = nouveau_crtc(crtc);
+			fpc = &dev_priv->mode_reg.crtc_reg[nv_crtc->index].fp_control;
+
+			nv_crtc->fp_users &= ~(1 << nouveau_encoder(encoder)->dcb->index);
+			if (!is_fpc_off(*fpc) && !nv_crtc->fp_users) {
+				nv_crtc->dpms_saved_fp_control = *fpc;
+				/* cut the FP output */
+				*fpc &= ~FP_TG_CONTROL_ON;
+				*fpc |= FP_TG_CONTROL_OFF;
+				NVWriteRAMDAC(dev, nv_crtc->index,
+					      NV_PRAMDAC_FP_TG_CONTROL, *fpc);
+			}
+		}
+	}
+}
+
+static bool nv04_dfp_mode_fixup(struct drm_encoder *encoder,
+				struct drm_display_mode *mode,
+				struct drm_display_mode *adjusted_mode)
+{
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	struct nouveau_connector *nv_connector = nouveau_encoder_connector_get(nv_encoder);
+
+	/* For internal panels and gpu scaling on DVI we need the native mode */
+	if (nv_connector->scaling_mode != DRM_MODE_SCALE_NONE) {
+		if (!nv_connector->native_mode)
+			return false;
+		nv_encoder->mode = *nv_connector->native_mode;
+		adjusted_mode->clock = nv_connector->native_mode->clock;
+	} else {
+		nv_encoder->mode = *adjusted_mode;
+	}
+
+	return true;
+}
+
+static void nv04_dfp_prepare_sel_clk(struct drm_device *dev,
+				     struct nouveau_encoder *nv_encoder, int head)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nv04_mode_state *state = &dev_priv->mode_reg;
+	uint32_t bits1618 = nv_encoder->dcb->or & OUTPUT_A ? 0x10000 : 0x40000;
+
+	if (nv_encoder->dcb->location != DCB_LOC_ON_CHIP)
+		return;
+
+	/* SEL_CLK is only used on the primary ramdac
+	 * It toggles spread spectrum PLL output and sets the bindings of PLLs
+	 * to heads on digital outputs
+	 */
+	if (head)
+		state->sel_clk |= bits1618;
+	else
+		state->sel_clk &= ~bits1618;
+
+	/* nv30:
+	 *	bit 0		NVClk spread spectrum on/off
+	 *	bit 2		MemClk spread spectrum on/off
+	 * 	bit 4		PixClk1 spread spectrum on/off toggle
+	 * 	bit 6		PixClk2 spread spectrum on/off toggle
+	 *
+	 * nv40 (observations from bios behaviour and mmio traces):
+	 * 	bits 4&6	as for nv30
+	 * 	bits 5&7	head dependent as for bits 4&6, but do not appear with 4&6;
+	 * 			maybe a different spread mode
+	 * 	bits 8&10	seen on dual-link dvi outputs, purpose unknown (set by POST scripts)
+	 * 	The logic behind turning spread spectrum on/off in the first place,
+	 * 	and which bit-pair to use, is unclear on nv40 (for earlier cards, the fp table
+	 * 	entry has the necessary info)
+	 */
+	if (nv_encoder->dcb->type == OUTPUT_LVDS && dev_priv->saved_reg.sel_clk & 0xf0) {
+		int shift = (dev_priv->saved_reg.sel_clk & 0x50) ? 0 : 1;
+
+		state->sel_clk &= ~0xf0;
+		state->sel_clk |= (head ? 0x40 : 0x10) << shift;
+	}
+}
+
+static void nv04_dfp_prepare(struct drm_encoder *encoder)
+{
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	struct drm_encoder_helper_funcs *helper = encoder->helper_private;
+	struct drm_device *dev = encoder->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	int head = nouveau_crtc(encoder->crtc)->index;
+	struct nv04_crtc_reg *crtcstate = dev_priv->mode_reg.crtc_reg;
+	uint8_t *cr_lcd = &crtcstate[head].CRTC[NV_CIO_CRE_LCD__INDEX];
+	uint8_t *cr_lcd_oth = &crtcstate[head ^ 1].CRTC[NV_CIO_CRE_LCD__INDEX];
+
+	helper->dpms(encoder, DRM_MODE_DPMS_OFF);
+
+	nv04_dfp_prepare_sel_clk(dev, nv_encoder, head);
+
+	/* Some NV4x have unknown values (0x3f, 0x50, 0x54, 0x6b, 0x79, 0x7f)
+	 * at LCD__INDEX which we don't alter
+	 */
+	if (!(*cr_lcd & 0x44)) {
+		*cr_lcd = 0x3;
+
+		if (nv_two_heads(dev)) {
+			if (nv_encoder->dcb->location == DCB_LOC_ON_CHIP)
+				*cr_lcd |= head ? 0x0 : 0x8;
+			else {
+				*cr_lcd |= (nv_encoder->dcb->or << 4) & 0x30;
+				if (nv_encoder->dcb->type == OUTPUT_LVDS)
+					*cr_lcd |= 0x30;
+				if ((*cr_lcd & 0x30) == (*cr_lcd_oth & 0x30)) {
+					/* avoid being connected to both crtcs */
+					*cr_lcd_oth &= ~0x30;
+					NVWriteVgaCrtc(dev, head ^ 1,
+						       NV_CIO_CRE_LCD__INDEX,
+						       *cr_lcd_oth);
+				}
+			}
+		}
+	}
+}
+
+
+static void nv04_dfp_mode_set(struct drm_encoder *encoder,
+			      struct drm_display_mode *mode,
+			      struct drm_display_mode *adjusted_mode)
+{
+	struct drm_device *dev = encoder->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(encoder->crtc);
+	struct nv04_crtc_reg *regp = &dev_priv->mode_reg.crtc_reg[nv_crtc->index];
+	struct nv04_crtc_reg *savep = &dev_priv->saved_reg.crtc_reg[nv_crtc->index];
+	struct nouveau_connector *nv_connector = nouveau_crtc_connector_get(nv_crtc);
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	struct drm_display_mode *output_mode = &nv_encoder->mode;
+	uint32_t mode_ratio, panel_ratio;
+
+	NV_DEBUG_KMS(dev, "Output mode on CRTC %d:\n", nv_crtc->index);
+	drm_mode_debug_printmodeline(output_mode);
+
+	/* Initialize the FP registers in this CRTC. */
+	regp->fp_horiz_regs[FP_DISPLAY_END] = output_mode->hdisplay - 1;
+	regp->fp_horiz_regs[FP_TOTAL] = output_mode->htotal - 1;
+	if (!nv_gf4_disp_arch(dev) ||
+	    (output_mode->hsync_start - output_mode->hdisplay) >=
+					dev_priv->vbios.digital_min_front_porch)
+		regp->fp_horiz_regs[FP_CRTC] = output_mode->hdisplay;
+	else
+		regp->fp_horiz_regs[FP_CRTC] = output_mode->hsync_start - dev_priv->vbios.digital_min_front_porch - 1;
+	regp->fp_horiz_regs[FP_SYNC_START] = output_mode->hsync_start - 1;
+	regp->fp_horiz_regs[FP_SYNC_END] = output_mode->hsync_end - 1;
+	regp->fp_horiz_regs[FP_VALID_START] = output_mode->hskew;
+	regp->fp_horiz_regs[FP_VALID_END] = output_mode->hdisplay - 1;
+
+	regp->fp_vert_regs[FP_DISPLAY_END] = output_mode->vdisplay - 1;
+	regp->fp_vert_regs[FP_TOTAL] = output_mode->vtotal - 1;
+	regp->fp_vert_regs[FP_CRTC] = output_mode->vtotal - 5 - 1;
+	regp->fp_vert_regs[FP_SYNC_START] = output_mode->vsync_start - 1;
+	regp->fp_vert_regs[FP_SYNC_END] = output_mode->vsync_end - 1;
+	regp->fp_vert_regs[FP_VALID_START] = 0;
+	regp->fp_vert_regs[FP_VALID_END] = output_mode->vdisplay - 1;
+
+	/* bit26: a bit seen on some g7x, no as yet discernable purpose */
+	regp->fp_control = NV_PRAMDAC_FP_TG_CONTROL_DISPEN_POS |
+			   (savep->fp_control & (1 << 26 | NV_PRAMDAC_FP_TG_CONTROL_READ_PROG));
+	/* Deal with vsync/hsync polarity */
+	/* LVDS screens do set this, but modes with +ve syncs are very rare */
+	if (output_mode->flags & DRM_MODE_FLAG_PVSYNC)
+		regp->fp_control |= NV_PRAMDAC_FP_TG_CONTROL_VSYNC_POS;
+	if (output_mode->flags & DRM_MODE_FLAG_PHSYNC)
+		regp->fp_control |= NV_PRAMDAC_FP_TG_CONTROL_HSYNC_POS;
+	/* panel scaling first, as native would get set otherwise */
+	if (nv_connector->scaling_mode == DRM_MODE_SCALE_NONE ||
+	    nv_connector->scaling_mode == DRM_MODE_SCALE_CENTER)	/* panel handles it */
+		regp->fp_control |= NV_PRAMDAC_FP_TG_CONTROL_MODE_CENTER;
+	else if (adjusted_mode->hdisplay == output_mode->hdisplay &&
+		 adjusted_mode->vdisplay == output_mode->vdisplay) /* native mode */
+		regp->fp_control |= NV_PRAMDAC_FP_TG_CONTROL_MODE_NATIVE;
+	else /* gpu needs to scale */
+		regp->fp_control |= NV_PRAMDAC_FP_TG_CONTROL_MODE_SCALE;
+	if (nvReadEXTDEV(dev, NV_PEXTDEV_BOOT_0) & NV_PEXTDEV_BOOT_0_STRAP_FP_IFACE_12BIT)
+		regp->fp_control |= NV_PRAMDAC_FP_TG_CONTROL_WIDTH_12;
+	if (nv_encoder->dcb->location != DCB_LOC_ON_CHIP &&
+	    output_mode->clock > 165000)
+		regp->fp_control |= (2 << 24);
+	if (nv_encoder->dcb->type == OUTPUT_LVDS) {
+		bool duallink, dummy;
+
+		nouveau_bios_parse_lvds_table(dev, nv_connector->native_mode->
+					      clock, &duallink, &dummy);
+		if (duallink)
+			regp->fp_control |= (8 << 28);
+	} else
+	if (output_mode->clock > 165000)
+		regp->fp_control |= (8 << 28);
+
+	regp->fp_debug_0 = NV_PRAMDAC_FP_DEBUG_0_YWEIGHT_ROUND |
+			   NV_PRAMDAC_FP_DEBUG_0_XWEIGHT_ROUND |
+			   NV_PRAMDAC_FP_DEBUG_0_YINTERP_BILINEAR |
+			   NV_PRAMDAC_FP_DEBUG_0_XINTERP_BILINEAR |
+			   NV_RAMDAC_FP_DEBUG_0_TMDS_ENABLED |
+			   NV_PRAMDAC_FP_DEBUG_0_YSCALE_ENABLE |
+			   NV_PRAMDAC_FP_DEBUG_0_XSCALE_ENABLE;
+
+	/* We want automatic scaling */
+	regp->fp_debug_1 = 0;
+	/* This can override HTOTAL and VTOTAL */
+	regp->fp_debug_2 = 0;
+
+	/* Use 20.12 fixed point format to avoid floats */
+	mode_ratio = (1 << 12) * adjusted_mode->hdisplay / adjusted_mode->vdisplay;
+	panel_ratio = (1 << 12) * output_mode->hdisplay / output_mode->vdisplay;
+	/* if ratios are equal, SCALE_ASPECT will automatically (and correctly)
+	 * get treated the same as SCALE_FULLSCREEN */
+	if (nv_connector->scaling_mode == DRM_MODE_SCALE_ASPECT &&
+	    mode_ratio != panel_ratio) {
+		uint32_t diff, scale;
+		bool divide_by_2 = nv_gf4_disp_arch(dev);
+
+		if (mode_ratio < panel_ratio) {
+			/* vertical needs to expand to glass size (automatic)
+			 * horizontal needs to be scaled at vertical scale factor
+			 * to maintain aspect */
+
+			scale = (1 << 12) * adjusted_mode->vdisplay / output_mode->vdisplay;
+			regp->fp_debug_1 = NV_PRAMDAC_FP_DEBUG_1_XSCALE_TESTMODE_ENABLE |
+					   XLATE(scale, divide_by_2, NV_PRAMDAC_FP_DEBUG_1_XSCALE_VALUE);
+
+			/* restrict area of screen used, horizontally */
+			diff = output_mode->hdisplay -
+			       output_mode->vdisplay * mode_ratio / (1 << 12);
+			regp->fp_horiz_regs[FP_VALID_START] += diff / 2;
+			regp->fp_horiz_regs[FP_VALID_END] -= diff / 2;
+		}
+
+		if (mode_ratio > panel_ratio) {
+			/* horizontal needs to expand to glass size (automatic)
+			 * vertical needs to be scaled at horizontal scale factor
+			 * to maintain aspect */
+
+			scale = (1 << 12) * adjusted_mode->hdisplay / output_mode->hdisplay;
+			regp->fp_debug_1 = NV_PRAMDAC_FP_DEBUG_1_YSCALE_TESTMODE_ENABLE |
+					   XLATE(scale, divide_by_2, NV_PRAMDAC_FP_DEBUG_1_YSCALE_VALUE);
+
+			/* restrict area of screen used, vertically */
+			diff = output_mode->vdisplay -
+			       (1 << 12) * output_mode->hdisplay / mode_ratio;
+			regp->fp_vert_regs[FP_VALID_START] += diff / 2;
+			regp->fp_vert_regs[FP_VALID_END] -= diff / 2;
+		}
+	}
+
+	/* Output property. */
+	if (nv_connector->use_dithering) {
+		if (dev_priv->chipset == 0x11)
+			regp->dither = savep->dither | 0x00010000;
+		else {
+			int i;
+			regp->dither = savep->dither | 0x00000001;
+			for (i = 0; i < 3; i++) {
+				regp->dither_regs[i] = 0xe4e4e4e4;
+				regp->dither_regs[i + 3] = 0x44444444;
+			}
+		}
+	} else {
+		if (dev_priv->chipset != 0x11) {
+			/* reset them */
+			int i;
+			for (i = 0; i < 3; i++) {
+				regp->dither_regs[i] = savep->dither_regs[i];
+				regp->dither_regs[i + 3] = savep->dither_regs[i + 3];
+			}
+		}
+		regp->dither = savep->dither;
+	}
+
+	regp->fp_margin_color = 0;
+}
+
+static void nv04_dfp_commit(struct drm_encoder *encoder)
+{
+	struct drm_device *dev = encoder->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct drm_encoder_helper_funcs *helper = encoder->helper_private;
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(encoder->crtc);
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	struct dcb_entry *dcbe = nv_encoder->dcb;
+	int head = nouveau_crtc(encoder->crtc)->index;
+
+	NV_INFO(dev, "Output %s is running on CRTC %d using output %c\n",
+		drm_get_connector_name(&nouveau_encoder_connector_get(nv_encoder)->base),
+		nv_crtc->index, '@' + ffs(nv_encoder->dcb->or));
+
+	if (dcbe->type == OUTPUT_TMDS)
+		run_tmds_table(dev, dcbe, head, nv_encoder->mode.clock);
+	else if (dcbe->type == OUTPUT_LVDS)
+		call_lvds_script(dev, dcbe, head, LVDS_RESET, nv_encoder->mode.clock);
+
+	/* update fp_control state for any changes made by scripts,
+	 * so correct value is written at DPMS on */
+	dev_priv->mode_reg.crtc_reg[head].fp_control =
+		NVReadRAMDAC(dev, head, NV_PRAMDAC_FP_TG_CONTROL);
+
+	/* This could use refinement for flatpanels, but it should work this way */
+	if (dev_priv->chipset < 0x44)
+		NVWriteRAMDAC(dev, 0, NV_PRAMDAC_TEST_CONTROL + nv04_dac_output_offset(encoder), 0xf0000000);
+	else
+		NVWriteRAMDAC(dev, 0, NV_PRAMDAC_TEST_CONTROL + nv04_dac_output_offset(encoder), 0x00100000);
+
+	helper->dpms(encoder, DRM_MODE_DPMS_ON);
+
+	NV_INFO(dev, "Output %s is running on CRTC %d using output %c\n",
+		drm_get_connector_name(&nouveau_encoder_connector_get(nv_encoder)->base),
+		nv_crtc->index, '@' + ffs(nv_encoder->dcb->or));
+}
+
+static inline bool is_powersaving_dpms(int mode)
+{
+	return (mode != DRM_MODE_DPMS_ON);
+}
+
+static void nv04_lvds_dpms(struct drm_encoder *encoder, int mode)
+{
+	struct drm_device *dev = encoder->dev;
+	struct drm_crtc *crtc = encoder->crtc;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	bool was_powersaving = is_powersaving_dpms(nv_encoder->last_dpms);
+
+	if (nv_encoder->last_dpms == mode)
+		return;
+	nv_encoder->last_dpms = mode;
+
+	NV_INFO(dev, "Setting dpms mode %d on lvds encoder (output %d)\n",
+		     mode, nv_encoder->dcb->index);
+
+	if (was_powersaving && is_powersaving_dpms(mode))
+		return;
+
+	if (nv_encoder->dcb->lvdsconf.use_power_scripts) {
+		struct nouveau_connector *nv_connector = nouveau_encoder_connector_get(nv_encoder);
+
+		/* when removing an output, crtc may not be set, but PANEL_OFF
+		 * must still be run
+		 */
+		int head = crtc ? nouveau_crtc(crtc)->index :
+			   nv04_dfp_get_bound_head(dev, nv_encoder->dcb);
+
+		if (mode == DRM_MODE_DPMS_ON) {
+			if (!nv_connector->native_mode) {
+				NV_ERROR(dev, "Not turning on LVDS without native mode\n");
+				return;
+			}
+			call_lvds_script(dev, nv_encoder->dcb, head,
+					 LVDS_PANEL_ON, nv_connector->native_mode->clock);
+		} else
+			/* pxclk of 0 is fine for PANEL_OFF, and for a
+			 * disconnected LVDS encoder there is no native_mode
+			 */
+			call_lvds_script(dev, nv_encoder->dcb, head,
+					 LVDS_PANEL_OFF, 0);
+	}
+
+	nv04_dfp_update_fp_control(encoder, mode);
+
+	if (mode == DRM_MODE_DPMS_ON)
+		nv04_dfp_prepare_sel_clk(dev, nv_encoder, nouveau_crtc(crtc)->index);
+	else {
+		dev_priv->mode_reg.sel_clk = NVReadRAMDAC(dev, 0, NV_PRAMDAC_SEL_CLK);
+		dev_priv->mode_reg.sel_clk &= ~0xf0;
+	}
+	NVWriteRAMDAC(dev, 0, NV_PRAMDAC_SEL_CLK, dev_priv->mode_reg.sel_clk);
+}
+
+static void nv04_tmds_dpms(struct drm_encoder *encoder, int mode)
+{
+	struct drm_device *dev = encoder->dev;
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+
+	if (nv_encoder->last_dpms == mode)
+		return;
+	nv_encoder->last_dpms = mode;
+
+	NV_INFO(dev, "Setting dpms mode %d on tmds encoder (output %d)\n",
+		     mode, nv_encoder->dcb->index);
+
+	nv04_dfp_update_fp_control(encoder, mode);
+}
+
+static void nv04_dfp_save(struct drm_encoder *encoder)
+{
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	struct drm_device *dev = encoder->dev;
+
+	if (nv_two_heads(dev))
+		nv_encoder->restore.head =
+			nv04_dfp_get_bound_head(dev, nv_encoder->dcb);
+}
+
+static void nv04_dfp_restore(struct drm_encoder *encoder)
+{
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	struct drm_device *dev = encoder->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	int head = nv_encoder->restore.head;
+
+	if (nv_encoder->dcb->type == OUTPUT_LVDS) {
+		struct drm_display_mode *native_mode = nouveau_encoder_connector_get(nv_encoder)->native_mode;
+		if (native_mode)
+			call_lvds_script(dev, nv_encoder->dcb, head, LVDS_PANEL_ON,
+					 native_mode->clock);
+		else
+			NV_ERROR(dev, "Not restoring LVDS without native mode\n");
+
+	} else if (nv_encoder->dcb->type == OUTPUT_TMDS) {
+		int clock = nouveau_hw_pllvals_to_clk
+					(&dev_priv->saved_reg.crtc_reg[head].pllvals);
+
+		run_tmds_table(dev, nv_encoder->dcb, head, clock);
+	}
+
+	nv_encoder->last_dpms = NV_DPMS_CLEARED;
+}
+
+static void nv04_dfp_destroy(struct drm_encoder *encoder)
+{
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+
+	NV_DEBUG_KMS(encoder->dev, "\n");
+
+	drm_encoder_cleanup(encoder);
+	kfree(nv_encoder);
+}
+
+static const struct drm_encoder_helper_funcs nv04_lvds_helper_funcs = {
+	.dpms = nv04_lvds_dpms,
+	.save = nv04_dfp_save,
+	.restore = nv04_dfp_restore,
+	.mode_fixup = nv04_dfp_mode_fixup,
+	.prepare = nv04_dfp_prepare,
+	.commit = nv04_dfp_commit,
+	.mode_set = nv04_dfp_mode_set,
+	.detect = NULL,
+};
+
+static const struct drm_encoder_helper_funcs nv04_tmds_helper_funcs = {
+	.dpms = nv04_tmds_dpms,
+	.save = nv04_dfp_save,
+	.restore = nv04_dfp_restore,
+	.mode_fixup = nv04_dfp_mode_fixup,
+	.prepare = nv04_dfp_prepare,
+	.commit = nv04_dfp_commit,
+	.mode_set = nv04_dfp_mode_set,
+	.detect = NULL,
+};
+
+static const struct drm_encoder_funcs nv04_dfp_funcs = {
+	.destroy = nv04_dfp_destroy,
+};
+
+int nv04_dfp_create(struct drm_device *dev, struct dcb_entry *entry)
+{
+	const struct drm_encoder_helper_funcs *helper;
+	struct drm_encoder *encoder;
+	struct nouveau_encoder *nv_encoder = NULL;
+	int type;
+
+	switch (entry->type) {
+	case OUTPUT_TMDS:
+		type = DRM_MODE_ENCODER_TMDS;
+		helper = &nv04_tmds_helper_funcs;
+		break;
+	case OUTPUT_LVDS:
+		type = DRM_MODE_ENCODER_LVDS;
+		helper = &nv04_lvds_helper_funcs;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	nv_encoder = kzalloc(sizeof(*nv_encoder), GFP_KERNEL);
+	if (!nv_encoder)
+		return -ENOMEM;
+
+	encoder = to_drm_encoder(nv_encoder);
+
+	nv_encoder->dcb = entry;
+	nv_encoder->or = ffs(entry->or) - 1;
+
+	drm_encoder_init(dev, encoder, &nv04_dfp_funcs, type);
+	drm_encoder_helper_add(encoder, helper);
+
+	encoder->possible_crtcs = entry->heads;
+	encoder->possible_clones = 0;
+
+	return 0;
+}
diff --git a/drivers/gpu/drm/nouveau/nv04_display.c b/drivers/gpu/drm/nouveau/nv04_display.c
new file mode 100644
index 000000000000..c7898b4f6dfb
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv04_display.c
@@ -0,0 +1,244 @@
+/*
+ * Copyright 2009 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Author: Ben Skeggs
+ */
+
+#include "drmP.h"
+#include "drm.h"
+#include "drm_crtc_helper.h"
+
+#include "nouveau_drv.h"
+#include "nouveau_fb.h"
+#include "nouveau_hw.h"
+#include "nouveau_encoder.h"
+#include "nouveau_connector.h"
+
+#define MULTIPLE_ENCODERS(e) (e & (e - 1))
+
+static void
+nv04_display_store_initial_head_owner(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	if (dev_priv->chipset != 0x11) {
+		dev_priv->crtc_owner = NVReadVgaCrtc(dev, 0, NV_CIO_CRE_44);
+		goto ownerknown;
+	}
+
+	/* reading CR44 is broken on nv11, so we attempt to infer it */
+	if (nvReadMC(dev, NV_PBUS_DEBUG_1) & (1 << 28))	/* heads tied, restore both */
+		dev_priv->crtc_owner = 0x4;
+	else {
+		uint8_t slaved_on_A, slaved_on_B;
+		bool tvA = false;
+		bool tvB = false;
+
+		NVLockVgaCrtcs(dev, false);
+
+		slaved_on_B = NVReadVgaCrtc(dev, 1, NV_CIO_CRE_PIXEL_INDEX) &
+									0x80;
+		if (slaved_on_B)
+			tvB = !(NVReadVgaCrtc(dev, 1, NV_CIO_CRE_LCD__INDEX) &
+					MASK(NV_CIO_CRE_LCD_LCD_SELECT));
+
+		slaved_on_A = NVReadVgaCrtc(dev, 0, NV_CIO_CRE_PIXEL_INDEX) &
+									0x80;
+		if (slaved_on_A)
+			tvA = !(NVReadVgaCrtc(dev, 0, NV_CIO_CRE_LCD__INDEX) &
+					MASK(NV_CIO_CRE_LCD_LCD_SELECT));
+
+		NVLockVgaCrtcs(dev, true);
+
+		if (slaved_on_A && !tvA)
+			dev_priv->crtc_owner = 0x0;
+		else if (slaved_on_B && !tvB)
+			dev_priv->crtc_owner = 0x3;
+		else if (slaved_on_A)
+			dev_priv->crtc_owner = 0x0;
+		else if (slaved_on_B)
+			dev_priv->crtc_owner = 0x3;
+		else
+			dev_priv->crtc_owner = 0x0;
+	}
+
+ownerknown:
+	NV_INFO(dev, "Initial CRTC_OWNER is %d\n", dev_priv->crtc_owner);
+
+	/* we need to ensure the heads are not tied henceforth, or reading any
+	 * 8 bit reg on head B will fail
+	 * setting a single arbitrary head solves that */
+	NVSetOwner(dev, 0);
+}
+
+int
+nv04_display_create(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct dcb_table *dcb = &dev_priv->vbios.dcb;
+	struct drm_encoder *encoder;
+	struct drm_crtc *crtc;
+	int i, ret;
+
+	NV_DEBUG_KMS(dev, "\n");
+
+	if (nv_two_heads(dev))
+		nv04_display_store_initial_head_owner(dev);
+	nouveau_hw_save_vga_fonts(dev, 1);
+
+	drm_mode_config_init(dev);
+	drm_mode_create_scaling_mode_property(dev);
+	drm_mode_create_dithering_property(dev);
+
+	dev->mode_config.funcs = (void *)&nouveau_mode_config_funcs;
+
+	dev->mode_config.min_width = 0;
+	dev->mode_config.min_height = 0;
+	switch (dev_priv->card_type) {
+	case NV_04:
+		dev->mode_config.max_width = 2048;
+		dev->mode_config.max_height = 2048;
+		break;
+	default:
+		dev->mode_config.max_width = 4096;
+		dev->mode_config.max_height = 4096;
+		break;
+	}
+
+	dev->mode_config.fb_base = dev_priv->fb_phys;
+
+	nv04_crtc_create(dev, 0);
+	if (nv_two_heads(dev))
+		nv04_crtc_create(dev, 1);
+
+	for (i = 0; i < dcb->entries; i++) {
+		struct dcb_entry *dcbent = &dcb->entry[i];
+
+		switch (dcbent->type) {
+		case OUTPUT_ANALOG:
+			ret = nv04_dac_create(dev, dcbent);
+			break;
+		case OUTPUT_LVDS:
+		case OUTPUT_TMDS:
+			ret = nv04_dfp_create(dev, dcbent);
+			break;
+		case OUTPUT_TV:
+			if (dcbent->location == DCB_LOC_ON_CHIP)
+				ret = nv17_tv_create(dev, dcbent);
+			else
+				ret = nv04_tv_create(dev, dcbent);
+			break;
+		default:
+			NV_WARN(dev, "DCB type %d not known\n", dcbent->type);
+			continue;
+		}
+
+		if (ret)
+			continue;
+	}
+
+	for (i = 0; i < dcb->connector.entries; i++)
+		nouveau_connector_create(dev, &dcb->connector.entry[i]);
+
+	/* Save previous state */
+	NVLockVgaCrtcs(dev, false);
+
+	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head)
+		crtc->funcs->save(crtc);
+
+	list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) {
+		struct drm_encoder_helper_funcs *func = encoder->helper_private;
+
+		func->save(encoder);
+	}
+
+	return 0;
+}
+
+void
+nv04_display_destroy(struct drm_device *dev)
+{
+	struct drm_encoder *encoder;
+	struct drm_crtc *crtc;
+
+	NV_DEBUG_KMS(dev, "\n");
+
+	/* Turn every CRTC off. */
+	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
+		struct drm_mode_set modeset = {
+			.crtc = crtc,
+		};
+
+		crtc->funcs->set_config(&modeset);
+	}
+
+	/* Restore state */
+	NVLockVgaCrtcs(dev, false);
+
+	list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) {
+		struct drm_encoder_helper_funcs *func = encoder->helper_private;
+
+		func->restore(encoder);
+	}
+
+	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head)
+		crtc->funcs->restore(crtc);
+
+	drm_mode_config_cleanup(dev);
+
+	nouveau_hw_save_vga_fonts(dev, 0);
+}
+
+void
+nv04_display_restore(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct drm_encoder *encoder;
+	struct drm_crtc *crtc;
+
+	NVLockVgaCrtcs(dev, false);
+
+	/* meh.. modeset apparently doesn't setup all the regs and depends
+	 * on pre-existing state, for now load the state of the card *before*
+	 * nouveau was loaded, and then do a modeset.
+	 *
+	 * best thing to do probably is to make save/restore routines not
+	 * save/restore "pre-load" state, but more general so we can save
+	 * on suspend too.
+	 */
+	list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) {
+		struct drm_encoder_helper_funcs *func = encoder->helper_private;
+
+		func->restore(encoder);
+	}
+
+	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head)
+		crtc->funcs->restore(crtc);
+
+	if (nv_two_heads(dev)) {
+		NV_INFO(dev, "Restoring CRTC_OWNER to %d.\n",
+			dev_priv->crtc_owner);
+		NVSetOwner(dev, dev_priv->crtc_owner);
+	}
+
+	NVLockVgaCrtcs(dev, true);
+}
+
diff --git a/drivers/gpu/drm/nouveau/nv04_fb.c b/drivers/gpu/drm/nouveau/nv04_fb.c
new file mode 100644
index 000000000000..638cf601c427
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv04_fb.c
@@ -0,0 +1,21 @@
+#include "drmP.h"
+#include "drm.h"
+#include "nouveau_drv.h"
+#include "nouveau_drm.h"
+
+int
+nv04_fb_init(struct drm_device *dev)
+{
+	/* This is what the DDX did for NV_ARCH_04, but a mmio-trace shows
+	 * nvidia reading PFB_CFG_0, then writing back its original value.
+	 * (which was 0x701114 in this case)
+	 */
+
+	nv_wr32(dev, NV04_PFB_CFG0, 0x1114);
+	return 0;
+}
+
+void
+nv04_fb_takedown(struct drm_device *dev)
+{
+}
diff --git a/drivers/gpu/drm/nouveau/nv04_fbcon.c b/drivers/gpu/drm/nouveau/nv04_fbcon.c
new file mode 100644
index 000000000000..3da90c2c4e63
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv04_fbcon.c
@@ -0,0 +1,312 @@
+/*
+ * Copyright 2009 Ben Skeggs
+ * Copyright 2008 Stuart Bennett
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include "drmP.h"
+#include "nouveau_drv.h"
+#include "nouveau_dma.h"
+#include "nouveau_fbcon.h"
+
+void
+nv04_fbcon_copyarea(struct fb_info *info, const struct fb_copyarea *region)
+{
+	struct nouveau_fbcon_par *par = info->par;
+	struct drm_device *dev = par->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_channel *chan = dev_priv->channel;
+
+	if (info->state != FBINFO_STATE_RUNNING)
+		return;
+
+	if (!(info->flags & FBINFO_HWACCEL_DISABLED) && RING_SPACE(chan, 4)) {
+		nouveau_fbcon_gpu_lockup(info);
+	}
+
+	if (info->flags & FBINFO_HWACCEL_DISABLED) {
+		cfb_copyarea(info, region);
+		return;
+	}
+
+	BEGIN_RING(chan, NvSubImageBlit, 0x0300, 3);
+	OUT_RING(chan, (region->sy << 16) | region->sx);
+	OUT_RING(chan, (region->dy << 16) | region->dx);
+	OUT_RING(chan, (region->height << 16) | region->width);
+	FIRE_RING(chan);
+}
+
+void
+nv04_fbcon_fillrect(struct fb_info *info, const struct fb_fillrect *rect)
+{
+	struct nouveau_fbcon_par *par = info->par;
+	struct drm_device *dev = par->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_channel *chan = dev_priv->channel;
+
+	if (info->state != FBINFO_STATE_RUNNING)
+		return;
+
+	if (!(info->flags & FBINFO_HWACCEL_DISABLED) && RING_SPACE(chan, 7)) {
+		nouveau_fbcon_gpu_lockup(info);
+	}
+
+	if (info->flags & FBINFO_HWACCEL_DISABLED) {
+		cfb_fillrect(info, rect);
+		return;
+	}
+
+	BEGIN_RING(chan, NvSubGdiRect, 0x02fc, 1);
+	OUT_RING(chan, (rect->rop != ROP_COPY) ? 1 : 3);
+	BEGIN_RING(chan, NvSubGdiRect, 0x03fc, 1);
+	if (info->fix.visual == FB_VISUAL_TRUECOLOR ||
+	    info->fix.visual == FB_VISUAL_DIRECTCOLOR)
+		OUT_RING(chan, ((uint32_t *)info->pseudo_palette)[rect->color]);
+	else
+		OUT_RING(chan, rect->color);
+	BEGIN_RING(chan, NvSubGdiRect, 0x0400, 2);
+	OUT_RING(chan, (rect->dx << 16) | rect->dy);
+	OUT_RING(chan, (rect->width << 16) | rect->height);
+	FIRE_RING(chan);
+}
+
+void
+nv04_fbcon_imageblit(struct fb_info *info, const struct fb_image *image)
+{
+	struct nouveau_fbcon_par *par = info->par;
+	struct drm_device *dev = par->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_channel *chan = dev_priv->channel;
+	uint32_t fg;
+	uint32_t bg;
+	uint32_t dsize;
+	uint32_t width;
+	uint32_t *data = (uint32_t *)image->data;
+
+	if (info->state != FBINFO_STATE_RUNNING)
+		return;
+
+	if (image->depth != 1) {
+		cfb_imageblit(info, image);
+		return;
+	}
+
+	if (!(info->flags & FBINFO_HWACCEL_DISABLED) && RING_SPACE(chan, 8)) {
+		nouveau_fbcon_gpu_lockup(info);
+	}
+
+	if (info->flags & FBINFO_HWACCEL_DISABLED) {
+		cfb_imageblit(info, image);
+		return;
+	}
+
+	width = ALIGN(image->width, 32);
+	dsize = (width * image->height) >> 5;
+
+	if (info->fix.visual == FB_VISUAL_TRUECOLOR ||
+	    info->fix.visual == FB_VISUAL_DIRECTCOLOR) {
+		fg = ((uint32_t *) info->pseudo_palette)[image->fg_color];
+		bg = ((uint32_t *) info->pseudo_palette)[image->bg_color];
+	} else {
+		fg = image->fg_color;
+		bg = image->bg_color;
+	}
+
+	BEGIN_RING(chan, NvSubGdiRect, 0x0be4, 7);
+	OUT_RING(chan, (image->dy << 16) | (image->dx & 0xffff));
+	OUT_RING(chan, ((image->dy + image->height) << 16) |
+			 ((image->dx + image->width) & 0xffff));
+	OUT_RING(chan, bg);
+	OUT_RING(chan, fg);
+	OUT_RING(chan, (image->height << 16) | image->width);
+	OUT_RING(chan, (image->height << 16) | width);
+	OUT_RING(chan, (image->dy << 16) | (image->dx & 0xffff));
+
+	while (dsize) {
+		int iter_len = dsize > 128 ? 128 : dsize;
+
+		if (RING_SPACE(chan, iter_len + 1)) {
+			nouveau_fbcon_gpu_lockup(info);
+			cfb_imageblit(info, image);
+			return;
+		}
+
+		BEGIN_RING(chan, NvSubGdiRect, 0x0c00, iter_len);
+		OUT_RINGp(chan, data, iter_len);
+		data += iter_len;
+		dsize -= iter_len;
+	}
+
+	FIRE_RING(chan);
+}
+
+static int
+nv04_fbcon_grobj_new(struct drm_device *dev, int class, uint32_t handle)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_gpuobj *obj = NULL;
+	int ret;
+
+	ret = nouveau_gpuobj_gr_new(dev_priv->channel, class, &obj);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_ref_add(dev, dev_priv->channel, handle, obj, NULL);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+int
+nv04_fbcon_accel_init(struct fb_info *info)
+{
+	struct nouveau_fbcon_par *par = info->par;
+	struct drm_device *dev = par->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_channel *chan = dev_priv->channel;
+	const int sub = NvSubCtxSurf2D;
+	int surface_fmt, pattern_fmt, rect_fmt;
+	int ret;
+
+	switch (info->var.bits_per_pixel) {
+	case 8:
+		surface_fmt = 1;
+		pattern_fmt = 3;
+		rect_fmt = 3;
+		break;
+	case 16:
+		surface_fmt = 4;
+		pattern_fmt = 1;
+		rect_fmt = 1;
+		break;
+	case 32:
+		switch (info->var.transp.length) {
+		case 0: /* depth 24 */
+		case 8: /* depth 32 */
+			break;
+		default:
+			return -EINVAL;
+		}
+
+		surface_fmt = 6;
+		pattern_fmt = 3;
+		rect_fmt = 3;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	ret = nv04_fbcon_grobj_new(dev, dev_priv->card_type >= NV_10 ?
+				   0x0062 : 0x0042, NvCtxSurf2D);
+	if (ret)
+		return ret;
+
+	ret = nv04_fbcon_grobj_new(dev, 0x0019, NvClipRect);
+	if (ret)
+		return ret;
+
+	ret = nv04_fbcon_grobj_new(dev, 0x0043, NvRop);
+	if (ret)
+		return ret;
+
+	ret = nv04_fbcon_grobj_new(dev, 0x0044, NvImagePatt);
+	if (ret)
+		return ret;
+
+	ret = nv04_fbcon_grobj_new(dev, 0x004a, NvGdiRect);
+	if (ret)
+		return ret;
+
+	ret = nv04_fbcon_grobj_new(dev, dev_priv->card_type >= NV_10 ?
+				   0x009f : 0x005f, NvImageBlit);
+	if (ret)
+		return ret;
+
+	if (RING_SPACE(chan, 49)) {
+		nouveau_fbcon_gpu_lockup(info);
+		return 0;
+	}
+
+	BEGIN_RING(chan, sub, 0x0000, 1);
+	OUT_RING(chan, NvCtxSurf2D);
+	BEGIN_RING(chan, sub, 0x0184, 2);
+	OUT_RING(chan, NvDmaFB);
+	OUT_RING(chan, NvDmaFB);
+	BEGIN_RING(chan, sub, 0x0300, 4);
+	OUT_RING(chan, surface_fmt);
+	OUT_RING(chan, info->fix.line_length | (info->fix.line_length << 16));
+	OUT_RING(chan, info->fix.smem_start - dev->mode_config.fb_base);
+	OUT_RING(chan, info->fix.smem_start - dev->mode_config.fb_base);
+
+	BEGIN_RING(chan, sub, 0x0000, 1);
+	OUT_RING(chan, NvRop);
+	BEGIN_RING(chan, sub, 0x0300, 1);
+	OUT_RING(chan, 0x55);
+
+	BEGIN_RING(chan, sub, 0x0000, 1);
+	OUT_RING(chan, NvImagePatt);
+	BEGIN_RING(chan, sub, 0x0300, 8);
+	OUT_RING(chan, pattern_fmt);
+#ifdef __BIG_ENDIAN
+	OUT_RING(chan, 2);
+#else
+	OUT_RING(chan, 1);
+#endif
+	OUT_RING(chan, 0);
+	OUT_RING(chan, 1);
+	OUT_RING(chan, ~0);
+	OUT_RING(chan, ~0);
+	OUT_RING(chan, ~0);
+	OUT_RING(chan, ~0);
+
+	BEGIN_RING(chan, sub, 0x0000, 1);
+	OUT_RING(chan, NvClipRect);
+	BEGIN_RING(chan, sub, 0x0300, 2);
+	OUT_RING(chan, 0);
+	OUT_RING(chan, (info->var.yres_virtual << 16) | info->var.xres_virtual);
+
+	BEGIN_RING(chan, NvSubImageBlit, 0x0000, 1);
+	OUT_RING(chan, NvImageBlit);
+	BEGIN_RING(chan, NvSubImageBlit, 0x019c, 1);
+	OUT_RING(chan, NvCtxSurf2D);
+	BEGIN_RING(chan, NvSubImageBlit, 0x02fc, 1);
+	OUT_RING(chan, 3);
+
+	BEGIN_RING(chan, NvSubGdiRect, 0x0000, 1);
+	OUT_RING(chan, NvGdiRect);
+	BEGIN_RING(chan, NvSubGdiRect, 0x0198, 1);
+	OUT_RING(chan, NvCtxSurf2D);
+	BEGIN_RING(chan, NvSubGdiRect, 0x0188, 2);
+	OUT_RING(chan, NvImagePatt);
+	OUT_RING(chan, NvRop);
+	BEGIN_RING(chan, NvSubGdiRect, 0x0304, 1);
+	OUT_RING(chan, 1);
+	BEGIN_RING(chan, NvSubGdiRect, 0x0300, 1);
+	OUT_RING(chan, rect_fmt);
+	BEGIN_RING(chan, NvSubGdiRect, 0x02fc, 1);
+	OUT_RING(chan, 3);
+
+	FIRE_RING(chan);
+
+	return 0;
+}
+
diff --git a/drivers/gpu/drm/nouveau/nv04_fifo.c b/drivers/gpu/drm/nouveau/nv04_fifo.c
new file mode 100644
index 000000000000..66fe55983b6e
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv04_fifo.c
@@ -0,0 +1,310 @@
+/*
+ * Copyright (C) 2007 Ben Skeggs.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "drmP.h"
+#include "drm.h"
+#include "nouveau_drv.h"
+
+#define NV04_RAMFC(c) (dev_priv->ramfc_offset + ((c) * NV04_RAMFC__SIZE))
+#define NV04_RAMFC__SIZE 32
+#define NV04_RAMFC_DMA_PUT                                       0x00
+#define NV04_RAMFC_DMA_GET                                       0x04
+#define NV04_RAMFC_DMA_INSTANCE                                  0x08
+#define NV04_RAMFC_DMA_STATE                                     0x0C
+#define NV04_RAMFC_DMA_FETCH                                     0x10
+#define NV04_RAMFC_ENGINE                                        0x14
+#define NV04_RAMFC_PULL1_ENGINE                                  0x18
+
+#define RAMFC_WR(offset, val) nv_wo32(dev, chan->ramfc->gpuobj, \
+					 NV04_RAMFC_##offset/4, (val))
+#define RAMFC_RD(offset)      nv_ro32(dev, chan->ramfc->gpuobj, \
+					 NV04_RAMFC_##offset/4)
+
+void
+nv04_fifo_disable(struct drm_device *dev)
+{
+	uint32_t tmp;
+
+	tmp = nv_rd32(dev, NV04_PFIFO_CACHE1_DMA_PUSH);
+	nv_wr32(dev, NV04_PFIFO_CACHE1_DMA_PUSH, tmp & ~1);
+	nv_wr32(dev, NV03_PFIFO_CACHE1_PUSH0, 0);
+	tmp = nv_rd32(dev, NV03_PFIFO_CACHE1_PULL1);
+	nv_wr32(dev, NV04_PFIFO_CACHE1_PULL0, tmp & ~1);
+}
+
+void
+nv04_fifo_enable(struct drm_device *dev)
+{
+	nv_wr32(dev, NV03_PFIFO_CACHE1_PUSH0, 1);
+	nv_wr32(dev, NV04_PFIFO_CACHE1_PULL0, 1);
+}
+
+bool
+nv04_fifo_reassign(struct drm_device *dev, bool enable)
+{
+	uint32_t reassign = nv_rd32(dev, NV03_PFIFO_CACHES);
+
+	nv_wr32(dev, NV03_PFIFO_CACHES, enable ? 1 : 0);
+	return (reassign == 1);
+}
+
+bool
+nv04_fifo_cache_flush(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_timer_engine *ptimer = &dev_priv->engine.timer;
+	uint64_t start = ptimer->read(dev);
+
+	do {
+		if (nv_rd32(dev, NV03_PFIFO_CACHE1_GET) ==
+		    nv_rd32(dev, NV03_PFIFO_CACHE1_PUT))
+			return true;
+
+	} while (ptimer->read(dev) - start < 100000000);
+
+	NV_ERROR(dev, "Timeout flushing the PFIFO cache.\n");
+
+	return false;
+}
+
+bool
+nv04_fifo_cache_pull(struct drm_device *dev, bool enable)
+{
+	uint32_t pull = nv_rd32(dev, NV04_PFIFO_CACHE1_PULL0);
+
+	if (enable) {
+		nv_wr32(dev, NV04_PFIFO_CACHE1_PULL0, pull | 1);
+	} else {
+		nv_wr32(dev, NV04_PFIFO_CACHE1_PULL0, pull & ~1);
+		nv_wr32(dev, NV04_PFIFO_CACHE1_HASH, 0);
+	}
+
+	return !!(pull & 1);
+}
+
+int
+nv04_fifo_channel_id(struct drm_device *dev)
+{
+	return nv_rd32(dev, NV03_PFIFO_CACHE1_PUSH1) &
+			NV03_PFIFO_CACHE1_PUSH1_CHID_MASK;
+}
+
+int
+nv04_fifo_create_context(struct nouveau_channel *chan)
+{
+	struct drm_device *dev = chan->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	unsigned long flags;
+	int ret;
+
+	ret = nouveau_gpuobj_new_fake(dev, NV04_RAMFC(chan->id), ~0,
+						NV04_RAMFC__SIZE,
+						NVOBJ_FLAG_ZERO_ALLOC |
+						NVOBJ_FLAG_ZERO_FREE,
+						NULL, &chan->ramfc);
+	if (ret)
+		return ret;
+
+	spin_lock_irqsave(&dev_priv->context_switch_lock, flags);
+
+	/* Setup initial state */
+	dev_priv->engine.instmem.prepare_access(dev, true);
+	RAMFC_WR(DMA_PUT, chan->pushbuf_base);
+	RAMFC_WR(DMA_GET, chan->pushbuf_base);
+	RAMFC_WR(DMA_INSTANCE, chan->pushbuf->instance >> 4);
+	RAMFC_WR(DMA_FETCH, (NV_PFIFO_CACHE1_DMA_FETCH_TRIG_128_BYTES |
+			     NV_PFIFO_CACHE1_DMA_FETCH_SIZE_128_BYTES |
+			     NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_8 |
+#ifdef __BIG_ENDIAN
+			     NV_PFIFO_CACHE1_BIG_ENDIAN |
+#endif
+			     0));
+	dev_priv->engine.instmem.finish_access(dev);
+
+	/* enable the fifo dma operation */
+	nv_wr32(dev, NV04_PFIFO_MODE,
+		nv_rd32(dev, NV04_PFIFO_MODE) | (1 << chan->id));
+
+	spin_unlock_irqrestore(&dev_priv->context_switch_lock, flags);
+	return 0;
+}
+
+void
+nv04_fifo_destroy_context(struct nouveau_channel *chan)
+{
+	struct drm_device *dev = chan->dev;
+
+	nv_wr32(dev, NV04_PFIFO_MODE,
+		nv_rd32(dev, NV04_PFIFO_MODE) & ~(1 << chan->id));
+
+	nouveau_gpuobj_ref_del(dev, &chan->ramfc);
+}
+
+static void
+nv04_fifo_do_load_context(struct drm_device *dev, int chid)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t fc = NV04_RAMFC(chid), tmp;
+
+	dev_priv->engine.instmem.prepare_access(dev, false);
+
+	nv_wr32(dev, NV04_PFIFO_CACHE1_DMA_PUT, nv_ri32(dev, fc + 0));
+	nv_wr32(dev, NV04_PFIFO_CACHE1_DMA_GET, nv_ri32(dev, fc + 4));
+	tmp = nv_ri32(dev, fc + 8);
+	nv_wr32(dev, NV04_PFIFO_CACHE1_DMA_INSTANCE, tmp & 0xFFFF);
+	nv_wr32(dev, NV04_PFIFO_CACHE1_DMA_DCOUNT, tmp >> 16);
+	nv_wr32(dev, NV04_PFIFO_CACHE1_DMA_STATE, nv_ri32(dev, fc + 12));
+	nv_wr32(dev, NV04_PFIFO_CACHE1_DMA_FETCH, nv_ri32(dev, fc + 16));
+	nv_wr32(dev, NV04_PFIFO_CACHE1_ENGINE, nv_ri32(dev, fc + 20));
+	nv_wr32(dev, NV04_PFIFO_CACHE1_PULL1, nv_ri32(dev, fc + 24));
+
+	dev_priv->engine.instmem.finish_access(dev);
+
+	nv_wr32(dev, NV03_PFIFO_CACHE1_GET, 0);
+	nv_wr32(dev, NV03_PFIFO_CACHE1_PUT, 0);
+}
+
+int
+nv04_fifo_load_context(struct nouveau_channel *chan)
+{
+	uint32_t tmp;
+
+	nv_wr32(chan->dev, NV03_PFIFO_CACHE1_PUSH1,
+			   NV03_PFIFO_CACHE1_PUSH1_DMA | chan->id);
+	nv04_fifo_do_load_context(chan->dev, chan->id);
+	nv_wr32(chan->dev, NV04_PFIFO_CACHE1_DMA_PUSH, 1);
+
+	/* Reset NV04_PFIFO_CACHE1_DMA_CTL_AT_INFO to INVALID */
+	tmp = nv_rd32(chan->dev, NV04_PFIFO_CACHE1_DMA_CTL) & ~(1 << 31);
+	nv_wr32(chan->dev, NV04_PFIFO_CACHE1_DMA_CTL, tmp);
+
+	return 0;
+}
+
+int
+nv04_fifo_unload_context(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_fifo_engine *pfifo = &dev_priv->engine.fifo;
+	struct nouveau_channel *chan = NULL;
+	uint32_t tmp;
+	int chid;
+
+	chid = pfifo->channel_id(dev);
+	if (chid < 0 || chid >= dev_priv->engine.fifo.channels)
+		return 0;
+
+	chan = dev_priv->fifos[chid];
+	if (!chan) {
+		NV_ERROR(dev, "Inactive channel on PFIFO: %d\n", chid);
+		return -EINVAL;
+	}
+
+	dev_priv->engine.instmem.prepare_access(dev, true);
+	RAMFC_WR(DMA_PUT, nv_rd32(dev, NV04_PFIFO_CACHE1_DMA_PUT));
+	RAMFC_WR(DMA_GET, nv_rd32(dev, NV04_PFIFO_CACHE1_DMA_GET));
+	tmp  = nv_rd32(dev, NV04_PFIFO_CACHE1_DMA_DCOUNT) << 16;
+	tmp |= nv_rd32(dev, NV04_PFIFO_CACHE1_DMA_INSTANCE);
+	RAMFC_WR(DMA_INSTANCE, tmp);
+	RAMFC_WR(DMA_STATE, nv_rd32(dev, NV04_PFIFO_CACHE1_DMA_STATE));
+	RAMFC_WR(DMA_FETCH, nv_rd32(dev, NV04_PFIFO_CACHE1_DMA_FETCH));
+	RAMFC_WR(ENGINE, nv_rd32(dev, NV04_PFIFO_CACHE1_ENGINE));
+	RAMFC_WR(PULL1_ENGINE, nv_rd32(dev, NV04_PFIFO_CACHE1_PULL1));
+	dev_priv->engine.instmem.finish_access(dev);
+
+	nv04_fifo_do_load_context(dev, pfifo->channels - 1);
+	nv_wr32(dev, NV03_PFIFO_CACHE1_PUSH1, pfifo->channels - 1);
+	return 0;
+}
+
+static void
+nv04_fifo_init_reset(struct drm_device *dev)
+{
+	nv_wr32(dev, NV03_PMC_ENABLE,
+		nv_rd32(dev, NV03_PMC_ENABLE) & ~NV_PMC_ENABLE_PFIFO);
+	nv_wr32(dev, NV03_PMC_ENABLE,
+		nv_rd32(dev, NV03_PMC_ENABLE) |  NV_PMC_ENABLE_PFIFO);
+
+	nv_wr32(dev, 0x003224, 0x000f0078);
+	nv_wr32(dev, 0x002044, 0x0101ffff);
+	nv_wr32(dev, 0x002040, 0x000000ff);
+	nv_wr32(dev, 0x002500, 0x00000000);
+	nv_wr32(dev, 0x003000, 0x00000000);
+	nv_wr32(dev, 0x003050, 0x00000000);
+	nv_wr32(dev, 0x003200, 0x00000000);
+	nv_wr32(dev, 0x003250, 0x00000000);
+	nv_wr32(dev, 0x003220, 0x00000000);
+
+	nv_wr32(dev, 0x003250, 0x00000000);
+	nv_wr32(dev, 0x003270, 0x00000000);
+	nv_wr32(dev, 0x003210, 0x00000000);
+}
+
+static void
+nv04_fifo_init_ramxx(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	nv_wr32(dev, NV03_PFIFO_RAMHT, (0x03 << 24) /* search 128 */ |
+				       ((dev_priv->ramht_bits - 9) << 16) |
+				       (dev_priv->ramht_offset >> 8));
+	nv_wr32(dev, NV03_PFIFO_RAMRO, dev_priv->ramro_offset>>8);
+	nv_wr32(dev, NV03_PFIFO_RAMFC, dev_priv->ramfc_offset >> 8);
+}
+
+static void
+nv04_fifo_init_intr(struct drm_device *dev)
+{
+	nv_wr32(dev, 0x002100, 0xffffffff);
+	nv_wr32(dev, 0x002140, 0xffffffff);
+}
+
+int
+nv04_fifo_init(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_fifo_engine *pfifo = &dev_priv->engine.fifo;
+	int i;
+
+	nv04_fifo_init_reset(dev);
+	nv04_fifo_init_ramxx(dev);
+
+	nv04_fifo_do_load_context(dev, pfifo->channels - 1);
+	nv_wr32(dev, NV03_PFIFO_CACHE1_PUSH1, pfifo->channels - 1);
+
+	nv04_fifo_init_intr(dev);
+	pfifo->enable(dev);
+
+	for (i = 0; i < dev_priv->engine.fifo.channels; i++) {
+		if (dev_priv->fifos[i]) {
+			uint32_t mode = nv_rd32(dev, NV04_PFIFO_MODE);
+			nv_wr32(dev, NV04_PFIFO_MODE, mode | (1 << i));
+		}
+	}
+
+	return 0;
+}
+
diff --git a/drivers/gpu/drm/nouveau/nv04_graph.c b/drivers/gpu/drm/nouveau/nv04_graph.c
new file mode 100644
index 000000000000..e260986ea65a
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv04_graph.c
@@ -0,0 +1,584 @@
+/*
+ * Copyright 2007 Stephane Marchesin
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include "drmP.h"
+#include "drm.h"
+#include "nouveau_drm.h"
+#include "nouveau_drv.h"
+
+static uint32_t nv04_graph_ctx_regs[] = {
+	0x0040053c,
+	0x00400544,
+	0x00400540,
+	0x00400548,
+	NV04_PGRAPH_CTX_SWITCH1,
+	NV04_PGRAPH_CTX_SWITCH2,
+	NV04_PGRAPH_CTX_SWITCH3,
+	NV04_PGRAPH_CTX_SWITCH4,
+	NV04_PGRAPH_CTX_CACHE1,
+	NV04_PGRAPH_CTX_CACHE2,
+	NV04_PGRAPH_CTX_CACHE3,
+	NV04_PGRAPH_CTX_CACHE4,
+	0x00400184,
+	0x004001a4,
+	0x004001c4,
+	0x004001e4,
+	0x00400188,
+	0x004001a8,
+	0x004001c8,
+	0x004001e8,
+	0x0040018c,
+	0x004001ac,
+	0x004001cc,
+	0x004001ec,
+	0x00400190,
+	0x004001b0,
+	0x004001d0,
+	0x004001f0,
+	0x00400194,
+	0x004001b4,
+	0x004001d4,
+	0x004001f4,
+	0x00400198,
+	0x004001b8,
+	0x004001d8,
+	0x004001f8,
+	0x0040019c,
+	0x004001bc,
+	0x004001dc,
+	0x004001fc,
+	0x00400174,
+	NV04_PGRAPH_DMA_START_0,
+	NV04_PGRAPH_DMA_START_1,
+	NV04_PGRAPH_DMA_LENGTH,
+	NV04_PGRAPH_DMA_MISC,
+	NV04_PGRAPH_DMA_PITCH,
+	NV04_PGRAPH_BOFFSET0,
+	NV04_PGRAPH_BBASE0,
+	NV04_PGRAPH_BLIMIT0,
+	NV04_PGRAPH_BOFFSET1,
+	NV04_PGRAPH_BBASE1,
+	NV04_PGRAPH_BLIMIT1,
+	NV04_PGRAPH_BOFFSET2,
+	NV04_PGRAPH_BBASE2,
+	NV04_PGRAPH_BLIMIT2,
+	NV04_PGRAPH_BOFFSET3,
+	NV04_PGRAPH_BBASE3,
+	NV04_PGRAPH_BLIMIT3,
+	NV04_PGRAPH_BOFFSET4,
+	NV04_PGRAPH_BBASE4,
+	NV04_PGRAPH_BLIMIT4,
+	NV04_PGRAPH_BOFFSET5,
+	NV04_PGRAPH_BBASE5,
+	NV04_PGRAPH_BLIMIT5,
+	NV04_PGRAPH_BPITCH0,
+	NV04_PGRAPH_BPITCH1,
+	NV04_PGRAPH_BPITCH2,
+	NV04_PGRAPH_BPITCH3,
+	NV04_PGRAPH_BPITCH4,
+	NV04_PGRAPH_SURFACE,
+	NV04_PGRAPH_STATE,
+	NV04_PGRAPH_BSWIZZLE2,
+	NV04_PGRAPH_BSWIZZLE5,
+	NV04_PGRAPH_BPIXEL,
+	NV04_PGRAPH_NOTIFY,
+	NV04_PGRAPH_PATT_COLOR0,
+	NV04_PGRAPH_PATT_COLOR1,
+	NV04_PGRAPH_PATT_COLORRAM+0x00,
+	NV04_PGRAPH_PATT_COLORRAM+0x04,
+	NV04_PGRAPH_PATT_COLORRAM+0x08,
+	NV04_PGRAPH_PATT_COLORRAM+0x0c,
+	NV04_PGRAPH_PATT_COLORRAM+0x10,
+	NV04_PGRAPH_PATT_COLORRAM+0x14,
+	NV04_PGRAPH_PATT_COLORRAM+0x18,
+	NV04_PGRAPH_PATT_COLORRAM+0x1c,
+	NV04_PGRAPH_PATT_COLORRAM+0x20,
+	NV04_PGRAPH_PATT_COLORRAM+0x24,
+	NV04_PGRAPH_PATT_COLORRAM+0x28,
+	NV04_PGRAPH_PATT_COLORRAM+0x2c,
+	NV04_PGRAPH_PATT_COLORRAM+0x30,
+	NV04_PGRAPH_PATT_COLORRAM+0x34,
+	NV04_PGRAPH_PATT_COLORRAM+0x38,
+	NV04_PGRAPH_PATT_COLORRAM+0x3c,
+	NV04_PGRAPH_PATT_COLORRAM+0x40,
+	NV04_PGRAPH_PATT_COLORRAM+0x44,
+	NV04_PGRAPH_PATT_COLORRAM+0x48,
+	NV04_PGRAPH_PATT_COLORRAM+0x4c,
+	NV04_PGRAPH_PATT_COLORRAM+0x50,
+	NV04_PGRAPH_PATT_COLORRAM+0x54,
+	NV04_PGRAPH_PATT_COLORRAM+0x58,
+	NV04_PGRAPH_PATT_COLORRAM+0x5c,
+	NV04_PGRAPH_PATT_COLORRAM+0x60,
+	NV04_PGRAPH_PATT_COLORRAM+0x64,
+	NV04_PGRAPH_PATT_COLORRAM+0x68,
+	NV04_PGRAPH_PATT_COLORRAM+0x6c,
+	NV04_PGRAPH_PATT_COLORRAM+0x70,
+	NV04_PGRAPH_PATT_COLORRAM+0x74,
+	NV04_PGRAPH_PATT_COLORRAM+0x78,
+	NV04_PGRAPH_PATT_COLORRAM+0x7c,
+	NV04_PGRAPH_PATT_COLORRAM+0x80,
+	NV04_PGRAPH_PATT_COLORRAM+0x84,
+	NV04_PGRAPH_PATT_COLORRAM+0x88,
+	NV04_PGRAPH_PATT_COLORRAM+0x8c,
+	NV04_PGRAPH_PATT_COLORRAM+0x90,
+	NV04_PGRAPH_PATT_COLORRAM+0x94,
+	NV04_PGRAPH_PATT_COLORRAM+0x98,
+	NV04_PGRAPH_PATT_COLORRAM+0x9c,
+	NV04_PGRAPH_PATT_COLORRAM+0xa0,
+	NV04_PGRAPH_PATT_COLORRAM+0xa4,
+	NV04_PGRAPH_PATT_COLORRAM+0xa8,
+	NV04_PGRAPH_PATT_COLORRAM+0xac,
+	NV04_PGRAPH_PATT_COLORRAM+0xb0,
+	NV04_PGRAPH_PATT_COLORRAM+0xb4,
+	NV04_PGRAPH_PATT_COLORRAM+0xb8,
+	NV04_PGRAPH_PATT_COLORRAM+0xbc,
+	NV04_PGRAPH_PATT_COLORRAM+0xc0,
+	NV04_PGRAPH_PATT_COLORRAM+0xc4,
+	NV04_PGRAPH_PATT_COLORRAM+0xc8,
+	NV04_PGRAPH_PATT_COLORRAM+0xcc,
+	NV04_PGRAPH_PATT_COLORRAM+0xd0,
+	NV04_PGRAPH_PATT_COLORRAM+0xd4,
+	NV04_PGRAPH_PATT_COLORRAM+0xd8,
+	NV04_PGRAPH_PATT_COLORRAM+0xdc,
+	NV04_PGRAPH_PATT_COLORRAM+0xe0,
+	NV04_PGRAPH_PATT_COLORRAM+0xe4,
+	NV04_PGRAPH_PATT_COLORRAM+0xe8,
+	NV04_PGRAPH_PATT_COLORRAM+0xec,
+	NV04_PGRAPH_PATT_COLORRAM+0xf0,
+	NV04_PGRAPH_PATT_COLORRAM+0xf4,
+	NV04_PGRAPH_PATT_COLORRAM+0xf8,
+	NV04_PGRAPH_PATT_COLORRAM+0xfc,
+	NV04_PGRAPH_PATTERN,
+	0x0040080c,
+	NV04_PGRAPH_PATTERN_SHAPE,
+	0x00400600,
+	NV04_PGRAPH_ROP3,
+	NV04_PGRAPH_CHROMA,
+	NV04_PGRAPH_BETA_AND,
+	NV04_PGRAPH_BETA_PREMULT,
+	NV04_PGRAPH_CONTROL0,
+	NV04_PGRAPH_CONTROL1,
+	NV04_PGRAPH_CONTROL2,
+	NV04_PGRAPH_BLEND,
+	NV04_PGRAPH_STORED_FMT,
+	NV04_PGRAPH_SOURCE_COLOR,
+	0x00400560,
+	0x00400568,
+	0x00400564,
+	0x0040056c,
+	0x00400400,
+	0x00400480,
+	0x00400404,
+	0x00400484,
+	0x00400408,
+	0x00400488,
+	0x0040040c,
+	0x0040048c,
+	0x00400410,
+	0x00400490,
+	0x00400414,
+	0x00400494,
+	0x00400418,
+	0x00400498,
+	0x0040041c,
+	0x0040049c,
+	0x00400420,
+	0x004004a0,
+	0x00400424,
+	0x004004a4,
+	0x00400428,
+	0x004004a8,
+	0x0040042c,
+	0x004004ac,
+	0x00400430,
+	0x004004b0,
+	0x00400434,
+	0x004004b4,
+	0x00400438,
+	0x004004b8,
+	0x0040043c,
+	0x004004bc,
+	0x00400440,
+	0x004004c0,
+	0x00400444,
+	0x004004c4,
+	0x00400448,
+	0x004004c8,
+	0x0040044c,
+	0x004004cc,
+	0x00400450,
+	0x004004d0,
+	0x00400454,
+	0x004004d4,
+	0x00400458,
+	0x004004d8,
+	0x0040045c,
+	0x004004dc,
+	0x00400460,
+	0x004004e0,
+	0x00400464,
+	0x004004e4,
+	0x00400468,
+	0x004004e8,
+	0x0040046c,
+	0x004004ec,
+	0x00400470,
+	0x004004f0,
+	0x00400474,
+	0x004004f4,
+	0x00400478,
+	0x004004f8,
+	0x0040047c,
+	0x004004fc,
+	0x00400534,
+	0x00400538,
+	0x00400514,
+	0x00400518,
+	0x0040051c,
+	0x00400520,
+	0x00400524,
+	0x00400528,
+	0x0040052c,
+	0x00400530,
+	0x00400d00,
+	0x00400d40,
+	0x00400d80,
+	0x00400d04,
+	0x00400d44,
+	0x00400d84,
+	0x00400d08,
+	0x00400d48,
+	0x00400d88,
+	0x00400d0c,
+	0x00400d4c,
+	0x00400d8c,
+	0x00400d10,
+	0x00400d50,
+	0x00400d90,
+	0x00400d14,
+	0x00400d54,
+	0x00400d94,
+	0x00400d18,
+	0x00400d58,
+	0x00400d98,
+	0x00400d1c,
+	0x00400d5c,
+	0x00400d9c,
+	0x00400d20,
+	0x00400d60,
+	0x00400da0,
+	0x00400d24,
+	0x00400d64,
+	0x00400da4,
+	0x00400d28,
+	0x00400d68,
+	0x00400da8,
+	0x00400d2c,
+	0x00400d6c,
+	0x00400dac,
+	0x00400d30,
+	0x00400d70,
+	0x00400db0,
+	0x00400d34,
+	0x00400d74,
+	0x00400db4,
+	0x00400d38,
+	0x00400d78,
+	0x00400db8,
+	0x00400d3c,
+	0x00400d7c,
+	0x00400dbc,
+	0x00400590,
+	0x00400594,
+	0x00400598,
+	0x0040059c,
+	0x004005a8,
+	0x004005ac,
+	0x004005b0,
+	0x004005b4,
+	0x004005c0,
+	0x004005c4,
+	0x004005c8,
+	0x004005cc,
+	0x004005d0,
+	0x004005d4,
+	0x004005d8,
+	0x004005dc,
+	0x004005e0,
+	NV04_PGRAPH_PASSTHRU_0,
+	NV04_PGRAPH_PASSTHRU_1,
+	NV04_PGRAPH_PASSTHRU_2,
+	NV04_PGRAPH_DVD_COLORFMT,
+	NV04_PGRAPH_SCALED_FORMAT,
+	NV04_PGRAPH_MISC24_0,
+	NV04_PGRAPH_MISC24_1,
+	NV04_PGRAPH_MISC24_2,
+	0x00400500,
+	0x00400504,
+	NV04_PGRAPH_VALID1,
+	NV04_PGRAPH_VALID2,
+	NV04_PGRAPH_DEBUG_3
+};
+
+struct graph_state {
+	int nv04[ARRAY_SIZE(nv04_graph_ctx_regs)];
+};
+
+struct nouveau_channel *
+nv04_graph_channel(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	int chid = dev_priv->engine.fifo.channels;
+
+	if (nv_rd32(dev, NV04_PGRAPH_CTX_CONTROL) & 0x00010000)
+		chid = nv_rd32(dev, NV04_PGRAPH_CTX_USER) >> 24;
+
+	if (chid >= dev_priv->engine.fifo.channels)
+		return NULL;
+
+	return dev_priv->fifos[chid];
+}
+
+void
+nv04_graph_context_switch(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
+	struct nouveau_channel *chan = NULL;
+	int chid;
+
+	pgraph->fifo_access(dev, false);
+	nouveau_wait_for_idle(dev);
+
+	/* If previous context is valid, we need to save it */
+	pgraph->unload_context(dev);
+
+	/* Load context for next channel */
+	chid = dev_priv->engine.fifo.channel_id(dev);
+	chan = dev_priv->fifos[chid];
+	if (chan)
+		nv04_graph_load_context(chan);
+
+	pgraph->fifo_access(dev, true);
+}
+
+static uint32_t *ctx_reg(struct graph_state *ctx, uint32_t reg)
+{
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(nv04_graph_ctx_regs); i++) {
+		if (nv04_graph_ctx_regs[i] == reg)
+			return &ctx->nv04[i];
+	}
+
+	return NULL;
+}
+
+int nv04_graph_create_context(struct nouveau_channel *chan)
+{
+	struct graph_state *pgraph_ctx;
+	NV_DEBUG(chan->dev, "nv04_graph_context_create %d\n", chan->id);
+
+	chan->pgraph_ctx = pgraph_ctx = kzalloc(sizeof(*pgraph_ctx),
+						GFP_KERNEL);
+	if (pgraph_ctx == NULL)
+		return -ENOMEM;
+
+	*ctx_reg(pgraph_ctx, NV04_PGRAPH_DEBUG_3) = 0xfad4ff31;
+
+	return 0;
+}
+
+void nv04_graph_destroy_context(struct nouveau_channel *chan)
+{
+	struct graph_state *pgraph_ctx = chan->pgraph_ctx;
+
+	kfree(pgraph_ctx);
+	chan->pgraph_ctx = NULL;
+}
+
+int nv04_graph_load_context(struct nouveau_channel *chan)
+{
+	struct drm_device *dev = chan->dev;
+	struct graph_state *pgraph_ctx = chan->pgraph_ctx;
+	uint32_t tmp;
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(nv04_graph_ctx_regs); i++)
+		nv_wr32(dev, nv04_graph_ctx_regs[i], pgraph_ctx->nv04[i]);
+
+	nv_wr32(dev, NV04_PGRAPH_CTX_CONTROL, 0x10010100);
+
+	tmp  = nv_rd32(dev, NV04_PGRAPH_CTX_USER) & 0x00ffffff;
+	nv_wr32(dev, NV04_PGRAPH_CTX_USER, tmp | chan->id << 24);
+
+	tmp = nv_rd32(dev, NV04_PGRAPH_FFINTFC_ST2);
+	nv_wr32(dev, NV04_PGRAPH_FFINTFC_ST2, tmp & 0x000fffff);
+
+	return 0;
+}
+
+int
+nv04_graph_unload_context(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
+	struct nouveau_channel *chan = NULL;
+	struct graph_state *ctx;
+	uint32_t tmp;
+	int i;
+
+	chan = pgraph->channel(dev);
+	if (!chan)
+		return 0;
+	ctx = chan->pgraph_ctx;
+
+	for (i = 0; i < ARRAY_SIZE(nv04_graph_ctx_regs); i++)
+		ctx->nv04[i] = nv_rd32(dev, nv04_graph_ctx_regs[i]);
+
+	nv_wr32(dev, NV04_PGRAPH_CTX_CONTROL, 0x10000000);
+	tmp  = nv_rd32(dev, NV04_PGRAPH_CTX_USER) & 0x00ffffff;
+	tmp |= (dev_priv->engine.fifo.channels - 1) << 24;
+	nv_wr32(dev, NV04_PGRAPH_CTX_USER, tmp);
+	return 0;
+}
+
+int nv04_graph_init(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t tmp;
+
+	nv_wr32(dev, NV03_PMC_ENABLE, nv_rd32(dev, NV03_PMC_ENABLE) &
+			~NV_PMC_ENABLE_PGRAPH);
+	nv_wr32(dev, NV03_PMC_ENABLE, nv_rd32(dev, NV03_PMC_ENABLE) |
+			 NV_PMC_ENABLE_PGRAPH);
+
+	/* Enable PGRAPH interrupts */
+	nv_wr32(dev, NV03_PGRAPH_INTR, 0xFFFFFFFF);
+	nv_wr32(dev, NV03_PGRAPH_INTR_EN, 0xFFFFFFFF);
+
+	nv_wr32(dev, NV04_PGRAPH_VALID1, 0);
+	nv_wr32(dev, NV04_PGRAPH_VALID2, 0);
+	/*nv_wr32(dev, NV04_PGRAPH_DEBUG_0, 0x000001FF);
+	nv_wr32(dev, NV04_PGRAPH_DEBUG_0, 0x001FFFFF);*/
+	nv_wr32(dev, NV04_PGRAPH_DEBUG_0, 0x1231c000);
+	/*1231C000 blob, 001 haiku*/
+	//*V_WRITE(NV04_PGRAPH_DEBUG_1, 0xf2d91100);*/
+	nv_wr32(dev, NV04_PGRAPH_DEBUG_1, 0x72111100);
+	/*0x72111100 blob , 01 haiku*/
+	/*nv_wr32(dev, NV04_PGRAPH_DEBUG_2, 0x11d5f870);*/
+	nv_wr32(dev, NV04_PGRAPH_DEBUG_2, 0x11d5f071);
+	/*haiku same*/
+
+	/*nv_wr32(dev, NV04_PGRAPH_DEBUG_3, 0xfad4ff31);*/
+	nv_wr32(dev, NV04_PGRAPH_DEBUG_3, 0xf0d4ff31);
+	/*haiku and blob 10d4*/
+
+	nv_wr32(dev, NV04_PGRAPH_STATE        , 0xFFFFFFFF);
+	nv_wr32(dev, NV04_PGRAPH_CTX_CONTROL  , 0x10000100);
+	tmp  = nv_rd32(dev, NV04_PGRAPH_CTX_USER) & 0x00ffffff;
+	tmp |= (dev_priv->engine.fifo.channels - 1) << 24;
+	nv_wr32(dev, NV04_PGRAPH_CTX_USER, tmp);
+
+	/* These don't belong here, they're part of a per-channel context */
+	nv_wr32(dev, NV04_PGRAPH_PATTERN_SHAPE, 0x00000000);
+	nv_wr32(dev, NV04_PGRAPH_BETA_AND     , 0xFFFFFFFF);
+
+	return 0;
+}
+
+void nv04_graph_takedown(struct drm_device *dev)
+{
+}
+
+void
+nv04_graph_fifo_access(struct drm_device *dev, bool enabled)
+{
+	if (enabled)
+		nv_wr32(dev, NV04_PGRAPH_FIFO,
+					nv_rd32(dev, NV04_PGRAPH_FIFO) | 1);
+	else
+		nv_wr32(dev, NV04_PGRAPH_FIFO,
+					nv_rd32(dev, NV04_PGRAPH_FIFO) & ~1);
+}
+
+static int
+nv04_graph_mthd_set_ref(struct nouveau_channel *chan, int grclass,
+			int mthd, uint32_t data)
+{
+	chan->fence.last_sequence_irq = data;
+	nouveau_fence_handler(chan->dev, chan->id);
+	return 0;
+}
+
+static int
+nv04_graph_mthd_set_operation(struct nouveau_channel *chan, int grclass,
+			      int mthd, uint32_t data)
+{
+	struct drm_device *dev = chan->dev;
+	uint32_t instance = (nv_rd32(dev, NV04_PGRAPH_CTX_SWITCH4) & 0xffff) << 4;
+	int subc = (nv_rd32(dev, NV04_PGRAPH_TRAPPED_ADDR) >> 13) & 0x7;
+	uint32_t tmp;
+
+	tmp  = nv_ri32(dev, instance);
+	tmp &= ~0x00038000;
+	tmp |= ((data & 7) << 15);
+
+	nv_wi32(dev, instance, tmp);
+	nv_wr32(dev, NV04_PGRAPH_CTX_SWITCH1, tmp);
+	nv_wr32(dev, NV04_PGRAPH_CTX_CACHE1 + (subc<<2), tmp);
+	return 0;
+}
+
+static struct nouveau_pgraph_object_method nv04_graph_mthds_sw[] = {
+	{ 0x0150, nv04_graph_mthd_set_ref },
+	{}
+};
+
+static struct nouveau_pgraph_object_method nv04_graph_mthds_set_operation[] = {
+	{ 0x02fc, nv04_graph_mthd_set_operation },
+	{},
+};
+
+struct nouveau_pgraph_object_class nv04_graph_grclass[] = {
+	{ 0x0039, false, NULL },
+	{ 0x004a, false, nv04_graph_mthds_set_operation }, /* gdirect */
+	{ 0x005f, false, nv04_graph_mthds_set_operation }, /* imageblit */
+	{ 0x0061, false, nv04_graph_mthds_set_operation }, /* ifc */
+	{ 0x0077, false, nv04_graph_mthds_set_operation }, /* sifm */
+	{ 0x0030, false, NULL }, /* null */
+	{ 0x0042, false, NULL }, /* surf2d */
+	{ 0x0043, false, NULL }, /* rop */
+	{ 0x0012, false, NULL }, /* beta1 */
+	{ 0x0072, false, NULL }, /* beta4 */
+	{ 0x0019, false, NULL }, /* cliprect */
+	{ 0x0044, false, NULL }, /* pattern */
+	{ 0x0052, false, NULL }, /* swzsurf */
+	{ 0x0053, false, NULL }, /* surf3d */
+	{ 0x0054, false, NULL }, /* tex_tri */
+	{ 0x0055, false, NULL }, /* multitex_tri */
+	{ 0x506e, true, nv04_graph_mthds_sw },
+	{}
+};
+
diff --git a/drivers/gpu/drm/nouveau/nv04_instmem.c b/drivers/gpu/drm/nouveau/nv04_instmem.c
new file mode 100644
index 000000000000..a3b9563a6f60
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv04_instmem.c
@@ -0,0 +1,208 @@
+#include "drmP.h"
+#include "drm.h"
+#include "nouveau_drv.h"
+
+/* returns the size of fifo context */
+static int
+nouveau_fifo_ctx_size(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	if (dev_priv->chipset >= 0x40)
+		return 128;
+	else
+	if (dev_priv->chipset >= 0x17)
+		return 64;
+
+	return 32;
+}
+
+static void
+nv04_instmem_determine_amount(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	int i;
+
+	/* Figure out how much instance memory we need */
+	if (dev_priv->card_type >= NV_40) {
+		/* We'll want more instance memory than this on some NV4x cards.
+		 * There's a 16MB aperture to play with that maps onto the end
+		 * of vram.  For now, only reserve a small piece until we know
+		 * more about what each chipset requires.
+		 */
+		switch (dev_priv->chipset) {
+		case 0x40:
+		case 0x47:
+		case 0x49:
+		case 0x4b:
+			dev_priv->ramin_rsvd_vram = (2 * 1024 * 1024);
+			break;
+		default:
+			dev_priv->ramin_rsvd_vram = (1 * 1024 * 1024);
+			break;
+		}
+	} else {
+		/*XXX: what *are* the limits on <NV40 cards?
+		 */
+		dev_priv->ramin_rsvd_vram = (512 * 1024);
+	}
+	NV_DEBUG(dev, "RAMIN size: %dKiB\n", dev_priv->ramin_rsvd_vram >> 10);
+
+	/* Clear all of it, except the BIOS image that's in the first 64KiB */
+	dev_priv->engine.instmem.prepare_access(dev, true);
+	for (i = 64 * 1024; i < dev_priv->ramin_rsvd_vram; i += 4)
+		nv_wi32(dev, i, 0x00000000);
+	dev_priv->engine.instmem.finish_access(dev);
+}
+
+static void
+nv04_instmem_configure_fixed_tables(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_engine *engine = &dev_priv->engine;
+
+	/* FIFO hash table (RAMHT)
+	 *   use 4k hash table at RAMIN+0x10000
+	 *   TODO: extend the hash table
+	 */
+	dev_priv->ramht_offset = 0x10000;
+	dev_priv->ramht_bits   = 9;
+	dev_priv->ramht_size   = (1 << dev_priv->ramht_bits); /* nr entries */
+	dev_priv->ramht_size  *= 8; /* 2 32-bit values per entry in RAMHT */
+	NV_DEBUG(dev, "RAMHT offset=0x%x, size=%d\n", dev_priv->ramht_offset,
+						      dev_priv->ramht_size);
+
+	/* FIFO runout table (RAMRO) - 512k at 0x11200 */
+	dev_priv->ramro_offset = 0x11200;
+	dev_priv->ramro_size   = 512;
+	NV_DEBUG(dev, "RAMRO offset=0x%x, size=%d\n", dev_priv->ramro_offset,
+						      dev_priv->ramro_size);
+
+	/* FIFO context table (RAMFC)
+	 *   NV40  : Not sure exactly how to position RAMFC on some cards,
+	 *           0x30002 seems to position it at RAMIN+0x20000 on these
+	 *           cards.  RAMFC is 4kb (32 fifos, 128byte entries).
+	 *   Others: Position RAMFC at RAMIN+0x11400
+	 */
+	dev_priv->ramfc_size = engine->fifo.channels *
+						nouveau_fifo_ctx_size(dev);
+	switch (dev_priv->card_type) {
+	case NV_40:
+		dev_priv->ramfc_offset = 0x20000;
+		break;
+	case NV_30:
+	case NV_20:
+	case NV_10:
+	case NV_04:
+	default:
+		dev_priv->ramfc_offset = 0x11400;
+		break;
+	}
+	NV_DEBUG(dev, "RAMFC offset=0x%x, size=%d\n", dev_priv->ramfc_offset,
+						      dev_priv->ramfc_size);
+}
+
+int nv04_instmem_init(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t offset;
+	int ret = 0;
+
+	nv04_instmem_determine_amount(dev);
+	nv04_instmem_configure_fixed_tables(dev);
+
+	/* Create a heap to manage RAMIN allocations, we don't allocate
+	 * the space that was reserved for RAMHT/FC/RO.
+	 */
+	offset = dev_priv->ramfc_offset + dev_priv->ramfc_size;
+
+	/* It appears RAMRO (or something?) is controlled by 0x2220/0x2230
+	 * on certain NV4x chipsets as well as RAMFC.  When 0x2230 == 0
+	 * ("new style" control) the upper 16-bits of 0x2220 points at this
+	 * other mysterious table that's clobbering important things.
+	 *
+	 * We're now pointing this at RAMIN+0x30000 to avoid RAMFC getting
+	 * smashed to pieces on us, so reserve 0x30000-0x40000 too..
+	 */
+	if (dev_priv->card_type >= NV_40) {
+		if (offset < 0x40000)
+			offset = 0x40000;
+	}
+
+	ret = nouveau_mem_init_heap(&dev_priv->ramin_heap,
+				    offset, dev_priv->ramin_rsvd_vram - offset);
+	if (ret) {
+		dev_priv->ramin_heap = NULL;
+		NV_ERROR(dev, "Failed to init RAMIN heap\n");
+	}
+
+	return ret;
+}
+
+void
+nv04_instmem_takedown(struct drm_device *dev)
+{
+}
+
+int
+nv04_instmem_populate(struct drm_device *dev, struct nouveau_gpuobj *gpuobj, uint32_t *sz)
+{
+	if (gpuobj->im_backing)
+		return -EINVAL;
+
+	return 0;
+}
+
+void
+nv04_instmem_clear(struct drm_device *dev, struct nouveau_gpuobj *gpuobj)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	if (gpuobj && gpuobj->im_backing) {
+		if (gpuobj->im_bound)
+			dev_priv->engine.instmem.unbind(dev, gpuobj);
+		gpuobj->im_backing = NULL;
+	}
+}
+
+int
+nv04_instmem_bind(struct drm_device *dev, struct nouveau_gpuobj *gpuobj)
+{
+	if (!gpuobj->im_pramin || gpuobj->im_bound)
+		return -EINVAL;
+
+	gpuobj->im_bound = 1;
+	return 0;
+}
+
+int
+nv04_instmem_unbind(struct drm_device *dev, struct nouveau_gpuobj *gpuobj)
+{
+	if (gpuobj->im_bound == 0)
+		return -EINVAL;
+
+	gpuobj->im_bound = 0;
+	return 0;
+}
+
+void
+nv04_instmem_prepare_access(struct drm_device *dev, bool write)
+{
+}
+
+void
+nv04_instmem_finish_access(struct drm_device *dev)
+{
+}
+
+int
+nv04_instmem_suspend(struct drm_device *dev)
+{
+	return 0;
+}
+
+void
+nv04_instmem_resume(struct drm_device *dev)
+{
+}
+
diff --git a/drivers/gpu/drm/nouveau/nv04_mc.c b/drivers/gpu/drm/nouveau/nv04_mc.c
new file mode 100644
index 000000000000..617ed1e05269
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv04_mc.c
@@ -0,0 +1,20 @@
+#include "drmP.h"
+#include "drm.h"
+#include "nouveau_drv.h"
+#include "nouveau_drm.h"
+
+int
+nv04_mc_init(struct drm_device *dev)
+{
+	/* Power up everything, resetting each individual unit will
+	 * be done later if needed.
+	 */
+
+	nv_wr32(dev, NV03_PMC_ENABLE, 0xFFFFFFFF);
+	return 0;
+}
+
+void
+nv04_mc_takedown(struct drm_device *dev)
+{
+}
diff --git a/drivers/gpu/drm/nouveau/nv04_timer.c b/drivers/gpu/drm/nouveau/nv04_timer.c
new file mode 100644
index 000000000000..1d09ddd57399
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv04_timer.c
@@ -0,0 +1,51 @@
+#include "drmP.h"
+#include "drm.h"
+#include "nouveau_drv.h"
+#include "nouveau_drm.h"
+
+int
+nv04_timer_init(struct drm_device *dev)
+{
+	nv_wr32(dev, NV04_PTIMER_INTR_EN_0, 0x00000000);
+	nv_wr32(dev, NV04_PTIMER_INTR_0, 0xFFFFFFFF);
+
+	/* Just use the pre-existing values when possible for now; these regs
+	 * are not written in nv (driver writer missed a /4 on the address), and
+	 * writing 8 and 3 to the correct regs breaks the timings on the LVDS
+	 * hardware sequencing microcode.
+	 * A correct solution (involving calculations with the GPU PLL) can
+	 * be done when kernel modesetting lands
+	 */
+	if (!nv_rd32(dev, NV04_PTIMER_NUMERATOR) ||
+				!nv_rd32(dev, NV04_PTIMER_DENOMINATOR)) {
+		nv_wr32(dev, NV04_PTIMER_NUMERATOR, 0x00000008);
+		nv_wr32(dev, NV04_PTIMER_DENOMINATOR, 0x00000003);
+	}
+
+	return 0;
+}
+
+uint64_t
+nv04_timer_read(struct drm_device *dev)
+{
+	uint32_t low;
+	/* From kmmio dumps on nv28 this looks like how the blob does this.
+	 * It reads the high dword twice, before and after.
+	 * The only explanation seems to be that the 64-bit timer counter
+	 * advances between high and low dword reads and may corrupt the
+	 * result. Not confirmed.
+	 */
+	uint32_t high2 = nv_rd32(dev, NV04_PTIMER_TIME_1);
+	uint32_t high1;
+	do {
+		high1 = high2;
+		low = nv_rd32(dev, NV04_PTIMER_TIME_0);
+		high2 = nv_rd32(dev, NV04_PTIMER_TIME_1);
+	} while (high1 != high2);
+	return (((uint64_t)high2) << 32) | (uint64_t)low;
+}
+
+void
+nv04_timer_takedown(struct drm_device *dev)
+{
+}
diff --git a/drivers/gpu/drm/nouveau/nv04_tv.c b/drivers/gpu/drm/nouveau/nv04_tv.c
new file mode 100644
index 000000000000..c4e3404337d4
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv04_tv.c
@@ -0,0 +1,305 @@
+/*
+ * Copyright (C) 2009 Francisco Jerez.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "drmP.h"
+#include "nouveau_drv.h"
+#include "nouveau_encoder.h"
+#include "nouveau_connector.h"
+#include "nouveau_crtc.h"
+#include "nouveau_hw.h"
+#include "drm_crtc_helper.h"
+
+#include "i2c/ch7006.h"
+
+static struct {
+	struct i2c_board_info board_info;
+	struct drm_encoder_funcs funcs;
+	struct drm_encoder_helper_funcs hfuncs;
+	void *params;
+
+} nv04_tv_encoder_info[] = {
+	{
+		.board_info = { I2C_BOARD_INFO("ch7006", 0x75) },
+		.params = &(struct ch7006_encoder_params) {
+			CH7006_FORMAT_RGB24m12I, CH7006_CLOCK_MASTER,
+			0, 0, 0,
+			CH7006_SYNC_SLAVE, CH7006_SYNC_SEPARATED,
+			CH7006_POUT_3_3V, CH7006_ACTIVE_HSYNC
+		},
+	},
+};
+
+static bool probe_i2c_addr(struct i2c_adapter *adapter, int addr)
+{
+	struct i2c_msg msg = {
+		.addr = addr,
+		.len = 0,
+	};
+
+	return i2c_transfer(adapter, &msg, 1) == 1;
+}
+
+int nv04_tv_identify(struct drm_device *dev, int i2c_index)
+{
+	struct nouveau_i2c_chan *i2c;
+	bool was_locked;
+	int i, ret;
+
+	NV_TRACE(dev, "Probing TV encoders on I2C bus: %d\n", i2c_index);
+
+	i2c = nouveau_i2c_find(dev, i2c_index);
+	if (!i2c)
+		return -ENODEV;
+
+	was_locked = NVLockVgaCrtcs(dev, false);
+
+	for (i = 0; i < ARRAY_SIZE(nv04_tv_encoder_info); i++) {
+		if (probe_i2c_addr(&i2c->adapter,
+				   nv04_tv_encoder_info[i].board_info.addr)) {
+			ret = i;
+			break;
+		}
+	}
+
+	if (i < ARRAY_SIZE(nv04_tv_encoder_info)) {
+		NV_TRACE(dev, "Detected TV encoder: %s\n",
+			 nv04_tv_encoder_info[i].board_info.type);
+
+	} else {
+		NV_TRACE(dev, "No TV encoders found.\n");
+		i = -ENODEV;
+	}
+
+	NVLockVgaCrtcs(dev, was_locked);
+	return i;
+}
+
+#define PLLSEL_TV_CRTC1_MASK				\
+	(NV_PRAMDAC_PLL_COEFF_SELECT_TV_VSCLK1		\
+	 | NV_PRAMDAC_PLL_COEFF_SELECT_TV_PCLK1)
+#define PLLSEL_TV_CRTC2_MASK				\
+	(NV_PRAMDAC_PLL_COEFF_SELECT_TV_VSCLK2		\
+	 | NV_PRAMDAC_PLL_COEFF_SELECT_TV_PCLK2)
+
+static void nv04_tv_dpms(struct drm_encoder *encoder, int mode)
+{
+	struct drm_device *dev = encoder->dev;
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nv04_mode_state *state = &dev_priv->mode_reg;
+	uint8_t crtc1A;
+
+	NV_INFO(dev, "Setting dpms mode %d on TV encoder (output %d)\n",
+		mode, nv_encoder->dcb->index);
+
+	state->pllsel &= ~(PLLSEL_TV_CRTC1_MASK | PLLSEL_TV_CRTC2_MASK);
+
+	if (mode == DRM_MODE_DPMS_ON) {
+		int head = nouveau_crtc(encoder->crtc)->index;
+		crtc1A = NVReadVgaCrtc(dev, head, NV_CIO_CRE_RPC1_INDEX);
+
+		state->pllsel |= head ? PLLSEL_TV_CRTC2_MASK :
+					PLLSEL_TV_CRTC1_MASK;
+
+		/* Inhibit hsync */
+		crtc1A |= 0x80;
+
+		NVWriteVgaCrtc(dev, head, NV_CIO_CRE_RPC1_INDEX, crtc1A);
+	}
+
+	NVWriteRAMDAC(dev, 0, NV_PRAMDAC_PLL_COEFF_SELECT, state->pllsel);
+
+	to_encoder_slave(encoder)->slave_funcs->dpms(encoder, mode);
+}
+
+static void nv04_tv_bind(struct drm_device *dev, int head, bool bind)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nv04_crtc_reg *state = &dev_priv->mode_reg.crtc_reg[head];
+
+	state->tv_setup = 0;
+
+	if (bind) {
+		state->CRTC[NV_CIO_CRE_LCD__INDEX] = 0;
+		state->CRTC[NV_CIO_CRE_49] |= 0x10;
+	} else {
+		state->CRTC[NV_CIO_CRE_49] &= ~0x10;
+	}
+
+	NVWriteVgaCrtc(dev, head, NV_CIO_CRE_LCD__INDEX,
+		       state->CRTC[NV_CIO_CRE_LCD__INDEX]);
+	NVWriteVgaCrtc(dev, head, NV_CIO_CRE_49,
+		       state->CRTC[NV_CIO_CRE_49]);
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_TV_SETUP,
+		      state->tv_setup);
+}
+
+static void nv04_tv_prepare(struct drm_encoder *encoder)
+{
+	struct drm_device *dev = encoder->dev;
+	int head = nouveau_crtc(encoder->crtc)->index;
+	struct drm_encoder_helper_funcs *helper = encoder->helper_private;
+
+	helper->dpms(encoder, DRM_MODE_DPMS_OFF);
+
+	nv04_dfp_disable(dev, head);
+
+	if (nv_two_heads(dev))
+		nv04_tv_bind(dev, head ^ 1, false);
+
+	nv04_tv_bind(dev, head, true);
+}
+
+static void nv04_tv_mode_set(struct drm_encoder *encoder,
+			     struct drm_display_mode *mode,
+			     struct drm_display_mode *adjusted_mode)
+{
+	struct drm_device *dev = encoder->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(encoder->crtc);
+	struct nv04_crtc_reg *regp = &dev_priv->mode_reg.crtc_reg[nv_crtc->index];
+
+	regp->tv_htotal = adjusted_mode->htotal;
+	regp->tv_vtotal = adjusted_mode->vtotal;
+
+	/* These delay the TV signals with respect to the VGA port,
+	 * they might be useful if we ever allow a CRTC to drive
+	 * multiple outputs.
+	 */
+	regp->tv_hskew = 1;
+	regp->tv_hsync_delay = 1;
+	regp->tv_hsync_delay2 = 64;
+	regp->tv_vskew = 1;
+	regp->tv_vsync_delay = 1;
+
+	to_encoder_slave(encoder)->slave_funcs->mode_set(encoder, mode, adjusted_mode);
+}
+
+static void nv04_tv_commit(struct drm_encoder *encoder)
+{
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	struct drm_device *dev = encoder->dev;
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(encoder->crtc);
+	struct drm_encoder_helper_funcs *helper = encoder->helper_private;
+
+	helper->dpms(encoder, DRM_MODE_DPMS_ON);
+
+	NV_INFO(dev, "Output %s is running on CRTC %d using output %c\n",
+		      drm_get_connector_name(&nouveau_encoder_connector_get(nv_encoder)->base), nv_crtc->index,
+		      '@' + ffs(nv_encoder->dcb->or));
+}
+
+static void nv04_tv_destroy(struct drm_encoder *encoder)
+{
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+
+	to_encoder_slave(encoder)->slave_funcs->destroy(encoder);
+
+	drm_encoder_cleanup(encoder);
+
+	kfree(nv_encoder);
+}
+
+int nv04_tv_create(struct drm_device *dev, struct dcb_entry *entry)
+{
+	struct nouveau_encoder *nv_encoder;
+	struct drm_encoder *encoder;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct i2c_adapter *adap;
+	struct drm_encoder_funcs *funcs = NULL;
+	struct drm_encoder_helper_funcs *hfuncs = NULL;
+	struct drm_encoder_slave_funcs *sfuncs = NULL;
+	int i2c_index = entry->i2c_index;
+	int type, ret;
+	bool was_locked;
+
+	/* Ensure that we can talk to this encoder */
+	type = nv04_tv_identify(dev, i2c_index);
+	if (type < 0)
+		return type;
+
+	/* Allocate the necessary memory */
+	nv_encoder = kzalloc(sizeof(*nv_encoder), GFP_KERNEL);
+	if (!nv_encoder)
+		return -ENOMEM;
+
+	/* Initialize the common members */
+	encoder = to_drm_encoder(nv_encoder);
+
+	funcs = &nv04_tv_encoder_info[type].funcs;
+	hfuncs = &nv04_tv_encoder_info[type].hfuncs;
+
+	drm_encoder_init(dev, encoder, funcs, DRM_MODE_ENCODER_TVDAC);
+	drm_encoder_helper_add(encoder, hfuncs);
+
+	encoder->possible_crtcs = entry->heads;
+	encoder->possible_clones = 0;
+
+	nv_encoder->dcb = entry;
+	nv_encoder->or = ffs(entry->or) - 1;
+
+	/* Run the slave-specific initialization */
+	adap = &dev_priv->vbios.dcb.i2c[i2c_index].chan->adapter;
+
+	was_locked = NVLockVgaCrtcs(dev, false);
+
+	ret = drm_i2c_encoder_init(encoder->dev, to_encoder_slave(encoder), adap,
+				   &nv04_tv_encoder_info[type].board_info);
+
+	NVLockVgaCrtcs(dev, was_locked);
+
+	if (ret < 0)
+		goto fail;
+
+	/* Fill the function pointers */
+	sfuncs = to_encoder_slave(encoder)->slave_funcs;
+
+	*funcs = (struct drm_encoder_funcs) {
+		.destroy = nv04_tv_destroy,
+	};
+
+	*hfuncs = (struct drm_encoder_helper_funcs) {
+		.dpms = nv04_tv_dpms,
+		.save = sfuncs->save,
+		.restore = sfuncs->restore,
+		.mode_fixup = sfuncs->mode_fixup,
+		.prepare = nv04_tv_prepare,
+		.commit = nv04_tv_commit,
+		.mode_set = nv04_tv_mode_set,
+		.detect = sfuncs->detect,
+	};
+
+	/* Set the slave encoder configuration */
+	sfuncs->set_config(encoder, nv04_tv_encoder_info[type].params);
+
+	return 0;
+
+fail:
+	drm_encoder_cleanup(encoder);
+
+	kfree(nv_encoder);
+	return ret;
+}
diff --git a/drivers/gpu/drm/nouveau/nv10_fb.c b/drivers/gpu/drm/nouveau/nv10_fb.c
new file mode 100644
index 000000000000..cc5cda44e501
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv10_fb.c
@@ -0,0 +1,44 @@
+#include "drmP.h"
+#include "drm.h"
+#include "nouveau_drv.h"
+#include "nouveau_drm.h"
+
+void
+nv10_fb_set_region_tiling(struct drm_device *dev, int i, uint32_t addr,
+			  uint32_t size, uint32_t pitch)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t limit = max(1u, addr + size) - 1;
+
+	if (pitch) {
+		if (dev_priv->card_type >= NV_20)
+			addr |= 1;
+		else
+			addr |= 1 << 31;
+	}
+
+	nv_wr32(dev, NV10_PFB_TLIMIT(i), limit);
+	nv_wr32(dev, NV10_PFB_TSIZE(i), pitch);
+	nv_wr32(dev, NV10_PFB_TILE(i), addr);
+}
+
+int
+nv10_fb_init(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_fb_engine *pfb = &dev_priv->engine.fb;
+	int i;
+
+	pfb->num_tiles = NV10_PFB_TILE__SIZE;
+
+	/* Turn all the tiling regions off. */
+	for (i = 0; i < pfb->num_tiles; i++)
+		pfb->set_region_tiling(dev, i, 0, 0, 0);
+
+	return 0;
+}
+
+void
+nv10_fb_takedown(struct drm_device *dev)
+{
+}
diff --git a/drivers/gpu/drm/nouveau/nv10_fifo.c b/drivers/gpu/drm/nouveau/nv10_fifo.c
new file mode 100644
index 000000000000..7aeabf262bc0
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv10_fifo.c
@@ -0,0 +1,260 @@
+/*
+ * Copyright (C) 2007 Ben Skeggs.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "drmP.h"
+#include "drm.h"
+#include "nouveau_drv.h"
+
+#define NV10_RAMFC(c) (dev_priv->ramfc_offset + ((c) * NV10_RAMFC__SIZE))
+#define NV10_RAMFC__SIZE ((dev_priv->chipset) >= 0x17 ? 64 : 32)
+
+int
+nv10_fifo_channel_id(struct drm_device *dev)
+{
+	return nv_rd32(dev, NV03_PFIFO_CACHE1_PUSH1) &
+			NV10_PFIFO_CACHE1_PUSH1_CHID_MASK;
+}
+
+int
+nv10_fifo_create_context(struct nouveau_channel *chan)
+{
+	struct drm_nouveau_private *dev_priv = chan->dev->dev_private;
+	struct drm_device *dev = chan->dev;
+	uint32_t fc = NV10_RAMFC(chan->id);
+	int ret;
+
+	ret = nouveau_gpuobj_new_fake(dev, NV10_RAMFC(chan->id), ~0,
+				      NV10_RAMFC__SIZE, NVOBJ_FLAG_ZERO_ALLOC |
+				      NVOBJ_FLAG_ZERO_FREE, NULL, &chan->ramfc);
+	if (ret)
+		return ret;
+
+	/* Fill entries that are seen filled in dumps of nvidia driver just
+	 * after channel's is put into DMA mode
+	 */
+	dev_priv->engine.instmem.prepare_access(dev, true);
+	nv_wi32(dev, fc +  0, chan->pushbuf_base);
+	nv_wi32(dev, fc +  4, chan->pushbuf_base);
+	nv_wi32(dev, fc + 12, chan->pushbuf->instance >> 4);
+	nv_wi32(dev, fc + 20, NV_PFIFO_CACHE1_DMA_FETCH_TRIG_128_BYTES |
+			      NV_PFIFO_CACHE1_DMA_FETCH_SIZE_128_BYTES |
+			      NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_8 |
+#ifdef __BIG_ENDIAN
+			      NV_PFIFO_CACHE1_BIG_ENDIAN |
+#endif
+			      0);
+	dev_priv->engine.instmem.finish_access(dev);
+
+	/* enable the fifo dma operation */
+	nv_wr32(dev, NV04_PFIFO_MODE,
+		nv_rd32(dev, NV04_PFIFO_MODE) | (1 << chan->id));
+	return 0;
+}
+
+void
+nv10_fifo_destroy_context(struct nouveau_channel *chan)
+{
+	struct drm_device *dev = chan->dev;
+
+	nv_wr32(dev, NV04_PFIFO_MODE,
+			nv_rd32(dev, NV04_PFIFO_MODE) & ~(1 << chan->id));
+
+	nouveau_gpuobj_ref_del(dev, &chan->ramfc);
+}
+
+static void
+nv10_fifo_do_load_context(struct drm_device *dev, int chid)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t fc = NV10_RAMFC(chid), tmp;
+
+	dev_priv->engine.instmem.prepare_access(dev, false);
+
+	nv_wr32(dev, NV04_PFIFO_CACHE1_DMA_PUT, nv_ri32(dev, fc + 0));
+	nv_wr32(dev, NV04_PFIFO_CACHE1_DMA_GET, nv_ri32(dev, fc + 4));
+	nv_wr32(dev, NV10_PFIFO_CACHE1_REF_CNT, nv_ri32(dev, fc + 8));
+
+	tmp = nv_ri32(dev, fc + 12);
+	nv_wr32(dev, NV04_PFIFO_CACHE1_DMA_INSTANCE, tmp & 0xFFFF);
+	nv_wr32(dev, NV04_PFIFO_CACHE1_DMA_DCOUNT, tmp >> 16);
+
+	nv_wr32(dev, NV04_PFIFO_CACHE1_DMA_STATE, nv_ri32(dev, fc + 16));
+	nv_wr32(dev, NV04_PFIFO_CACHE1_DMA_FETCH, nv_ri32(dev, fc + 20));
+	nv_wr32(dev, NV04_PFIFO_CACHE1_ENGINE, nv_ri32(dev, fc + 24));
+	nv_wr32(dev, NV04_PFIFO_CACHE1_PULL1, nv_ri32(dev, fc + 28));
+
+	if (dev_priv->chipset < 0x17)
+		goto out;
+
+	nv_wr32(dev, NV10_PFIFO_CACHE1_ACQUIRE_VALUE, nv_ri32(dev, fc + 32));
+	tmp = nv_ri32(dev, fc + 36);
+	nv_wr32(dev, NV10_PFIFO_CACHE1_ACQUIRE_TIMESTAMP, tmp);
+	nv_wr32(dev, NV10_PFIFO_CACHE1_ACQUIRE_TIMEOUT, nv_ri32(dev, fc + 40));
+	nv_wr32(dev, NV10_PFIFO_CACHE1_SEMAPHORE, nv_ri32(dev, fc + 44));
+	nv_wr32(dev, NV10_PFIFO_CACHE1_DMA_SUBROUTINE, nv_ri32(dev, fc + 48));
+
+out:
+	dev_priv->engine.instmem.finish_access(dev);
+
+	nv_wr32(dev, NV03_PFIFO_CACHE1_GET, 0);
+	nv_wr32(dev, NV03_PFIFO_CACHE1_PUT, 0);
+}
+
+int
+nv10_fifo_load_context(struct nouveau_channel *chan)
+{
+	struct drm_device *dev = chan->dev;
+	uint32_t tmp;
+
+	nv10_fifo_do_load_context(dev, chan->id);
+
+	nv_wr32(dev, NV03_PFIFO_CACHE1_PUSH1,
+		     NV03_PFIFO_CACHE1_PUSH1_DMA | chan->id);
+	nv_wr32(dev, NV04_PFIFO_CACHE1_DMA_PUSH, 1);
+
+	/* Reset NV04_PFIFO_CACHE1_DMA_CTL_AT_INFO to INVALID */
+	tmp = nv_rd32(dev, NV04_PFIFO_CACHE1_DMA_CTL) & ~(1 << 31);
+	nv_wr32(dev, NV04_PFIFO_CACHE1_DMA_CTL, tmp);
+
+	return 0;
+}
+
+int
+nv10_fifo_unload_context(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_fifo_engine *pfifo = &dev_priv->engine.fifo;
+	uint32_t fc, tmp;
+	int chid;
+
+	chid = pfifo->channel_id(dev);
+	if (chid < 0 || chid >= dev_priv->engine.fifo.channels)
+		return 0;
+	fc = NV10_RAMFC(chid);
+
+	dev_priv->engine.instmem.prepare_access(dev, true);
+
+	nv_wi32(dev, fc +  0, nv_rd32(dev, NV04_PFIFO_CACHE1_DMA_PUT));
+	nv_wi32(dev, fc +  4, nv_rd32(dev, NV04_PFIFO_CACHE1_DMA_GET));
+	nv_wi32(dev, fc +  8, nv_rd32(dev, NV10_PFIFO_CACHE1_REF_CNT));
+	tmp  = nv_rd32(dev, NV04_PFIFO_CACHE1_DMA_INSTANCE) & 0xFFFF;
+	tmp |= (nv_rd32(dev, NV04_PFIFO_CACHE1_DMA_DCOUNT) << 16);
+	nv_wi32(dev, fc + 12, tmp);
+	nv_wi32(dev, fc + 16, nv_rd32(dev, NV04_PFIFO_CACHE1_DMA_STATE));
+	nv_wi32(dev, fc + 20, nv_rd32(dev, NV04_PFIFO_CACHE1_DMA_FETCH));
+	nv_wi32(dev, fc + 24, nv_rd32(dev, NV04_PFIFO_CACHE1_ENGINE));
+	nv_wi32(dev, fc + 28, nv_rd32(dev, NV04_PFIFO_CACHE1_PULL1));
+
+	if (dev_priv->chipset < 0x17)
+		goto out;
+
+	nv_wi32(dev, fc + 32, nv_rd32(dev, NV10_PFIFO_CACHE1_ACQUIRE_VALUE));
+	tmp = nv_rd32(dev, NV10_PFIFO_CACHE1_ACQUIRE_TIMESTAMP);
+	nv_wi32(dev, fc + 36, tmp);
+	nv_wi32(dev, fc + 40, nv_rd32(dev, NV10_PFIFO_CACHE1_ACQUIRE_TIMEOUT));
+	nv_wi32(dev, fc + 44, nv_rd32(dev, NV10_PFIFO_CACHE1_SEMAPHORE));
+	nv_wi32(dev, fc + 48, nv_rd32(dev, NV04_PFIFO_CACHE1_DMA_GET));
+
+out:
+	dev_priv->engine.instmem.finish_access(dev);
+
+	nv10_fifo_do_load_context(dev, pfifo->channels - 1);
+	nv_wr32(dev, NV03_PFIFO_CACHE1_PUSH1, pfifo->channels - 1);
+	return 0;
+}
+
+static void
+nv10_fifo_init_reset(struct drm_device *dev)
+{
+	nv_wr32(dev, NV03_PMC_ENABLE,
+		nv_rd32(dev, NV03_PMC_ENABLE) & ~NV_PMC_ENABLE_PFIFO);
+	nv_wr32(dev, NV03_PMC_ENABLE,
+		nv_rd32(dev, NV03_PMC_ENABLE) |  NV_PMC_ENABLE_PFIFO);
+
+	nv_wr32(dev, 0x003224, 0x000f0078);
+	nv_wr32(dev, 0x002044, 0x0101ffff);
+	nv_wr32(dev, 0x002040, 0x000000ff);
+	nv_wr32(dev, 0x002500, 0x00000000);
+	nv_wr32(dev, 0x003000, 0x00000000);
+	nv_wr32(dev, 0x003050, 0x00000000);
+
+	nv_wr32(dev, 0x003258, 0x00000000);
+	nv_wr32(dev, 0x003210, 0x00000000);
+	nv_wr32(dev, 0x003270, 0x00000000);
+}
+
+static void
+nv10_fifo_init_ramxx(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	nv_wr32(dev, NV03_PFIFO_RAMHT, (0x03 << 24) /* search 128 */ |
+				       ((dev_priv->ramht_bits - 9) << 16) |
+				       (dev_priv->ramht_offset >> 8));
+	nv_wr32(dev, NV03_PFIFO_RAMRO, dev_priv->ramro_offset>>8);
+
+	if (dev_priv->chipset < 0x17) {
+		nv_wr32(dev, NV03_PFIFO_RAMFC, dev_priv->ramfc_offset >> 8);
+	} else {
+		nv_wr32(dev, NV03_PFIFO_RAMFC, (dev_priv->ramfc_offset >> 8) |
+					       (1 << 16) /* 64 Bytes entry*/);
+		/* XXX nvidia blob set bit 18, 21,23 for nv20 & nv30 */
+	}
+}
+
+static void
+nv10_fifo_init_intr(struct drm_device *dev)
+{
+	nv_wr32(dev, 0x002100, 0xffffffff);
+	nv_wr32(dev, 0x002140, 0xffffffff);
+}
+
+int
+nv10_fifo_init(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_fifo_engine *pfifo = &dev_priv->engine.fifo;
+	int i;
+
+	nv10_fifo_init_reset(dev);
+	nv10_fifo_init_ramxx(dev);
+
+	nv10_fifo_do_load_context(dev, pfifo->channels - 1);
+	nv_wr32(dev, NV03_PFIFO_CACHE1_PUSH1, pfifo->channels - 1);
+
+	nv10_fifo_init_intr(dev);
+	pfifo->enable(dev);
+	pfifo->reassign(dev, true);
+
+	for (i = 0; i < dev_priv->engine.fifo.channels; i++) {
+		if (dev_priv->fifos[i]) {
+			uint32_t mode = nv_rd32(dev, NV04_PFIFO_MODE);
+			nv_wr32(dev, NV04_PFIFO_MODE, mode | (1 << i));
+		}
+	}
+
+	return 0;
+}
diff --git a/drivers/gpu/drm/nouveau/nv10_graph.c b/drivers/gpu/drm/nouveau/nv10_graph.c
new file mode 100644
index 000000000000..fcf2cdd19493
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv10_graph.c
@@ -0,0 +1,1009 @@
+/*
+ * Copyright 2007 Matthieu CASTET <castet.matthieu@free.fr>
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include "drmP.h"
+#include "drm.h"
+#include "nouveau_drm.h"
+#include "nouveau_drv.h"
+
+#define NV10_FIFO_NUMBER 32
+
+struct pipe_state {
+	uint32_t pipe_0x0000[0x040/4];
+	uint32_t pipe_0x0040[0x010/4];
+	uint32_t pipe_0x0200[0x0c0/4];
+	uint32_t pipe_0x4400[0x080/4];
+	uint32_t pipe_0x6400[0x3b0/4];
+	uint32_t pipe_0x6800[0x2f0/4];
+	uint32_t pipe_0x6c00[0x030/4];
+	uint32_t pipe_0x7000[0x130/4];
+	uint32_t pipe_0x7400[0x0c0/4];
+	uint32_t pipe_0x7800[0x0c0/4];
+};
+
+static int nv10_graph_ctx_regs[] = {
+	NV10_PGRAPH_CTX_SWITCH1,
+	NV10_PGRAPH_CTX_SWITCH2,
+	NV10_PGRAPH_CTX_SWITCH3,
+	NV10_PGRAPH_CTX_SWITCH4,
+	NV10_PGRAPH_CTX_SWITCH5,
+	NV10_PGRAPH_CTX_CACHE1,	/* 8 values from 0x400160 to 0x40017c */
+	NV10_PGRAPH_CTX_CACHE2,	/* 8 values from 0x400180 to 0x40019c */
+	NV10_PGRAPH_CTX_CACHE3,	/* 8 values from 0x4001a0 to 0x4001bc */
+	NV10_PGRAPH_CTX_CACHE4,	/* 8 values from 0x4001c0 to 0x4001dc */
+	NV10_PGRAPH_CTX_CACHE5,	/* 8 values from 0x4001e0 to 0x4001fc */
+	0x00400164,
+	0x00400184,
+	0x004001a4,
+	0x004001c4,
+	0x004001e4,
+	0x00400168,
+	0x00400188,
+	0x004001a8,
+	0x004001c8,
+	0x004001e8,
+	0x0040016c,
+	0x0040018c,
+	0x004001ac,
+	0x004001cc,
+	0x004001ec,
+	0x00400170,
+	0x00400190,
+	0x004001b0,
+	0x004001d0,
+	0x004001f0,
+	0x00400174,
+	0x00400194,
+	0x004001b4,
+	0x004001d4,
+	0x004001f4,
+	0x00400178,
+	0x00400198,
+	0x004001b8,
+	0x004001d8,
+	0x004001f8,
+	0x0040017c,
+	0x0040019c,
+	0x004001bc,
+	0x004001dc,
+	0x004001fc,
+	NV10_PGRAPH_CTX_USER,
+	NV04_PGRAPH_DMA_START_0,
+	NV04_PGRAPH_DMA_START_1,
+	NV04_PGRAPH_DMA_LENGTH,
+	NV04_PGRAPH_DMA_MISC,
+	NV10_PGRAPH_DMA_PITCH,
+	NV04_PGRAPH_BOFFSET0,
+	NV04_PGRAPH_BBASE0,
+	NV04_PGRAPH_BLIMIT0,
+	NV04_PGRAPH_BOFFSET1,
+	NV04_PGRAPH_BBASE1,
+	NV04_PGRAPH_BLIMIT1,
+	NV04_PGRAPH_BOFFSET2,
+	NV04_PGRAPH_BBASE2,
+	NV04_PGRAPH_BLIMIT2,
+	NV04_PGRAPH_BOFFSET3,
+	NV04_PGRAPH_BBASE3,
+	NV04_PGRAPH_BLIMIT3,
+	NV04_PGRAPH_BOFFSET4,
+	NV04_PGRAPH_BBASE4,
+	NV04_PGRAPH_BLIMIT4,
+	NV04_PGRAPH_BOFFSET5,
+	NV04_PGRAPH_BBASE5,
+	NV04_PGRAPH_BLIMIT5,
+	NV04_PGRAPH_BPITCH0,
+	NV04_PGRAPH_BPITCH1,
+	NV04_PGRAPH_BPITCH2,
+	NV04_PGRAPH_BPITCH3,
+	NV04_PGRAPH_BPITCH4,
+	NV10_PGRAPH_SURFACE,
+	NV10_PGRAPH_STATE,
+	NV04_PGRAPH_BSWIZZLE2,
+	NV04_PGRAPH_BSWIZZLE5,
+	NV04_PGRAPH_BPIXEL,
+	NV10_PGRAPH_NOTIFY,
+	NV04_PGRAPH_PATT_COLOR0,
+	NV04_PGRAPH_PATT_COLOR1,
+	NV04_PGRAPH_PATT_COLORRAM, /* 64 values from 0x400900 to 0x4009fc */
+	0x00400904,
+	0x00400908,
+	0x0040090c,
+	0x00400910,
+	0x00400914,
+	0x00400918,
+	0x0040091c,
+	0x00400920,
+	0x00400924,
+	0x00400928,
+	0x0040092c,
+	0x00400930,
+	0x00400934,
+	0x00400938,
+	0x0040093c,
+	0x00400940,
+	0x00400944,
+	0x00400948,
+	0x0040094c,
+	0x00400950,
+	0x00400954,
+	0x00400958,
+	0x0040095c,
+	0x00400960,
+	0x00400964,
+	0x00400968,
+	0x0040096c,
+	0x00400970,
+	0x00400974,
+	0x00400978,
+	0x0040097c,
+	0x00400980,
+	0x00400984,
+	0x00400988,
+	0x0040098c,
+	0x00400990,
+	0x00400994,
+	0x00400998,
+	0x0040099c,
+	0x004009a0,
+	0x004009a4,
+	0x004009a8,
+	0x004009ac,
+	0x004009b0,
+	0x004009b4,
+	0x004009b8,
+	0x004009bc,
+	0x004009c0,
+	0x004009c4,
+	0x004009c8,
+	0x004009cc,
+	0x004009d0,
+	0x004009d4,
+	0x004009d8,
+	0x004009dc,
+	0x004009e0,
+	0x004009e4,
+	0x004009e8,
+	0x004009ec,
+	0x004009f0,
+	0x004009f4,
+	0x004009f8,
+	0x004009fc,
+	NV04_PGRAPH_PATTERN,	/* 2 values from 0x400808 to 0x40080c */
+	0x0040080c,
+	NV04_PGRAPH_PATTERN_SHAPE,
+	NV03_PGRAPH_MONO_COLOR0,
+	NV04_PGRAPH_ROP3,
+	NV04_PGRAPH_CHROMA,
+	NV04_PGRAPH_BETA_AND,
+	NV04_PGRAPH_BETA_PREMULT,
+	0x00400e70,
+	0x00400e74,
+	0x00400e78,
+	0x00400e7c,
+	0x00400e80,
+	0x00400e84,
+	0x00400e88,
+	0x00400e8c,
+	0x00400ea0,
+	0x00400ea4,
+	0x00400ea8,
+	0x00400e90,
+	0x00400e94,
+	0x00400e98,
+	0x00400e9c,
+	NV10_PGRAPH_WINDOWCLIP_HORIZONTAL, /* 8 values from 0x400f00-0x400f1c */
+	NV10_PGRAPH_WINDOWCLIP_VERTICAL,   /* 8 values from 0x400f20-0x400f3c */
+	0x00400f04,
+	0x00400f24,
+	0x00400f08,
+	0x00400f28,
+	0x00400f0c,
+	0x00400f2c,
+	0x00400f10,
+	0x00400f30,
+	0x00400f14,
+	0x00400f34,
+	0x00400f18,
+	0x00400f38,
+	0x00400f1c,
+	0x00400f3c,
+	NV10_PGRAPH_XFMODE0,
+	NV10_PGRAPH_XFMODE1,
+	NV10_PGRAPH_GLOBALSTATE0,
+	NV10_PGRAPH_GLOBALSTATE1,
+	NV04_PGRAPH_STORED_FMT,
+	NV04_PGRAPH_SOURCE_COLOR,
+	NV03_PGRAPH_ABS_X_RAM,	/* 32 values from 0x400400 to 0x40047c */
+	NV03_PGRAPH_ABS_Y_RAM,	/* 32 values from 0x400480 to 0x4004fc */
+	0x00400404,
+	0x00400484,
+	0x00400408,
+	0x00400488,
+	0x0040040c,
+	0x0040048c,
+	0x00400410,
+	0x00400490,
+	0x00400414,
+	0x00400494,
+	0x00400418,
+	0x00400498,
+	0x0040041c,
+	0x0040049c,
+	0x00400420,
+	0x004004a0,
+	0x00400424,
+	0x004004a4,
+	0x00400428,
+	0x004004a8,
+	0x0040042c,
+	0x004004ac,
+	0x00400430,
+	0x004004b0,
+	0x00400434,
+	0x004004b4,
+	0x00400438,
+	0x004004b8,
+	0x0040043c,
+	0x004004bc,
+	0x00400440,
+	0x004004c0,
+	0x00400444,
+	0x004004c4,
+	0x00400448,
+	0x004004c8,
+	0x0040044c,
+	0x004004cc,
+	0x00400450,
+	0x004004d0,
+	0x00400454,
+	0x004004d4,
+	0x00400458,
+	0x004004d8,
+	0x0040045c,
+	0x004004dc,
+	0x00400460,
+	0x004004e0,
+	0x00400464,
+	0x004004e4,
+	0x00400468,
+	0x004004e8,
+	0x0040046c,
+	0x004004ec,
+	0x00400470,
+	0x004004f0,
+	0x00400474,
+	0x004004f4,
+	0x00400478,
+	0x004004f8,
+	0x0040047c,
+	0x004004fc,
+	NV03_PGRAPH_ABS_UCLIP_XMIN,
+	NV03_PGRAPH_ABS_UCLIP_XMAX,
+	NV03_PGRAPH_ABS_UCLIP_YMIN,
+	NV03_PGRAPH_ABS_UCLIP_YMAX,
+	0x00400550,
+	0x00400558,
+	0x00400554,
+	0x0040055c,
+	NV03_PGRAPH_ABS_UCLIPA_XMIN,
+	NV03_PGRAPH_ABS_UCLIPA_XMAX,
+	NV03_PGRAPH_ABS_UCLIPA_YMIN,
+	NV03_PGRAPH_ABS_UCLIPA_YMAX,
+	NV03_PGRAPH_ABS_ICLIP_XMAX,
+	NV03_PGRAPH_ABS_ICLIP_YMAX,
+	NV03_PGRAPH_XY_LOGIC_MISC0,
+	NV03_PGRAPH_XY_LOGIC_MISC1,
+	NV03_PGRAPH_XY_LOGIC_MISC2,
+	NV03_PGRAPH_XY_LOGIC_MISC3,
+	NV03_PGRAPH_CLIPX_0,
+	NV03_PGRAPH_CLIPX_1,
+	NV03_PGRAPH_CLIPY_0,
+	NV03_PGRAPH_CLIPY_1,
+	NV10_PGRAPH_COMBINER0_IN_ALPHA,
+	NV10_PGRAPH_COMBINER1_IN_ALPHA,
+	NV10_PGRAPH_COMBINER0_IN_RGB,
+	NV10_PGRAPH_COMBINER1_IN_RGB,
+	NV10_PGRAPH_COMBINER_COLOR0,
+	NV10_PGRAPH_COMBINER_COLOR1,
+	NV10_PGRAPH_COMBINER0_OUT_ALPHA,
+	NV10_PGRAPH_COMBINER1_OUT_ALPHA,
+	NV10_PGRAPH_COMBINER0_OUT_RGB,
+	NV10_PGRAPH_COMBINER1_OUT_RGB,
+	NV10_PGRAPH_COMBINER_FINAL0,
+	NV10_PGRAPH_COMBINER_FINAL1,
+	0x00400e00,
+	0x00400e04,
+	0x00400e08,
+	0x00400e0c,
+	0x00400e10,
+	0x00400e14,
+	0x00400e18,
+	0x00400e1c,
+	0x00400e20,
+	0x00400e24,
+	0x00400e28,
+	0x00400e2c,
+	0x00400e30,
+	0x00400e34,
+	0x00400e38,
+	0x00400e3c,
+	NV04_PGRAPH_PASSTHRU_0,
+	NV04_PGRAPH_PASSTHRU_1,
+	NV04_PGRAPH_PASSTHRU_2,
+	NV10_PGRAPH_DIMX_TEXTURE,
+	NV10_PGRAPH_WDIMX_TEXTURE,
+	NV10_PGRAPH_DVD_COLORFMT,
+	NV10_PGRAPH_SCALED_FORMAT,
+	NV04_PGRAPH_MISC24_0,
+	NV04_PGRAPH_MISC24_1,
+	NV04_PGRAPH_MISC24_2,
+	NV03_PGRAPH_X_MISC,
+	NV03_PGRAPH_Y_MISC,
+	NV04_PGRAPH_VALID1,
+	NV04_PGRAPH_VALID2,
+};
+
+static int nv17_graph_ctx_regs[] = {
+	NV10_PGRAPH_DEBUG_4,
+	0x004006b0,
+	0x00400eac,
+	0x00400eb0,
+	0x00400eb4,
+	0x00400eb8,
+	0x00400ebc,
+	0x00400ec0,
+	0x00400ec4,
+	0x00400ec8,
+	0x00400ecc,
+	0x00400ed0,
+	0x00400ed4,
+	0x00400ed8,
+	0x00400edc,
+	0x00400ee0,
+	0x00400a00,
+	0x00400a04,
+};
+
+struct graph_state {
+	int nv10[ARRAY_SIZE(nv10_graph_ctx_regs)];
+	int nv17[ARRAY_SIZE(nv17_graph_ctx_regs)];
+	struct pipe_state pipe_state;
+	uint32_t lma_window[4];
+};
+
+#define PIPE_SAVE(dev, state, addr)					\
+	do {								\
+		int __i;						\
+		nv_wr32(dev, NV10_PGRAPH_PIPE_ADDRESS, addr);		\
+		for (__i = 0; __i < ARRAY_SIZE(state); __i++)		\
+			state[__i] = nv_rd32(dev, NV10_PGRAPH_PIPE_DATA); \
+	} while (0)
+
+#define PIPE_RESTORE(dev, state, addr)					\
+	do {								\
+		int __i;						\
+		nv_wr32(dev, NV10_PGRAPH_PIPE_ADDRESS, addr);		\
+		for (__i = 0; __i < ARRAY_SIZE(state); __i++)		\
+			nv_wr32(dev, NV10_PGRAPH_PIPE_DATA, state[__i]); \
+	} while (0)
+
+static void nv10_graph_save_pipe(struct nouveau_channel *chan)
+{
+	struct drm_device *dev = chan->dev;
+	struct graph_state *pgraph_ctx = chan->pgraph_ctx;
+	struct pipe_state *pipe = &pgraph_ctx->pipe_state;
+
+	PIPE_SAVE(dev, pipe->pipe_0x4400, 0x4400);
+	PIPE_SAVE(dev, pipe->pipe_0x0200, 0x0200);
+	PIPE_SAVE(dev, pipe->pipe_0x6400, 0x6400);
+	PIPE_SAVE(dev, pipe->pipe_0x6800, 0x6800);
+	PIPE_SAVE(dev, pipe->pipe_0x6c00, 0x6c00);
+	PIPE_SAVE(dev, pipe->pipe_0x7000, 0x7000);
+	PIPE_SAVE(dev, pipe->pipe_0x7400, 0x7400);
+	PIPE_SAVE(dev, pipe->pipe_0x7800, 0x7800);
+	PIPE_SAVE(dev, pipe->pipe_0x0040, 0x0040);
+	PIPE_SAVE(dev, pipe->pipe_0x0000, 0x0000);
+}
+
+static void nv10_graph_load_pipe(struct nouveau_channel *chan)
+{
+	struct drm_device *dev = chan->dev;
+	struct graph_state *pgraph_ctx = chan->pgraph_ctx;
+	struct pipe_state *pipe = &pgraph_ctx->pipe_state;
+	uint32_t xfmode0, xfmode1;
+	int i;
+
+	nouveau_wait_for_idle(dev);
+	/* XXX check haiku comments */
+	xfmode0 = nv_rd32(dev, NV10_PGRAPH_XFMODE0);
+	xfmode1 = nv_rd32(dev, NV10_PGRAPH_XFMODE1);
+	nv_wr32(dev, NV10_PGRAPH_XFMODE0, 0x10000000);
+	nv_wr32(dev, NV10_PGRAPH_XFMODE1, 0x00000000);
+	nv_wr32(dev, NV10_PGRAPH_PIPE_ADDRESS, 0x000064c0);
+	for (i = 0; i < 4; i++)
+		nv_wr32(dev, NV10_PGRAPH_PIPE_DATA, 0x3f800000);
+	for (i = 0; i < 4; i++)
+		nv_wr32(dev, NV10_PGRAPH_PIPE_DATA, 0x00000000);
+
+	nv_wr32(dev, NV10_PGRAPH_PIPE_ADDRESS, 0x00006ab0);
+	for (i = 0; i < 3; i++)
+		nv_wr32(dev, NV10_PGRAPH_PIPE_DATA, 0x3f800000);
+
+	nv_wr32(dev, NV10_PGRAPH_PIPE_ADDRESS, 0x00006a80);
+	for (i = 0; i < 3; i++)
+		nv_wr32(dev, NV10_PGRAPH_PIPE_DATA, 0x00000000);
+
+	nv_wr32(dev, NV10_PGRAPH_PIPE_ADDRESS, 0x00000040);
+	nv_wr32(dev, NV10_PGRAPH_PIPE_DATA, 0x00000008);
+
+
+	PIPE_RESTORE(dev, pipe->pipe_0x0200, 0x0200);
+	nouveau_wait_for_idle(dev);
+
+	/* restore XFMODE */
+	nv_wr32(dev, NV10_PGRAPH_XFMODE0, xfmode0);
+	nv_wr32(dev, NV10_PGRAPH_XFMODE1, xfmode1);
+	PIPE_RESTORE(dev, pipe->pipe_0x6400, 0x6400);
+	PIPE_RESTORE(dev, pipe->pipe_0x6800, 0x6800);
+	PIPE_RESTORE(dev, pipe->pipe_0x6c00, 0x6c00);
+	PIPE_RESTORE(dev, pipe->pipe_0x7000, 0x7000);
+	PIPE_RESTORE(dev, pipe->pipe_0x7400, 0x7400);
+	PIPE_RESTORE(dev, pipe->pipe_0x7800, 0x7800);
+	PIPE_RESTORE(dev, pipe->pipe_0x4400, 0x4400);
+	PIPE_RESTORE(dev, pipe->pipe_0x0000, 0x0000);
+	PIPE_RESTORE(dev, pipe->pipe_0x0040, 0x0040);
+	nouveau_wait_for_idle(dev);
+}
+
+static void nv10_graph_create_pipe(struct nouveau_channel *chan)
+{
+	struct drm_device *dev = chan->dev;
+	struct graph_state *pgraph_ctx = chan->pgraph_ctx;
+	struct pipe_state *fifo_pipe_state = &pgraph_ctx->pipe_state;
+	uint32_t *fifo_pipe_state_addr;
+	int i;
+#define PIPE_INIT(addr) \
+	do { \
+		fifo_pipe_state_addr = fifo_pipe_state->pipe_##addr; \
+	} while (0)
+#define PIPE_INIT_END(addr) \
+	do { \
+		uint32_t *__end_addr = fifo_pipe_state->pipe_##addr + \
+				ARRAY_SIZE(fifo_pipe_state->pipe_##addr); \
+		if (fifo_pipe_state_addr != __end_addr) \
+			NV_ERROR(dev, "incomplete pipe init for 0x%x :  %p/%p\n", \
+				addr, fifo_pipe_state_addr, __end_addr); \
+	} while (0)
+#define NV_WRITE_PIPE_INIT(value) *(fifo_pipe_state_addr++) = value
+
+	PIPE_INIT(0x0200);
+	for (i = 0; i < 48; i++)
+		NV_WRITE_PIPE_INIT(0x00000000);
+	PIPE_INIT_END(0x0200);
+
+	PIPE_INIT(0x6400);
+	for (i = 0; i < 211; i++)
+		NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x3f800000);
+	NV_WRITE_PIPE_INIT(0x40000000);
+	NV_WRITE_PIPE_INIT(0x40000000);
+	NV_WRITE_PIPE_INIT(0x40000000);
+	NV_WRITE_PIPE_INIT(0x40000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x3f800000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x3f000000);
+	NV_WRITE_PIPE_INIT(0x3f000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x3f800000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x3f800000);
+	NV_WRITE_PIPE_INIT(0x3f800000);
+	NV_WRITE_PIPE_INIT(0x3f800000);
+	NV_WRITE_PIPE_INIT(0x3f800000);
+	PIPE_INIT_END(0x6400);
+
+	PIPE_INIT(0x6800);
+	for (i = 0; i < 162; i++)
+		NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x3f800000);
+	for (i = 0; i < 25; i++)
+		NV_WRITE_PIPE_INIT(0x00000000);
+	PIPE_INIT_END(0x6800);
+
+	PIPE_INIT(0x6c00);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0xbf800000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	PIPE_INIT_END(0x6c00);
+
+	PIPE_INIT(0x7000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x7149f2ca);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x7149f2ca);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x7149f2ca);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x7149f2ca);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x7149f2ca);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x7149f2ca);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x7149f2ca);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x00000000);
+	NV_WRITE_PIPE_INIT(0x7149f2ca);
+	for (i = 0; i < 35; i++)
+		NV_WRITE_PIPE_INIT(0x00000000);
+	PIPE_INIT_END(0x7000);
+
+	PIPE_INIT(0x7400);
+	for (i = 0; i < 48; i++)
+		NV_WRITE_PIPE_INIT(0x00000000);
+	PIPE_INIT_END(0x7400);
+
+	PIPE_INIT(0x7800);
+	for (i = 0; i < 48; i++)
+		NV_WRITE_PIPE_INIT(0x00000000);
+	PIPE_INIT_END(0x7800);
+
+	PIPE_INIT(0x4400);
+	for (i = 0; i < 32; i++)
+		NV_WRITE_PIPE_INIT(0x00000000);
+	PIPE_INIT_END(0x4400);
+
+	PIPE_INIT(0x0000);
+	for (i = 0; i < 16; i++)
+		NV_WRITE_PIPE_INIT(0x00000000);
+	PIPE_INIT_END(0x0000);
+
+	PIPE_INIT(0x0040);
+	for (i = 0; i < 4; i++)
+		NV_WRITE_PIPE_INIT(0x00000000);
+	PIPE_INIT_END(0x0040);
+
+#undef PIPE_INIT
+#undef PIPE_INIT_END
+#undef NV_WRITE_PIPE_INIT
+}
+
+static int nv10_graph_ctx_regs_find_offset(struct drm_device *dev, int reg)
+{
+	int i;
+	for (i = 0; i < ARRAY_SIZE(nv10_graph_ctx_regs); i++) {
+		if (nv10_graph_ctx_regs[i] == reg)
+			return i;
+	}
+	NV_ERROR(dev, "unknow offset nv10_ctx_regs %d\n", reg);
+	return -1;
+}
+
+static int nv17_graph_ctx_regs_find_offset(struct drm_device *dev, int reg)
+{
+	int i;
+	for (i = 0; i < ARRAY_SIZE(nv17_graph_ctx_regs); i++) {
+		if (nv17_graph_ctx_regs[i] == reg)
+			return i;
+	}
+	NV_ERROR(dev, "unknow offset nv17_ctx_regs %d\n", reg);
+	return -1;
+}
+
+int nv10_graph_load_context(struct nouveau_channel *chan)
+{
+	struct drm_device *dev = chan->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct graph_state *pgraph_ctx = chan->pgraph_ctx;
+	uint32_t tmp;
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(nv10_graph_ctx_regs); i++)
+		nv_wr32(dev, nv10_graph_ctx_regs[i], pgraph_ctx->nv10[i]);
+	if (dev_priv->chipset >= 0x17) {
+		for (i = 0; i < ARRAY_SIZE(nv17_graph_ctx_regs); i++)
+			nv_wr32(dev, nv17_graph_ctx_regs[i],
+						pgraph_ctx->nv17[i]);
+	}
+
+	nv10_graph_load_pipe(chan);
+
+	nv_wr32(dev, NV10_PGRAPH_CTX_CONTROL, 0x10010100);
+	tmp = nv_rd32(dev, NV10_PGRAPH_CTX_USER);
+	nv_wr32(dev, NV10_PGRAPH_CTX_USER, (tmp & 0xffffff) | chan->id << 24);
+	tmp = nv_rd32(dev, NV10_PGRAPH_FFINTFC_ST2);
+	nv_wr32(dev, NV10_PGRAPH_FFINTFC_ST2, tmp & 0xcfffffff);
+	return 0;
+}
+
+int
+nv10_graph_unload_context(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
+	struct nouveau_fifo_engine *pfifo = &dev_priv->engine.fifo;
+	struct nouveau_channel *chan;
+	struct graph_state *ctx;
+	uint32_t tmp;
+	int i;
+
+	chan = pgraph->channel(dev);
+	if (!chan)
+		return 0;
+	ctx = chan->pgraph_ctx;
+
+	for (i = 0; i < ARRAY_SIZE(nv10_graph_ctx_regs); i++)
+		ctx->nv10[i] = nv_rd32(dev, nv10_graph_ctx_regs[i]);
+
+	if (dev_priv->chipset >= 0x17) {
+		for (i = 0; i < ARRAY_SIZE(nv17_graph_ctx_regs); i++)
+			ctx->nv17[i] = nv_rd32(dev, nv17_graph_ctx_regs[i]);
+	}
+
+	nv10_graph_save_pipe(chan);
+
+	nv_wr32(dev, NV10_PGRAPH_CTX_CONTROL, 0x10000000);
+	tmp  = nv_rd32(dev, NV10_PGRAPH_CTX_USER) & 0x00ffffff;
+	tmp |= (pfifo->channels - 1) << 24;
+	nv_wr32(dev, NV10_PGRAPH_CTX_USER, tmp);
+	return 0;
+}
+
+void
+nv10_graph_context_switch(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
+	struct nouveau_channel *chan = NULL;
+	int chid;
+
+	pgraph->fifo_access(dev, false);
+	nouveau_wait_for_idle(dev);
+
+	/* If previous context is valid, we need to save it */
+	nv10_graph_unload_context(dev);
+
+	/* Load context for next channel */
+	chid = (nv_rd32(dev, NV04_PGRAPH_TRAPPED_ADDR) >> 20) & 0x1f;
+	chan = dev_priv->fifos[chid];
+	if (chan)
+		nv10_graph_load_context(chan);
+
+	pgraph->fifo_access(dev, true);
+}
+
+#define NV_WRITE_CTX(reg, val) do { \
+	int offset = nv10_graph_ctx_regs_find_offset(dev, reg); \
+	if (offset > 0) \
+		pgraph_ctx->nv10[offset] = val; \
+	} while (0)
+
+#define NV17_WRITE_CTX(reg, val) do { \
+	int offset = nv17_graph_ctx_regs_find_offset(dev, reg); \
+	if (offset > 0) \
+		pgraph_ctx->nv17[offset] = val; \
+	} while (0)
+
+struct nouveau_channel *
+nv10_graph_channel(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	int chid = dev_priv->engine.fifo.channels;
+
+	if (nv_rd32(dev, NV10_PGRAPH_CTX_CONTROL) & 0x00010000)
+		chid = nv_rd32(dev, NV10_PGRAPH_CTX_USER) >> 24;
+
+	if (chid >= dev_priv->engine.fifo.channels)
+		return NULL;
+
+	return dev_priv->fifos[chid];
+}
+
+int nv10_graph_create_context(struct nouveau_channel *chan)
+{
+	struct drm_device *dev = chan->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct graph_state *pgraph_ctx;
+
+	NV_DEBUG(dev, "nv10_graph_context_create %d\n", chan->id);
+
+	chan->pgraph_ctx = pgraph_ctx = kzalloc(sizeof(*pgraph_ctx),
+						GFP_KERNEL);
+	if (pgraph_ctx == NULL)
+		return -ENOMEM;
+
+
+	NV_WRITE_CTX(0x00400e88, 0x08000000);
+	NV_WRITE_CTX(0x00400e9c, 0x4b7fffff);
+	NV_WRITE_CTX(NV03_PGRAPH_XY_LOGIC_MISC0, 0x0001ffff);
+	NV_WRITE_CTX(0x00400e10, 0x00001000);
+	NV_WRITE_CTX(0x00400e14, 0x00001000);
+	NV_WRITE_CTX(0x00400e30, 0x00080008);
+	NV_WRITE_CTX(0x00400e34, 0x00080008);
+	if (dev_priv->chipset >= 0x17) {
+		/* is it really needed ??? */
+		NV17_WRITE_CTX(NV10_PGRAPH_DEBUG_4,
+					nv_rd32(dev, NV10_PGRAPH_DEBUG_4));
+		NV17_WRITE_CTX(0x004006b0, nv_rd32(dev, 0x004006b0));
+		NV17_WRITE_CTX(0x00400eac, 0x0fff0000);
+		NV17_WRITE_CTX(0x00400eb0, 0x0fff0000);
+		NV17_WRITE_CTX(0x00400ec0, 0x00000080);
+		NV17_WRITE_CTX(0x00400ed0, 0x00000080);
+	}
+	NV_WRITE_CTX(NV10_PGRAPH_CTX_USER, chan->id << 24);
+
+	nv10_graph_create_pipe(chan);
+	return 0;
+}
+
+void nv10_graph_destroy_context(struct nouveau_channel *chan)
+{
+	struct graph_state *pgraph_ctx = chan->pgraph_ctx;
+
+	kfree(pgraph_ctx);
+	chan->pgraph_ctx = NULL;
+}
+
+void
+nv10_graph_set_region_tiling(struct drm_device *dev, int i, uint32_t addr,
+			     uint32_t size, uint32_t pitch)
+{
+	uint32_t limit = max(1u, addr + size) - 1;
+
+	if (pitch)
+		addr |= 1 << 31;
+
+	nv_wr32(dev, NV10_PGRAPH_TLIMIT(i), limit);
+	nv_wr32(dev, NV10_PGRAPH_TSIZE(i), pitch);
+	nv_wr32(dev, NV10_PGRAPH_TILE(i), addr);
+}
+
+int nv10_graph_init(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t tmp;
+	int i;
+
+	nv_wr32(dev, NV03_PMC_ENABLE, nv_rd32(dev, NV03_PMC_ENABLE) &
+			~NV_PMC_ENABLE_PGRAPH);
+	nv_wr32(dev, NV03_PMC_ENABLE, nv_rd32(dev, NV03_PMC_ENABLE) |
+			 NV_PMC_ENABLE_PGRAPH);
+
+	nv_wr32(dev, NV03_PGRAPH_INTR   , 0xFFFFFFFF);
+	nv_wr32(dev, NV03_PGRAPH_INTR_EN, 0xFFFFFFFF);
+
+	nv_wr32(dev, NV04_PGRAPH_DEBUG_0, 0xFFFFFFFF);
+	nv_wr32(dev, NV04_PGRAPH_DEBUG_0, 0x00000000);
+	nv_wr32(dev, NV04_PGRAPH_DEBUG_1, 0x00118700);
+	/* nv_wr32(dev, NV04_PGRAPH_DEBUG_2, 0x24E00810); */ /* 0x25f92ad9 */
+	nv_wr32(dev, NV04_PGRAPH_DEBUG_2, 0x25f92ad9);
+	nv_wr32(dev, NV04_PGRAPH_DEBUG_3, 0x55DE0830 |
+				      (1<<29) |
+				      (1<<31));
+	if (dev_priv->chipset >= 0x17) {
+		nv_wr32(dev, NV10_PGRAPH_DEBUG_4, 0x1f000000);
+		nv_wr32(dev, 0x400a10, 0x3ff3fb6);
+		nv_wr32(dev, 0x400838, 0x2f8684);
+		nv_wr32(dev, 0x40083c, 0x115f3f);
+		nv_wr32(dev, 0x004006b0, 0x40000020);
+	} else
+		nv_wr32(dev, NV10_PGRAPH_DEBUG_4, 0x00000000);
+
+	/* Turn all the tiling regions off. */
+	for (i = 0; i < NV10_PFB_TILE__SIZE; i++)
+		nv10_graph_set_region_tiling(dev, i, 0, 0, 0);
+
+	nv_wr32(dev, NV10_PGRAPH_CTX_SWITCH1, 0x00000000);
+	nv_wr32(dev, NV10_PGRAPH_CTX_SWITCH2, 0x00000000);
+	nv_wr32(dev, NV10_PGRAPH_CTX_SWITCH3, 0x00000000);
+	nv_wr32(dev, NV10_PGRAPH_CTX_SWITCH4, 0x00000000);
+	nv_wr32(dev, NV10_PGRAPH_STATE      , 0xFFFFFFFF);
+
+	tmp  = nv_rd32(dev, NV10_PGRAPH_CTX_USER) & 0x00ffffff;
+	tmp |= (dev_priv->engine.fifo.channels - 1) << 24;
+	nv_wr32(dev, NV10_PGRAPH_CTX_USER, tmp);
+	nv_wr32(dev, NV10_PGRAPH_CTX_CONTROL, 0x10000100);
+	nv_wr32(dev, NV10_PGRAPH_FFINTFC_ST2, 0x08000000);
+
+	return 0;
+}
+
+void nv10_graph_takedown(struct drm_device *dev)
+{
+}
+
+static int
+nv17_graph_mthd_lma_window(struct nouveau_channel *chan, int grclass,
+			   int mthd, uint32_t data)
+{
+	struct drm_device *dev = chan->dev;
+	struct graph_state *ctx = chan->pgraph_ctx;
+	struct pipe_state *pipe = &ctx->pipe_state;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
+	uint32_t pipe_0x0040[1], pipe_0x64c0[8], pipe_0x6a80[3], pipe_0x6ab0[3];
+	uint32_t xfmode0, xfmode1;
+	int i;
+
+	ctx->lma_window[(mthd - 0x1638) / 4] = data;
+
+	if (mthd != 0x1644)
+		return 0;
+
+	nouveau_wait_for_idle(dev);
+
+	PIPE_SAVE(dev, pipe_0x0040, 0x0040);
+	PIPE_SAVE(dev, pipe->pipe_0x0200, 0x0200);
+
+	PIPE_RESTORE(dev, ctx->lma_window, 0x6790);
+
+	nouveau_wait_for_idle(dev);
+
+	xfmode0 = nv_rd32(dev, NV10_PGRAPH_XFMODE0);
+	xfmode1 = nv_rd32(dev, NV10_PGRAPH_XFMODE1);
+
+	PIPE_SAVE(dev, pipe->pipe_0x4400, 0x4400);
+	PIPE_SAVE(dev, pipe_0x64c0, 0x64c0);
+	PIPE_SAVE(dev, pipe_0x6ab0, 0x6ab0);
+	PIPE_SAVE(dev, pipe_0x6a80, 0x6a80);
+
+	nouveau_wait_for_idle(dev);
+
+	nv_wr32(dev, NV10_PGRAPH_XFMODE0, 0x10000000);
+	nv_wr32(dev, NV10_PGRAPH_XFMODE1, 0x00000000);
+	nv_wr32(dev, NV10_PGRAPH_PIPE_ADDRESS, 0x000064c0);
+	for (i = 0; i < 4; i++)
+		nv_wr32(dev, NV10_PGRAPH_PIPE_DATA, 0x3f800000);
+	for (i = 0; i < 4; i++)
+		nv_wr32(dev, NV10_PGRAPH_PIPE_DATA, 0x00000000);
+
+	nv_wr32(dev, NV10_PGRAPH_PIPE_ADDRESS, 0x00006ab0);
+	for (i = 0; i < 3; i++)
+		nv_wr32(dev, NV10_PGRAPH_PIPE_DATA, 0x3f800000);
+
+	nv_wr32(dev, NV10_PGRAPH_PIPE_ADDRESS, 0x00006a80);
+	for (i = 0; i < 3; i++)
+		nv_wr32(dev, NV10_PGRAPH_PIPE_DATA, 0x00000000);
+
+	nv_wr32(dev, NV10_PGRAPH_PIPE_ADDRESS, 0x00000040);
+	nv_wr32(dev, NV10_PGRAPH_PIPE_DATA, 0x00000008);
+
+	PIPE_RESTORE(dev, pipe->pipe_0x0200, 0x0200);
+
+	nouveau_wait_for_idle(dev);
+
+	PIPE_RESTORE(dev, pipe_0x0040, 0x0040);
+
+	nv_wr32(dev, NV10_PGRAPH_XFMODE0, xfmode0);
+	nv_wr32(dev, NV10_PGRAPH_XFMODE1, xfmode1);
+
+	PIPE_RESTORE(dev, pipe_0x64c0, 0x64c0);
+	PIPE_RESTORE(dev, pipe_0x6ab0, 0x6ab0);
+	PIPE_RESTORE(dev, pipe_0x6a80, 0x6a80);
+	PIPE_RESTORE(dev, pipe->pipe_0x4400, 0x4400);
+
+	nv_wr32(dev, NV10_PGRAPH_PIPE_ADDRESS, 0x000000c0);
+	nv_wr32(dev, NV10_PGRAPH_PIPE_DATA, 0x00000000);
+
+	nouveau_wait_for_idle(dev);
+
+	pgraph->fifo_access(dev, true);
+
+	return 0;
+}
+
+static int
+nv17_graph_mthd_lma_enable(struct nouveau_channel *chan, int grclass,
+			   int mthd, uint32_t data)
+{
+	struct drm_device *dev = chan->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
+
+	nouveau_wait_for_idle(dev);
+
+	nv_wr32(dev, NV10_PGRAPH_DEBUG_4,
+		nv_rd32(dev, NV10_PGRAPH_DEBUG_4) | 0x1 << 8);
+	nv_wr32(dev, 0x004006b0,
+		nv_rd32(dev, 0x004006b0) | 0x8 << 24);
+
+	pgraph->fifo_access(dev, true);
+
+	return 0;
+}
+
+static struct nouveau_pgraph_object_method nv17_graph_celsius_mthds[] = {
+	{ 0x1638, nv17_graph_mthd_lma_window },
+	{ 0x163c, nv17_graph_mthd_lma_window },
+	{ 0x1640, nv17_graph_mthd_lma_window },
+	{ 0x1644, nv17_graph_mthd_lma_window },
+	{ 0x1658, nv17_graph_mthd_lma_enable },
+	{}
+};
+
+struct nouveau_pgraph_object_class nv10_graph_grclass[] = {
+	{ 0x0030, false, NULL }, /* null */
+	{ 0x0039, false, NULL }, /* m2mf */
+	{ 0x004a, false, NULL }, /* gdirect */
+	{ 0x005f, false, NULL }, /* imageblit */
+	{ 0x009f, false, NULL }, /* imageblit (nv12) */
+	{ 0x008a, false, NULL }, /* ifc */
+	{ 0x0089, false, NULL }, /* sifm */
+	{ 0x0062, false, NULL }, /* surf2d */
+	{ 0x0043, false, NULL }, /* rop */
+	{ 0x0012, false, NULL }, /* beta1 */
+	{ 0x0072, false, NULL }, /* beta4 */
+	{ 0x0019, false, NULL }, /* cliprect */
+	{ 0x0044, false, NULL }, /* pattern */
+	{ 0x0052, false, NULL }, /* swzsurf */
+	{ 0x0093, false, NULL }, /* surf3d */
+	{ 0x0094, false, NULL }, /* tex_tri */
+	{ 0x0095, false, NULL }, /* multitex_tri */
+	{ 0x0056, false, NULL }, /* celcius (nv10) */
+	{ 0x0096, false, NULL }, /* celcius (nv11) */
+	{ 0x0099, false, nv17_graph_celsius_mthds }, /* celcius (nv17) */
+	{}
+};
diff --git a/drivers/gpu/drm/nouveau/nv17_gpio.c b/drivers/gpu/drm/nouveau/nv17_gpio.c
new file mode 100644
index 000000000000..2e58c331e9b7
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv17_gpio.c
@@ -0,0 +1,92 @@
+/*
+ * Copyright (C) 2009 Francisco Jerez.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "drmP.h"
+#include "nouveau_drv.h"
+#include "nouveau_hw.h"
+
+static bool
+get_gpio_location(struct dcb_gpio_entry *ent, uint32_t *reg, uint32_t *shift,
+		  uint32_t *mask)
+{
+	if (ent->line < 2) {
+		*reg = NV_PCRTC_GPIO;
+		*shift = ent->line * 16;
+		*mask = 0x11;
+
+	} else if (ent->line < 10) {
+		*reg = NV_PCRTC_GPIO_EXT;
+		*shift = (ent->line - 2) * 4;
+		*mask = 0x3;
+
+	} else if (ent->line < 14) {
+		*reg = NV_PCRTC_850;
+		*shift = (ent->line - 10) * 4;
+		*mask = 0x3;
+
+	} else {
+		return false;
+	}
+
+	return true;
+}
+
+int
+nv17_gpio_get(struct drm_device *dev, enum dcb_gpio_tag tag)
+{
+	struct dcb_gpio_entry *ent = nouveau_bios_gpio_entry(dev, tag);
+	uint32_t reg, shift, mask, value;
+
+	if (!ent)
+		return -ENODEV;
+
+	if (!get_gpio_location(ent, &reg, &shift, &mask))
+		return -ENODEV;
+
+	value = NVReadCRTC(dev, 0, reg) >> shift;
+
+	return (ent->invert ? 1 : 0) ^ (value & 1);
+}
+
+int
+nv17_gpio_set(struct drm_device *dev, enum dcb_gpio_tag tag, int state)
+{
+	struct dcb_gpio_entry *ent = nouveau_bios_gpio_entry(dev, tag);
+	uint32_t reg, shift, mask, value;
+
+	if (!ent)
+		return -ENODEV;
+
+	if (!get_gpio_location(ent, &reg, &shift, &mask))
+		return -ENODEV;
+
+	value = ((ent->invert ? 1 : 0) ^ (state ? 1 : 0)) << shift;
+	mask = ~(mask << shift);
+
+	NVWriteCRTC(dev, 0, reg, value | (NVReadCRTC(dev, 0, reg) & mask));
+
+	return 0;
+}
diff --git a/drivers/gpu/drm/nouveau/nv17_tv.c b/drivers/gpu/drm/nouveau/nv17_tv.c
new file mode 100644
index 000000000000..74c880374fb9
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv17_tv.c
@@ -0,0 +1,778 @@
+/*
+ * Copyright (C) 2009 Francisco Jerez.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "drmP.h"
+#include "drm_crtc_helper.h"
+#include "nouveau_drv.h"
+#include "nouveau_encoder.h"
+#include "nouveau_connector.h"
+#include "nouveau_crtc.h"
+#include "nouveau_hw.h"
+#include "nv17_tv.h"
+
+static uint32_t nv42_tv_sample_load(struct drm_encoder *encoder)
+{
+	struct drm_device *dev = encoder->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t testval, regoffset = nv04_dac_output_offset(encoder);
+	uint32_t gpio0, gpio1, fp_htotal, fp_hsync_start, fp_hsync_end,
+		fp_control, test_ctrl, dacclk, ctv_14, ctv_1c, ctv_6c;
+	uint32_t sample = 0;
+	int head;
+
+#define RGB_TEST_DATA(r, g, b) (r << 0 | g << 10 | b << 20)
+	testval = RGB_TEST_DATA(0x82, 0xeb, 0x82);
+	if (dev_priv->vbios.tvdactestval)
+		testval = dev_priv->vbios.tvdactestval;
+
+	dacclk = NVReadRAMDAC(dev, 0, NV_PRAMDAC_DACCLK + regoffset);
+	head = (dacclk & 0x100) >> 8;
+
+	/* Save the previous state. */
+	gpio1 = nv17_gpio_get(dev, DCB_GPIO_TVDAC1);
+	gpio0 = nv17_gpio_get(dev, DCB_GPIO_TVDAC0);
+	fp_htotal = NVReadRAMDAC(dev, head, NV_PRAMDAC_FP_HTOTAL);
+	fp_hsync_start = NVReadRAMDAC(dev, head, NV_PRAMDAC_FP_HSYNC_START);
+	fp_hsync_end = NVReadRAMDAC(dev, head, NV_PRAMDAC_FP_HSYNC_END);
+	fp_control = NVReadRAMDAC(dev, head, NV_PRAMDAC_FP_TG_CONTROL);
+	test_ctrl = NVReadRAMDAC(dev, 0, NV_PRAMDAC_TEST_CONTROL + regoffset);
+	ctv_1c = NVReadRAMDAC(dev, head, 0x680c1c);
+	ctv_14 = NVReadRAMDAC(dev, head, 0x680c14);
+	ctv_6c = NVReadRAMDAC(dev, head, 0x680c6c);
+
+	/* Prepare the DAC for load detection.  */
+	nv17_gpio_set(dev, DCB_GPIO_TVDAC1, true);
+	nv17_gpio_set(dev, DCB_GPIO_TVDAC0, true);
+
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_FP_HTOTAL, 1343);
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_FP_HSYNC_START, 1047);
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_FP_HSYNC_END, 1183);
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_FP_TG_CONTROL,
+		      NV_PRAMDAC_FP_TG_CONTROL_DISPEN_POS |
+		      NV_PRAMDAC_FP_TG_CONTROL_WIDTH_12 |
+		      NV_PRAMDAC_FP_TG_CONTROL_READ_PROG |
+		      NV_PRAMDAC_FP_TG_CONTROL_HSYNC_POS |
+		      NV_PRAMDAC_FP_TG_CONTROL_VSYNC_POS);
+
+	NVWriteRAMDAC(dev, 0, NV_PRAMDAC_TEST_CONTROL + regoffset, 0);
+
+	NVWriteRAMDAC(dev, 0, NV_PRAMDAC_DACCLK + regoffset,
+		      (dacclk & ~0xff) | 0x22);
+	msleep(1);
+	NVWriteRAMDAC(dev, 0, NV_PRAMDAC_DACCLK + regoffset,
+		      (dacclk & ~0xff) | 0x21);
+
+	NVWriteRAMDAC(dev, head, 0x680c1c, 1 << 20);
+	NVWriteRAMDAC(dev, head, 0x680c14, 4 << 16);
+
+	/* Sample pin 0x4 (usually S-video luma). */
+	NVWriteRAMDAC(dev, head, 0x680c6c, testval >> 10 & 0x3ff);
+	msleep(20);
+	sample |= NVReadRAMDAC(dev, 0, NV_PRAMDAC_TEST_CONTROL + regoffset)
+		& 0x4 << 28;
+
+	/* Sample the remaining pins. */
+	NVWriteRAMDAC(dev, head, 0x680c6c, testval & 0x3ff);
+	msleep(20);
+	sample |= NVReadRAMDAC(dev, 0, NV_PRAMDAC_TEST_CONTROL + regoffset)
+		& 0xa << 28;
+
+	/* Restore the previous state. */
+	NVWriteRAMDAC(dev, head, 0x680c1c, ctv_1c);
+	NVWriteRAMDAC(dev, head, 0x680c14, ctv_14);
+	NVWriteRAMDAC(dev, head, 0x680c6c, ctv_6c);
+	NVWriteRAMDAC(dev, 0, NV_PRAMDAC_DACCLK + regoffset, dacclk);
+	NVWriteRAMDAC(dev, 0, NV_PRAMDAC_TEST_CONTROL + regoffset, test_ctrl);
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_FP_TG_CONTROL, fp_control);
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_FP_HSYNC_END, fp_hsync_end);
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_FP_HSYNC_START, fp_hsync_start);
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_FP_HTOTAL, fp_htotal);
+	nv17_gpio_set(dev, DCB_GPIO_TVDAC1, gpio1);
+	nv17_gpio_set(dev, DCB_GPIO_TVDAC0, gpio0);
+
+	return sample;
+}
+
+static enum drm_connector_status
+nv17_tv_detect(struct drm_encoder *encoder, struct drm_connector *connector)
+{
+	struct drm_device *dev = encoder->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct drm_mode_config *conf = &dev->mode_config;
+	struct nv17_tv_encoder *tv_enc = to_tv_enc(encoder);
+	struct dcb_entry *dcb = tv_enc->base.dcb;
+
+	if (dev_priv->chipset == 0x42 ||
+	    dev_priv->chipset == 0x43)
+		tv_enc->pin_mask = nv42_tv_sample_load(encoder) >> 28 & 0xe;
+	else
+		tv_enc->pin_mask = nv17_dac_sample_load(encoder) >> 28 & 0xe;
+
+	switch (tv_enc->pin_mask) {
+	case 0x2:
+	case 0x4:
+		tv_enc->subconnector = DRM_MODE_SUBCONNECTOR_Composite;
+		break;
+	case 0xc:
+		tv_enc->subconnector = DRM_MODE_SUBCONNECTOR_SVIDEO;
+		break;
+	case 0xe:
+		if (dcb->tvconf.has_component_output)
+			tv_enc->subconnector = DRM_MODE_SUBCONNECTOR_Component;
+		else
+			tv_enc->subconnector = DRM_MODE_SUBCONNECTOR_SCART;
+		break;
+	default:
+		tv_enc->subconnector = DRM_MODE_SUBCONNECTOR_Unknown;
+		break;
+	}
+
+	drm_connector_property_set_value(connector,
+					 conf->tv_subconnector_property,
+					 tv_enc->subconnector);
+
+	if (tv_enc->subconnector) {
+		NV_INFO(dev, "Load detected on output %c\n",
+			'@' + ffs(dcb->or));
+		return connector_status_connected;
+	} else {
+		return connector_status_disconnected;
+	}
+}
+
+static const struct {
+	int hdisplay;
+	int vdisplay;
+} modes[] = {
+	{ 640, 400 },
+	{ 640, 480 },
+	{ 720, 480 },
+	{ 720, 576 },
+	{ 800, 600 },
+	{ 1024, 768 },
+	{ 1280, 720 },
+	{ 1280, 1024 },
+	{ 1920, 1080 }
+};
+
+static int nv17_tv_get_modes(struct drm_encoder *encoder,
+			     struct drm_connector *connector)
+{
+	struct nv17_tv_norm_params *tv_norm = get_tv_norm(encoder);
+	struct drm_display_mode *mode;
+	struct drm_display_mode *output_mode;
+	int n = 0;
+	int i;
+
+	if (tv_norm->kind != CTV_ENC_MODE) {
+		struct drm_display_mode *tv_mode;
+
+		for (tv_mode = nv17_tv_modes; tv_mode->hdisplay; tv_mode++) {
+			mode = drm_mode_duplicate(encoder->dev, tv_mode);
+
+			mode->clock = tv_norm->tv_enc_mode.vrefresh *
+						mode->htotal / 1000 *
+						mode->vtotal / 1000;
+
+			if (mode->flags & DRM_MODE_FLAG_DBLSCAN)
+				mode->clock *= 2;
+
+			if (mode->hdisplay == tv_norm->tv_enc_mode.hdisplay &&
+			    mode->vdisplay == tv_norm->tv_enc_mode.vdisplay)
+				mode->type |= DRM_MODE_TYPE_PREFERRED;
+
+			drm_mode_probed_add(connector, mode);
+			n++;
+		}
+		return n;
+	}
+
+	/* tv_norm->kind == CTV_ENC_MODE */
+	output_mode = &tv_norm->ctv_enc_mode.mode;
+	for (i = 0; i < ARRAY_SIZE(modes); i++) {
+		if (modes[i].hdisplay > output_mode->hdisplay ||
+		    modes[i].vdisplay > output_mode->vdisplay)
+			continue;
+
+		if (modes[i].hdisplay == output_mode->hdisplay &&
+		    modes[i].vdisplay == output_mode->vdisplay) {
+			mode = drm_mode_duplicate(encoder->dev, output_mode);
+			mode->type |= DRM_MODE_TYPE_PREFERRED;
+		} else {
+			mode = drm_cvt_mode(encoder->dev, modes[i].hdisplay,
+				modes[i].vdisplay, 60, false,
+				output_mode->flags & DRM_MODE_FLAG_INTERLACE,
+				false);
+		}
+
+		/* CVT modes are sometimes unsuitable... */
+		if (output_mode->hdisplay <= 720
+		    || output_mode->hdisplay >= 1920) {
+			mode->htotal = output_mode->htotal;
+			mode->hsync_start = (mode->hdisplay + (mode->htotal
+					     - mode->hdisplay) * 9 / 10) & ~7;
+			mode->hsync_end = mode->hsync_start + 8;
+		}
+		if (output_mode->vdisplay >= 1024) {
+			mode->vtotal = output_mode->vtotal;
+			mode->vsync_start = output_mode->vsync_start;
+			mode->vsync_end = output_mode->vsync_end;
+		}
+
+		mode->type |= DRM_MODE_TYPE_DRIVER;
+		drm_mode_probed_add(connector, mode);
+		n++;
+	}
+	return n;
+}
+
+static int nv17_tv_mode_valid(struct drm_encoder *encoder,
+			      struct drm_display_mode *mode)
+{
+	struct nv17_tv_norm_params *tv_norm = get_tv_norm(encoder);
+
+	if (tv_norm->kind == CTV_ENC_MODE) {
+		struct drm_display_mode *output_mode =
+						&tv_norm->ctv_enc_mode.mode;
+
+		if (mode->clock > 400000)
+			return MODE_CLOCK_HIGH;
+
+		if (mode->hdisplay > output_mode->hdisplay ||
+		    mode->vdisplay > output_mode->vdisplay)
+			return MODE_BAD;
+
+		if ((mode->flags & DRM_MODE_FLAG_INTERLACE) !=
+		    (output_mode->flags & DRM_MODE_FLAG_INTERLACE))
+			return MODE_NO_INTERLACE;
+
+		if (mode->flags & DRM_MODE_FLAG_DBLSCAN)
+			return MODE_NO_DBLESCAN;
+
+	} else {
+		const int vsync_tolerance = 600;
+
+		if (mode->clock > 70000)
+			return MODE_CLOCK_HIGH;
+
+		if (abs(drm_mode_vrefresh(mode) * 1000 -
+			tv_norm->tv_enc_mode.vrefresh) > vsync_tolerance)
+			return MODE_VSYNC;
+
+		/* The encoder takes care of the actual interlacing */
+		if (mode->flags & DRM_MODE_FLAG_INTERLACE)
+			return MODE_NO_INTERLACE;
+	}
+
+	return MODE_OK;
+}
+
+static bool nv17_tv_mode_fixup(struct drm_encoder *encoder,
+			       struct drm_display_mode *mode,
+			       struct drm_display_mode *adjusted_mode)
+{
+	struct nv17_tv_norm_params *tv_norm = get_tv_norm(encoder);
+
+	if (tv_norm->kind == CTV_ENC_MODE)
+		adjusted_mode->clock = tv_norm->ctv_enc_mode.mode.clock;
+	else
+		adjusted_mode->clock = 90000;
+
+	return true;
+}
+
+static void  nv17_tv_dpms(struct drm_encoder *encoder, int mode)
+{
+	struct drm_device *dev = encoder->dev;
+	struct nv17_tv_state *regs = &to_tv_enc(encoder)->state;
+	struct nv17_tv_norm_params *tv_norm = get_tv_norm(encoder);
+
+	if (nouveau_encoder(encoder)->last_dpms == mode)
+		return;
+	nouveau_encoder(encoder)->last_dpms = mode;
+
+	NV_INFO(dev, "Setting dpms mode %d on TV encoder (output %d)\n",
+		 mode, nouveau_encoder(encoder)->dcb->index);
+
+	regs->ptv_200 &= ~1;
+
+	if (tv_norm->kind == CTV_ENC_MODE) {
+		nv04_dfp_update_fp_control(encoder, mode);
+
+	} else {
+		nv04_dfp_update_fp_control(encoder, DRM_MODE_DPMS_OFF);
+
+		if (mode == DRM_MODE_DPMS_ON)
+			regs->ptv_200 |= 1;
+	}
+
+	nv_load_ptv(dev, regs, 200);
+
+	nv17_gpio_set(dev, DCB_GPIO_TVDAC1, mode == DRM_MODE_DPMS_ON);
+	nv17_gpio_set(dev, DCB_GPIO_TVDAC0, mode == DRM_MODE_DPMS_ON);
+
+	nv04_dac_update_dacclk(encoder, mode == DRM_MODE_DPMS_ON);
+}
+
+static void nv17_tv_prepare(struct drm_encoder *encoder)
+{
+	struct drm_device *dev = encoder->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct drm_encoder_helper_funcs *helper = encoder->helper_private;
+	struct nv17_tv_norm_params *tv_norm = get_tv_norm(encoder);
+	int head = nouveau_crtc(encoder->crtc)->index;
+	uint8_t *cr_lcd = &dev_priv->mode_reg.crtc_reg[head].CRTC[
+							NV_CIO_CRE_LCD__INDEX];
+	uint32_t dacclk_off = NV_PRAMDAC_DACCLK +
+					nv04_dac_output_offset(encoder);
+	uint32_t dacclk;
+
+	helper->dpms(encoder, DRM_MODE_DPMS_OFF);
+
+	nv04_dfp_disable(dev, head);
+
+	/* Unbind any FP encoders from this head if we need the FP
+	 * stuff enabled. */
+	if (tv_norm->kind == CTV_ENC_MODE) {
+		struct drm_encoder *enc;
+
+		list_for_each_entry(enc, &dev->mode_config.encoder_list, head) {
+			struct dcb_entry *dcb = nouveau_encoder(enc)->dcb;
+
+			if ((dcb->type == OUTPUT_TMDS ||
+			     dcb->type == OUTPUT_LVDS) &&
+			     !enc->crtc &&
+			     nv04_dfp_get_bound_head(dev, dcb) == head) {
+				nv04_dfp_bind_head(dev, dcb, head ^ 1,
+						dev_priv->vbios.fp.dual_link);
+			}
+		}
+
+	}
+
+	/* Some NV4x have unknown values (0x3f, 0x50, 0x54, 0x6b, 0x79, 0x7f)
+	 * at LCD__INDEX which we don't alter
+	 */
+	if (!(*cr_lcd & 0x44)) {
+		if (tv_norm->kind == CTV_ENC_MODE)
+			*cr_lcd = 0x1 | (head ? 0x0 : 0x8);
+		else
+			*cr_lcd = 0;
+	}
+
+	/* Set the DACCLK register */
+	dacclk = (NVReadRAMDAC(dev, 0, dacclk_off) & ~0x30) | 0x1;
+
+	if (dev_priv->card_type == NV_40)
+		dacclk |= 0x1a << 16;
+
+	if (tv_norm->kind == CTV_ENC_MODE) {
+		dacclk |=  0x20;
+
+		if (head)
+			dacclk |= 0x100;
+		else
+			dacclk &= ~0x100;
+
+	} else {
+		dacclk |= 0x10;
+
+	}
+
+	NVWriteRAMDAC(dev, 0, dacclk_off, dacclk);
+}
+
+static void nv17_tv_mode_set(struct drm_encoder *encoder,
+			     struct drm_display_mode *drm_mode,
+			     struct drm_display_mode *adjusted_mode)
+{
+	struct drm_device *dev = encoder->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	int head = nouveau_crtc(encoder->crtc)->index;
+	struct nv04_crtc_reg *regs = &dev_priv->mode_reg.crtc_reg[head];
+	struct nv17_tv_state *tv_regs = &to_tv_enc(encoder)->state;
+	struct nv17_tv_norm_params *tv_norm = get_tv_norm(encoder);
+	int i;
+
+	regs->CRTC[NV_CIO_CRE_53] = 0x40; /* FP_HTIMING */
+	regs->CRTC[NV_CIO_CRE_54] = 0; /* FP_VTIMING */
+	regs->ramdac_630 = 0x2; /* turn off green mode (tv test pattern?) */
+	regs->tv_setup = 1;
+	regs->ramdac_8c0 = 0x0;
+
+	if (tv_norm->kind == TV_ENC_MODE) {
+		tv_regs->ptv_200 = 0x13111100;
+		if (head)
+			tv_regs->ptv_200 |= 0x10;
+
+		tv_regs->ptv_20c = 0x808010;
+		tv_regs->ptv_304 = 0x2d00000;
+		tv_regs->ptv_600 = 0x0;
+		tv_regs->ptv_60c = 0x0;
+		tv_regs->ptv_610 = 0x1e00000;
+
+		if (tv_norm->tv_enc_mode.vdisplay == 576) {
+			tv_regs->ptv_508 = 0x1200000;
+			tv_regs->ptv_614 = 0x33;
+
+		} else if (tv_norm->tv_enc_mode.vdisplay == 480) {
+			tv_regs->ptv_508 = 0xf00000;
+			tv_regs->ptv_614 = 0x13;
+		}
+
+		if (dev_priv->card_type >= NV_30) {
+			tv_regs->ptv_500 = 0xe8e0;
+			tv_regs->ptv_504 = 0x1710;
+			tv_regs->ptv_604 = 0x0;
+			tv_regs->ptv_608 = 0x0;
+		} else {
+			if (tv_norm->tv_enc_mode.vdisplay == 576) {
+				tv_regs->ptv_604 = 0x20;
+				tv_regs->ptv_608 = 0x10;
+				tv_regs->ptv_500 = 0x19710;
+				tv_regs->ptv_504 = 0x68f0;
+
+			} else if (tv_norm->tv_enc_mode.vdisplay == 480) {
+				tv_regs->ptv_604 = 0x10;
+				tv_regs->ptv_608 = 0x20;
+				tv_regs->ptv_500 = 0x4b90;
+				tv_regs->ptv_504 = 0x1b480;
+			}
+		}
+
+		for (i = 0; i < 0x40; i++)
+			tv_regs->tv_enc[i] = tv_norm->tv_enc_mode.tv_enc[i];
+
+	} else {
+		struct drm_display_mode *output_mode =
+						&tv_norm->ctv_enc_mode.mode;
+
+		/* The registers in PRAMDAC+0xc00 control some timings and CSC
+		 * parameters for the CTV encoder (It's only used for "HD" TV
+		 * modes, I don't think I have enough working to guess what
+		 * they exactly mean...), it's probably connected at the
+		 * output of the FP encoder, but it also needs the analog
+		 * encoder in its OR enabled and routed to the head it's
+		 * using. It's enabled with the DACCLK register, bits [5:4].
+		 */
+		for (i = 0; i < 38; i++)
+			regs->ctv_regs[i] = tv_norm->ctv_enc_mode.ctv_regs[i];
+
+		regs->fp_horiz_regs[FP_DISPLAY_END] = output_mode->hdisplay - 1;
+		regs->fp_horiz_regs[FP_TOTAL] = output_mode->htotal - 1;
+		regs->fp_horiz_regs[FP_SYNC_START] =
+						output_mode->hsync_start - 1;
+		regs->fp_horiz_regs[FP_SYNC_END] = output_mode->hsync_end - 1;
+		regs->fp_horiz_regs[FP_CRTC] = output_mode->hdisplay +
+			max((output_mode->hdisplay-600)/40 - 1, 1);
+
+		regs->fp_vert_regs[FP_DISPLAY_END] = output_mode->vdisplay - 1;
+		regs->fp_vert_regs[FP_TOTAL] = output_mode->vtotal - 1;
+		regs->fp_vert_regs[FP_SYNC_START] =
+						output_mode->vsync_start - 1;
+		regs->fp_vert_regs[FP_SYNC_END] = output_mode->vsync_end - 1;
+		regs->fp_vert_regs[FP_CRTC] = output_mode->vdisplay - 1;
+
+		regs->fp_control = NV_PRAMDAC_FP_TG_CONTROL_DISPEN_POS |
+			NV_PRAMDAC_FP_TG_CONTROL_READ_PROG |
+			NV_PRAMDAC_FP_TG_CONTROL_WIDTH_12;
+
+		if (output_mode->flags & DRM_MODE_FLAG_PVSYNC)
+			regs->fp_control |= NV_PRAMDAC_FP_TG_CONTROL_VSYNC_POS;
+		if (output_mode->flags & DRM_MODE_FLAG_PHSYNC)
+			regs->fp_control |= NV_PRAMDAC_FP_TG_CONTROL_HSYNC_POS;
+
+		regs->fp_debug_0 = NV_PRAMDAC_FP_DEBUG_0_YWEIGHT_ROUND |
+			NV_PRAMDAC_FP_DEBUG_0_XWEIGHT_ROUND |
+			NV_PRAMDAC_FP_DEBUG_0_YINTERP_BILINEAR |
+			NV_PRAMDAC_FP_DEBUG_0_XINTERP_BILINEAR |
+			NV_RAMDAC_FP_DEBUG_0_TMDS_ENABLED |
+			NV_PRAMDAC_FP_DEBUG_0_YSCALE_ENABLE |
+			NV_PRAMDAC_FP_DEBUG_0_XSCALE_ENABLE;
+
+		regs->fp_debug_2 = 0;
+
+		regs->fp_margin_color = 0x801080;
+
+	}
+}
+
+static void nv17_tv_commit(struct drm_encoder *encoder)
+{
+	struct drm_device *dev = encoder->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(encoder->crtc);
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	struct drm_encoder_helper_funcs *helper = encoder->helper_private;
+
+	if (get_tv_norm(encoder)->kind == TV_ENC_MODE) {
+		nv17_tv_update_rescaler(encoder);
+		nv17_tv_update_properties(encoder);
+	} else {
+		nv17_ctv_update_rescaler(encoder);
+	}
+
+	nv17_tv_state_load(dev, &to_tv_enc(encoder)->state);
+
+	/* This could use refinement for flatpanels, but it should work */
+	if (dev_priv->chipset < 0x44)
+		NVWriteRAMDAC(dev, 0, NV_PRAMDAC_TEST_CONTROL +
+					nv04_dac_output_offset(encoder),
+					0xf0000000);
+	else
+		NVWriteRAMDAC(dev, 0, NV_PRAMDAC_TEST_CONTROL +
+					nv04_dac_output_offset(encoder),
+					0x00100000);
+
+	helper->dpms(encoder, DRM_MODE_DPMS_ON);
+
+	NV_INFO(dev, "Output %s is running on CRTC %d using output %c\n",
+		drm_get_connector_name(
+			&nouveau_encoder_connector_get(nv_encoder)->base),
+		nv_crtc->index, '@' + ffs(nv_encoder->dcb->or));
+}
+
+static void nv17_tv_save(struct drm_encoder *encoder)
+{
+	struct drm_device *dev = encoder->dev;
+	struct nv17_tv_encoder *tv_enc = to_tv_enc(encoder);
+
+	nouveau_encoder(encoder)->restore.output =
+					NVReadRAMDAC(dev, 0,
+					NV_PRAMDAC_DACCLK +
+					nv04_dac_output_offset(encoder));
+
+	nv17_tv_state_save(dev, &tv_enc->saved_state);
+
+	tv_enc->state.ptv_200 = tv_enc->saved_state.ptv_200;
+}
+
+static void nv17_tv_restore(struct drm_encoder *encoder)
+{
+	struct drm_device *dev = encoder->dev;
+
+	NVWriteRAMDAC(dev, 0, NV_PRAMDAC_DACCLK +
+				nv04_dac_output_offset(encoder),
+				nouveau_encoder(encoder)->restore.output);
+
+	nv17_tv_state_load(dev, &to_tv_enc(encoder)->saved_state);
+
+	nouveau_encoder(encoder)->last_dpms = NV_DPMS_CLEARED;
+}
+
+static int nv17_tv_create_resources(struct drm_encoder *encoder,
+				    struct drm_connector *connector)
+{
+	struct drm_device *dev = encoder->dev;
+	struct drm_mode_config *conf = &dev->mode_config;
+	struct nv17_tv_encoder *tv_enc = to_tv_enc(encoder);
+	struct dcb_entry *dcb = nouveau_encoder(encoder)->dcb;
+	int num_tv_norms = dcb->tvconf.has_component_output ? NUM_TV_NORMS :
+							NUM_LD_TV_NORMS;
+	int i;
+
+	if (nouveau_tv_norm) {
+		for (i = 0; i < num_tv_norms; i++) {
+			if (!strcmp(nv17_tv_norm_names[i], nouveau_tv_norm)) {
+				tv_enc->tv_norm = i;
+				break;
+			}
+		}
+
+		if (i == num_tv_norms)
+			NV_WARN(dev, "Invalid TV norm setting \"%s\"\n",
+				nouveau_tv_norm);
+	}
+
+	drm_mode_create_tv_properties(dev, num_tv_norms, nv17_tv_norm_names);
+
+	drm_connector_attach_property(connector,
+					conf->tv_select_subconnector_property,
+					tv_enc->select_subconnector);
+	drm_connector_attach_property(connector,
+					conf->tv_subconnector_property,
+					tv_enc->subconnector);
+	drm_connector_attach_property(connector,
+					conf->tv_mode_property,
+					tv_enc->tv_norm);
+	drm_connector_attach_property(connector,
+					conf->tv_flicker_reduction_property,
+					tv_enc->flicker);
+	drm_connector_attach_property(connector,
+					conf->tv_saturation_property,
+					tv_enc->saturation);
+	drm_connector_attach_property(connector,
+					conf->tv_hue_property,
+					tv_enc->hue);
+	drm_connector_attach_property(connector,
+					conf->tv_overscan_property,
+					tv_enc->overscan);
+
+	return 0;
+}
+
+static int nv17_tv_set_property(struct drm_encoder *encoder,
+				struct drm_connector *connector,
+				struct drm_property *property,
+				uint64_t val)
+{
+	struct drm_mode_config *conf = &encoder->dev->mode_config;
+	struct drm_crtc *crtc = encoder->crtc;
+	struct nv17_tv_encoder *tv_enc = to_tv_enc(encoder);
+	struct nv17_tv_norm_params *tv_norm = get_tv_norm(encoder);
+	bool modes_changed = false;
+
+	if (property == conf->tv_overscan_property) {
+		tv_enc->overscan = val;
+		if (encoder->crtc) {
+			if (tv_norm->kind == CTV_ENC_MODE)
+				nv17_ctv_update_rescaler(encoder);
+			else
+				nv17_tv_update_rescaler(encoder);
+		}
+
+	} else if (property == conf->tv_saturation_property) {
+		if (tv_norm->kind != TV_ENC_MODE)
+			return -EINVAL;
+
+		tv_enc->saturation = val;
+		nv17_tv_update_properties(encoder);
+
+	} else if (property == conf->tv_hue_property) {
+		if (tv_norm->kind != TV_ENC_MODE)
+			return -EINVAL;
+
+		tv_enc->hue = val;
+		nv17_tv_update_properties(encoder);
+
+	} else if (property == conf->tv_flicker_reduction_property) {
+		if (tv_norm->kind != TV_ENC_MODE)
+			return -EINVAL;
+
+		tv_enc->flicker = val;
+		if (encoder->crtc)
+			nv17_tv_update_rescaler(encoder);
+
+	} else if (property == conf->tv_mode_property) {
+		if (connector->dpms != DRM_MODE_DPMS_OFF)
+			return -EINVAL;
+
+		tv_enc->tv_norm = val;
+
+		modes_changed = true;
+
+	} else if (property == conf->tv_select_subconnector_property) {
+		if (tv_norm->kind != TV_ENC_MODE)
+			return -EINVAL;
+
+		tv_enc->select_subconnector = val;
+		nv17_tv_update_properties(encoder);
+
+	} else {
+		return -EINVAL;
+	}
+
+	if (modes_changed) {
+		drm_helper_probe_single_connector_modes(connector, 0, 0);
+
+		/* Disable the crtc to ensure a full modeset is
+		 * performed whenever it's turned on again. */
+		if (crtc) {
+			struct drm_mode_set modeset = {
+				.crtc = crtc,
+			};
+
+			crtc->funcs->set_config(&modeset);
+		}
+	}
+
+	return 0;
+}
+
+static void nv17_tv_destroy(struct drm_encoder *encoder)
+{
+	struct nv17_tv_encoder *tv_enc = to_tv_enc(encoder);
+
+	NV_DEBUG_KMS(encoder->dev, "\n");
+
+	drm_encoder_cleanup(encoder);
+	kfree(tv_enc);
+}
+
+static struct drm_encoder_helper_funcs nv17_tv_helper_funcs = {
+	.dpms = nv17_tv_dpms,
+	.save = nv17_tv_save,
+	.restore = nv17_tv_restore,
+	.mode_fixup = nv17_tv_mode_fixup,
+	.prepare = nv17_tv_prepare,
+	.commit = nv17_tv_commit,
+	.mode_set = nv17_tv_mode_set,
+	.detect = nv17_tv_detect,
+};
+
+static struct drm_encoder_slave_funcs nv17_tv_slave_funcs = {
+	.get_modes = nv17_tv_get_modes,
+	.mode_valid = nv17_tv_mode_valid,
+	.create_resources = nv17_tv_create_resources,
+	.set_property = nv17_tv_set_property,
+};
+
+static struct drm_encoder_funcs nv17_tv_funcs = {
+	.destroy = nv17_tv_destroy,
+};
+
+int nv17_tv_create(struct drm_device *dev, struct dcb_entry *entry)
+{
+	struct drm_encoder *encoder;
+	struct nv17_tv_encoder *tv_enc = NULL;
+
+	tv_enc = kzalloc(sizeof(*tv_enc), GFP_KERNEL);
+	if (!tv_enc)
+		return -ENOMEM;
+
+	tv_enc->overscan = 50;
+	tv_enc->flicker = 50;
+	tv_enc->saturation = 50;
+	tv_enc->hue = 0;
+	tv_enc->tv_norm = TV_NORM_PAL;
+	tv_enc->subconnector = DRM_MODE_SUBCONNECTOR_Unknown;
+	tv_enc->select_subconnector = DRM_MODE_SUBCONNECTOR_Automatic;
+	tv_enc->pin_mask = 0;
+
+	encoder = to_drm_encoder(&tv_enc->base);
+
+	tv_enc->base.dcb = entry;
+	tv_enc->base.or = ffs(entry->or) - 1;
+
+	drm_encoder_init(dev, encoder, &nv17_tv_funcs, DRM_MODE_ENCODER_TVDAC);
+	drm_encoder_helper_add(encoder, &nv17_tv_helper_funcs);
+	to_encoder_slave(encoder)->slave_funcs = &nv17_tv_slave_funcs;
+
+	encoder->possible_crtcs = entry->heads;
+	encoder->possible_clones = 0;
+
+	return 0;
+}
diff --git a/drivers/gpu/drm/nouveau/nv17_tv.h b/drivers/gpu/drm/nouveau/nv17_tv.h
new file mode 100644
index 000000000000..c00977cedabd
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv17_tv.h
@@ -0,0 +1,156 @@
+/*
+ * Copyright (C) 2009 Francisco Jerez.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __NV17_TV_H__
+#define __NV17_TV_H__
+
+struct nv17_tv_state {
+	uint8_t tv_enc[0x40];
+
+	uint32_t hfilter[4][7];
+	uint32_t hfilter2[4][7];
+	uint32_t vfilter[4][7];
+
+	uint32_t ptv_200;
+	uint32_t ptv_204;
+	uint32_t ptv_208;
+	uint32_t ptv_20c;
+	uint32_t ptv_304;
+	uint32_t ptv_500;
+	uint32_t ptv_504;
+	uint32_t ptv_508;
+	uint32_t ptv_600;
+	uint32_t ptv_604;
+	uint32_t ptv_608;
+	uint32_t ptv_60c;
+	uint32_t ptv_610;
+	uint32_t ptv_614;
+};
+
+enum nv17_tv_norm{
+	TV_NORM_PAL,
+	TV_NORM_PAL_M,
+	TV_NORM_PAL_N,
+	TV_NORM_PAL_NC,
+	TV_NORM_NTSC_M,
+	TV_NORM_NTSC_J,
+	NUM_LD_TV_NORMS,
+	TV_NORM_HD480I = NUM_LD_TV_NORMS,
+	TV_NORM_HD480P,
+	TV_NORM_HD576I,
+	TV_NORM_HD576P,
+	TV_NORM_HD720P,
+	TV_NORM_HD1080I,
+	NUM_TV_NORMS
+};
+
+struct nv17_tv_encoder {
+	struct nouveau_encoder base;
+
+	struct nv17_tv_state state;
+	struct nv17_tv_state saved_state;
+
+	int overscan;
+	int flicker;
+	int saturation;
+	int hue;
+	enum nv17_tv_norm tv_norm;
+	int subconnector;
+	int select_subconnector;
+	uint32_t pin_mask;
+};
+#define to_tv_enc(x) container_of(nouveau_encoder(x),		\
+				  struct nv17_tv_encoder, base)
+
+extern char *nv17_tv_norm_names[NUM_TV_NORMS];
+
+extern struct nv17_tv_norm_params {
+	enum {
+		TV_ENC_MODE,
+		CTV_ENC_MODE,
+	} kind;
+
+	union {
+		struct {
+			int hdisplay;
+			int vdisplay;
+			int vrefresh; /* mHz */
+
+			uint8_t tv_enc[0x40];
+		} tv_enc_mode;
+
+		struct {
+			struct drm_display_mode mode;
+
+			uint32_t ctv_regs[38];
+		} ctv_enc_mode;
+	};
+
+} nv17_tv_norms[NUM_TV_NORMS];
+#define get_tv_norm(enc) (&nv17_tv_norms[to_tv_enc(enc)->tv_norm])
+
+extern struct drm_display_mode nv17_tv_modes[];
+
+static inline int interpolate(int y0, int y1, int y2, int x)
+{
+	return y1 + (x < 50 ? y1 - y0 : y2 - y1) * (x - 50) / 50;
+}
+
+void nv17_tv_state_save(struct drm_device *dev, struct nv17_tv_state *state);
+void nv17_tv_state_load(struct drm_device *dev, struct nv17_tv_state *state);
+void nv17_tv_update_properties(struct drm_encoder *encoder);
+void nv17_tv_update_rescaler(struct drm_encoder *encoder);
+void nv17_ctv_update_rescaler(struct drm_encoder *encoder);
+
+/* TV hardware access functions */
+
+static inline void nv_write_ptv(struct drm_device *dev, uint32_t reg, uint32_t val)
+{
+	nv_wr32(dev, reg, val);
+}
+
+static inline uint32_t nv_read_ptv(struct drm_device *dev, uint32_t reg)
+{
+	return nv_rd32(dev, reg);
+}
+
+static inline void nv_write_tv_enc(struct drm_device *dev, uint8_t reg, uint8_t val)
+{
+	nv_write_ptv(dev, NV_PTV_TV_INDEX, reg);
+	nv_write_ptv(dev, NV_PTV_TV_DATA, val);
+}
+
+static inline uint8_t nv_read_tv_enc(struct drm_device *dev, uint8_t reg)
+{
+	nv_write_ptv(dev, NV_PTV_TV_INDEX, reg);
+	return nv_read_ptv(dev, NV_PTV_TV_DATA);
+}
+
+#define nv_load_ptv(dev, state, reg) nv_write_ptv(dev, NV_PTV_OFFSET + 0x##reg, state->ptv_##reg)
+#define nv_save_ptv(dev, state, reg) state->ptv_##reg = nv_read_ptv(dev, NV_PTV_OFFSET + 0x##reg)
+#define nv_load_tv_enc(dev, state, reg) nv_write_tv_enc(dev, 0x##reg, state->tv_enc[0x##reg])
+
+#endif
diff --git a/drivers/gpu/drm/nouveau/nv17_tv_modes.c b/drivers/gpu/drm/nouveau/nv17_tv_modes.c
new file mode 100644
index 000000000000..d64683d97e0d
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv17_tv_modes.c
@@ -0,0 +1,583 @@
+/*
+ * Copyright (C) 2009 Francisco Jerez.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "drmP.h"
+#include "drm_crtc_helper.h"
+#include "nouveau_drv.h"
+#include "nouveau_encoder.h"
+#include "nouveau_crtc.h"
+#include "nouveau_hw.h"
+#include "nv17_tv.h"
+
+char *nv17_tv_norm_names[NUM_TV_NORMS] = {
+	[TV_NORM_PAL] = "PAL",
+	[TV_NORM_PAL_M] = "PAL-M",
+	[TV_NORM_PAL_N] = "PAL-N",
+	[TV_NORM_PAL_NC] = "PAL-Nc",
+	[TV_NORM_NTSC_M] = "NTSC-M",
+	[TV_NORM_NTSC_J] = "NTSC-J",
+	[TV_NORM_HD480I] = "hd480i",
+	[TV_NORM_HD480P] = "hd480p",
+	[TV_NORM_HD576I] = "hd576i",
+	[TV_NORM_HD576P] = "hd576p",
+	[TV_NORM_HD720P] = "hd720p",
+	[TV_NORM_HD1080I] = "hd1080i"
+};
+
+/* TV standard specific parameters */
+
+struct nv17_tv_norm_params nv17_tv_norms[NUM_TV_NORMS] = {
+	[TV_NORM_PAL] = { TV_ENC_MODE, {
+			.tv_enc_mode = { 720, 576, 50000, {
+					0x2a, 0x9, 0x8a, 0xcb, 0x0, 0x0, 0xb, 0x18,
+					0x7e, 0x40, 0x8a, 0x35, 0x27, 0x0, 0x34, 0x3,
+					0x3e, 0x3, 0x17, 0x21, 0x1b, 0x1b, 0x24, 0x9c,
+					0x1, 0x0, 0xf, 0xf, 0x60, 0x5, 0xd3, 0x3,
+					0xd3, 0x4, 0xd4, 0x1, 0x2, 0x0, 0xa, 0x5,
+					0x0, 0x1a, 0xff, 0x3, 0x18, 0xf, 0x78, 0x0,
+					0x0, 0xb4, 0x0, 0x15, 0x49, 0x10, 0x0, 0x9b,
+					0xbd, 0x15, 0x5, 0x15, 0x3e, 0x3, 0x0, 0x0
+				} } } },
+
+	[TV_NORM_PAL_M] = { TV_ENC_MODE, {
+			.tv_enc_mode = { 720, 480, 59940, {
+					0x21, 0xe6, 0xef, 0xe3, 0x0, 0x0, 0xb, 0x18,
+					0x7e, 0x44, 0x76, 0x32, 0x25, 0x0, 0x3c, 0x0,
+					0x3c, 0x0, 0x17, 0x21, 0x1b, 0x1b, 0x24, 0x83,
+					0x1, 0x0, 0xf, 0xf, 0x60, 0x5, 0xd3, 0x1,
+					0xc5, 0x4, 0xc5, 0x1, 0x2, 0x0, 0xa, 0x5,
+					0x0, 0x18, 0xff, 0x3, 0x20, 0xf, 0x78, 0x0,
+					0x0, 0xb4, 0x0, 0x15, 0x40, 0x10, 0x0, 0x9c,
+					0xc8, 0x15, 0x5, 0x15, 0x3c, 0x0, 0x0, 0x0
+				} } } },
+
+	[TV_NORM_PAL_N] = { TV_ENC_MODE, {
+			.tv_enc_mode = { 720, 576, 50000, {
+					0x2a, 0x9, 0x8a, 0xcb, 0x0, 0x0, 0xb, 0x18,
+					0x7e, 0x40, 0x8a, 0x32, 0x25, 0x0, 0x3c, 0x0,
+					0x3c, 0x0, 0x17, 0x21, 0x1b, 0x1b, 0x24, 0x9c,
+					0x1, 0x0, 0xf, 0xf, 0x60, 0x5, 0xd3, 0x1,
+					0xc5, 0x4, 0xc5, 0x1, 0x2, 0x0, 0xa, 0x5,
+					0x0, 0x1a, 0xff, 0x3, 0x18, 0xf, 0x78, 0x0,
+					0x0, 0xb4, 0x0, 0x15, 0x49, 0x10, 0x0, 0x9b,
+					0xbd, 0x15, 0x5, 0x15, 0x3c, 0x0, 0x0, 0x0
+				} } } },
+
+	[TV_NORM_PAL_NC] = { TV_ENC_MODE, {
+			.tv_enc_mode = { 720, 576, 50000, {
+					0x21, 0xf6, 0x94, 0x46, 0x0, 0x0, 0xb, 0x18,
+					0x7e, 0x44, 0x8a, 0x35, 0x27, 0x0, 0x34, 0x3,
+					0x3e, 0x3, 0x17, 0x21, 0x1b, 0x1b, 0x24, 0x9c,
+					0x1, 0x0, 0xf, 0xf, 0x60, 0x5, 0xd3, 0x3,
+					0xd3, 0x4, 0xd4, 0x1, 0x2, 0x0, 0xa, 0x5,
+					0x0, 0x1a, 0xff, 0x3, 0x18, 0xf, 0x78, 0x0,
+					0x0, 0xb4, 0x0, 0x15, 0x49, 0x10, 0x0, 0x9b,
+					0xbd, 0x15, 0x5, 0x15, 0x3e, 0x3, 0x0, 0x0
+				} } } },
+
+	[TV_NORM_NTSC_M] = { TV_ENC_MODE, {
+			.tv_enc_mode = { 720, 480, 59940, {
+					0x21, 0xf0, 0x7c, 0x1f, 0x0, 0x0, 0xb, 0x18,
+					0x7e, 0x44, 0x76, 0x48, 0x0, 0x0, 0x3c, 0x0,
+					0x3c, 0x0, 0x17, 0x21, 0x1b, 0x1b, 0x24, 0x83,
+					0x1, 0x0, 0xf, 0xf, 0x60, 0x5, 0xd3, 0x1,
+					0xc5, 0x4, 0xc5, 0x1, 0x2, 0x0, 0xa, 0x5,
+					0x0, 0x16, 0xff, 0x3, 0x20, 0xf, 0x78, 0x0,
+					0x0, 0xb4, 0x0, 0x15, 0x4, 0x10, 0x0, 0x9c,
+					0xc8, 0x15, 0x5, 0x15, 0x3c, 0x0, 0x0, 0x0
+				} } } },
+
+	[TV_NORM_NTSC_J] = { TV_ENC_MODE, {
+			.tv_enc_mode = { 720, 480, 59940, {
+					0x21, 0xf0, 0x7c, 0x1f, 0x0, 0x0, 0xb, 0x18,
+					0x7e, 0x44, 0x76, 0x48, 0x0, 0x0, 0x32, 0x0,
+					0x3c, 0x0, 0x17, 0x21, 0x1b, 0x1b, 0x24, 0x83,
+					0x1, 0x0, 0xf, 0xf, 0x60, 0x5, 0xd3, 0x1,
+					0xcf, 0x4, 0xcf, 0x1, 0x2, 0x0, 0xa, 0x5,
+					0x0, 0x16, 0xff, 0x3, 0x20, 0xf, 0x78, 0x0,
+					0x0, 0xb4, 0x0, 0x15, 0x4, 0x10, 0x0, 0xa4,
+					0xc8, 0x15, 0x5, 0x15, 0x3c, 0x0, 0x0, 0x0
+				} } } },
+
+	[TV_NORM_HD480I] = { TV_ENC_MODE, {
+			.tv_enc_mode = { 720, 480, 59940, {
+					0x21, 0xf0, 0x7c, 0x1f, 0x0, 0x0, 0xb, 0x18,
+					0x7e, 0x44, 0x76, 0x48, 0x0, 0x0, 0x32, 0x0,
+					0x3c, 0x0, 0x17, 0x21, 0x1b, 0x1b, 0x24, 0x83,
+					0x1, 0x0, 0xf, 0xf, 0x60, 0x5, 0xd3, 0x1,
+					0xcf, 0x4, 0xcf, 0x1, 0x2, 0x0, 0xa, 0x5,
+					0x0, 0x16, 0xff, 0x3, 0x20, 0xf, 0x78, 0x0,
+					0x0, 0xb4, 0x0, 0x15, 0x4, 0x10, 0x0, 0xa4,
+					0xc8, 0x15, 0x5, 0x15, 0x3c, 0x0, 0x0, 0x0
+				} } } },
+
+	[TV_NORM_HD576I] = { TV_ENC_MODE, {
+			.tv_enc_mode = { 720, 576, 50000, {
+					0x2a, 0x9, 0x8a, 0xcb, 0x0, 0x0, 0xb, 0x18,
+					0x7e, 0x40, 0x8a, 0x35, 0x27, 0x0, 0x34, 0x3,
+					0x3e, 0x3, 0x17, 0x21, 0x1b, 0x1b, 0x24, 0x9c,
+					0x1, 0x0, 0xf, 0xf, 0x60, 0x5, 0xd3, 0x3,
+					0xd3, 0x4, 0xd4, 0x1, 0x2, 0x0, 0xa, 0x5,
+					0x0, 0x1a, 0xff, 0x3, 0x18, 0xf, 0x78, 0x0,
+					0x0, 0xb4, 0x0, 0x15, 0x49, 0x10, 0x0, 0x9b,
+					0xbd, 0x15, 0x5, 0x15, 0x3e, 0x3, 0x0, 0x0
+				} } } },
+
+
+	[TV_NORM_HD480P] = { CTV_ENC_MODE, {
+			.ctv_enc_mode = {
+				.mode = { DRM_MODE("720x480", DRM_MODE_TYPE_DRIVER, 27000,
+						   720, 735, 743, 858, 0, 480, 490, 494, 525, 0,
+						   DRM_MODE_FLAG_PHSYNC | DRM_MODE_FLAG_PVSYNC) },
+				.ctv_regs = { 0x3540000, 0x0, 0x0, 0x314,
+					      0x354003a, 0x40000, 0x6f0344, 0x18100000,
+					      0x10160004, 0x10060005, 0x1006000c, 0x10060020,
+					      0x10060021, 0x140e0022, 0x10060202, 0x1802020a,
+					      0x1810020b, 0x10000fff, 0x10000fff, 0x10000fff,
+					      0x10000fff, 0x10000fff, 0x10000fff, 0x70,
+					      0x3ff0000, 0x57, 0x2e001e, 0x258012c,
+					      0xa0aa04ec, 0x30, 0x80960019, 0x12c0300,
+					      0x2019, 0x600, 0x32060019, 0x0, 0x0, 0x400
+				} } } },
+
+	[TV_NORM_HD576P] = { CTV_ENC_MODE, {
+			.ctv_enc_mode = {
+				.mode = { DRM_MODE("720x576", DRM_MODE_TYPE_DRIVER, 27000,
+						   720, 730, 738, 864, 0, 576, 581, 585, 625, 0,
+						   DRM_MODE_FLAG_PHSYNC | DRM_MODE_FLAG_PVSYNC) },
+				.ctv_regs = { 0x3540000, 0x0, 0x0, 0x314,
+					      0x354003a, 0x40000, 0x6f0344, 0x18100000,
+					      0x10060001, 0x10060009, 0x10060026, 0x10060027,
+					      0x140e0028, 0x10060268, 0x1810026d, 0x10000fff,
+					      0x10000fff, 0x10000fff, 0x10000fff, 0x10000fff,
+					      0x10000fff, 0x10000fff, 0x10000fff, 0x69,
+					      0x3ff0000, 0x57, 0x2e001e, 0x258012c,
+					      0xa0aa04ec, 0x30, 0x80960019, 0x12c0300,
+					      0x2019, 0x600, 0x32060019, 0x0, 0x0, 0x400
+				} } } },
+
+	[TV_NORM_HD720P] = { CTV_ENC_MODE, {
+			.ctv_enc_mode = {
+				.mode = { DRM_MODE("1280x720", DRM_MODE_TYPE_DRIVER, 74250,
+						   1280, 1349, 1357, 1650, 0, 720, 725, 730, 750, 0,
+						   DRM_MODE_FLAG_PHSYNC | DRM_MODE_FLAG_PVSYNC) },
+				.ctv_regs = { 0x1260394, 0x0, 0x0, 0x622,
+					      0x66b0021, 0x6004a, 0x1210626, 0x8170000,
+					      0x70004, 0x70016, 0x70017, 0x40f0018,
+					      0x702e8, 0x81702ed, 0xfff, 0xfff,
+					      0xfff, 0xfff, 0xfff, 0xfff,
+					      0xfff, 0xfff, 0xfff, 0x0,
+					      0x2e40001, 0x58, 0x2e001e, 0x258012c,
+					      0xa0aa04ec, 0x30, 0x810c0039, 0x12c0300,
+					      0xc0002039, 0x600, 0x32060039, 0x0, 0x0, 0x0
+				} } } },
+
+	[TV_NORM_HD1080I] = { CTV_ENC_MODE, {
+			.ctv_enc_mode = {
+				.mode = { DRM_MODE("1920x1080", DRM_MODE_TYPE_DRIVER, 74250,
+						   1920, 1961, 2049, 2200, 0, 1080, 1084, 1088, 1125, 0,
+						   DRM_MODE_FLAG_PHSYNC | DRM_MODE_FLAG_PVSYNC
+						   | DRM_MODE_FLAG_INTERLACE) },
+				.ctv_regs = { 0xac0420, 0x44c0478, 0x4a4, 0x4fc0868,
+					      0x8940028, 0x60054, 0xe80870, 0xbf70000,
+					      0xbc70004, 0x70005, 0x70012, 0x70013,
+					      0x40f0014, 0x70230, 0xbf70232, 0xbf70233,
+					      0x1c70237, 0x70238, 0x70244, 0x70245,
+					      0x40f0246, 0x70462, 0x1f70464, 0x0,
+					      0x2e40001, 0x58, 0x2e001e, 0x258012c,
+					      0xa0aa04ec, 0x30, 0x815f004c, 0x12c0300,
+					      0xc000204c, 0x600, 0x3206004c, 0x0, 0x0, 0x0
+				} } } }
+};
+
+/*
+ * The following is some guesswork on how the TV encoder flicker
+ * filter/rescaler works:
+ *
+ * It seems to use some sort of resampling filter, it is controlled
+ * through the registers at NV_PTV_HFILTER and NV_PTV_VFILTER, they
+ * control the horizontal and vertical stage respectively, there is
+ * also NV_PTV_HFILTER2 the blob fills identically to NV_PTV_HFILTER,
+ * but they seem to do nothing. A rough guess might be that they could
+ * be used to independently control the filtering of each interlaced
+ * field, but I don't know how they are enabled. The whole filtering
+ * process seems to be disabled with bits 26:27 of PTV_200, but we
+ * aren't doing that.
+ *
+ * The layout of both register sets is the same:
+ *
+ * A: [BASE+0x18]...[BASE+0x0] [BASE+0x58]..[BASE+0x40]
+ * B: [BASE+0x34]...[BASE+0x1c] [BASE+0x74]..[BASE+0x5c]
+ *
+ * Each coefficient is stored in bits [31],[15:9] in two's complement
+ * format. They seem to be some kind of weights used in a low-pass
+ * filter. Both A and B coefficients are applied to the 14 nearest
+ * samples on each side (Listed from nearest to furthermost.  They
+ * roughly cover 2 framebuffer pixels on each side).  They are
+ * probably multiplied with some more hardwired weights before being
+ * used: B-coefficients are applied the same on both sides,
+ * A-coefficients are inverted before being applied to the opposite
+ * side.
+ *
+ * After all the hassle, I got the following formula by empirical
+ * means...
+ */
+
+#define calc_overscan(o) interpolate(0x100, 0xe1, 0xc1, o)
+
+#define id1 (1LL << 8)
+#define id2 (1LL << 16)
+#define id3 (1LL << 24)
+#define id4 (1LL << 32)
+#define id5 (1LL << 48)
+
+static struct filter_params{
+	int64_t k1;
+	int64_t ki;
+	int64_t ki2;
+	int64_t ki3;
+	int64_t kr;
+	int64_t kir;
+	int64_t ki2r;
+	int64_t ki3r;
+	int64_t kf;
+	int64_t kif;
+	int64_t ki2f;
+	int64_t ki3f;
+	int64_t krf;
+	int64_t kirf;
+	int64_t ki2rf;
+	int64_t ki3rf;
+} fparams[2][4] = {
+	/* Horizontal filter parameters */
+	{
+		{64.311690 * id5, -39.516924 * id5, 6.586143 * id5, 0.000002 * id5,
+		 0.051285 * id4, 26.168746 * id4, -4.361449 * id4, -0.000001 * id4,
+		 9.308169 * id3, 78.180965 * id3, -13.030158 * id3, -0.000001 * id3,
+		 -8.801540 * id1, -46.572890 * id1, 7.762145 * id1, -0.000000 * id1},
+		{-44.565569 * id5, -68.081246 * id5, 39.812074 * id5, -4.009316 * id5,
+		 29.832207 * id4, 50.047322 * id4, -25.380017 * id4, 2.546422 * id4,
+		 104.605622 * id3, 141.908641 * id3, -74.322319 * id3, 7.484316 * id3,
+		 -37.081621 * id1, -90.397510 * id1, 42.784229 * id1, -4.289952 * id1},
+		{-56.793244 * id5, 31.153584 * id5, -5.192247 * id5, -0.000003 * id5,
+		 33.541131 * id4, -34.149302 * id4, 5.691537 * id4, 0.000002 * id4,
+		 87.196610 * id3, -88.995169 * id3, 14.832456 * id3, 0.000012 * id3,
+		 17.288138 * id1, 71.864786 * id1, -11.977408 * id1, -0.000009 * id1},
+		{51.787796 * id5, 21.211771 * id5, -18.993730 * id5, 1.853310 * id5,
+		 -41.470726 * id4, -17.775823 * id4, 13.057821 * id4, -1.15823 * id4,
+		 -154.235673 * id3, -44.878641 * id3, 40.656077 * id3, -3.695595 * id3,
+		 112.201065 * id1, 39.992155 * id1, -25.155714 * id1, 2.113984 * id1},
+	},
+
+	/* Vertical filter parameters */
+	{
+		{67.601979 * id5, 0.428319 * id5, -0.071318 * id5, -0.000012 * id5,
+		 -3.402339 * id4, 0.000209 * id4, -0.000092 * id4, 0.000010 * id4,
+		 -9.180996 * id3, 6.111270 * id3, -1.024457 * id3, 0.001043 * id3,
+		 6.060315 * id1, -0.017425 * id1, 0.007830 * id1, -0.000869 * id1},
+		{6.755647 * id5, 5.841348 * id5, 1.469734 * id5, -0.149656 * id5,
+		 8.293120 * id4, -1.192888 * id4, -0.947652 * id4, 0.094507 * id4,
+		 37.526655 * id3, 10.257875 * id3, -10.823275 * id3, 1.081497 * id3,
+		 -2.361928 * id1, -2.059432 * id1, 1.840671 * id1, -0.168100 * id1},
+		{-14.780391 * id5, -16.042148 * id5, 2.673692 * id5, -0.000000 * id5,
+		 39.541978 * id4, 5.680053 * id4, -0.946676 * id4, 0.000000 * id4,
+		 152.994486 * id3, 12.625439 * id3, -2.119579 * id3, 0.002708 * id3,
+		 -38.125089 * id1, -0.855880 * id1, 0.155359 * id1, -0.002245 * id1},
+		{-27.476193 * id5, -1.454976 * id5, 1.286557 * id5, 0.025346 * id5,
+		 20.687300 * id4, 3.014003 * id4, -0.557786 * id4, -0.01311 * id4,
+		 60.008737 * id3, -0.738273 * id3, 5.408217 * id3, -0.796798 * id3,
+		 -17.296835 * id1, 4.438577 * id1, -2.809420 * id1, 0.385491 * id1},
+	}
+};
+
+static void tv_setup_filter(struct drm_encoder *encoder)
+{
+	struct nv17_tv_encoder *tv_enc = to_tv_enc(encoder);
+	struct nv17_tv_norm_params *tv_norm = get_tv_norm(encoder);
+	struct drm_display_mode *mode = &encoder->crtc->mode;
+	uint32_t (*filters[])[4][7] = {&tv_enc->state.hfilter,
+				       &tv_enc->state.vfilter};
+	int i, j, k;
+	int32_t overscan = calc_overscan(tv_enc->overscan);
+	int64_t flicker = (tv_enc->flicker - 50) * (id3 / 100);
+	uint64_t rs[] = {mode->hdisplay * id3,
+			 mode->vdisplay * id3};
+
+	do_div(rs[0], overscan * tv_norm->tv_enc_mode.hdisplay);
+	do_div(rs[1], overscan * tv_norm->tv_enc_mode.vdisplay);
+
+	for (k = 0; k < 2; k++) {
+		rs[k] = max((int64_t)rs[k], id2);
+
+		for (j = 0; j < 4; j++) {
+			struct filter_params *p = &fparams[k][j];
+
+			for (i = 0; i < 7; i++) {
+				int64_t c = (p->k1 + p->ki*i + p->ki2*i*i + p->ki3*i*i*i)
+					+ (p->kr + p->kir*i + p->ki2r*i*i + p->ki3r*i*i*i)*rs[k]
+					+ (p->kf + p->kif*i + p->ki2f*i*i + p->ki3f*i*i*i)*flicker
+					+ (p->krf + p->kirf*i + p->ki2rf*i*i + p->ki3rf*i*i*i)*flicker*rs[k];
+
+				(*filters[k])[j][i] = (c + id5/2) >> 39 & (0x1 << 31 | 0x7f << 9);
+			}
+		}
+	}
+}
+
+/* Hardware state saving/restoring */
+
+static void tv_save_filter(struct drm_device *dev, uint32_t base, uint32_t regs[4][7])
+{
+	int i, j;
+	uint32_t offsets[] = { base, base + 0x1c, base + 0x40, base + 0x5c };
+
+	for (i = 0; i < 4; i++) {
+		for (j = 0; j < 7; j++)
+			regs[i][j] = nv_read_ptv(dev, offsets[i]+4*j);
+	}
+}
+
+static void tv_load_filter(struct drm_device *dev, uint32_t base, uint32_t regs[4][7])
+{
+	int i, j;
+	uint32_t offsets[] = { base, base + 0x1c, base + 0x40, base + 0x5c };
+
+	for (i = 0; i < 4; i++) {
+		for (j = 0; j < 7; j++)
+			nv_write_ptv(dev, offsets[i]+4*j, regs[i][j]);
+	}
+}
+
+void nv17_tv_state_save(struct drm_device *dev, struct nv17_tv_state *state)
+{
+	int i;
+
+	for (i = 0; i < 0x40; i++)
+		state->tv_enc[i] = nv_read_tv_enc(dev, i);
+
+	tv_save_filter(dev, NV_PTV_HFILTER, state->hfilter);
+	tv_save_filter(dev, NV_PTV_HFILTER2, state->hfilter2);
+	tv_save_filter(dev, NV_PTV_VFILTER, state->vfilter);
+
+	nv_save_ptv(dev, state, 200);
+	nv_save_ptv(dev, state, 204);
+	nv_save_ptv(dev, state, 208);
+	nv_save_ptv(dev, state, 20c);
+	nv_save_ptv(dev, state, 304);
+	nv_save_ptv(dev, state, 500);
+	nv_save_ptv(dev, state, 504);
+	nv_save_ptv(dev, state, 508);
+	nv_save_ptv(dev, state, 600);
+	nv_save_ptv(dev, state, 604);
+	nv_save_ptv(dev, state, 608);
+	nv_save_ptv(dev, state, 60c);
+	nv_save_ptv(dev, state, 610);
+	nv_save_ptv(dev, state, 614);
+}
+
+void nv17_tv_state_load(struct drm_device *dev, struct nv17_tv_state *state)
+{
+	int i;
+
+	for (i = 0; i < 0x40; i++)
+		nv_write_tv_enc(dev, i, state->tv_enc[i]);
+
+	tv_load_filter(dev, NV_PTV_HFILTER, state->hfilter);
+	tv_load_filter(dev, NV_PTV_HFILTER2, state->hfilter2);
+	tv_load_filter(dev, NV_PTV_VFILTER, state->vfilter);
+
+	nv_load_ptv(dev, state, 200);
+	nv_load_ptv(dev, state, 204);
+	nv_load_ptv(dev, state, 208);
+	nv_load_ptv(dev, state, 20c);
+	nv_load_ptv(dev, state, 304);
+	nv_load_ptv(dev, state, 500);
+	nv_load_ptv(dev, state, 504);
+	nv_load_ptv(dev, state, 508);
+	nv_load_ptv(dev, state, 600);
+	nv_load_ptv(dev, state, 604);
+	nv_load_ptv(dev, state, 608);
+	nv_load_ptv(dev, state, 60c);
+	nv_load_ptv(dev, state, 610);
+	nv_load_ptv(dev, state, 614);
+
+	/* This is required for some settings to kick in. */
+	nv_write_tv_enc(dev, 0x3e, 1);
+	nv_write_tv_enc(dev, 0x3e, 0);
+}
+
+/* Timings similar to the ones the blob sets */
+
+struct drm_display_mode nv17_tv_modes[] = {
+	{ DRM_MODE("320x200", DRM_MODE_TYPE_DRIVER, 0,
+		   320, 344, 392, 560, 0, 200, 200, 202, 220, 0,
+		   DRM_MODE_FLAG_NHSYNC | DRM_MODE_FLAG_NVSYNC
+		   | DRM_MODE_FLAG_DBLSCAN | DRM_MODE_FLAG_CLKDIV2) },
+	{ DRM_MODE("320x240", DRM_MODE_TYPE_DRIVER, 0,
+		   320, 344, 392, 560, 0, 240, 240, 246, 263, 0,
+		   DRM_MODE_FLAG_NHSYNC | DRM_MODE_FLAG_NVSYNC
+		   | DRM_MODE_FLAG_DBLSCAN | DRM_MODE_FLAG_CLKDIV2) },
+	{ DRM_MODE("400x300", DRM_MODE_TYPE_DRIVER, 0,
+		   400, 432, 496, 640, 0, 300, 300, 303, 314, 0,
+		   DRM_MODE_FLAG_PHSYNC | DRM_MODE_FLAG_PVSYNC
+		   | DRM_MODE_FLAG_DBLSCAN | DRM_MODE_FLAG_CLKDIV2) },
+	{ DRM_MODE("640x480", DRM_MODE_TYPE_DRIVER, 0,
+		   640, 672, 768, 880, 0, 480, 480, 492, 525, 0,
+		   DRM_MODE_FLAG_NHSYNC | DRM_MODE_FLAG_NVSYNC) },
+	{ DRM_MODE("720x480", DRM_MODE_TYPE_DRIVER, 0,
+		   720, 752, 872, 960, 0, 480, 480, 493, 525, 0,
+		   DRM_MODE_FLAG_NHSYNC | DRM_MODE_FLAG_NVSYNC) },
+	{ DRM_MODE("720x576", DRM_MODE_TYPE_DRIVER, 0,
+		   720, 776, 856, 960, 0, 576, 576, 588, 597, 0,
+		   DRM_MODE_FLAG_NHSYNC | DRM_MODE_FLAG_NVSYNC) },
+	{ DRM_MODE("800x600", DRM_MODE_TYPE_DRIVER, 0,
+		   800, 840, 920, 1040, 0, 600, 600, 604, 618, 0,
+		   DRM_MODE_FLAG_PHSYNC | DRM_MODE_FLAG_PVSYNC) },
+	{ DRM_MODE("1024x768", DRM_MODE_TYPE_DRIVER, 0,
+		   1024, 1064, 1200, 1344, 0, 768, 768, 777, 806, 0,
+		   DRM_MODE_FLAG_NHSYNC | DRM_MODE_FLAG_NVSYNC) },
+	{}
+};
+
+void nv17_tv_update_properties(struct drm_encoder *encoder)
+{
+	struct drm_device *dev = encoder->dev;
+	struct nv17_tv_encoder *tv_enc = to_tv_enc(encoder);
+	struct nv17_tv_state *regs = &tv_enc->state;
+	struct nv17_tv_norm_params *tv_norm = get_tv_norm(encoder);
+	int subconnector = tv_enc->select_subconnector ?
+						tv_enc->select_subconnector :
+						tv_enc->subconnector;
+
+	switch (subconnector) {
+	case DRM_MODE_SUBCONNECTOR_Composite:
+	{
+		regs->ptv_204 = 0x2;
+
+		/* The composite connector may be found on either pin. */
+		if (tv_enc->pin_mask & 0x4)
+			regs->ptv_204 |= 0x010000;
+		else if (tv_enc->pin_mask & 0x2)
+			regs->ptv_204 |= 0x100000;
+		else
+			regs->ptv_204 |= 0x110000;
+
+		regs->tv_enc[0x7] = 0x10;
+		break;
+	}
+	case DRM_MODE_SUBCONNECTOR_SVIDEO:
+		regs->ptv_204 = 0x11012;
+		regs->tv_enc[0x7] = 0x18;
+		break;
+
+	case DRM_MODE_SUBCONNECTOR_Component:
+		regs->ptv_204 = 0x111333;
+		regs->tv_enc[0x7] = 0x14;
+		break;
+
+	case DRM_MODE_SUBCONNECTOR_SCART:
+		regs->ptv_204 = 0x111012;
+		regs->tv_enc[0x7] = 0x18;
+		break;
+	}
+
+	regs->tv_enc[0x20] = interpolate(0, tv_norm->tv_enc_mode.tv_enc[0x20], 255,
+					 tv_enc->saturation);
+	regs->tv_enc[0x22] = interpolate(0, tv_norm->tv_enc_mode.tv_enc[0x22], 255,
+					 tv_enc->saturation);
+	regs->tv_enc[0x25] = tv_enc->hue * 255 / 100;
+
+	nv_load_ptv(dev, regs, 204);
+	nv_load_tv_enc(dev, regs, 7);
+	nv_load_tv_enc(dev, regs, 20);
+	nv_load_tv_enc(dev, regs, 22);
+	nv_load_tv_enc(dev, regs, 25);
+}
+
+void nv17_tv_update_rescaler(struct drm_encoder *encoder)
+{
+	struct drm_device *dev = encoder->dev;
+	struct nv17_tv_encoder *tv_enc = to_tv_enc(encoder);
+	struct nv17_tv_state *regs = &tv_enc->state;
+
+	regs->ptv_208 = 0x40 | (calc_overscan(tv_enc->overscan) << 8);
+
+	tv_setup_filter(encoder);
+
+	nv_load_ptv(dev, regs, 208);
+	tv_load_filter(dev, NV_PTV_HFILTER, regs->hfilter);
+	tv_load_filter(dev, NV_PTV_HFILTER2, regs->hfilter2);
+	tv_load_filter(dev, NV_PTV_VFILTER, regs->vfilter);
+}
+
+void nv17_ctv_update_rescaler(struct drm_encoder *encoder)
+{
+	struct drm_device *dev = encoder->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nv17_tv_encoder *tv_enc = to_tv_enc(encoder);
+	int head = nouveau_crtc(encoder->crtc)->index;
+	struct nv04_crtc_reg *regs = &dev_priv->mode_reg.crtc_reg[head];
+	struct drm_display_mode *crtc_mode = &encoder->crtc->mode;
+	struct drm_display_mode *output_mode = &get_tv_norm(encoder)->ctv_enc_mode.mode;
+	int overscan, hmargin, vmargin, hratio, vratio;
+
+	/* The rescaler doesn't do the right thing for interlaced modes. */
+	if (output_mode->flags & DRM_MODE_FLAG_INTERLACE)
+		overscan = 100;
+	else
+		overscan = tv_enc->overscan;
+
+	hmargin = (output_mode->hdisplay - crtc_mode->hdisplay) / 2;
+	vmargin = (output_mode->vdisplay - crtc_mode->vdisplay) / 2;
+
+	hmargin = interpolate(0, min(hmargin, output_mode->hdisplay/20), hmargin,
+			      overscan);
+	vmargin = interpolate(0, min(vmargin, output_mode->vdisplay/20), vmargin,
+			      overscan);
+
+	hratio = crtc_mode->hdisplay * 0x800 / (output_mode->hdisplay - 2*hmargin);
+	vratio = crtc_mode->vdisplay * 0x800 / (output_mode->vdisplay - 2*vmargin) & ~3;
+
+	regs->fp_horiz_regs[FP_VALID_START] = hmargin;
+	regs->fp_horiz_regs[FP_VALID_END] = output_mode->hdisplay - hmargin - 1;
+	regs->fp_vert_regs[FP_VALID_START] = vmargin;
+	regs->fp_vert_regs[FP_VALID_END] = output_mode->vdisplay - vmargin - 1;
+
+	regs->fp_debug_1 = NV_PRAMDAC_FP_DEBUG_1_YSCALE_TESTMODE_ENABLE |
+		XLATE(vratio, 0, NV_PRAMDAC_FP_DEBUG_1_YSCALE_VALUE) |
+		NV_PRAMDAC_FP_DEBUG_1_XSCALE_TESTMODE_ENABLE |
+		XLATE(hratio, 0, NV_PRAMDAC_FP_DEBUG_1_XSCALE_VALUE);
+
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_FP_HVALID_START,
+		      regs->fp_horiz_regs[FP_VALID_START]);
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_FP_HVALID_END,
+		      regs->fp_horiz_regs[FP_VALID_END]);
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_FP_VVALID_START,
+		      regs->fp_vert_regs[FP_VALID_START]);
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_FP_VVALID_END,
+		      regs->fp_vert_regs[FP_VALID_END]);
+	NVWriteRAMDAC(dev, head, NV_PRAMDAC_FP_DEBUG_1, regs->fp_debug_1);
+}
diff --git a/drivers/gpu/drm/nouveau/nv20_graph.c b/drivers/gpu/drm/nouveau/nv20_graph.c
new file mode 100644
index 000000000000..d6fc0a82f03d
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv20_graph.c
@@ -0,0 +1,775 @@
+#include "drmP.h"
+#include "drm.h"
+#include "nouveau_drv.h"
+#include "nouveau_drm.h"
+
+/*
+ * NV20
+ * -----
+ * There are 3 families :
+ * NV20 is 0x10de:0x020*
+ * NV25/28 is 0x10de:0x025* / 0x10de:0x028*
+ * NV2A is 0x10de:0x02A0
+ *
+ * NV30
+ * -----
+ * There are 3 families :
+ * NV30/31 is 0x10de:0x030* / 0x10de:0x031*
+ * NV34 is 0x10de:0x032*
+ * NV35/36 is 0x10de:0x033* / 0x10de:0x034*
+ *
+ * Not seen in the wild, no dumps (probably NV35) :
+ * NV37 is 0x10de:0x00fc, 0x10de:0x00fd
+ * NV38 is 0x10de:0x0333, 0x10de:0x00fe
+ *
+ */
+
+#define NV20_GRCTX_SIZE (3580*4)
+#define NV25_GRCTX_SIZE (3529*4)
+#define NV2A_GRCTX_SIZE (3500*4)
+
+#define NV30_31_GRCTX_SIZE (24392)
+#define NV34_GRCTX_SIZE    (18140)
+#define NV35_36_GRCTX_SIZE (22396)
+
+static void
+nv20_graph_context_init(struct drm_device *dev, struct nouveau_gpuobj *ctx)
+{
+	int i;
+
+	nv_wo32(dev, ctx, 0x033c/4, 0xffff0000);
+	nv_wo32(dev, ctx, 0x03a0/4, 0x0fff0000);
+	nv_wo32(dev, ctx, 0x03a4/4, 0x0fff0000);
+	nv_wo32(dev, ctx, 0x047c/4, 0x00000101);
+	nv_wo32(dev, ctx, 0x0490/4, 0x00000111);
+	nv_wo32(dev, ctx, 0x04a8/4, 0x44400000);
+	for (i = 0x04d4; i <= 0x04e0; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x00030303);
+	for (i = 0x04f4; i <= 0x0500; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x00080000);
+	for (i = 0x050c; i <= 0x0518; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x01012000);
+	for (i = 0x051c; i <= 0x0528; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x000105b8);
+	for (i = 0x052c; i <= 0x0538; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x00080008);
+	for (i = 0x055c; i <= 0x0598; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x07ff0000);
+	nv_wo32(dev, ctx, 0x05a4/4, 0x4b7fffff);
+	nv_wo32(dev, ctx, 0x05fc/4, 0x00000001);
+	nv_wo32(dev, ctx, 0x0604/4, 0x00004000);
+	nv_wo32(dev, ctx, 0x0610/4, 0x00000001);
+	nv_wo32(dev, ctx, 0x0618/4, 0x00040000);
+	nv_wo32(dev, ctx, 0x061c/4, 0x00010000);
+	for (i = 0x1c1c; i <= 0x248c; i += 16) {
+		nv_wo32(dev, ctx, (i + 0)/4, 0x10700ff9);
+		nv_wo32(dev, ctx, (i + 4)/4, 0x0436086c);
+		nv_wo32(dev, ctx, (i + 8)/4, 0x000c001b);
+	}
+	nv_wo32(dev, ctx, 0x281c/4, 0x3f800000);
+	nv_wo32(dev, ctx, 0x2830/4, 0x3f800000);
+	nv_wo32(dev, ctx, 0x285c/4, 0x40000000);
+	nv_wo32(dev, ctx, 0x2860/4, 0x3f800000);
+	nv_wo32(dev, ctx, 0x2864/4, 0x3f000000);
+	nv_wo32(dev, ctx, 0x286c/4, 0x40000000);
+	nv_wo32(dev, ctx, 0x2870/4, 0x3f800000);
+	nv_wo32(dev, ctx, 0x2878/4, 0xbf800000);
+	nv_wo32(dev, ctx, 0x2880/4, 0xbf800000);
+	nv_wo32(dev, ctx, 0x34a4/4, 0x000fe000);
+	nv_wo32(dev, ctx, 0x3530/4, 0x000003f8);
+	nv_wo32(dev, ctx, 0x3540/4, 0x002fe000);
+	for (i = 0x355c; i <= 0x3578; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x001c527c);
+}
+
+static void
+nv25_graph_context_init(struct drm_device *dev, struct nouveau_gpuobj *ctx)
+{
+	int i;
+
+	nv_wo32(dev, ctx, 0x035c/4, 0xffff0000);
+	nv_wo32(dev, ctx, 0x03c0/4, 0x0fff0000);
+	nv_wo32(dev, ctx, 0x03c4/4, 0x0fff0000);
+	nv_wo32(dev, ctx, 0x049c/4, 0x00000101);
+	nv_wo32(dev, ctx, 0x04b0/4, 0x00000111);
+	nv_wo32(dev, ctx, 0x04c8/4, 0x00000080);
+	nv_wo32(dev, ctx, 0x04cc/4, 0xffff0000);
+	nv_wo32(dev, ctx, 0x04d0/4, 0x00000001);
+	nv_wo32(dev, ctx, 0x04e4/4, 0x44400000);
+	nv_wo32(dev, ctx, 0x04fc/4, 0x4b800000);
+	for (i = 0x0510; i <= 0x051c; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x00030303);
+	for (i = 0x0530; i <= 0x053c; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x00080000);
+	for (i = 0x0548; i <= 0x0554; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x01012000);
+	for (i = 0x0558; i <= 0x0564; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x000105b8);
+	for (i = 0x0568; i <= 0x0574; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x00080008);
+	for (i = 0x0598; i <= 0x05d4; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x07ff0000);
+	nv_wo32(dev, ctx, 0x05e0/4, 0x4b7fffff);
+	nv_wo32(dev, ctx, 0x0620/4, 0x00000080);
+	nv_wo32(dev, ctx, 0x0624/4, 0x30201000);
+	nv_wo32(dev, ctx, 0x0628/4, 0x70605040);
+	nv_wo32(dev, ctx, 0x062c/4, 0xb0a09080);
+	nv_wo32(dev, ctx, 0x0630/4, 0xf0e0d0c0);
+	nv_wo32(dev, ctx, 0x0664/4, 0x00000001);
+	nv_wo32(dev, ctx, 0x066c/4, 0x00004000);
+	nv_wo32(dev, ctx, 0x0678/4, 0x00000001);
+	nv_wo32(dev, ctx, 0x0680/4, 0x00040000);
+	nv_wo32(dev, ctx, 0x0684/4, 0x00010000);
+	for (i = 0x1b04; i <= 0x2374; i += 16) {
+		nv_wo32(dev, ctx, (i + 0)/4, 0x10700ff9);
+		nv_wo32(dev, ctx, (i + 4)/4, 0x0436086c);
+		nv_wo32(dev, ctx, (i + 8)/4, 0x000c001b);
+	}
+	nv_wo32(dev, ctx, 0x2704/4, 0x3f800000);
+	nv_wo32(dev, ctx, 0x2718/4, 0x3f800000);
+	nv_wo32(dev, ctx, 0x2744/4, 0x40000000);
+	nv_wo32(dev, ctx, 0x2748/4, 0x3f800000);
+	nv_wo32(dev, ctx, 0x274c/4, 0x3f000000);
+	nv_wo32(dev, ctx, 0x2754/4, 0x40000000);
+	nv_wo32(dev, ctx, 0x2758/4, 0x3f800000);
+	nv_wo32(dev, ctx, 0x2760/4, 0xbf800000);
+	nv_wo32(dev, ctx, 0x2768/4, 0xbf800000);
+	nv_wo32(dev, ctx, 0x308c/4, 0x000fe000);
+	nv_wo32(dev, ctx, 0x3108/4, 0x000003f8);
+	nv_wo32(dev, ctx, 0x3468/4, 0x002fe000);
+	for (i = 0x3484; i <= 0x34a0; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x001c527c);
+}
+
+static void
+nv2a_graph_context_init(struct drm_device *dev, struct nouveau_gpuobj *ctx)
+{
+	int i;
+
+	nv_wo32(dev, ctx, 0x033c/4, 0xffff0000);
+	nv_wo32(dev, ctx, 0x03a0/4, 0x0fff0000);
+	nv_wo32(dev, ctx, 0x03a4/4, 0x0fff0000);
+	nv_wo32(dev, ctx, 0x047c/4, 0x00000101);
+	nv_wo32(dev, ctx, 0x0490/4, 0x00000111);
+	nv_wo32(dev, ctx, 0x04a8/4, 0x44400000);
+	for (i = 0x04d4; i <= 0x04e0; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x00030303);
+	for (i = 0x04f4; i <= 0x0500; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x00080000);
+	for (i = 0x050c; i <= 0x0518; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x01012000);
+	for (i = 0x051c; i <= 0x0528; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x000105b8);
+	for (i = 0x052c; i <= 0x0538; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x00080008);
+	for (i = 0x055c; i <= 0x0598; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x07ff0000);
+	nv_wo32(dev, ctx, 0x05a4/4, 0x4b7fffff);
+	nv_wo32(dev, ctx, 0x05fc/4, 0x00000001);
+	nv_wo32(dev, ctx, 0x0604/4, 0x00004000);
+	nv_wo32(dev, ctx, 0x0610/4, 0x00000001);
+	nv_wo32(dev, ctx, 0x0618/4, 0x00040000);
+	nv_wo32(dev, ctx, 0x061c/4, 0x00010000);
+	for (i = 0x1a9c; i <= 0x22fc; i += 16) { /*XXX: check!! */
+		nv_wo32(dev, ctx, (i + 0)/4, 0x10700ff9);
+		nv_wo32(dev, ctx, (i + 4)/4, 0x0436086c);
+		nv_wo32(dev, ctx, (i + 8)/4, 0x000c001b);
+	}
+	nv_wo32(dev, ctx, 0x269c/4, 0x3f800000);
+	nv_wo32(dev, ctx, 0x26b0/4, 0x3f800000);
+	nv_wo32(dev, ctx, 0x26dc/4, 0x40000000);
+	nv_wo32(dev, ctx, 0x26e0/4, 0x3f800000);
+	nv_wo32(dev, ctx, 0x26e4/4, 0x3f000000);
+	nv_wo32(dev, ctx, 0x26ec/4, 0x40000000);
+	nv_wo32(dev, ctx, 0x26f0/4, 0x3f800000);
+	nv_wo32(dev, ctx, 0x26f8/4, 0xbf800000);
+	nv_wo32(dev, ctx, 0x2700/4, 0xbf800000);
+	nv_wo32(dev, ctx, 0x3024/4, 0x000fe000);
+	nv_wo32(dev, ctx, 0x30a0/4, 0x000003f8);
+	nv_wo32(dev, ctx, 0x33fc/4, 0x002fe000);
+	for (i = 0x341c; i <= 0x3438; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x001c527c);
+}
+
+static void
+nv30_31_graph_context_init(struct drm_device *dev, struct nouveau_gpuobj *ctx)
+{
+	int i;
+
+	nv_wo32(dev, ctx, 0x0410/4, 0x00000101);
+	nv_wo32(dev, ctx, 0x0424/4, 0x00000111);
+	nv_wo32(dev, ctx, 0x0428/4, 0x00000060);
+	nv_wo32(dev, ctx, 0x0444/4, 0x00000080);
+	nv_wo32(dev, ctx, 0x0448/4, 0xffff0000);
+	nv_wo32(dev, ctx, 0x044c/4, 0x00000001);
+	nv_wo32(dev, ctx, 0x0460/4, 0x44400000);
+	nv_wo32(dev, ctx, 0x048c/4, 0xffff0000);
+	for (i = 0x04e0; i < 0x04e8; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x0fff0000);
+	nv_wo32(dev, ctx, 0x04ec/4, 0x00011100);
+	for (i = 0x0508; i < 0x0548; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x07ff0000);
+	nv_wo32(dev, ctx, 0x0550/4, 0x4b7fffff);
+	nv_wo32(dev, ctx, 0x058c/4, 0x00000080);
+	nv_wo32(dev, ctx, 0x0590/4, 0x30201000);
+	nv_wo32(dev, ctx, 0x0594/4, 0x70605040);
+	nv_wo32(dev, ctx, 0x0598/4, 0xb8a89888);
+	nv_wo32(dev, ctx, 0x059c/4, 0xf8e8d8c8);
+	nv_wo32(dev, ctx, 0x05b0/4, 0xb0000000);
+	for (i = 0x0600; i < 0x0640; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x00010588);
+	for (i = 0x0640; i < 0x0680; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x00030303);
+	for (i = 0x06c0; i < 0x0700; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x0008aae4);
+	for (i = 0x0700; i < 0x0740; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x01012000);
+	for (i = 0x0740; i < 0x0780; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x00080008);
+	nv_wo32(dev, ctx, 0x085c/4, 0x00040000);
+	nv_wo32(dev, ctx, 0x0860/4, 0x00010000);
+	for (i = 0x0864; i < 0x0874; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x00040004);
+	for (i = 0x1f18; i <= 0x3088 ; i += 16) {
+		nv_wo32(dev, ctx, i/4 + 0, 0x10700ff9);
+		nv_wo32(dev, ctx, i/4 + 1, 0x0436086c);
+		nv_wo32(dev, ctx, i/4 + 2, 0x000c001b);
+	}
+	for (i = 0x30b8; i < 0x30c8; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x0000ffff);
+	nv_wo32(dev, ctx, 0x344c/4, 0x3f800000);
+	nv_wo32(dev, ctx, 0x3808/4, 0x3f800000);
+	nv_wo32(dev, ctx, 0x381c/4, 0x3f800000);
+	nv_wo32(dev, ctx, 0x3848/4, 0x40000000);
+	nv_wo32(dev, ctx, 0x384c/4, 0x3f800000);
+	nv_wo32(dev, ctx, 0x3850/4, 0x3f000000);
+	nv_wo32(dev, ctx, 0x3858/4, 0x40000000);
+	nv_wo32(dev, ctx, 0x385c/4, 0x3f800000);
+	nv_wo32(dev, ctx, 0x3864/4, 0xbf800000);
+	nv_wo32(dev, ctx, 0x386c/4, 0xbf800000);
+}
+
+static void
+nv34_graph_context_init(struct drm_device *dev, struct nouveau_gpuobj *ctx)
+{
+	int i;
+
+	nv_wo32(dev, ctx, 0x040c/4, 0x01000101);
+	nv_wo32(dev, ctx, 0x0420/4, 0x00000111);
+	nv_wo32(dev, ctx, 0x0424/4, 0x00000060);
+	nv_wo32(dev, ctx, 0x0440/4, 0x00000080);
+	nv_wo32(dev, ctx, 0x0444/4, 0xffff0000);
+	nv_wo32(dev, ctx, 0x0448/4, 0x00000001);
+	nv_wo32(dev, ctx, 0x045c/4, 0x44400000);
+	nv_wo32(dev, ctx, 0x0480/4, 0xffff0000);
+	for (i = 0x04d4; i < 0x04dc; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x0fff0000);
+	nv_wo32(dev, ctx, 0x04e0/4, 0x00011100);
+	for (i = 0x04fc; i < 0x053c; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x07ff0000);
+	nv_wo32(dev, ctx, 0x0544/4, 0x4b7fffff);
+	nv_wo32(dev, ctx, 0x057c/4, 0x00000080);
+	nv_wo32(dev, ctx, 0x0580/4, 0x30201000);
+	nv_wo32(dev, ctx, 0x0584/4, 0x70605040);
+	nv_wo32(dev, ctx, 0x0588/4, 0xb8a89888);
+	nv_wo32(dev, ctx, 0x058c/4, 0xf8e8d8c8);
+	nv_wo32(dev, ctx, 0x05a0/4, 0xb0000000);
+	for (i = 0x05f0; i < 0x0630; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x00010588);
+	for (i = 0x0630; i < 0x0670; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x00030303);
+	for (i = 0x06b0; i < 0x06f0; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x0008aae4);
+	for (i = 0x06f0; i < 0x0730; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x01012000);
+	for (i = 0x0730; i < 0x0770; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x00080008);
+	nv_wo32(dev, ctx, 0x0850/4, 0x00040000);
+	nv_wo32(dev, ctx, 0x0854/4, 0x00010000);
+	for (i = 0x0858; i < 0x0868; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x00040004);
+	for (i = 0x15ac; i <= 0x271c ; i += 16) {
+		nv_wo32(dev, ctx, i/4 + 0, 0x10700ff9);
+		nv_wo32(dev, ctx, i/4 + 1, 0x0436086c);
+		nv_wo32(dev, ctx, i/4 + 2, 0x000c001b);
+	}
+	for (i = 0x274c; i < 0x275c; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x0000ffff);
+	nv_wo32(dev, ctx, 0x2ae0/4, 0x3f800000);
+	nv_wo32(dev, ctx, 0x2e9c/4, 0x3f800000);
+	nv_wo32(dev, ctx, 0x2eb0/4, 0x3f800000);
+	nv_wo32(dev, ctx, 0x2edc/4, 0x40000000);
+	nv_wo32(dev, ctx, 0x2ee0/4, 0x3f800000);
+	nv_wo32(dev, ctx, 0x2ee4/4, 0x3f000000);
+	nv_wo32(dev, ctx, 0x2eec/4, 0x40000000);
+	nv_wo32(dev, ctx, 0x2ef0/4, 0x3f800000);
+	nv_wo32(dev, ctx, 0x2ef8/4, 0xbf800000);
+	nv_wo32(dev, ctx, 0x2f00/4, 0xbf800000);
+}
+
+static void
+nv35_36_graph_context_init(struct drm_device *dev, struct nouveau_gpuobj *ctx)
+{
+	int i;
+
+	nv_wo32(dev, ctx, 0x040c/4, 0x00000101);
+	nv_wo32(dev, ctx, 0x0420/4, 0x00000111);
+	nv_wo32(dev, ctx, 0x0424/4, 0x00000060);
+	nv_wo32(dev, ctx, 0x0440/4, 0x00000080);
+	nv_wo32(dev, ctx, 0x0444/4, 0xffff0000);
+	nv_wo32(dev, ctx, 0x0448/4, 0x00000001);
+	nv_wo32(dev, ctx, 0x045c/4, 0x44400000);
+	nv_wo32(dev, ctx, 0x0488/4, 0xffff0000);
+	for (i = 0x04dc; i < 0x04e4; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x0fff0000);
+	nv_wo32(dev, ctx, 0x04e8/4, 0x00011100);
+	for (i = 0x0504; i < 0x0544; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x07ff0000);
+	nv_wo32(dev, ctx, 0x054c/4, 0x4b7fffff);
+	nv_wo32(dev, ctx, 0x0588/4, 0x00000080);
+	nv_wo32(dev, ctx, 0x058c/4, 0x30201000);
+	nv_wo32(dev, ctx, 0x0590/4, 0x70605040);
+	nv_wo32(dev, ctx, 0x0594/4, 0xb8a89888);
+	nv_wo32(dev, ctx, 0x0598/4, 0xf8e8d8c8);
+	nv_wo32(dev, ctx, 0x05ac/4, 0xb0000000);
+	for (i = 0x0604; i < 0x0644; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x00010588);
+	for (i = 0x0644; i < 0x0684; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x00030303);
+	for (i = 0x06c4; i < 0x0704; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x0008aae4);
+	for (i = 0x0704; i < 0x0744; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x01012000);
+	for (i = 0x0744; i < 0x0784; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x00080008);
+	nv_wo32(dev, ctx, 0x0860/4, 0x00040000);
+	nv_wo32(dev, ctx, 0x0864/4, 0x00010000);
+	for (i = 0x0868; i < 0x0878; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x00040004);
+	for (i = 0x1f1c; i <= 0x308c ; i += 16) {
+		nv_wo32(dev, ctx, i/4 + 0, 0x10700ff9);
+		nv_wo32(dev, ctx, i/4 + 1, 0x0436086c);
+		nv_wo32(dev, ctx, i/4 + 2, 0x000c001b);
+	}
+	for (i = 0x30bc; i < 0x30cc; i += 4)
+		nv_wo32(dev, ctx, i/4, 0x0000ffff);
+	nv_wo32(dev, ctx, 0x3450/4, 0x3f800000);
+	nv_wo32(dev, ctx, 0x380c/4, 0x3f800000);
+	nv_wo32(dev, ctx, 0x3820/4, 0x3f800000);
+	nv_wo32(dev, ctx, 0x384c/4, 0x40000000);
+	nv_wo32(dev, ctx, 0x3850/4, 0x3f800000);
+	nv_wo32(dev, ctx, 0x3854/4, 0x3f000000);
+	nv_wo32(dev, ctx, 0x385c/4, 0x40000000);
+	nv_wo32(dev, ctx, 0x3860/4, 0x3f800000);
+	nv_wo32(dev, ctx, 0x3868/4, 0xbf800000);
+	nv_wo32(dev, ctx, 0x3870/4, 0xbf800000);
+}
+
+int
+nv20_graph_create_context(struct nouveau_channel *chan)
+{
+	struct drm_device *dev = chan->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	void (*ctx_init)(struct drm_device *, struct nouveau_gpuobj *);
+	unsigned int ctx_size;
+	unsigned int idoffs = 0x28/4;
+	int ret;
+
+	switch (dev_priv->chipset) {
+	case 0x20:
+		ctx_size = NV20_GRCTX_SIZE;
+		ctx_init = nv20_graph_context_init;
+		idoffs = 0;
+		break;
+	case 0x25:
+	case 0x28:
+		ctx_size = NV25_GRCTX_SIZE;
+		ctx_init = nv25_graph_context_init;
+		break;
+	case 0x2a:
+		ctx_size = NV2A_GRCTX_SIZE;
+		ctx_init = nv2a_graph_context_init;
+		idoffs = 0;
+		break;
+	case 0x30:
+	case 0x31:
+		ctx_size = NV30_31_GRCTX_SIZE;
+		ctx_init = nv30_31_graph_context_init;
+		break;
+	case 0x34:
+		ctx_size = NV34_GRCTX_SIZE;
+		ctx_init = nv34_graph_context_init;
+		break;
+	case 0x35:
+	case 0x36:
+		ctx_size = NV35_36_GRCTX_SIZE;
+		ctx_init = nv35_36_graph_context_init;
+		break;
+	default:
+		ctx_size = 0;
+		ctx_init = nv35_36_graph_context_init;
+		NV_ERROR(dev, "Please contact the devs if you want your NV%x"
+			      " card to work\n", dev_priv->chipset);
+		return -ENOSYS;
+		break;
+	}
+
+	ret = nouveau_gpuobj_new_ref(dev, chan, NULL, 0, ctx_size, 16,
+					  NVOBJ_FLAG_ZERO_ALLOC,
+					  &chan->ramin_grctx);
+	if (ret)
+		return ret;
+
+	/* Initialise default context values */
+	dev_priv->engine.instmem.prepare_access(dev, true);
+	ctx_init(dev, chan->ramin_grctx->gpuobj);
+
+	/* nv20: nv_wo32(dev, chan->ramin_grctx->gpuobj, 10, chan->id<<24); */
+	nv_wo32(dev, chan->ramin_grctx->gpuobj, idoffs,
+					(chan->id << 24) | 0x1); /* CTX_USER */
+
+	nv_wo32(dev, dev_priv->ctx_table->gpuobj, chan->id,
+			chan->ramin_grctx->instance >> 4);
+
+	dev_priv->engine.instmem.finish_access(dev);
+	return 0;
+}
+
+void
+nv20_graph_destroy_context(struct nouveau_channel *chan)
+{
+	struct drm_device *dev = chan->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	if (chan->ramin_grctx)
+		nouveau_gpuobj_ref_del(dev, &chan->ramin_grctx);
+
+	dev_priv->engine.instmem.prepare_access(dev, true);
+	nv_wo32(dev, dev_priv->ctx_table->gpuobj, chan->id, 0);
+	dev_priv->engine.instmem.finish_access(dev);
+}
+
+int
+nv20_graph_load_context(struct nouveau_channel *chan)
+{
+	struct drm_device *dev = chan->dev;
+	uint32_t inst;
+
+	if (!chan->ramin_grctx)
+		return -EINVAL;
+	inst = chan->ramin_grctx->instance >> 4;
+
+	nv_wr32(dev, NV20_PGRAPH_CHANNEL_CTX_POINTER, inst);
+	nv_wr32(dev, NV20_PGRAPH_CHANNEL_CTX_XFER,
+		     NV20_PGRAPH_CHANNEL_CTX_XFER_LOAD);
+	nv_wr32(dev, NV10_PGRAPH_CTX_CONTROL, 0x10010100);
+
+	nouveau_wait_for_idle(dev);
+	return 0;
+}
+
+int
+nv20_graph_unload_context(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
+	struct nouveau_fifo_engine *pfifo = &dev_priv->engine.fifo;
+	struct nouveau_channel *chan;
+	uint32_t inst, tmp;
+
+	chan = pgraph->channel(dev);
+	if (!chan)
+		return 0;
+	inst = chan->ramin_grctx->instance >> 4;
+
+	nv_wr32(dev, NV20_PGRAPH_CHANNEL_CTX_POINTER, inst);
+	nv_wr32(dev, NV20_PGRAPH_CHANNEL_CTX_XFER,
+		     NV20_PGRAPH_CHANNEL_CTX_XFER_SAVE);
+
+	nouveau_wait_for_idle(dev);
+
+	nv_wr32(dev, NV10_PGRAPH_CTX_CONTROL, 0x10000000);
+	tmp  = nv_rd32(dev, NV10_PGRAPH_CTX_USER) & 0x00ffffff;
+	tmp |= (pfifo->channels - 1) << 24;
+	nv_wr32(dev, NV10_PGRAPH_CTX_USER, tmp);
+	return 0;
+}
+
+static void
+nv20_graph_rdi(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	int i, writecount = 32;
+	uint32_t rdi_index = 0x2c80000;
+
+	if (dev_priv->chipset == 0x20) {
+		rdi_index = 0x3d0000;
+		writecount = 15;
+	}
+
+	nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, rdi_index);
+	for (i = 0; i < writecount; i++)
+		nv_wr32(dev, NV10_PGRAPH_RDI_DATA, 0);
+
+	nouveau_wait_for_idle(dev);
+}
+
+void
+nv20_graph_set_region_tiling(struct drm_device *dev, int i, uint32_t addr,
+			     uint32_t size, uint32_t pitch)
+{
+	uint32_t limit = max(1u, addr + size) - 1;
+
+	if (pitch)
+		addr |= 1;
+
+	nv_wr32(dev, NV20_PGRAPH_TLIMIT(i), limit);
+	nv_wr32(dev, NV20_PGRAPH_TSIZE(i), pitch);
+	nv_wr32(dev, NV20_PGRAPH_TILE(i), addr);
+
+	nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00EA0030 + 4 * i);
+	nv_wr32(dev, NV10_PGRAPH_RDI_DATA, limit);
+	nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00EA0050 + 4 * i);
+	nv_wr32(dev, NV10_PGRAPH_RDI_DATA, pitch);
+	nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00EA0010 + 4 * i);
+	nv_wr32(dev, NV10_PGRAPH_RDI_DATA, addr);
+}
+
+int
+nv20_graph_init(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv =
+		(struct drm_nouveau_private *)dev->dev_private;
+	uint32_t tmp, vramsz;
+	int ret, i;
+
+	nv_wr32(dev, NV03_PMC_ENABLE,
+		nv_rd32(dev, NV03_PMC_ENABLE) & ~NV_PMC_ENABLE_PGRAPH);
+	nv_wr32(dev, NV03_PMC_ENABLE,
+		nv_rd32(dev, NV03_PMC_ENABLE) |  NV_PMC_ENABLE_PGRAPH);
+
+	if (!dev_priv->ctx_table) {
+		/* Create Context Pointer Table */
+		dev_priv->ctx_table_size = 32 * 4;
+		ret = nouveau_gpuobj_new_ref(dev, NULL, NULL, 0,
+						  dev_priv->ctx_table_size, 16,
+						  NVOBJ_FLAG_ZERO_ALLOC,
+						  &dev_priv->ctx_table);
+		if (ret)
+			return ret;
+	}
+
+	nv_wr32(dev, NV20_PGRAPH_CHANNEL_CTX_TABLE,
+		 dev_priv->ctx_table->instance >> 4);
+
+	nv20_graph_rdi(dev);
+
+	nv_wr32(dev, NV03_PGRAPH_INTR   , 0xFFFFFFFF);
+	nv_wr32(dev, NV03_PGRAPH_INTR_EN, 0xFFFFFFFF);
+
+	nv_wr32(dev, NV04_PGRAPH_DEBUG_0, 0xFFFFFFFF);
+	nv_wr32(dev, NV04_PGRAPH_DEBUG_0, 0x00000000);
+	nv_wr32(dev, NV04_PGRAPH_DEBUG_1, 0x00118700);
+	nv_wr32(dev, NV04_PGRAPH_DEBUG_3, 0xF3CE0475); /* 0x4 = auto ctx switch */
+	nv_wr32(dev, NV10_PGRAPH_DEBUG_4, 0x00000000);
+	nv_wr32(dev, 0x40009C           , 0x00000040);
+
+	if (dev_priv->chipset >= 0x25) {
+		nv_wr32(dev, 0x400890, 0x00080000);
+		nv_wr32(dev, 0x400610, 0x304B1FB6);
+		nv_wr32(dev, 0x400B80, 0x18B82880);
+		nv_wr32(dev, 0x400B84, 0x44000000);
+		nv_wr32(dev, 0x400098, 0x40000080);
+		nv_wr32(dev, 0x400B88, 0x000000ff);
+	} else {
+		nv_wr32(dev, 0x400880, 0x00080000); /* 0x0008c7df */
+		nv_wr32(dev, 0x400094, 0x00000005);
+		nv_wr32(dev, 0x400B80, 0x45CAA208); /* 0x45eae20e */
+		nv_wr32(dev, 0x400B84, 0x24000000);
+		nv_wr32(dev, 0x400098, 0x00000040);
+		nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00E00038);
+		nv_wr32(dev, NV10_PGRAPH_RDI_DATA , 0x00000030);
+		nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00E10038);
+		nv_wr32(dev, NV10_PGRAPH_RDI_DATA , 0x00000030);
+	}
+
+	/* Turn all the tiling regions off. */
+	for (i = 0; i < NV10_PFB_TILE__SIZE; i++)
+		nv20_graph_set_region_tiling(dev, i, 0, 0, 0);
+
+	for (i = 0; i < 8; i++) {
+		nv_wr32(dev, 0x400980 + i * 4, nv_rd32(dev, 0x100300 + i * 4));
+		nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00EA0090 + i * 4);
+		nv_wr32(dev, NV10_PGRAPH_RDI_DATA,
+					nv_rd32(dev, 0x100300 + i * 4));
+	}
+	nv_wr32(dev, 0x4009a0, nv_rd32(dev, 0x100324));
+	nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00EA000C);
+	nv_wr32(dev, NV10_PGRAPH_RDI_DATA, nv_rd32(dev, 0x100324));
+
+	nv_wr32(dev, NV10_PGRAPH_CTX_CONTROL, 0x10000100);
+	nv_wr32(dev, NV10_PGRAPH_STATE      , 0xFFFFFFFF);
+
+	tmp = nv_rd32(dev, NV10_PGRAPH_SURFACE) & 0x0007ff00;
+	nv_wr32(dev, NV10_PGRAPH_SURFACE, tmp);
+	tmp = nv_rd32(dev, NV10_PGRAPH_SURFACE) | 0x00020100;
+	nv_wr32(dev, NV10_PGRAPH_SURFACE, tmp);
+
+	/* begin RAM config */
+	vramsz = drm_get_resource_len(dev, 0) - 1;
+	nv_wr32(dev, 0x4009A4, nv_rd32(dev, NV04_PFB_CFG0));
+	nv_wr32(dev, 0x4009A8, nv_rd32(dev, NV04_PFB_CFG1));
+	nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00EA0000);
+	nv_wr32(dev, NV10_PGRAPH_RDI_DATA , nv_rd32(dev, NV04_PFB_CFG0));
+	nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00EA0004);
+	nv_wr32(dev, NV10_PGRAPH_RDI_DATA , nv_rd32(dev, NV04_PFB_CFG1));
+	nv_wr32(dev, 0x400820, 0);
+	nv_wr32(dev, 0x400824, 0);
+	nv_wr32(dev, 0x400864, vramsz - 1);
+	nv_wr32(dev, 0x400868, vramsz - 1);
+
+	/* interesting.. the below overwrites some of the tile setup above.. */
+	nv_wr32(dev, 0x400B20, 0x00000000);
+	nv_wr32(dev, 0x400B04, 0xFFFFFFFF);
+
+	nv_wr32(dev, NV03_PGRAPH_ABS_UCLIP_XMIN, 0);
+	nv_wr32(dev, NV03_PGRAPH_ABS_UCLIP_YMIN, 0);
+	nv_wr32(dev, NV03_PGRAPH_ABS_UCLIP_XMAX, 0x7fff);
+	nv_wr32(dev, NV03_PGRAPH_ABS_UCLIP_YMAX, 0x7fff);
+
+	return 0;
+}
+
+void
+nv20_graph_takedown(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	nouveau_gpuobj_ref_del(dev, &dev_priv->ctx_table);
+}
+
+int
+nv30_graph_init(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	int ret, i;
+
+	nv_wr32(dev, NV03_PMC_ENABLE,
+		nv_rd32(dev, NV03_PMC_ENABLE) & ~NV_PMC_ENABLE_PGRAPH);
+	nv_wr32(dev, NV03_PMC_ENABLE,
+		nv_rd32(dev, NV03_PMC_ENABLE) |  NV_PMC_ENABLE_PGRAPH);
+
+	if (!dev_priv->ctx_table) {
+		/* Create Context Pointer Table */
+		dev_priv->ctx_table_size = 32 * 4;
+		ret = nouveau_gpuobj_new_ref(dev, NULL, NULL, 0,
+						  dev_priv->ctx_table_size, 16,
+						  NVOBJ_FLAG_ZERO_ALLOC,
+						  &dev_priv->ctx_table);
+		if (ret)
+			return ret;
+	}
+
+	nv_wr32(dev, NV20_PGRAPH_CHANNEL_CTX_TABLE,
+			dev_priv->ctx_table->instance >> 4);
+
+	nv_wr32(dev, NV03_PGRAPH_INTR   , 0xFFFFFFFF);
+	nv_wr32(dev, NV03_PGRAPH_INTR_EN, 0xFFFFFFFF);
+
+	nv_wr32(dev, NV04_PGRAPH_DEBUG_0, 0xFFFFFFFF);
+	nv_wr32(dev, NV04_PGRAPH_DEBUG_0, 0x00000000);
+	nv_wr32(dev, NV04_PGRAPH_DEBUG_1, 0x401287c0);
+	nv_wr32(dev, 0x400890, 0x01b463ff);
+	nv_wr32(dev, NV04_PGRAPH_DEBUG_3, 0xf2de0475);
+	nv_wr32(dev, NV10_PGRAPH_DEBUG_4, 0x00008000);
+	nv_wr32(dev, NV04_PGRAPH_LIMIT_VIOL_PIX, 0xf04bdff6);
+	nv_wr32(dev, 0x400B80, 0x1003d888);
+	nv_wr32(dev, 0x400B84, 0x0c000000);
+	nv_wr32(dev, 0x400098, 0x00000000);
+	nv_wr32(dev, 0x40009C, 0x0005ad00);
+	nv_wr32(dev, 0x400B88, 0x62ff00ff); /* suspiciously like PGRAPH_DEBUG_2 */
+	nv_wr32(dev, 0x4000a0, 0x00000000);
+	nv_wr32(dev, 0x4000a4, 0x00000008);
+	nv_wr32(dev, 0x4008a8, 0xb784a400);
+	nv_wr32(dev, 0x400ba0, 0x002f8685);
+	nv_wr32(dev, 0x400ba4, 0x00231f3f);
+	nv_wr32(dev, 0x4008a4, 0x40000020);
+
+	if (dev_priv->chipset == 0x34) {
+		nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00EA0004);
+		nv_wr32(dev, NV10_PGRAPH_RDI_DATA , 0x00200201);
+		nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00EA0008);
+		nv_wr32(dev, NV10_PGRAPH_RDI_DATA , 0x00000008);
+		nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00EA0000);
+		nv_wr32(dev, NV10_PGRAPH_RDI_DATA , 0x00000032);
+		nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00E00004);
+		nv_wr32(dev, NV10_PGRAPH_RDI_DATA , 0x00000002);
+	}
+
+	nv_wr32(dev, 0x4000c0, 0x00000016);
+
+	/* Turn all the tiling regions off. */
+	for (i = 0; i < NV10_PFB_TILE__SIZE; i++)
+		nv20_graph_set_region_tiling(dev, i, 0, 0, 0);
+
+	nv_wr32(dev, NV10_PGRAPH_CTX_CONTROL, 0x10000100);
+	nv_wr32(dev, NV10_PGRAPH_STATE      , 0xFFFFFFFF);
+	nv_wr32(dev, 0x0040075c             , 0x00000001);
+
+	/* begin RAM config */
+	/* vramsz = drm_get_resource_len(dev, 0) - 1; */
+	nv_wr32(dev, 0x4009A4, nv_rd32(dev, NV04_PFB_CFG0));
+	nv_wr32(dev, 0x4009A8, nv_rd32(dev, NV04_PFB_CFG1));
+	if (dev_priv->chipset != 0x34) {
+		nv_wr32(dev, 0x400750, 0x00EA0000);
+		nv_wr32(dev, 0x400754, nv_rd32(dev, NV04_PFB_CFG0));
+		nv_wr32(dev, 0x400750, 0x00EA0004);
+		nv_wr32(dev, 0x400754, nv_rd32(dev, NV04_PFB_CFG1));
+	}
+
+	return 0;
+}
+
+struct nouveau_pgraph_object_class nv20_graph_grclass[] = {
+	{ 0x0030, false, NULL }, /* null */
+	{ 0x0039, false, NULL }, /* m2mf */
+	{ 0x004a, false, NULL }, /* gdirect */
+	{ 0x009f, false, NULL }, /* imageblit (nv12) */
+	{ 0x008a, false, NULL }, /* ifc */
+	{ 0x0089, false, NULL }, /* sifm */
+	{ 0x0062, false, NULL }, /* surf2d */
+	{ 0x0043, false, NULL }, /* rop */
+	{ 0x0012, false, NULL }, /* beta1 */
+	{ 0x0072, false, NULL }, /* beta4 */
+	{ 0x0019, false, NULL }, /* cliprect */
+	{ 0x0044, false, NULL }, /* pattern */
+	{ 0x009e, false, NULL }, /* swzsurf */
+	{ 0x0096, false, NULL }, /* celcius */
+	{ 0x0097, false, NULL }, /* kelvin (nv20) */
+	{ 0x0597, false, NULL }, /* kelvin (nv25) */
+	{}
+};
+
+struct nouveau_pgraph_object_class nv30_graph_grclass[] = {
+	{ 0x0030, false, NULL }, /* null */
+	{ 0x0039, false, NULL }, /* m2mf */
+	{ 0x004a, false, NULL }, /* gdirect */
+	{ 0x009f, false, NULL }, /* imageblit (nv12) */
+	{ 0x008a, false, NULL }, /* ifc */
+	{ 0x038a, false, NULL }, /* ifc (nv30) */
+	{ 0x0089, false, NULL }, /* sifm */
+	{ 0x0389, false, NULL }, /* sifm (nv30) */
+	{ 0x0062, false, NULL }, /* surf2d */
+	{ 0x0362, false, NULL }, /* surf2d (nv30) */
+	{ 0x0043, false, NULL }, /* rop */
+	{ 0x0012, false, NULL }, /* beta1 */
+	{ 0x0072, false, NULL }, /* beta4 */
+	{ 0x0019, false, NULL }, /* cliprect */
+	{ 0x0044, false, NULL }, /* pattern */
+	{ 0x039e, false, NULL }, /* swzsurf */
+	{ 0x0397, false, NULL }, /* rankine (nv30) */
+	{ 0x0497, false, NULL }, /* rankine (nv35) */
+	{ 0x0697, false, NULL }, /* rankine (nv34) */
+	{}
+};
+
diff --git a/drivers/gpu/drm/nouveau/nv40_fb.c b/drivers/gpu/drm/nouveau/nv40_fb.c
new file mode 100644
index 000000000000..3cd07d8d5bd7
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv40_fb.c
@@ -0,0 +1,75 @@
+#include "drmP.h"
+#include "drm.h"
+#include "nouveau_drv.h"
+#include "nouveau_drm.h"
+
+void
+nv40_fb_set_region_tiling(struct drm_device *dev, int i, uint32_t addr,
+			  uint32_t size, uint32_t pitch)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t limit = max(1u, addr + size) - 1;
+
+	if (pitch)
+		addr |= 1;
+
+	switch (dev_priv->chipset) {
+	case 0x40:
+		nv_wr32(dev, NV10_PFB_TLIMIT(i), limit);
+		nv_wr32(dev, NV10_PFB_TSIZE(i), pitch);
+		nv_wr32(dev, NV10_PFB_TILE(i), addr);
+		break;
+
+	default:
+		nv_wr32(dev, NV40_PFB_TLIMIT(i), limit);
+		nv_wr32(dev, NV40_PFB_TSIZE(i), pitch);
+		nv_wr32(dev, NV40_PFB_TILE(i), addr);
+		break;
+	}
+}
+
+int
+nv40_fb_init(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_fb_engine *pfb = &dev_priv->engine.fb;
+	uint32_t tmp;
+	int i;
+
+	/* This is strictly a NV4x register (don't know about NV5x). */
+	/* The blob sets these to all kinds of values, and they mess up our setup. */
+	/* I got value 0x52802 instead. For some cards the blob even sets it back to 0x1. */
+	/* Note: the blob doesn't read this value, so i'm pretty sure this is safe for all cards. */
+	/* Any idea what this is? */
+	nv_wr32(dev, NV40_PFB_UNK_800, 0x1);
+
+	switch (dev_priv->chipset) {
+	case 0x40:
+	case 0x45:
+		tmp = nv_rd32(dev, NV10_PFB_CLOSE_PAGE2);
+		nv_wr32(dev, NV10_PFB_CLOSE_PAGE2, tmp & ~(1 << 15));
+		pfb->num_tiles = NV10_PFB_TILE__SIZE;
+		break;
+	case 0x46: /* G72 */
+	case 0x47: /* G70 */
+	case 0x49: /* G71 */
+	case 0x4b: /* G73 */
+	case 0x4c: /* C51 (G7X version) */
+		pfb->num_tiles = NV40_PFB_TILE__SIZE_1;
+		break;
+	default:
+		pfb->num_tiles = NV40_PFB_TILE__SIZE_0;
+		break;
+	}
+
+	/* Turn all the tiling regions off. */
+	for (i = 0; i < pfb->num_tiles; i++)
+		pfb->set_region_tiling(dev, i, 0, 0, 0);
+
+	return 0;
+}
+
+void
+nv40_fb_takedown(struct drm_device *dev)
+{
+}
diff --git a/drivers/gpu/drm/nouveau/nv40_fifo.c b/drivers/gpu/drm/nouveau/nv40_fifo.c
new file mode 100644
index 000000000000..6b2ef4a9fce1
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv40_fifo.c
@@ -0,0 +1,319 @@
+/*
+ * Copyright (C) 2007 Ben Skeggs.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "drmP.h"
+#include "nouveau_drv.h"
+#include "nouveau_drm.h"
+
+#define NV40_RAMFC(c) (dev_priv->ramfc_offset + ((c) * NV40_RAMFC__SIZE))
+#define NV40_RAMFC__SIZE 128
+
+int
+nv40_fifo_create_context(struct nouveau_channel *chan)
+{
+	struct drm_device *dev = chan->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t fc = NV40_RAMFC(chan->id);
+	unsigned long flags;
+	int ret;
+
+	ret = nouveau_gpuobj_new_fake(dev, NV40_RAMFC(chan->id), ~0,
+				      NV40_RAMFC__SIZE, NVOBJ_FLAG_ZERO_ALLOC |
+				      NVOBJ_FLAG_ZERO_FREE, NULL, &chan->ramfc);
+	if (ret)
+		return ret;
+
+	spin_lock_irqsave(&dev_priv->context_switch_lock, flags);
+
+	dev_priv->engine.instmem.prepare_access(dev, true);
+	nv_wi32(dev, fc +  0, chan->pushbuf_base);
+	nv_wi32(dev, fc +  4, chan->pushbuf_base);
+	nv_wi32(dev, fc + 12, chan->pushbuf->instance >> 4);
+	nv_wi32(dev, fc + 24, NV_PFIFO_CACHE1_DMA_FETCH_TRIG_128_BYTES |
+			      NV_PFIFO_CACHE1_DMA_FETCH_SIZE_128_BYTES |
+			      NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_8 |
+#ifdef __BIG_ENDIAN
+			      NV_PFIFO_CACHE1_BIG_ENDIAN |
+#endif
+			      0x30000000 /* no idea.. */);
+	nv_wi32(dev, fc + 56, chan->ramin_grctx->instance >> 4);
+	nv_wi32(dev, fc + 60, 0x0001FFFF);
+	dev_priv->engine.instmem.finish_access(dev);
+
+	/* enable the fifo dma operation */
+	nv_wr32(dev, NV04_PFIFO_MODE,
+		nv_rd32(dev, NV04_PFIFO_MODE) | (1 << chan->id));
+
+	spin_unlock_irqrestore(&dev_priv->context_switch_lock, flags);
+	return 0;
+}
+
+void
+nv40_fifo_destroy_context(struct nouveau_channel *chan)
+{
+	struct drm_device *dev = chan->dev;
+
+	nv_wr32(dev, NV04_PFIFO_MODE,
+		nv_rd32(dev, NV04_PFIFO_MODE) & ~(1 << chan->id));
+
+	if (chan->ramfc)
+		nouveau_gpuobj_ref_del(dev, &chan->ramfc);
+}
+
+static void
+nv40_fifo_do_load_context(struct drm_device *dev, int chid)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t fc = NV40_RAMFC(chid), tmp, tmp2;
+
+	dev_priv->engine.instmem.prepare_access(dev, false);
+
+	nv_wr32(dev, NV04_PFIFO_CACHE1_DMA_PUT, nv_ri32(dev, fc + 0));
+	nv_wr32(dev, NV04_PFIFO_CACHE1_DMA_GET, nv_ri32(dev, fc + 4));
+	nv_wr32(dev, NV10_PFIFO_CACHE1_REF_CNT, nv_ri32(dev, fc + 8));
+	nv_wr32(dev, NV04_PFIFO_CACHE1_DMA_INSTANCE, nv_ri32(dev, fc + 12));
+	nv_wr32(dev, NV04_PFIFO_CACHE1_DMA_DCOUNT, nv_ri32(dev, fc + 16));
+	nv_wr32(dev, NV04_PFIFO_CACHE1_DMA_STATE, nv_ri32(dev, fc + 20));
+
+	/* No idea what 0x2058 is.. */
+	tmp   = nv_ri32(dev, fc + 24);
+	tmp2  = nv_rd32(dev, 0x2058) & 0xFFF;
+	tmp2 |= (tmp & 0x30000000);
+	nv_wr32(dev, 0x2058, tmp2);
+	tmp  &= ~0x30000000;
+	nv_wr32(dev, NV04_PFIFO_CACHE1_DMA_FETCH, tmp);
+
+	nv_wr32(dev, NV04_PFIFO_CACHE1_ENGINE, nv_ri32(dev, fc + 28));
+	nv_wr32(dev, NV04_PFIFO_CACHE1_PULL1, nv_ri32(dev, fc + 32));
+	nv_wr32(dev, NV10_PFIFO_CACHE1_ACQUIRE_VALUE, nv_ri32(dev, fc + 36));
+	tmp = nv_ri32(dev, fc + 40);
+	nv_wr32(dev, NV10_PFIFO_CACHE1_ACQUIRE_TIMESTAMP, tmp);
+	nv_wr32(dev, NV10_PFIFO_CACHE1_ACQUIRE_TIMEOUT, nv_ri32(dev, fc + 44));
+	nv_wr32(dev, NV10_PFIFO_CACHE1_SEMAPHORE, nv_ri32(dev, fc + 48));
+	nv_wr32(dev, NV10_PFIFO_CACHE1_DMA_SUBROUTINE, nv_ri32(dev, fc + 52));
+	nv_wr32(dev, NV40_PFIFO_GRCTX_INSTANCE, nv_ri32(dev, fc + 56));
+
+	/* Don't clobber the TIMEOUT_ENABLED flag when restoring from RAMFC */
+	tmp  = nv_rd32(dev, NV04_PFIFO_DMA_TIMESLICE) & ~0x1FFFF;
+	tmp |= nv_ri32(dev, fc + 60) & 0x1FFFF;
+	nv_wr32(dev, NV04_PFIFO_DMA_TIMESLICE, tmp);
+
+	nv_wr32(dev, 0x32e4, nv_ri32(dev, fc + 64));
+	/* NVIDIA does this next line twice... */
+	nv_wr32(dev, 0x32e8, nv_ri32(dev, fc + 68));
+	nv_wr32(dev, 0x2088, nv_ri32(dev, fc + 76));
+	nv_wr32(dev, 0x3300, nv_ri32(dev, fc + 80));
+
+	dev_priv->engine.instmem.finish_access(dev);
+
+	nv_wr32(dev, NV03_PFIFO_CACHE1_GET, 0);
+	nv_wr32(dev, NV03_PFIFO_CACHE1_PUT, 0);
+}
+
+int
+nv40_fifo_load_context(struct nouveau_channel *chan)
+{
+	struct drm_device *dev = chan->dev;
+	uint32_t tmp;
+
+	nv40_fifo_do_load_context(dev, chan->id);
+
+	/* Set channel active, and in DMA mode */
+	nv_wr32(dev, NV03_PFIFO_CACHE1_PUSH1,
+		     NV40_PFIFO_CACHE1_PUSH1_DMA | chan->id);
+	nv_wr32(dev, NV04_PFIFO_CACHE1_DMA_PUSH, 1);
+
+	/* Reset DMA_CTL_AT_INFO to INVALID */
+	tmp = nv_rd32(dev, NV04_PFIFO_CACHE1_DMA_CTL) & ~(1 << 31);
+	nv_wr32(dev, NV04_PFIFO_CACHE1_DMA_CTL, tmp);
+
+	return 0;
+}
+
+int
+nv40_fifo_unload_context(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_fifo_engine *pfifo = &dev_priv->engine.fifo;
+	uint32_t fc, tmp;
+	int chid;
+
+	chid = pfifo->channel_id(dev);
+	if (chid < 0 || chid >= dev_priv->engine.fifo.channels)
+		return 0;
+	fc = NV40_RAMFC(chid);
+
+	dev_priv->engine.instmem.prepare_access(dev, true);
+	nv_wi32(dev, fc + 0, nv_rd32(dev, NV04_PFIFO_CACHE1_DMA_PUT));
+	nv_wi32(dev, fc + 4, nv_rd32(dev, NV04_PFIFO_CACHE1_DMA_GET));
+	nv_wi32(dev, fc + 8, nv_rd32(dev, NV10_PFIFO_CACHE1_REF_CNT));
+	nv_wi32(dev, fc + 12, nv_rd32(dev, NV04_PFIFO_CACHE1_DMA_INSTANCE));
+	nv_wi32(dev, fc + 16, nv_rd32(dev, NV04_PFIFO_CACHE1_DMA_DCOUNT));
+	nv_wi32(dev, fc + 20, nv_rd32(dev, NV04_PFIFO_CACHE1_DMA_STATE));
+	tmp  = nv_rd32(dev, NV04_PFIFO_CACHE1_DMA_FETCH);
+	tmp |= nv_rd32(dev, 0x2058) & 0x30000000;
+	nv_wi32(dev, fc + 24, tmp);
+	nv_wi32(dev, fc + 28, nv_rd32(dev, NV04_PFIFO_CACHE1_ENGINE));
+	nv_wi32(dev, fc + 32, nv_rd32(dev, NV04_PFIFO_CACHE1_PULL1));
+	nv_wi32(dev, fc + 36, nv_rd32(dev, NV10_PFIFO_CACHE1_ACQUIRE_VALUE));
+	tmp = nv_rd32(dev, NV10_PFIFO_CACHE1_ACQUIRE_TIMESTAMP);
+	nv_wi32(dev, fc + 40, tmp);
+	nv_wi32(dev, fc + 44, nv_rd32(dev, NV10_PFIFO_CACHE1_ACQUIRE_TIMEOUT));
+	nv_wi32(dev, fc + 48, nv_rd32(dev, NV10_PFIFO_CACHE1_SEMAPHORE));
+	/* NVIDIA read 0x3228 first, then write DMA_GET here.. maybe something
+	 * more involved depending on the value of 0x3228?
+	 */
+	nv_wi32(dev, fc + 52, nv_rd32(dev, NV04_PFIFO_CACHE1_DMA_GET));
+	nv_wi32(dev, fc + 56, nv_rd32(dev, NV40_PFIFO_GRCTX_INSTANCE));
+	nv_wi32(dev, fc + 60, nv_rd32(dev, NV04_PFIFO_DMA_TIMESLICE) & 0x1ffff);
+	/* No idea what the below is for exactly, ripped from a mmio-trace */
+	nv_wi32(dev, fc + 64, nv_rd32(dev, NV40_PFIFO_UNK32E4));
+	/* NVIDIA do this next line twice.. bug? */
+	nv_wi32(dev, fc + 68, nv_rd32(dev, 0x32e8));
+	nv_wi32(dev, fc + 76, nv_rd32(dev, 0x2088));
+	nv_wi32(dev, fc + 80, nv_rd32(dev, 0x3300));
+#if 0 /* no real idea which is PUT/GET in UNK_48.. */
+	tmp  = nv_rd32(dev, NV04_PFIFO_CACHE1_GET);
+	tmp |= (nv_rd32(dev, NV04_PFIFO_CACHE1_PUT) << 16);
+	nv_wi32(dev, fc + 72, tmp);
+#endif
+	dev_priv->engine.instmem.finish_access(dev);
+
+	nv40_fifo_do_load_context(dev, pfifo->channels - 1);
+	nv_wr32(dev, NV03_PFIFO_CACHE1_PUSH1,
+		     NV40_PFIFO_CACHE1_PUSH1_DMA | (pfifo->channels - 1));
+	return 0;
+}
+
+static void
+nv40_fifo_init_reset(struct drm_device *dev)
+{
+	int i;
+
+	nv_wr32(dev, NV03_PMC_ENABLE,
+		nv_rd32(dev, NV03_PMC_ENABLE) & ~NV_PMC_ENABLE_PFIFO);
+	nv_wr32(dev, NV03_PMC_ENABLE,
+		nv_rd32(dev, NV03_PMC_ENABLE) |  NV_PMC_ENABLE_PFIFO);
+
+	nv_wr32(dev, 0x003224, 0x000f0078);
+	nv_wr32(dev, 0x003210, 0x00000000);
+	nv_wr32(dev, 0x003270, 0x00000000);
+	nv_wr32(dev, 0x003240, 0x00000000);
+	nv_wr32(dev, 0x003244, 0x00000000);
+	nv_wr32(dev, 0x003258, 0x00000000);
+	nv_wr32(dev, 0x002504, 0x00000000);
+	for (i = 0; i < 16; i++)
+		nv_wr32(dev, 0x002510 + (i * 4), 0x00000000);
+	nv_wr32(dev, 0x00250c, 0x0000ffff);
+	nv_wr32(dev, 0x002048, 0x00000000);
+	nv_wr32(dev, 0x003228, 0x00000000);
+	nv_wr32(dev, 0x0032e8, 0x00000000);
+	nv_wr32(dev, 0x002410, 0x00000000);
+	nv_wr32(dev, 0x002420, 0x00000000);
+	nv_wr32(dev, 0x002058, 0x00000001);
+	nv_wr32(dev, 0x00221c, 0x00000000);
+	/* something with 0x2084, read/modify/write, no change */
+	nv_wr32(dev, 0x002040, 0x000000ff);
+	nv_wr32(dev, 0x002500, 0x00000000);
+	nv_wr32(dev, 0x003200, 0x00000000);
+
+	nv_wr32(dev, NV04_PFIFO_DMA_TIMESLICE, 0x2101ffff);
+}
+
+static void
+nv40_fifo_init_ramxx(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	nv_wr32(dev, NV03_PFIFO_RAMHT, (0x03 << 24) /* search 128 */ |
+				       ((dev_priv->ramht_bits - 9) << 16) |
+				       (dev_priv->ramht_offset >> 8));
+	nv_wr32(dev, NV03_PFIFO_RAMRO, dev_priv->ramro_offset>>8);
+
+	switch (dev_priv->chipset) {
+	case 0x47:
+	case 0x49:
+	case 0x4b:
+		nv_wr32(dev, 0x2230, 1);
+		break;
+	default:
+		break;
+	}
+
+	switch (dev_priv->chipset) {
+	case 0x40:
+	case 0x41:
+	case 0x42:
+	case 0x43:
+	case 0x45:
+	case 0x47:
+	case 0x48:
+	case 0x49:
+	case 0x4b:
+		nv_wr32(dev, NV40_PFIFO_RAMFC, 0x30002);
+		break;
+	default:
+		nv_wr32(dev, 0x2230, 0);
+		nv_wr32(dev, NV40_PFIFO_RAMFC,
+			((nouveau_mem_fb_amount(dev) - 512 * 1024 +
+			  dev_priv->ramfc_offset) >> 16) | (3 << 16));
+		break;
+	}
+}
+
+static void
+nv40_fifo_init_intr(struct drm_device *dev)
+{
+	nv_wr32(dev, 0x002100, 0xffffffff);
+	nv_wr32(dev, 0x002140, 0xffffffff);
+}
+
+int
+nv40_fifo_init(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_fifo_engine *pfifo = &dev_priv->engine.fifo;
+	int i;
+
+	nv40_fifo_init_reset(dev);
+	nv40_fifo_init_ramxx(dev);
+
+	nv40_fifo_do_load_context(dev, pfifo->channels - 1);
+	nv_wr32(dev, NV03_PFIFO_CACHE1_PUSH1, pfifo->channels - 1);
+
+	nv40_fifo_init_intr(dev);
+	pfifo->enable(dev);
+	pfifo->reassign(dev, true);
+
+	for (i = 0; i < dev_priv->engine.fifo.channels; i++) {
+		if (dev_priv->fifos[i]) {
+			uint32_t mode = nv_rd32(dev, NV04_PFIFO_MODE);
+			nv_wr32(dev, NV04_PFIFO_MODE, mode | (1 << i));
+		}
+	}
+
+	return 0;
+}
diff --git a/drivers/gpu/drm/nouveau/nv40_graph.c b/drivers/gpu/drm/nouveau/nv40_graph.c
new file mode 100644
index 000000000000..53e8afe1dcd1
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv40_graph.c
@@ -0,0 +1,406 @@
+/*
+ * Copyright (C) 2007 Ben Skeggs.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "drmP.h"
+#include "drm.h"
+#include "nouveau_drv.h"
+#include "nouveau_grctx.h"
+
+struct nouveau_channel *
+nv40_graph_channel(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t inst;
+	int i;
+
+	inst = nv_rd32(dev, NV40_PGRAPH_CTXCTL_CUR);
+	if (!(inst & NV40_PGRAPH_CTXCTL_CUR_LOADED))
+		return NULL;
+	inst = (inst & NV40_PGRAPH_CTXCTL_CUR_INSTANCE) << 4;
+
+	for (i = 0; i < dev_priv->engine.fifo.channels; i++) {
+		struct nouveau_channel *chan = dev_priv->fifos[i];
+
+		if (chan && chan->ramin_grctx &&
+		    chan->ramin_grctx->instance == inst)
+			return chan;
+	}
+
+	return NULL;
+}
+
+int
+nv40_graph_create_context(struct nouveau_channel *chan)
+{
+	struct drm_device *dev = chan->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
+	int ret;
+
+	ret = nouveau_gpuobj_new_ref(dev, chan, NULL, 0, pgraph->grctx_size,
+				     16, NVOBJ_FLAG_ZERO_ALLOC,
+				     &chan->ramin_grctx);
+	if (ret)
+		return ret;
+
+	/* Initialise default context values */
+	dev_priv->engine.instmem.prepare_access(dev, true);
+	if (!pgraph->ctxprog) {
+		struct nouveau_grctx ctx = {};
+
+		ctx.dev = chan->dev;
+		ctx.mode = NOUVEAU_GRCTX_VALS;
+		ctx.data = chan->ramin_grctx->gpuobj;
+		nv40_grctx_init(&ctx);
+	} else {
+		nouveau_grctx_vals_load(dev, chan->ramin_grctx->gpuobj);
+	}
+	nv_wo32(dev, chan->ramin_grctx->gpuobj, 0,
+		     chan->ramin_grctx->gpuobj->im_pramin->start);
+	dev_priv->engine.instmem.finish_access(dev);
+	return 0;
+}
+
+void
+nv40_graph_destroy_context(struct nouveau_channel *chan)
+{
+	nouveau_gpuobj_ref_del(chan->dev, &chan->ramin_grctx);
+}
+
+static int
+nv40_graph_transfer_context(struct drm_device *dev, uint32_t inst, int save)
+{
+	uint32_t old_cp, tv = 1000, tmp;
+	int i;
+
+	old_cp = nv_rd32(dev, NV20_PGRAPH_CHANNEL_CTX_POINTER);
+	nv_wr32(dev, NV20_PGRAPH_CHANNEL_CTX_POINTER, inst);
+
+	tmp  = nv_rd32(dev, NV40_PGRAPH_CTXCTL_0310);
+	tmp |= save ? NV40_PGRAPH_CTXCTL_0310_XFER_SAVE :
+		      NV40_PGRAPH_CTXCTL_0310_XFER_LOAD;
+	nv_wr32(dev, NV40_PGRAPH_CTXCTL_0310, tmp);
+
+	tmp  = nv_rd32(dev, NV40_PGRAPH_CTXCTL_0304);
+	tmp |= NV40_PGRAPH_CTXCTL_0304_XFER_CTX;
+	nv_wr32(dev, NV40_PGRAPH_CTXCTL_0304, tmp);
+
+	nouveau_wait_for_idle(dev);
+
+	for (i = 0; i < tv; i++) {
+		if (nv_rd32(dev, NV40_PGRAPH_CTXCTL_030C) == 0)
+			break;
+	}
+
+	nv_wr32(dev, NV20_PGRAPH_CHANNEL_CTX_POINTER, old_cp);
+
+	if (i == tv) {
+		uint32_t ucstat = nv_rd32(dev, NV40_PGRAPH_CTXCTL_UCODE_STAT);
+		NV_ERROR(dev, "Failed: Instance=0x%08x Save=%d\n", inst, save);
+		NV_ERROR(dev, "IP: 0x%02x, Opcode: 0x%08x\n",
+			 ucstat >> NV40_PGRAPH_CTXCTL_UCODE_STAT_IP_SHIFT,
+			 ucstat  & NV40_PGRAPH_CTXCTL_UCODE_STAT_OP_MASK);
+		NV_ERROR(dev, "0x40030C = 0x%08x\n",
+			 nv_rd32(dev, NV40_PGRAPH_CTXCTL_030C));
+		return -EBUSY;
+	}
+
+	return 0;
+}
+
+/* Restore the context for a specific channel into PGRAPH */
+int
+nv40_graph_load_context(struct nouveau_channel *chan)
+{
+	struct drm_device *dev = chan->dev;
+	uint32_t inst;
+	int ret;
+
+	if (!chan->ramin_grctx)
+		return -EINVAL;
+	inst = chan->ramin_grctx->instance >> 4;
+
+	ret = nv40_graph_transfer_context(dev, inst, 0);
+	if (ret)
+		return ret;
+
+	/* 0x40032C, no idea of it's exact function.  Could simply be a
+	 * record of the currently active PGRAPH context.  It's currently
+	 * unknown as to what bit 24 does.  The nv ddx has it set, so we will
+	 * set it here too.
+	 */
+	nv_wr32(dev, NV20_PGRAPH_CHANNEL_CTX_POINTER, inst);
+	nv_wr32(dev, NV40_PGRAPH_CTXCTL_CUR,
+		 (inst & NV40_PGRAPH_CTXCTL_CUR_INSTANCE) |
+		  NV40_PGRAPH_CTXCTL_CUR_LOADED);
+	/* 0x32E0 records the instance address of the active FIFO's PGRAPH
+	 * context.  If at any time this doesn't match 0x40032C, you will
+	 * recieve PGRAPH_INTR_CONTEXT_SWITCH
+	 */
+	nv_wr32(dev, NV40_PFIFO_GRCTX_INSTANCE, inst);
+	return 0;
+}
+
+int
+nv40_graph_unload_context(struct drm_device *dev)
+{
+	uint32_t inst;
+	int ret;
+
+	inst = nv_rd32(dev, NV40_PGRAPH_CTXCTL_CUR);
+	if (!(inst & NV40_PGRAPH_CTXCTL_CUR_LOADED))
+		return 0;
+	inst &= NV40_PGRAPH_CTXCTL_CUR_INSTANCE;
+
+	ret = nv40_graph_transfer_context(dev, inst, 1);
+
+	nv_wr32(dev, NV40_PGRAPH_CTXCTL_CUR, inst);
+	return ret;
+}
+
+void
+nv40_graph_set_region_tiling(struct drm_device *dev, int i, uint32_t addr,
+			     uint32_t size, uint32_t pitch)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t limit = max(1u, addr + size) - 1;
+
+	if (pitch)
+		addr |= 1;
+
+	switch (dev_priv->chipset) {
+	case 0x44:
+	case 0x4a:
+	case 0x4e:
+		nv_wr32(dev, NV20_PGRAPH_TSIZE(i), pitch);
+		nv_wr32(dev, NV20_PGRAPH_TLIMIT(i), limit);
+		nv_wr32(dev, NV20_PGRAPH_TILE(i), addr);
+		break;
+
+	case 0x46:
+	case 0x47:
+	case 0x49:
+	case 0x4b:
+		nv_wr32(dev, NV47_PGRAPH_TSIZE(i), pitch);
+		nv_wr32(dev, NV47_PGRAPH_TLIMIT(i), limit);
+		nv_wr32(dev, NV47_PGRAPH_TILE(i), addr);
+		nv_wr32(dev, NV40_PGRAPH_TSIZE1(i), pitch);
+		nv_wr32(dev, NV40_PGRAPH_TLIMIT1(i), limit);
+		nv_wr32(dev, NV40_PGRAPH_TILE1(i), addr);
+		break;
+
+	default:
+		nv_wr32(dev, NV20_PGRAPH_TSIZE(i), pitch);
+		nv_wr32(dev, NV20_PGRAPH_TLIMIT(i), limit);
+		nv_wr32(dev, NV20_PGRAPH_TILE(i), addr);
+		nv_wr32(dev, NV40_PGRAPH_TSIZE1(i), pitch);
+		nv_wr32(dev, NV40_PGRAPH_TLIMIT1(i), limit);
+		nv_wr32(dev, NV40_PGRAPH_TILE1(i), addr);
+		break;
+	}
+}
+
+/*
+ * G70		0x47
+ * G71		0x49
+ * NV45		0x48
+ * G72[M]	0x46
+ * G73		0x4b
+ * C51_G7X	0x4c
+ * C51		0x4e
+ */
+int
+nv40_graph_init(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv =
+		(struct drm_nouveau_private *)dev->dev_private;
+	struct nouveau_fb_engine *pfb = &dev_priv->engine.fb;
+	uint32_t vramsz;
+	int i, j;
+
+	nv_wr32(dev, NV03_PMC_ENABLE, nv_rd32(dev, NV03_PMC_ENABLE) &
+			~NV_PMC_ENABLE_PGRAPH);
+	nv_wr32(dev, NV03_PMC_ENABLE, nv_rd32(dev, NV03_PMC_ENABLE) |
+			 NV_PMC_ENABLE_PGRAPH);
+
+	if (nouveau_ctxfw) {
+		nouveau_grctx_prog_load(dev);
+		dev_priv->engine.graph.grctx_size = 175 * 1024;
+	}
+
+	if (!dev_priv->engine.graph.ctxprog) {
+		struct nouveau_grctx ctx = {};
+		uint32_t cp[256];
+
+		ctx.dev = dev;
+		ctx.mode = NOUVEAU_GRCTX_PROG;
+		ctx.data = cp;
+		ctx.ctxprog_max = 256;
+		nv40_grctx_init(&ctx);
+		dev_priv->engine.graph.grctx_size = ctx.ctxvals_pos * 4;
+
+		nv_wr32(dev, NV40_PGRAPH_CTXCTL_UCODE_INDEX, 0);
+		for (i = 0; i < ctx.ctxprog_len; i++)
+			nv_wr32(dev, NV40_PGRAPH_CTXCTL_UCODE_DATA, cp[i]);
+	}
+
+	/* No context present currently */
+	nv_wr32(dev, NV40_PGRAPH_CTXCTL_CUR, 0x00000000);
+
+	nv_wr32(dev, NV03_PGRAPH_INTR   , 0xFFFFFFFF);
+	nv_wr32(dev, NV40_PGRAPH_INTR_EN, 0xFFFFFFFF);
+
+	nv_wr32(dev, NV04_PGRAPH_DEBUG_0, 0xFFFFFFFF);
+	nv_wr32(dev, NV04_PGRAPH_DEBUG_0, 0x00000000);
+	nv_wr32(dev, NV04_PGRAPH_DEBUG_1, 0x401287c0);
+	nv_wr32(dev, NV04_PGRAPH_DEBUG_3, 0xe0de8055);
+	nv_wr32(dev, NV10_PGRAPH_DEBUG_4, 0x00008000);
+	nv_wr32(dev, NV04_PGRAPH_LIMIT_VIOL_PIX, 0x00be3c5f);
+
+	nv_wr32(dev, NV10_PGRAPH_CTX_CONTROL, 0x10010100);
+	nv_wr32(dev, NV10_PGRAPH_STATE      , 0xFFFFFFFF);
+
+	j = nv_rd32(dev, 0x1540) & 0xff;
+	if (j) {
+		for (i = 0; !(j & 1); j >>= 1, i++)
+			;
+		nv_wr32(dev, 0x405000, i);
+	}
+
+	if (dev_priv->chipset == 0x40) {
+		nv_wr32(dev, 0x4009b0, 0x83280fff);
+		nv_wr32(dev, 0x4009b4, 0x000000a0);
+	} else {
+		nv_wr32(dev, 0x400820, 0x83280eff);
+		nv_wr32(dev, 0x400824, 0x000000a0);
+	}
+
+	switch (dev_priv->chipset) {
+	case 0x40:
+	case 0x45:
+		nv_wr32(dev, 0x4009b8, 0x0078e366);
+		nv_wr32(dev, 0x4009bc, 0x0000014c);
+		break;
+	case 0x41:
+	case 0x42: /* pciid also 0x00Cx */
+	/* case 0x0120: XXX (pciid) */
+		nv_wr32(dev, 0x400828, 0x007596ff);
+		nv_wr32(dev, 0x40082c, 0x00000108);
+		break;
+	case 0x43:
+		nv_wr32(dev, 0x400828, 0x0072cb77);
+		nv_wr32(dev, 0x40082c, 0x00000108);
+		break;
+	case 0x44:
+	case 0x46: /* G72 */
+	case 0x4a:
+	case 0x4c: /* G7x-based C51 */
+	case 0x4e:
+		nv_wr32(dev, 0x400860, 0);
+		nv_wr32(dev, 0x400864, 0);
+		break;
+	case 0x47: /* G70 */
+	case 0x49: /* G71 */
+	case 0x4b: /* G73 */
+		nv_wr32(dev, 0x400828, 0x07830610);
+		nv_wr32(dev, 0x40082c, 0x0000016A);
+		break;
+	default:
+		break;
+	}
+
+	nv_wr32(dev, 0x400b38, 0x2ffff800);
+	nv_wr32(dev, 0x400b3c, 0x00006000);
+
+	/* Turn all the tiling regions off. */
+	for (i = 0; i < pfb->num_tiles; i++)
+		nv40_graph_set_region_tiling(dev, i, 0, 0, 0);
+
+	/* begin RAM config */
+	vramsz = drm_get_resource_len(dev, 0) - 1;
+	switch (dev_priv->chipset) {
+	case 0x40:
+		nv_wr32(dev, 0x4009A4, nv_rd32(dev, NV04_PFB_CFG0));
+		nv_wr32(dev, 0x4009A8, nv_rd32(dev, NV04_PFB_CFG1));
+		nv_wr32(dev, 0x4069A4, nv_rd32(dev, NV04_PFB_CFG0));
+		nv_wr32(dev, 0x4069A8, nv_rd32(dev, NV04_PFB_CFG1));
+		nv_wr32(dev, 0x400820, 0);
+		nv_wr32(dev, 0x400824, 0);
+		nv_wr32(dev, 0x400864, vramsz);
+		nv_wr32(dev, 0x400868, vramsz);
+		break;
+	default:
+		switch (dev_priv->chipset) {
+		case 0x46:
+		case 0x47:
+		case 0x49:
+		case 0x4b:
+			nv_wr32(dev, 0x400DF0, nv_rd32(dev, NV04_PFB_CFG0));
+			nv_wr32(dev, 0x400DF4, nv_rd32(dev, NV04_PFB_CFG1));
+			break;
+		default:
+			nv_wr32(dev, 0x4009F0, nv_rd32(dev, NV04_PFB_CFG0));
+			nv_wr32(dev, 0x4009F4, nv_rd32(dev, NV04_PFB_CFG1));
+			break;
+		}
+		nv_wr32(dev, 0x4069F0, nv_rd32(dev, NV04_PFB_CFG0));
+		nv_wr32(dev, 0x4069F4, nv_rd32(dev, NV04_PFB_CFG1));
+		nv_wr32(dev, 0x400840, 0);
+		nv_wr32(dev, 0x400844, 0);
+		nv_wr32(dev, 0x4008A0, vramsz);
+		nv_wr32(dev, 0x4008A4, vramsz);
+		break;
+	}
+
+	return 0;
+}
+
+void nv40_graph_takedown(struct drm_device *dev)
+{
+	nouveau_grctx_fini(dev);
+}
+
+struct nouveau_pgraph_object_class nv40_graph_grclass[] = {
+	{ 0x0030, false, NULL }, /* null */
+	{ 0x0039, false, NULL }, /* m2mf */
+	{ 0x004a, false, NULL }, /* gdirect */
+	{ 0x009f, false, NULL }, /* imageblit (nv12) */
+	{ 0x008a, false, NULL }, /* ifc */
+	{ 0x0089, false, NULL }, /* sifm */
+	{ 0x3089, false, NULL }, /* sifm (nv40) */
+	{ 0x0062, false, NULL }, /* surf2d */
+	{ 0x3062, false, NULL }, /* surf2d (nv40) */
+	{ 0x0043, false, NULL }, /* rop */
+	{ 0x0012, false, NULL }, /* beta1 */
+	{ 0x0072, false, NULL }, /* beta4 */
+	{ 0x0019, false, NULL }, /* cliprect */
+	{ 0x0044, false, NULL }, /* pattern */
+	{ 0x309e, false, NULL }, /* swzsurf */
+	{ 0x4097, false, NULL }, /* curie (nv40) */
+	{ 0x4497, false, NULL }, /* curie (nv44) */
+	{}
+};
+
diff --git a/drivers/gpu/drm/nouveau/nv40_grctx.c b/drivers/gpu/drm/nouveau/nv40_grctx.c
new file mode 100644
index 000000000000..11b11c31f543
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv40_grctx.c
@@ -0,0 +1,678 @@
+/*
+ * Copyright 2009 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Ben Skeggs
+ */
+
+/* NVIDIA context programs handle a number of other conditions which are
+ * not implemented in our versions.  It's not clear why NVIDIA context
+ * programs have this code, nor whether it's strictly necessary for
+ * correct operation.  We'll implement additional handling if/when we
+ * discover it's necessary.
+ *
+ * - On context save, NVIDIA set 0x400314 bit 0 to 1 if the "3D state"
+ *   flag is set, this gets saved into the context.
+ * - On context save, the context program for all cards load nsource
+ *   into a flag register and check for ILLEGAL_MTHD.  If it's set,
+ *   opcode 0x60000d is called before resuming normal operation.
+ * - Some context programs check more conditions than the above.  NV44
+ *   checks: ((nsource & 0x0857) || (0x400718 & 0x0100) || (intr & 0x0001))
+ *   and calls 0x60000d before resuming normal operation.
+ * - At the very beginning of NVIDIA's context programs, flag 9 is checked
+ *   and if true 0x800001 is called with count=0, pos=0, the flag is cleared
+ *   and then the ctxprog is aborted.  It looks like a complicated NOP,
+ *   its purpose is unknown.
+ * - In the section of code that loads the per-vs state, NVIDIA check
+ *   flag 10.  If it's set, they only transfer the small 0x300 byte block
+ *   of state + the state for a single vs as opposed to the state for
+ *   all vs units.  It doesn't seem likely that it'll occur in normal
+ *   operation, especially seeing as it appears NVIDIA may have screwed
+ *   up the ctxprogs for some cards and have an invalid instruction
+ *   rather than a cp_lsr(ctx, dwords_for_1_vs_unit) instruction.
+ * - There's a number of places where context offset 0 (where we place
+ *   the PRAMIN offset of the context) is loaded into either 0x408000,
+ *   0x408004 or 0x408008.  Not sure what's up there either.
+ * - The ctxprogs for some cards save 0x400a00 again during the cleanup
+ *   path for auto-loadctx.
+ */
+
+#define CP_FLAG_CLEAR                 0
+#define CP_FLAG_SET                   1
+#define CP_FLAG_SWAP_DIRECTION        ((0 * 32) + 0)
+#define CP_FLAG_SWAP_DIRECTION_LOAD   0
+#define CP_FLAG_SWAP_DIRECTION_SAVE   1
+#define CP_FLAG_USER_SAVE             ((0 * 32) + 5)
+#define CP_FLAG_USER_SAVE_NOT_PENDING 0
+#define CP_FLAG_USER_SAVE_PENDING     1
+#define CP_FLAG_USER_LOAD             ((0 * 32) + 6)
+#define CP_FLAG_USER_LOAD_NOT_PENDING 0
+#define CP_FLAG_USER_LOAD_PENDING     1
+#define CP_FLAG_STATUS                ((3 * 32) + 0)
+#define CP_FLAG_STATUS_IDLE           0
+#define CP_FLAG_STATUS_BUSY           1
+#define CP_FLAG_AUTO_SAVE             ((3 * 32) + 4)
+#define CP_FLAG_AUTO_SAVE_NOT_PENDING 0
+#define CP_FLAG_AUTO_SAVE_PENDING     1
+#define CP_FLAG_AUTO_LOAD             ((3 * 32) + 5)
+#define CP_FLAG_AUTO_LOAD_NOT_PENDING 0
+#define CP_FLAG_AUTO_LOAD_PENDING     1
+#define CP_FLAG_UNK54                 ((3 * 32) + 6)
+#define CP_FLAG_UNK54_CLEAR           0
+#define CP_FLAG_UNK54_SET             1
+#define CP_FLAG_ALWAYS                ((3 * 32) + 8)
+#define CP_FLAG_ALWAYS_FALSE          0
+#define CP_FLAG_ALWAYS_TRUE           1
+#define CP_FLAG_UNK57                 ((3 * 32) + 9)
+#define CP_FLAG_UNK57_CLEAR           0
+#define CP_FLAG_UNK57_SET             1
+
+#define CP_CTX                   0x00100000
+#define CP_CTX_COUNT             0x000fc000
+#define CP_CTX_COUNT_SHIFT               14
+#define CP_CTX_REG               0x00003fff
+#define CP_LOAD_SR               0x00200000
+#define CP_LOAD_SR_VALUE         0x000fffff
+#define CP_BRA                   0x00400000
+#define CP_BRA_IP                0x0000ff00
+#define CP_BRA_IP_SHIFT                   8
+#define CP_BRA_IF_CLEAR          0x00000080
+#define CP_BRA_FLAG              0x0000007f
+#define CP_WAIT                  0x00500000
+#define CP_WAIT_SET              0x00000080
+#define CP_WAIT_FLAG             0x0000007f
+#define CP_SET                   0x00700000
+#define CP_SET_1                 0x00000080
+#define CP_SET_FLAG              0x0000007f
+#define CP_NEXT_TO_SWAP          0x00600007
+#define CP_NEXT_TO_CURRENT       0x00600009
+#define CP_SET_CONTEXT_POINTER   0x0060000a
+#define CP_END                   0x0060000e
+#define CP_LOAD_MAGIC_UNK01      0x00800001 /* unknown */
+#define CP_LOAD_MAGIC_NV44TCL    0x00800029 /* per-vs state (0x4497) */
+#define CP_LOAD_MAGIC_NV40TCL    0x00800041 /* per-vs state (0x4097) */
+
+#include "drmP.h"
+#include "nouveau_drv.h"
+#include "nouveau_grctx.h"
+
+/* TODO:
+ *  - get vs count from 0x1540
+ *  - document unimplemented bits compared to nvidia
+ *    - nsource handling
+ *    - R0 & 0x0200 handling
+ *    - single-vs handling
+ *    - 400314 bit 0
+ */
+
+static int
+nv40_graph_4097(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	if ((dev_priv->chipset & 0xf0) == 0x60)
+		return 0;
+
+	return !!(0x0baf & (1 << dev_priv->chipset));
+}
+
+static int
+nv40_graph_vs_count(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	switch (dev_priv->chipset) {
+	case 0x47:
+	case 0x49:
+	case 0x4b:
+		return 8;
+	case 0x40:
+		return 6;
+	case 0x41:
+	case 0x42:
+		return 5;
+	case 0x43:
+	case 0x44:
+	case 0x46:
+	case 0x4a:
+		return 3;
+	case 0x4c:
+	case 0x4e:
+	case 0x67:
+	default:
+		return 1;
+	}
+}
+
+
+enum cp_label {
+	cp_check_load = 1,
+	cp_setup_auto_load,
+	cp_setup_load,
+	cp_setup_save,
+	cp_swap_state,
+	cp_swap_state3d_3_is_save,
+	cp_prepare_exit,
+	cp_exit,
+};
+
+static void
+nv40_graph_construct_general(struct nouveau_grctx *ctx)
+{
+	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
+	int i;
+
+	cp_ctx(ctx, 0x4000a4, 1);
+	gr_def(ctx, 0x4000a4, 0x00000008);
+	cp_ctx(ctx, 0x400144, 58);
+	gr_def(ctx, 0x400144, 0x00000001);
+	cp_ctx(ctx, 0x400314, 1);
+	gr_def(ctx, 0x400314, 0x00000000);
+	cp_ctx(ctx, 0x400400, 10);
+	cp_ctx(ctx, 0x400480, 10);
+	cp_ctx(ctx, 0x400500, 19);
+	gr_def(ctx, 0x400514, 0x00040000);
+	gr_def(ctx, 0x400524, 0x55555555);
+	gr_def(ctx, 0x400528, 0x55555555);
+	gr_def(ctx, 0x40052c, 0x55555555);
+	gr_def(ctx, 0x400530, 0x55555555);
+	cp_ctx(ctx, 0x400560, 6);
+	gr_def(ctx, 0x400568, 0x0000ffff);
+	gr_def(ctx, 0x40056c, 0x0000ffff);
+	cp_ctx(ctx, 0x40057c, 5);
+	cp_ctx(ctx, 0x400710, 3);
+	gr_def(ctx, 0x400710, 0x20010001);
+	gr_def(ctx, 0x400714, 0x0f73ef00);
+	cp_ctx(ctx, 0x400724, 1);
+	gr_def(ctx, 0x400724, 0x02008821);
+	cp_ctx(ctx, 0x400770, 3);
+	if (dev_priv->chipset == 0x40) {
+		cp_ctx(ctx, 0x400814, 4);
+		cp_ctx(ctx, 0x400828, 5);
+		cp_ctx(ctx, 0x400840, 5);
+		gr_def(ctx, 0x400850, 0x00000040);
+		cp_ctx(ctx, 0x400858, 4);
+		gr_def(ctx, 0x400858, 0x00000040);
+		gr_def(ctx, 0x40085c, 0x00000040);
+		gr_def(ctx, 0x400864, 0x80000000);
+		cp_ctx(ctx, 0x40086c, 9);
+		gr_def(ctx, 0x40086c, 0x80000000);
+		gr_def(ctx, 0x400870, 0x80000000);
+		gr_def(ctx, 0x400874, 0x80000000);
+		gr_def(ctx, 0x400878, 0x80000000);
+		gr_def(ctx, 0x400888, 0x00000040);
+		gr_def(ctx, 0x40088c, 0x80000000);
+		cp_ctx(ctx, 0x4009c0, 8);
+		gr_def(ctx, 0x4009cc, 0x80000000);
+		gr_def(ctx, 0x4009dc, 0x80000000);
+	} else {
+		cp_ctx(ctx, 0x400840, 20);
+		if (!nv40_graph_4097(ctx->dev)) {
+			for (i = 0; i < 8; i++)
+				gr_def(ctx, 0x400860 + (i * 4), 0x00000001);
+		}
+		gr_def(ctx, 0x400880, 0x00000040);
+		gr_def(ctx, 0x400884, 0x00000040);
+		gr_def(ctx, 0x400888, 0x00000040);
+		cp_ctx(ctx, 0x400894, 11);
+		gr_def(ctx, 0x400894, 0x00000040);
+		if (nv40_graph_4097(ctx->dev)) {
+			for (i = 0; i < 8; i++)
+				gr_def(ctx, 0x4008a0 + (i * 4), 0x80000000);
+		}
+		cp_ctx(ctx, 0x4008e0, 2);
+		cp_ctx(ctx, 0x4008f8, 2);
+		if (dev_priv->chipset == 0x4c ||
+		    (dev_priv->chipset & 0xf0) == 0x60)
+			cp_ctx(ctx, 0x4009f8, 1);
+	}
+	cp_ctx(ctx, 0x400a00, 73);
+	gr_def(ctx, 0x400b0c, 0x0b0b0b0c);
+	cp_ctx(ctx, 0x401000, 4);
+	cp_ctx(ctx, 0x405004, 1);
+	switch (dev_priv->chipset) {
+	case 0x47:
+	case 0x49:
+	case 0x4b:
+		cp_ctx(ctx, 0x403448, 1);
+		gr_def(ctx, 0x403448, 0x00001010);
+		break;
+	default:
+		cp_ctx(ctx, 0x403440, 1);
+		switch (dev_priv->chipset) {
+		case 0x40:
+			gr_def(ctx, 0x403440, 0x00000010);
+			break;
+		case 0x44:
+		case 0x46:
+		case 0x4a:
+			gr_def(ctx, 0x403440, 0x00003010);
+			break;
+		case 0x41:
+		case 0x42:
+		case 0x43:
+		case 0x4c:
+		case 0x4e:
+		case 0x67:
+		default:
+			gr_def(ctx, 0x403440, 0x00001010);
+			break;
+		}
+		break;
+	}
+}
+
+static void
+nv40_graph_construct_state3d(struct nouveau_grctx *ctx)
+{
+	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
+	int i;
+
+	if (dev_priv->chipset == 0x40) {
+		cp_ctx(ctx, 0x401880, 51);
+		gr_def(ctx, 0x401940, 0x00000100);
+	} else
+	if (dev_priv->chipset == 0x46 || dev_priv->chipset == 0x47 ||
+	    dev_priv->chipset == 0x49 || dev_priv->chipset == 0x4b) {
+		cp_ctx(ctx, 0x401880, 32);
+		for (i = 0; i < 16; i++)
+			gr_def(ctx, 0x401880 + (i * 4), 0x00000111);
+		if (dev_priv->chipset == 0x46)
+			cp_ctx(ctx, 0x401900, 16);
+		cp_ctx(ctx, 0x401940, 3);
+	}
+	cp_ctx(ctx, 0x40194c, 18);
+	gr_def(ctx, 0x401954, 0x00000111);
+	gr_def(ctx, 0x401958, 0x00080060);
+	gr_def(ctx, 0x401974, 0x00000080);
+	gr_def(ctx, 0x401978, 0xffff0000);
+	gr_def(ctx, 0x40197c, 0x00000001);
+	gr_def(ctx, 0x401990, 0x46400000);
+	if (dev_priv->chipset == 0x40) {
+		cp_ctx(ctx, 0x4019a0, 2);
+		cp_ctx(ctx, 0x4019ac, 5);
+	} else {
+		cp_ctx(ctx, 0x4019a0, 1);
+		cp_ctx(ctx, 0x4019b4, 3);
+	}
+	gr_def(ctx, 0x4019bc, 0xffff0000);
+	switch (dev_priv->chipset) {
+	case 0x46:
+	case 0x47:
+	case 0x49:
+	case 0x4b:
+		cp_ctx(ctx, 0x4019c0, 18);
+		for (i = 0; i < 16; i++)
+			gr_def(ctx, 0x4019c0 + (i * 4), 0x88888888);
+		break;
+	}
+	cp_ctx(ctx, 0x401a08, 8);
+	gr_def(ctx, 0x401a10, 0x0fff0000);
+	gr_def(ctx, 0x401a14, 0x0fff0000);
+	gr_def(ctx, 0x401a1c, 0x00011100);
+	cp_ctx(ctx, 0x401a2c, 4);
+	cp_ctx(ctx, 0x401a44, 26);
+	for (i = 0; i < 16; i++)
+		gr_def(ctx, 0x401a44 + (i * 4), 0x07ff0000);
+	gr_def(ctx, 0x401a8c, 0x4b7fffff);
+	if (dev_priv->chipset == 0x40) {
+		cp_ctx(ctx, 0x401ab8, 3);
+	} else {
+		cp_ctx(ctx, 0x401ab8, 1);
+		cp_ctx(ctx, 0x401ac0, 1);
+	}
+	cp_ctx(ctx, 0x401ad0, 8);
+	gr_def(ctx, 0x401ad0, 0x30201000);
+	gr_def(ctx, 0x401ad4, 0x70605040);
+	gr_def(ctx, 0x401ad8, 0xb8a89888);
+	gr_def(ctx, 0x401adc, 0xf8e8d8c8);
+	cp_ctx(ctx, 0x401b10, dev_priv->chipset == 0x40 ? 2 : 1);
+	gr_def(ctx, 0x401b10, 0x40100000);
+	cp_ctx(ctx, 0x401b18, dev_priv->chipset == 0x40 ? 6 : 5);
+	gr_def(ctx, 0x401b28, dev_priv->chipset == 0x40 ?
+			      0x00000004 : 0x00000000);
+	cp_ctx(ctx, 0x401b30, 25);
+	gr_def(ctx, 0x401b34, 0x0000ffff);
+	gr_def(ctx, 0x401b68, 0x435185d6);
+	gr_def(ctx, 0x401b6c, 0x2155b699);
+	gr_def(ctx, 0x401b70, 0xfedcba98);
+	gr_def(ctx, 0x401b74, 0x00000098);
+	gr_def(ctx, 0x401b84, 0xffffffff);
+	gr_def(ctx, 0x401b88, 0x00ff7000);
+	gr_def(ctx, 0x401b8c, 0x0000ffff);
+	if (dev_priv->chipset != 0x44 && dev_priv->chipset != 0x4a &&
+	    dev_priv->chipset != 0x4e)
+		cp_ctx(ctx, 0x401b94, 1);
+	cp_ctx(ctx, 0x401b98, 8);
+	gr_def(ctx, 0x401b9c, 0x00ff0000);
+	cp_ctx(ctx, 0x401bc0, 9);
+	gr_def(ctx, 0x401be0, 0x00ffff00);
+	cp_ctx(ctx, 0x401c00, 192);
+	for (i = 0; i < 16; i++) { /* fragment texture units */
+		gr_def(ctx, 0x401c40 + (i * 4), 0x00018488);
+		gr_def(ctx, 0x401c80 + (i * 4), 0x00028202);
+		gr_def(ctx, 0x401d00 + (i * 4), 0x0000aae4);
+		gr_def(ctx, 0x401d40 + (i * 4), 0x01012000);
+		gr_def(ctx, 0x401d80 + (i * 4), 0x00080008);
+		gr_def(ctx, 0x401e00 + (i * 4), 0x00100008);
+	}
+	for (i = 0; i < 4; i++) { /* vertex texture units */
+		gr_def(ctx, 0x401e90 + (i * 4), 0x0001bc80);
+		gr_def(ctx, 0x401ea0 + (i * 4), 0x00000202);
+		gr_def(ctx, 0x401ec0 + (i * 4), 0x00000008);
+		gr_def(ctx, 0x401ee0 + (i * 4), 0x00080008);
+	}
+	cp_ctx(ctx, 0x400f5c, 3);
+	gr_def(ctx, 0x400f5c, 0x00000002);
+	cp_ctx(ctx, 0x400f84, 1);
+}
+
+static void
+nv40_graph_construct_state3d_2(struct nouveau_grctx *ctx)
+{
+	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
+	int i;
+
+	cp_ctx(ctx, 0x402000, 1);
+	cp_ctx(ctx, 0x402404, dev_priv->chipset == 0x40 ? 1 : 2);
+	switch (dev_priv->chipset) {
+	case 0x40:
+		gr_def(ctx, 0x402404, 0x00000001);
+		break;
+	case 0x4c:
+	case 0x4e:
+	case 0x67:
+		gr_def(ctx, 0x402404, 0x00000020);
+		break;
+	case 0x46:
+	case 0x49:
+	case 0x4b:
+		gr_def(ctx, 0x402404, 0x00000421);
+		break;
+	default:
+		gr_def(ctx, 0x402404, 0x00000021);
+	}
+	if (dev_priv->chipset != 0x40)
+		gr_def(ctx, 0x402408, 0x030c30c3);
+	switch (dev_priv->chipset) {
+	case 0x44:
+	case 0x46:
+	case 0x4a:
+	case 0x4c:
+	case 0x4e:
+	case 0x67:
+		cp_ctx(ctx, 0x402440, 1);
+		gr_def(ctx, 0x402440, 0x00011001);
+		break;
+	default:
+		break;
+	}
+	cp_ctx(ctx, 0x402480, dev_priv->chipset == 0x40 ? 8 : 9);
+	gr_def(ctx, 0x402488, 0x3e020200);
+	gr_def(ctx, 0x40248c, 0x00ffffff);
+	switch (dev_priv->chipset) {
+	case 0x40:
+		gr_def(ctx, 0x402490, 0x60103f00);
+		break;
+	case 0x47:
+		gr_def(ctx, 0x402490, 0x40103f00);
+		break;
+	case 0x41:
+	case 0x42:
+	case 0x49:
+	case 0x4b:
+		gr_def(ctx, 0x402490, 0x20103f00);
+		break;
+	default:
+		gr_def(ctx, 0x402490, 0x0c103f00);
+		break;
+	}
+	gr_def(ctx, 0x40249c, dev_priv->chipset <= 0x43 ?
+			      0x00020000 : 0x00040000);
+	cp_ctx(ctx, 0x402500, 31);
+	gr_def(ctx, 0x402530, 0x00008100);
+	if (dev_priv->chipset == 0x40)
+		cp_ctx(ctx, 0x40257c, 6);
+	cp_ctx(ctx, 0x402594, 16);
+	cp_ctx(ctx, 0x402800, 17);
+	gr_def(ctx, 0x402800, 0x00000001);
+	switch (dev_priv->chipset) {
+	case 0x47:
+	case 0x49:
+	case 0x4b:
+		cp_ctx(ctx, 0x402864, 1);
+		gr_def(ctx, 0x402864, 0x00001001);
+		cp_ctx(ctx, 0x402870, 3);
+		gr_def(ctx, 0x402878, 0x00000003);
+		if (dev_priv->chipset != 0x47) { /* belong at end!! */
+			cp_ctx(ctx, 0x402900, 1);
+			cp_ctx(ctx, 0x402940, 1);
+			cp_ctx(ctx, 0x402980, 1);
+			cp_ctx(ctx, 0x4029c0, 1);
+			cp_ctx(ctx, 0x402a00, 1);
+			cp_ctx(ctx, 0x402a40, 1);
+			cp_ctx(ctx, 0x402a80, 1);
+			cp_ctx(ctx, 0x402ac0, 1);
+		}
+		break;
+	case 0x40:
+		cp_ctx(ctx, 0x402844, 1);
+		gr_def(ctx, 0x402844, 0x00000001);
+		cp_ctx(ctx, 0x402850, 1);
+		break;
+	default:
+		cp_ctx(ctx, 0x402844, 1);
+		gr_def(ctx, 0x402844, 0x00001001);
+		cp_ctx(ctx, 0x402850, 2);
+		gr_def(ctx, 0x402854, 0x00000003);
+		break;
+	}
+
+	cp_ctx(ctx, 0x402c00, 4);
+	gr_def(ctx, 0x402c00, dev_priv->chipset == 0x40 ?
+			      0x80800001 : 0x00888001);
+	switch (dev_priv->chipset) {
+	case 0x47:
+	case 0x49:
+	case 0x4b:
+		cp_ctx(ctx, 0x402c20, 40);
+		for (i = 0; i < 32; i++)
+			gr_def(ctx, 0x402c40 + (i * 4), 0xffffffff);
+		cp_ctx(ctx, 0x4030b8, 13);
+		gr_def(ctx, 0x4030dc, 0x00000005);
+		gr_def(ctx, 0x4030e8, 0x0000ffff);
+		break;
+	default:
+		cp_ctx(ctx, 0x402c10, 4);
+		if (dev_priv->chipset == 0x40)
+			cp_ctx(ctx, 0x402c20, 36);
+		else
+		if (dev_priv->chipset <= 0x42)
+			cp_ctx(ctx, 0x402c20, 24);
+		else
+		if (dev_priv->chipset <= 0x4a)
+			cp_ctx(ctx, 0x402c20, 16);
+		else
+			cp_ctx(ctx, 0x402c20, 8);
+		cp_ctx(ctx, 0x402cb0, dev_priv->chipset == 0x40 ? 12 : 13);
+		gr_def(ctx, 0x402cd4, 0x00000005);
+		if (dev_priv->chipset != 0x40)
+			gr_def(ctx, 0x402ce0, 0x0000ffff);
+		break;
+	}
+
+	cp_ctx(ctx, 0x403400, dev_priv->chipset == 0x40 ? 4 : 3);
+	cp_ctx(ctx, 0x403410, dev_priv->chipset == 0x40 ? 4 : 3);
+	cp_ctx(ctx, 0x403420, nv40_graph_vs_count(ctx->dev));
+	for (i = 0; i < nv40_graph_vs_count(ctx->dev); i++)
+		gr_def(ctx, 0x403420 + (i * 4), 0x00005555);
+
+	if (dev_priv->chipset != 0x40) {
+		cp_ctx(ctx, 0x403600, 1);
+		gr_def(ctx, 0x403600, 0x00000001);
+	}
+	cp_ctx(ctx, 0x403800, 1);
+
+	cp_ctx(ctx, 0x403c18, 1);
+	gr_def(ctx, 0x403c18, 0x00000001);
+	switch (dev_priv->chipset) {
+	case 0x46:
+	case 0x47:
+	case 0x49:
+	case 0x4b:
+		cp_ctx(ctx, 0x405018, 1);
+		gr_def(ctx, 0x405018, 0x08e00001);
+		cp_ctx(ctx, 0x405c24, 1);
+		gr_def(ctx, 0x405c24, 0x000e3000);
+		break;
+	}
+	if (dev_priv->chipset != 0x4e)
+		cp_ctx(ctx, 0x405800, 11);
+	cp_ctx(ctx, 0x407000, 1);
+}
+
+static void
+nv40_graph_construct_state3d_3(struct nouveau_grctx *ctx)
+{
+	int len = nv40_graph_4097(ctx->dev) ? 0x0684 : 0x0084;
+
+	cp_out (ctx, 0x300000);
+	cp_lsr (ctx, len - 4);
+	cp_bra (ctx, SWAP_DIRECTION, SAVE, cp_swap_state3d_3_is_save);
+	cp_lsr (ctx, len);
+	cp_name(ctx, cp_swap_state3d_3_is_save);
+	cp_out (ctx, 0x800001);
+
+	ctx->ctxvals_pos += len;
+}
+
+static void
+nv40_graph_construct_shader(struct nouveau_grctx *ctx)
+{
+	struct drm_device *dev = ctx->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_gpuobj *obj = ctx->data;
+	int vs, vs_nr, vs_len, vs_nr_b0, vs_nr_b1, b0_offset, b1_offset;
+	int offset, i;
+
+	vs_nr    = nv40_graph_vs_count(ctx->dev);
+	vs_nr_b0 = 363;
+	vs_nr_b1 = dev_priv->chipset == 0x40 ? 128 : 64;
+	if (dev_priv->chipset == 0x40) {
+		b0_offset = 0x2200/4; /* 33a0 */
+		b1_offset = 0x55a0/4; /* 1500 */
+		vs_len = 0x6aa0/4;
+	} else
+	if (dev_priv->chipset == 0x41 || dev_priv->chipset == 0x42) {
+		b0_offset = 0x2200/4; /* 2200 */
+		b1_offset = 0x4400/4; /* 0b00 */
+		vs_len = 0x4f00/4;
+	} else {
+		b0_offset = 0x1d40/4; /* 2200 */
+		b1_offset = 0x3f40/4; /* 0b00 : 0a40 */
+		vs_len = nv40_graph_4097(dev) ? 0x4a40/4 : 0x4980/4;
+	}
+
+	cp_lsr(ctx, vs_len * vs_nr + 0x300/4);
+	cp_out(ctx, nv40_graph_4097(dev) ? 0x800041 : 0x800029);
+
+	offset = ctx->ctxvals_pos;
+	ctx->ctxvals_pos += (0x0300/4 + (vs_nr * vs_len));
+
+	if (ctx->mode != NOUVEAU_GRCTX_VALS)
+		return;
+
+	offset += 0x0280/4;
+	for (i = 0; i < 16; i++, offset += 2)
+		nv_wo32(dev, obj, offset, 0x3f800000);
+
+	for (vs = 0; vs < vs_nr; vs++, offset += vs_len) {
+		for (i = 0; i < vs_nr_b0 * 6; i += 6)
+			nv_wo32(dev, obj, offset + b0_offset + i, 0x00000001);
+		for (i = 0; i < vs_nr_b1 * 4; i += 4)
+			nv_wo32(dev, obj, offset + b1_offset + i, 0x3f800000);
+	}
+}
+
+void
+nv40_grctx_init(struct nouveau_grctx *ctx)
+{
+	/* decide whether we're loading/unloading the context */
+	cp_bra (ctx, AUTO_SAVE, PENDING, cp_setup_save);
+	cp_bra (ctx, USER_SAVE, PENDING, cp_setup_save);
+
+	cp_name(ctx, cp_check_load);
+	cp_bra (ctx, AUTO_LOAD, PENDING, cp_setup_auto_load);
+	cp_bra (ctx, USER_LOAD, PENDING, cp_setup_load);
+	cp_bra (ctx, ALWAYS, TRUE, cp_exit);
+
+	/* setup for context load */
+	cp_name(ctx, cp_setup_auto_load);
+	cp_wait(ctx, STATUS, IDLE);
+	cp_out (ctx, CP_NEXT_TO_SWAP);
+	cp_name(ctx, cp_setup_load);
+	cp_wait(ctx, STATUS, IDLE);
+	cp_set (ctx, SWAP_DIRECTION, LOAD);
+	cp_out (ctx, 0x00910880); /* ?? */
+	cp_out (ctx, 0x00901ffe); /* ?? */
+	cp_out (ctx, 0x01940000); /* ?? */
+	cp_lsr (ctx, 0x20);
+	cp_out (ctx, 0x0060000b); /* ?? */
+	cp_wait(ctx, UNK57, CLEAR);
+	cp_out (ctx, 0x0060000c); /* ?? */
+	cp_bra (ctx, ALWAYS, TRUE, cp_swap_state);
+
+	/* setup for context save */
+	cp_name(ctx, cp_setup_save);
+	cp_set (ctx, SWAP_DIRECTION, SAVE);
+
+	/* general PGRAPH state */
+	cp_name(ctx, cp_swap_state);
+	cp_pos (ctx, 0x00020/4);
+	nv40_graph_construct_general(ctx);
+	cp_wait(ctx, STATUS, IDLE);
+
+	/* 3D state, block 1 */
+	cp_bra (ctx, UNK54, CLEAR, cp_prepare_exit);
+	nv40_graph_construct_state3d(ctx);
+	cp_wait(ctx, STATUS, IDLE);
+
+	/* 3D state, block 2 */
+	nv40_graph_construct_state3d_2(ctx);
+
+	/* Some other block of "random" state */
+	nv40_graph_construct_state3d_3(ctx);
+
+	/* Per-vertex shader state */
+	cp_pos (ctx, ctx->ctxvals_pos);
+	nv40_graph_construct_shader(ctx);
+
+	/* pre-exit state updates */
+	cp_name(ctx, cp_prepare_exit);
+	cp_bra (ctx, SWAP_DIRECTION, SAVE, cp_check_load);
+	cp_bra (ctx, USER_SAVE, PENDING, cp_exit);
+	cp_out (ctx, CP_NEXT_TO_CURRENT);
+
+	cp_name(ctx, cp_exit);
+	cp_set (ctx, USER_SAVE, NOT_PENDING);
+	cp_set (ctx, USER_LOAD, NOT_PENDING);
+	cp_out (ctx, CP_END);
+}
+
diff --git a/drivers/gpu/drm/nouveau/nv40_mc.c b/drivers/gpu/drm/nouveau/nv40_mc.c
new file mode 100644
index 000000000000..2a3495e848e9
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv40_mc.c
@@ -0,0 +1,38 @@
+#include "drmP.h"
+#include "drm.h"
+#include "nouveau_drv.h"
+#include "nouveau_drm.h"
+
+int
+nv40_mc_init(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t tmp;
+
+	/* Power up everything, resetting each individual unit will
+	 * be done later if needed.
+	 */
+	nv_wr32(dev, NV03_PMC_ENABLE, 0xFFFFFFFF);
+
+	switch (dev_priv->chipset) {
+	case 0x44:
+	case 0x46: /* G72 */
+	case 0x4e:
+	case 0x4c: /* C51_G7X */
+		tmp = nv_rd32(dev, NV40_PFB_020C);
+		nv_wr32(dev, NV40_PMC_1700, tmp);
+		nv_wr32(dev, NV40_PMC_1704, 0);
+		nv_wr32(dev, NV40_PMC_1708, 0);
+		nv_wr32(dev, NV40_PMC_170C, tmp);
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+void
+nv40_mc_takedown(struct drm_device *dev)
+{
+}
diff --git a/drivers/gpu/drm/nouveau/nv50_crtc.c b/drivers/gpu/drm/nouveau/nv50_crtc.c
new file mode 100644
index 000000000000..cfabeb974a56
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv50_crtc.c
@@ -0,0 +1,790 @@
+/*
+ * Copyright (C) 2008 Maarten Maathuis.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "drmP.h"
+#include "drm_mode.h"
+#include "drm_crtc_helper.h"
+
+#define NOUVEAU_DMA_DEBUG (nouveau_reg_debug & NOUVEAU_REG_DEBUG_EVO)
+#include "nouveau_reg.h"
+#include "nouveau_drv.h"
+#include "nouveau_hw.h"
+#include "nouveau_encoder.h"
+#include "nouveau_crtc.h"
+#include "nouveau_fb.h"
+#include "nouveau_connector.h"
+#include "nv50_display.h"
+
+static void
+nv50_crtc_lut_load(struct drm_crtc *crtc)
+{
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+	void __iomem *lut = nvbo_kmap_obj_iovirtual(nv_crtc->lut.nvbo);
+	int i;
+
+	NV_DEBUG_KMS(crtc->dev, "\n");
+
+	for (i = 0; i < 256; i++) {
+		writew(nv_crtc->lut.r[i] >> 2, lut + 8*i + 0);
+		writew(nv_crtc->lut.g[i] >> 2, lut + 8*i + 2);
+		writew(nv_crtc->lut.b[i] >> 2, lut + 8*i + 4);
+	}
+
+	if (nv_crtc->lut.depth == 30) {
+		writew(nv_crtc->lut.r[i - 1] >> 2, lut + 8*i + 0);
+		writew(nv_crtc->lut.g[i - 1] >> 2, lut + 8*i + 2);
+		writew(nv_crtc->lut.b[i - 1] >> 2, lut + 8*i + 4);
+	}
+}
+
+int
+nv50_crtc_blank(struct nouveau_crtc *nv_crtc, bool blanked)
+{
+	struct drm_device *dev = nv_crtc->base.dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_channel *evo = dev_priv->evo;
+	int index = nv_crtc->index, ret;
+
+	NV_DEBUG_KMS(dev, "index %d\n", nv_crtc->index);
+	NV_DEBUG_KMS(dev, "%s\n", blanked ? "blanked" : "unblanked");
+
+	if (blanked) {
+		nv_crtc->cursor.hide(nv_crtc, false);
+
+		ret = RING_SPACE(evo, dev_priv->chipset != 0x50 ? 7 : 5);
+		if (ret) {
+			NV_ERROR(dev, "no space while blanking crtc\n");
+			return ret;
+		}
+		BEGIN_RING(evo, 0, NV50_EVO_CRTC(index, CLUT_MODE), 2);
+		OUT_RING(evo, NV50_EVO_CRTC_CLUT_MODE_BLANK);
+		OUT_RING(evo, 0);
+		if (dev_priv->chipset != 0x50) {
+			BEGIN_RING(evo, 0, NV84_EVO_CRTC(index, CLUT_DMA), 1);
+			OUT_RING(evo, NV84_EVO_CRTC_CLUT_DMA_HANDLE_NONE);
+		}
+
+		BEGIN_RING(evo, 0, NV50_EVO_CRTC(index, FB_DMA), 1);
+		OUT_RING(evo, NV50_EVO_CRTC_FB_DMA_HANDLE_NONE);
+	} else {
+		if (nv_crtc->cursor.visible)
+			nv_crtc->cursor.show(nv_crtc, false);
+		else
+			nv_crtc->cursor.hide(nv_crtc, false);
+
+		ret = RING_SPACE(evo, dev_priv->chipset != 0x50 ? 10 : 8);
+		if (ret) {
+			NV_ERROR(dev, "no space while unblanking crtc\n");
+			return ret;
+		}
+		BEGIN_RING(evo, 0, NV50_EVO_CRTC(index, CLUT_MODE), 2);
+		OUT_RING(evo, nv_crtc->lut.depth == 8 ?
+				NV50_EVO_CRTC_CLUT_MODE_OFF :
+				NV50_EVO_CRTC_CLUT_MODE_ON);
+		OUT_RING(evo, (nv_crtc->lut.nvbo->bo.mem.mm_node->start <<
+				 PAGE_SHIFT) >> 8);
+		if (dev_priv->chipset != 0x50) {
+			BEGIN_RING(evo, 0, NV84_EVO_CRTC(index, CLUT_DMA), 1);
+			OUT_RING(evo, NvEvoVRAM);
+		}
+
+		BEGIN_RING(evo, 0, NV50_EVO_CRTC(index, FB_OFFSET), 2);
+		OUT_RING(evo, nv_crtc->fb.offset >> 8);
+		OUT_RING(evo, 0);
+		BEGIN_RING(evo, 0, NV50_EVO_CRTC(index, FB_DMA), 1);
+		if (dev_priv->chipset != 0x50)
+			if (nv_crtc->fb.tile_flags == 0x7a00)
+				OUT_RING(evo, NvEvoFB32);
+			else
+			if (nv_crtc->fb.tile_flags == 0x7000)
+				OUT_RING(evo, NvEvoFB16);
+			else
+				OUT_RING(evo, NvEvoVRAM);
+		else
+			OUT_RING(evo, NvEvoVRAM);
+	}
+
+	nv_crtc->fb.blanked = blanked;
+	return 0;
+}
+
+static int
+nv50_crtc_set_dither(struct nouveau_crtc *nv_crtc, bool on, bool update)
+{
+	struct drm_device *dev = nv_crtc->base.dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_channel *evo = dev_priv->evo;
+	int ret;
+
+	NV_DEBUG_KMS(dev, "\n");
+
+	ret = RING_SPACE(evo, 2 + (update ? 2 : 0));
+	if (ret) {
+		NV_ERROR(dev, "no space while setting dither\n");
+		return ret;
+	}
+
+	BEGIN_RING(evo, 0, NV50_EVO_CRTC(nv_crtc->index, DITHER_CTRL), 1);
+	if (on)
+		OUT_RING(evo, NV50_EVO_CRTC_DITHER_CTRL_ON);
+	else
+		OUT_RING(evo, NV50_EVO_CRTC_DITHER_CTRL_OFF);
+
+	if (update) {
+		BEGIN_RING(evo, 0, NV50_EVO_UPDATE, 1);
+		OUT_RING(evo, 0);
+		FIRE_RING(evo);
+	}
+
+	return 0;
+}
+
+struct nouveau_connector *
+nouveau_crtc_connector_get(struct nouveau_crtc *nv_crtc)
+{
+	struct drm_device *dev = nv_crtc->base.dev;
+	struct drm_connector *connector;
+	struct drm_crtc *crtc = to_drm_crtc(nv_crtc);
+
+	/* The safest approach is to find an encoder with the right crtc, that
+	 * is also linked to a connector. */
+	list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
+		if (connector->encoder)
+			if (connector->encoder->crtc == crtc)
+				return nouveau_connector(connector);
+	}
+
+	return NULL;
+}
+
+static int
+nv50_crtc_set_scale(struct nouveau_crtc *nv_crtc, int scaling_mode, bool update)
+{
+	struct nouveau_connector *nv_connector =
+		nouveau_crtc_connector_get(nv_crtc);
+	struct drm_device *dev = nv_crtc->base.dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_channel *evo = dev_priv->evo;
+	struct drm_display_mode *native_mode = NULL;
+	struct drm_display_mode *mode = &nv_crtc->base.mode;
+	uint32_t outX, outY, horiz, vert;
+	int ret;
+
+	NV_DEBUG_KMS(dev, "\n");
+
+	switch (scaling_mode) {
+	case DRM_MODE_SCALE_NONE:
+		break;
+	default:
+		if (!nv_connector || !nv_connector->native_mode) {
+			NV_ERROR(dev, "No native mode, forcing panel scaling\n");
+			scaling_mode = DRM_MODE_SCALE_NONE;
+		} else {
+			native_mode = nv_connector->native_mode;
+		}
+		break;
+	}
+
+	switch (scaling_mode) {
+	case DRM_MODE_SCALE_ASPECT:
+		horiz = (native_mode->hdisplay << 19) / mode->hdisplay;
+		vert = (native_mode->vdisplay << 19) / mode->vdisplay;
+
+		if (vert > horiz) {
+			outX = (mode->hdisplay * horiz) >> 19;
+			outY = (mode->vdisplay * horiz) >> 19;
+		} else {
+			outX = (mode->hdisplay * vert) >> 19;
+			outY = (mode->vdisplay * vert) >> 19;
+		}
+		break;
+	case DRM_MODE_SCALE_FULLSCREEN:
+		outX = native_mode->hdisplay;
+		outY = native_mode->vdisplay;
+		break;
+	case DRM_MODE_SCALE_CENTER:
+	case DRM_MODE_SCALE_NONE:
+	default:
+		outX = mode->hdisplay;
+		outY = mode->vdisplay;
+		break;
+	}
+
+	ret = RING_SPACE(evo, update ? 7 : 5);
+	if (ret)
+		return ret;
+
+	/* Got a better name for SCALER_ACTIVE? */
+	/* One day i've got to really figure out why this is needed. */
+	BEGIN_RING(evo, 0, NV50_EVO_CRTC(nv_crtc->index, SCALE_CTRL), 1);
+	if ((mode->flags & DRM_MODE_FLAG_DBLSCAN) ||
+	    (mode->flags & DRM_MODE_FLAG_INTERLACE) ||
+	    mode->hdisplay != outX || mode->vdisplay != outY) {
+		OUT_RING(evo, NV50_EVO_CRTC_SCALE_CTRL_ACTIVE);
+	} else {
+		OUT_RING(evo, NV50_EVO_CRTC_SCALE_CTRL_INACTIVE);
+	}
+
+	BEGIN_RING(evo, 0, NV50_EVO_CRTC(nv_crtc->index, SCALE_RES1), 2);
+	OUT_RING(evo, outY << 16 | outX);
+	OUT_RING(evo, outY << 16 | outX);
+
+	if (update) {
+		BEGIN_RING(evo, 0, NV50_EVO_UPDATE, 1);
+		OUT_RING(evo, 0);
+		FIRE_RING(evo);
+	}
+
+	return 0;
+}
+
+int
+nv50_crtc_set_clock(struct drm_device *dev, int head, int pclk)
+{
+	uint32_t pll_reg = NV50_PDISPLAY_CRTC_CLK_CTRL1(head);
+	struct nouveau_pll_vals pll;
+	struct pll_lims limits;
+	uint32_t reg1, reg2;
+	int ret;
+
+	ret = get_pll_limits(dev, pll_reg, &limits);
+	if (ret)
+		return ret;
+
+	ret = nouveau_calc_pll_mnp(dev, &limits, pclk, &pll);
+	if (ret <= 0)
+		return ret;
+
+	if (limits.vco2.maxfreq) {
+		reg1 = nv_rd32(dev, pll_reg + 4) & 0xff00ff00;
+		reg2 = nv_rd32(dev, pll_reg + 8) & 0x8000ff00;
+		nv_wr32(dev, pll_reg, 0x10000611);
+		nv_wr32(dev, pll_reg + 4, reg1 | (pll.M1 << 16) | pll.N1);
+		nv_wr32(dev, pll_reg + 8,
+			reg2 | (pll.log2P << 28) | (pll.M2 << 16) | pll.N2);
+	} else {
+		reg1 = nv_rd32(dev, pll_reg + 4) & 0xffc00000;
+		nv_wr32(dev, pll_reg, 0x50000610);
+		nv_wr32(dev, pll_reg + 4, reg1 |
+			(pll.log2P << 16) | (pll.M1 << 8) | pll.N1);
+	}
+
+	return 0;
+}
+
+static void
+nv50_crtc_destroy(struct drm_crtc *crtc)
+{
+	struct drm_device *dev;
+	struct nouveau_crtc *nv_crtc;
+
+	if (!crtc)
+		return;
+
+	dev = crtc->dev;
+	nv_crtc = nouveau_crtc(crtc);
+
+	NV_DEBUG_KMS(dev, "\n");
+
+	drm_crtc_cleanup(&nv_crtc->base);
+
+	nv50_cursor_fini(nv_crtc);
+
+	nouveau_bo_ref(NULL, &nv_crtc->lut.nvbo);
+	nouveau_bo_ref(NULL, &nv_crtc->cursor.nvbo);
+	kfree(nv_crtc->mode);
+	kfree(nv_crtc);
+}
+
+int
+nv50_crtc_cursor_set(struct drm_crtc *crtc, struct drm_file *file_priv,
+		     uint32_t buffer_handle, uint32_t width, uint32_t height)
+{
+	struct drm_device *dev = crtc->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+	struct nouveau_bo *cursor = NULL;
+	struct drm_gem_object *gem;
+	int ret = 0, i;
+
+	if (width != 64 || height != 64)
+		return -EINVAL;
+
+	if (!buffer_handle) {
+		nv_crtc->cursor.hide(nv_crtc, true);
+		return 0;
+	}
+
+	gem = drm_gem_object_lookup(dev, file_priv, buffer_handle);
+	if (!gem)
+		return -EINVAL;
+	cursor = nouveau_gem_object(gem);
+
+	ret = nouveau_bo_map(cursor);
+	if (ret)
+		goto out;
+
+	/* The simple will do for now. */
+	for (i = 0; i < 64 * 64; i++)
+		nouveau_bo_wr32(nv_crtc->cursor.nvbo, i, nouveau_bo_rd32(cursor, i));
+
+	nouveau_bo_unmap(cursor);
+
+	nv_crtc->cursor.set_offset(nv_crtc, nv_crtc->cursor.nvbo->bo.offset -
+					    dev_priv->vm_vram_base);
+	nv_crtc->cursor.show(nv_crtc, true);
+
+out:
+	drm_gem_object_unreference_unlocked(gem);
+	return ret;
+}
+
+int
+nv50_crtc_cursor_move(struct drm_crtc *crtc, int x, int y)
+{
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+
+	nv_crtc->cursor.set_pos(nv_crtc, x, y);
+	return 0;
+}
+
+static void
+nv50_crtc_gamma_set(struct drm_crtc *crtc, u16 *r, u16 *g, u16 *b,
+		    uint32_t size)
+{
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+	int i;
+
+	if (size != 256)
+		return;
+
+	for (i = 0; i < 256; i++) {
+		nv_crtc->lut.r[i] = r[i];
+		nv_crtc->lut.g[i] = g[i];
+		nv_crtc->lut.b[i] = b[i];
+	}
+
+	/* We need to know the depth before we upload, but it's possible to
+	 * get called before a framebuffer is bound.  If this is the case,
+	 * mark the lut values as dirty by setting depth==0, and it'll be
+	 * uploaded on the first mode_set_base()
+	 */
+	if (!nv_crtc->base.fb) {
+		nv_crtc->lut.depth = 0;
+		return;
+	}
+
+	nv50_crtc_lut_load(crtc);
+}
+
+static void
+nv50_crtc_save(struct drm_crtc *crtc)
+{
+	NV_ERROR(crtc->dev, "!!\n");
+}
+
+static void
+nv50_crtc_restore(struct drm_crtc *crtc)
+{
+	NV_ERROR(crtc->dev, "!!\n");
+}
+
+static const struct drm_crtc_funcs nv50_crtc_funcs = {
+	.save = nv50_crtc_save,
+	.restore = nv50_crtc_restore,
+	.cursor_set = nv50_crtc_cursor_set,
+	.cursor_move = nv50_crtc_cursor_move,
+	.gamma_set = nv50_crtc_gamma_set,
+	.set_config = drm_crtc_helper_set_config,
+	.destroy = nv50_crtc_destroy,
+};
+
+static void
+nv50_crtc_dpms(struct drm_crtc *crtc, int mode)
+{
+}
+
+static void
+nv50_crtc_prepare(struct drm_crtc *crtc)
+{
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+	struct drm_device *dev = crtc->dev;
+	struct drm_encoder *encoder;
+	uint32_t dac = 0, sor = 0;
+
+	NV_DEBUG_KMS(dev, "index %d\n", nv_crtc->index);
+
+	/* Disconnect all unused encoders. */
+	list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) {
+		struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+
+		if (!drm_helper_encoder_in_use(encoder))
+			continue;
+
+		if (nv_encoder->dcb->type == OUTPUT_ANALOG ||
+		    nv_encoder->dcb->type == OUTPUT_TV)
+			dac |= (1 << nv_encoder->or);
+		else
+			sor |= (1 << nv_encoder->or);
+	}
+
+	list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) {
+		struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+
+		if (nv_encoder->dcb->type == OUTPUT_ANALOG ||
+		    nv_encoder->dcb->type == OUTPUT_TV) {
+			if (dac & (1 << nv_encoder->or))
+				continue;
+		} else {
+			if (sor & (1 << nv_encoder->or))
+				continue;
+		}
+
+		nv_encoder->disconnect(nv_encoder);
+	}
+
+	nv50_crtc_blank(nv_crtc, true);
+}
+
+static void
+nv50_crtc_commit(struct drm_crtc *crtc)
+{
+	struct drm_crtc *crtc2;
+	struct drm_device *dev = crtc->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_channel *evo = dev_priv->evo;
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+	int ret;
+
+	NV_DEBUG_KMS(dev, "index %d\n", nv_crtc->index);
+
+	nv50_crtc_blank(nv_crtc, false);
+
+	/* Explicitly blank all unused crtc's. */
+	list_for_each_entry(crtc2, &dev->mode_config.crtc_list, head) {
+		if (!drm_helper_crtc_in_use(crtc2))
+			nv50_crtc_blank(nouveau_crtc(crtc2), true);
+	}
+
+	ret = RING_SPACE(evo, 2);
+	if (ret) {
+		NV_ERROR(dev, "no space while committing crtc\n");
+		return;
+	}
+	BEGIN_RING(evo, 0, NV50_EVO_UPDATE, 1);
+	OUT_RING(evo, 0);
+	FIRE_RING(evo);
+}
+
+static bool
+nv50_crtc_mode_fixup(struct drm_crtc *crtc, struct drm_display_mode *mode,
+		     struct drm_display_mode *adjusted_mode)
+{
+	return true;
+}
+
+static int
+nv50_crtc_do_mode_set_base(struct drm_crtc *crtc, int x, int y,
+			   struct drm_framebuffer *old_fb, bool update)
+{
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+	struct drm_device *dev = nv_crtc->base.dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_channel *evo = dev_priv->evo;
+	struct drm_framebuffer *drm_fb = nv_crtc->base.fb;
+	struct nouveau_framebuffer *fb = nouveau_framebuffer(drm_fb);
+	int ret, format;
+
+	NV_DEBUG_KMS(dev, "index %d\n", nv_crtc->index);
+
+	switch (drm_fb->depth) {
+	case  8:
+		format = NV50_EVO_CRTC_FB_DEPTH_8;
+		break;
+	case 15:
+		format = NV50_EVO_CRTC_FB_DEPTH_15;
+		break;
+	case 16:
+		format = NV50_EVO_CRTC_FB_DEPTH_16;
+		break;
+	case 24:
+	case 32:
+		format = NV50_EVO_CRTC_FB_DEPTH_24;
+		break;
+	case 30:
+		format = NV50_EVO_CRTC_FB_DEPTH_30;
+		break;
+	default:
+		 NV_ERROR(dev, "unknown depth %d\n", drm_fb->depth);
+		 return -EINVAL;
+	}
+
+	ret = nouveau_bo_pin(fb->nvbo, TTM_PL_FLAG_VRAM);
+	if (ret)
+		return ret;
+
+	if (old_fb) {
+		struct nouveau_framebuffer *ofb = nouveau_framebuffer(old_fb);
+		nouveau_bo_unpin(ofb->nvbo);
+	}
+
+	nv_crtc->fb.offset = fb->nvbo->bo.offset - dev_priv->vm_vram_base;
+	nv_crtc->fb.tile_flags = fb->nvbo->tile_flags;
+	nv_crtc->fb.cpp = drm_fb->bits_per_pixel / 8;
+	if (!nv_crtc->fb.blanked && dev_priv->chipset != 0x50) {
+		ret = RING_SPACE(evo, 2);
+		if (ret)
+			return ret;
+
+		BEGIN_RING(evo, 0, NV50_EVO_CRTC(nv_crtc->index, FB_DMA), 1);
+		if (nv_crtc->fb.tile_flags == 0x7a00)
+			OUT_RING(evo, NvEvoFB32);
+		else
+		if (nv_crtc->fb.tile_flags == 0x7000)
+			OUT_RING(evo, NvEvoFB16);
+		else
+			OUT_RING(evo, NvEvoVRAM);
+	}
+
+	ret = RING_SPACE(evo, 12);
+	if (ret)
+		return ret;
+
+	BEGIN_RING(evo, 0, NV50_EVO_CRTC(nv_crtc->index, FB_OFFSET), 5);
+	OUT_RING(evo, nv_crtc->fb.offset >> 8);
+	OUT_RING(evo, 0);
+	OUT_RING(evo, (drm_fb->height << 16) | drm_fb->width);
+	if (!nv_crtc->fb.tile_flags) {
+		OUT_RING(evo, drm_fb->pitch | (1 << 20));
+	} else {
+		OUT_RING(evo, ((drm_fb->pitch / 4) << 4) |
+				  fb->nvbo->tile_mode);
+	}
+	if (dev_priv->chipset == 0x50)
+		OUT_RING(evo, (fb->nvbo->tile_flags << 8) | format);
+	else
+		OUT_RING(evo, format);
+
+	BEGIN_RING(evo, 0, NV50_EVO_CRTC(nv_crtc->index, CLUT_MODE), 1);
+	OUT_RING(evo, fb->base.depth == 8 ?
+		 NV50_EVO_CRTC_CLUT_MODE_OFF : NV50_EVO_CRTC_CLUT_MODE_ON);
+
+	BEGIN_RING(evo, 0, NV50_EVO_CRTC(nv_crtc->index, COLOR_CTRL), 1);
+	OUT_RING(evo, NV50_EVO_CRTC_COLOR_CTRL_COLOR);
+	BEGIN_RING(evo, 0, NV50_EVO_CRTC(nv_crtc->index, FB_POS), 1);
+	OUT_RING(evo, (y << 16) | x);
+
+	if (nv_crtc->lut.depth != fb->base.depth) {
+		nv_crtc->lut.depth = fb->base.depth;
+		nv50_crtc_lut_load(crtc);
+	}
+
+	if (update) {
+		ret = RING_SPACE(evo, 2);
+		if (ret)
+			return ret;
+		BEGIN_RING(evo, 0, NV50_EVO_UPDATE, 1);
+		OUT_RING(evo, 0);
+		FIRE_RING(evo);
+	}
+
+	return 0;
+}
+
+static int
+nv50_crtc_mode_set(struct drm_crtc *crtc, struct drm_display_mode *mode,
+		   struct drm_display_mode *adjusted_mode, int x, int y,
+		   struct drm_framebuffer *old_fb)
+{
+	struct drm_device *dev = crtc->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_channel *evo = dev_priv->evo;
+	struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+	struct nouveau_connector *nv_connector = NULL;
+	uint32_t hsync_dur,  vsync_dur, hsync_start_to_end, vsync_start_to_end;
+	uint32_t hunk1, vunk1, vunk2a, vunk2b;
+	int ret;
+
+	/* Find the connector attached to this CRTC */
+	nv_connector = nouveau_crtc_connector_get(nv_crtc);
+
+	*nv_crtc->mode = *adjusted_mode;
+
+	NV_DEBUG_KMS(dev, "index %d\n", nv_crtc->index);
+
+	hsync_dur = adjusted_mode->hsync_end - adjusted_mode->hsync_start;
+	vsync_dur = adjusted_mode->vsync_end - adjusted_mode->vsync_start;
+	hsync_start_to_end = adjusted_mode->htotal - adjusted_mode->hsync_start;
+	vsync_start_to_end = adjusted_mode->vtotal - adjusted_mode->vsync_start;
+	/* I can't give this a proper name, anyone else can? */
+	hunk1 = adjusted_mode->htotal -
+		adjusted_mode->hsync_start + adjusted_mode->hdisplay;
+	vunk1 = adjusted_mode->vtotal -
+		adjusted_mode->vsync_start + adjusted_mode->vdisplay;
+	/* Another strange value, this time only for interlaced adjusted_modes. */
+	vunk2a = 2 * adjusted_mode->vtotal -
+		 adjusted_mode->vsync_start + adjusted_mode->vdisplay;
+	vunk2b = adjusted_mode->vtotal -
+		 adjusted_mode->vsync_start + adjusted_mode->vtotal;
+
+	if (adjusted_mode->flags & DRM_MODE_FLAG_INTERLACE) {
+		vsync_dur /= 2;
+		vsync_start_to_end  /= 2;
+		vunk1 /= 2;
+		vunk2a /= 2;
+		vunk2b /= 2;
+		/* magic */
+		if (adjusted_mode->flags & DRM_MODE_FLAG_DBLSCAN) {
+			vsync_start_to_end -= 1;
+			vunk1 -= 1;
+			vunk2a -= 1;
+			vunk2b -= 1;
+		}
+	}
+
+	ret = RING_SPACE(evo, 17);
+	if (ret)
+		return ret;
+
+	BEGIN_RING(evo, 0, NV50_EVO_CRTC(nv_crtc->index, CLOCK), 2);
+	OUT_RING(evo, adjusted_mode->clock | 0x800000);
+	OUT_RING(evo, (adjusted_mode->flags & DRM_MODE_FLAG_INTERLACE) ? 2 : 0);
+
+	BEGIN_RING(evo, 0, NV50_EVO_CRTC(nv_crtc->index, DISPLAY_START), 5);
+	OUT_RING(evo, 0);
+	OUT_RING(evo, (adjusted_mode->vtotal << 16) | adjusted_mode->htotal);
+	OUT_RING(evo, (vsync_dur - 1) << 16 | (hsync_dur - 1));
+	OUT_RING(evo, (vsync_start_to_end - 1) << 16 |
+			(hsync_start_to_end - 1));
+	OUT_RING(evo, (vunk1 - 1) << 16 | (hunk1 - 1));
+
+	if (adjusted_mode->flags & DRM_MODE_FLAG_INTERLACE) {
+		BEGIN_RING(evo, 0, NV50_EVO_CRTC(nv_crtc->index, UNK0824), 1);
+		OUT_RING(evo, (vunk2b - 1) << 16 | (vunk2a - 1));
+	} else {
+		OUT_RING(evo, 0);
+		OUT_RING(evo, 0);
+	}
+
+	BEGIN_RING(evo, 0, NV50_EVO_CRTC(nv_crtc->index, UNK082C), 1);
+	OUT_RING(evo, 0);
+
+	/* This is the actual resolution of the mode. */
+	BEGIN_RING(evo, 0, NV50_EVO_CRTC(nv_crtc->index, REAL_RES), 1);
+	OUT_RING(evo, (mode->vdisplay << 16) | mode->hdisplay);
+	BEGIN_RING(evo, 0, NV50_EVO_CRTC(nv_crtc->index, SCALE_CENTER_OFFSET), 1);
+	OUT_RING(evo, NV50_EVO_CRTC_SCALE_CENTER_OFFSET_VAL(0, 0));
+
+	nv_crtc->set_dither(nv_crtc, nv_connector->use_dithering, false);
+	nv_crtc->set_scale(nv_crtc, nv_connector->scaling_mode, false);
+
+	return nv50_crtc_do_mode_set_base(crtc, x, y, old_fb, false);
+}
+
+static int
+nv50_crtc_mode_set_base(struct drm_crtc *crtc, int x, int y,
+			struct drm_framebuffer *old_fb)
+{
+	return nv50_crtc_do_mode_set_base(crtc, x, y, old_fb, true);
+}
+
+static const struct drm_crtc_helper_funcs nv50_crtc_helper_funcs = {
+	.dpms = nv50_crtc_dpms,
+	.prepare = nv50_crtc_prepare,
+	.commit = nv50_crtc_commit,
+	.mode_fixup = nv50_crtc_mode_fixup,
+	.mode_set = nv50_crtc_mode_set,
+	.mode_set_base = nv50_crtc_mode_set_base,
+	.load_lut = nv50_crtc_lut_load,
+};
+
+int
+nv50_crtc_create(struct drm_device *dev, int index)
+{
+	struct nouveau_crtc *nv_crtc = NULL;
+	int ret, i;
+
+	NV_DEBUG_KMS(dev, "\n");
+
+	nv_crtc = kzalloc(sizeof(*nv_crtc), GFP_KERNEL);
+	if (!nv_crtc)
+		return -ENOMEM;
+
+	nv_crtc->mode = kzalloc(sizeof(*nv_crtc->mode), GFP_KERNEL);
+	if (!nv_crtc->mode) {
+		kfree(nv_crtc);
+		return -ENOMEM;
+	}
+
+	/* Default CLUT parameters, will be activated on the hw upon
+	 * first mode set.
+	 */
+	for (i = 0; i < 256; i++) {
+		nv_crtc->lut.r[i] = i << 8;
+		nv_crtc->lut.g[i] = i << 8;
+		nv_crtc->lut.b[i] = i << 8;
+	}
+	nv_crtc->lut.depth = 0;
+
+	ret = nouveau_bo_new(dev, NULL, 4096, 0x100, TTM_PL_FLAG_VRAM,
+			     0, 0x0000, false, true, &nv_crtc->lut.nvbo);
+	if (!ret) {
+		ret = nouveau_bo_pin(nv_crtc->lut.nvbo, TTM_PL_FLAG_VRAM);
+		if (!ret)
+			ret = nouveau_bo_map(nv_crtc->lut.nvbo);
+		if (ret)
+			nouveau_bo_ref(NULL, &nv_crtc->lut.nvbo);
+	}
+
+	if (ret) {
+		kfree(nv_crtc->mode);
+		kfree(nv_crtc);
+		return ret;
+	}
+
+	nv_crtc->index = index;
+
+	/* set function pointers */
+	nv_crtc->set_dither = nv50_crtc_set_dither;
+	nv_crtc->set_scale = nv50_crtc_set_scale;
+
+	drm_crtc_init(dev, &nv_crtc->base, &nv50_crtc_funcs);
+	drm_crtc_helper_add(&nv_crtc->base, &nv50_crtc_helper_funcs);
+	drm_mode_crtc_set_gamma_size(&nv_crtc->base, 256);
+
+	ret = nouveau_bo_new(dev, NULL, 64*64*4, 0x100, TTM_PL_FLAG_VRAM,
+			     0, 0x0000, false, true, &nv_crtc->cursor.nvbo);
+	if (!ret) {
+		ret = nouveau_bo_pin(nv_crtc->cursor.nvbo, TTM_PL_FLAG_VRAM);
+		if (!ret)
+			ret = nouveau_bo_map(nv_crtc->cursor.nvbo);
+		if (ret)
+			nouveau_bo_ref(NULL, &nv_crtc->cursor.nvbo);
+	}
+
+	nv50_cursor_init(nv_crtc);
+	return 0;
+}
diff --git a/drivers/gpu/drm/nouveau/nv50_cursor.c b/drivers/gpu/drm/nouveau/nv50_cursor.c
new file mode 100644
index 000000000000..753e723adb3a
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv50_cursor.c
@@ -0,0 +1,156 @@
+/*
+ * Copyright (C) 2008 Maarten Maathuis.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "drmP.h"
+#include "drm_mode.h"
+
+#define NOUVEAU_DMA_DEBUG (nouveau_reg_debug & NOUVEAU_REG_DEBUG_EVO)
+#include "nouveau_reg.h"
+#include "nouveau_drv.h"
+#include "nouveau_crtc.h"
+#include "nv50_display.h"
+
+static void
+nv50_cursor_show(struct nouveau_crtc *nv_crtc, bool update)
+{
+	struct drm_nouveau_private *dev_priv = nv_crtc->base.dev->dev_private;
+	struct nouveau_channel *evo = dev_priv->evo;
+	struct drm_device *dev = nv_crtc->base.dev;
+	int ret;
+
+	NV_DEBUG_KMS(dev, "\n");
+
+	if (update && nv_crtc->cursor.visible)
+		return;
+
+	ret = RING_SPACE(evo, (dev_priv->chipset != 0x50 ? 5 : 3) + update * 2);
+	if (ret) {
+		NV_ERROR(dev, "no space while unhiding cursor\n");
+		return;
+	}
+
+	if (dev_priv->chipset != 0x50) {
+		BEGIN_RING(evo, 0, NV84_EVO_CRTC(nv_crtc->index, CURSOR_DMA), 1);
+		OUT_RING(evo, NvEvoVRAM);
+	}
+	BEGIN_RING(evo, 0, NV50_EVO_CRTC(nv_crtc->index, CURSOR_CTRL), 2);
+	OUT_RING(evo, NV50_EVO_CRTC_CURSOR_CTRL_SHOW);
+	OUT_RING(evo, nv_crtc->cursor.offset >> 8);
+
+	if (update) {
+		BEGIN_RING(evo, 0, NV50_EVO_UPDATE, 1);
+		OUT_RING(evo, 0);
+		FIRE_RING(evo);
+		nv_crtc->cursor.visible = true;
+	}
+}
+
+static void
+nv50_cursor_hide(struct nouveau_crtc *nv_crtc, bool update)
+{
+	struct drm_nouveau_private *dev_priv = nv_crtc->base.dev->dev_private;
+	struct nouveau_channel *evo = dev_priv->evo;
+	struct drm_device *dev = nv_crtc->base.dev;
+	int ret;
+
+	NV_DEBUG_KMS(dev, "\n");
+
+	if (update && !nv_crtc->cursor.visible)
+		return;
+
+	ret = RING_SPACE(evo, (dev_priv->chipset != 0x50 ? 5 : 3) + update * 2);
+	if (ret) {
+		NV_ERROR(dev, "no space while hiding cursor\n");
+		return;
+	}
+	BEGIN_RING(evo, 0, NV50_EVO_CRTC(nv_crtc->index, CURSOR_CTRL), 2);
+	OUT_RING(evo, NV50_EVO_CRTC_CURSOR_CTRL_HIDE);
+	OUT_RING(evo, 0);
+	if (dev_priv->chipset != 0x50) {
+		BEGIN_RING(evo, 0, NV84_EVO_CRTC(nv_crtc->index, CURSOR_DMA), 1);
+		OUT_RING(evo, NV84_EVO_CRTC_CURSOR_DMA_HANDLE_NONE);
+	}
+
+	if (update) {
+		BEGIN_RING(evo, 0, NV50_EVO_UPDATE, 1);
+		OUT_RING(evo, 0);
+		FIRE_RING(evo);
+		nv_crtc->cursor.visible = false;
+	}
+}
+
+static void
+nv50_cursor_set_pos(struct nouveau_crtc *nv_crtc, int x, int y)
+{
+	struct drm_device *dev = nv_crtc->base.dev;
+
+	nv_wr32(dev, NV50_PDISPLAY_CURSOR_USER_POS(nv_crtc->index),
+		((y & 0xFFFF) << 16) | (x & 0xFFFF));
+	/* Needed to make the cursor move. */
+	nv_wr32(dev, NV50_PDISPLAY_CURSOR_USER_POS_CTRL(nv_crtc->index), 0);
+}
+
+static void
+nv50_cursor_set_offset(struct nouveau_crtc *nv_crtc, uint32_t offset)
+{
+	NV_DEBUG_KMS(nv_crtc->base.dev, "\n");
+	if (offset == nv_crtc->cursor.offset)
+		return;
+
+	nv_crtc->cursor.offset = offset;
+	if (nv_crtc->cursor.visible) {
+		nv_crtc->cursor.visible = false;
+		nv_crtc->cursor.show(nv_crtc, true);
+	}
+}
+
+int
+nv50_cursor_init(struct nouveau_crtc *nv_crtc)
+{
+	nv_crtc->cursor.set_offset = nv50_cursor_set_offset;
+	nv_crtc->cursor.set_pos = nv50_cursor_set_pos;
+	nv_crtc->cursor.hide = nv50_cursor_hide;
+	nv_crtc->cursor.show = nv50_cursor_show;
+	return 0;
+}
+
+void
+nv50_cursor_fini(struct nouveau_crtc *nv_crtc)
+{
+	struct drm_device *dev = nv_crtc->base.dev;
+	int idx = nv_crtc->index;
+
+	NV_DEBUG_KMS(dev, "\n");
+
+	nv_wr32(dev, NV50_PDISPLAY_CURSOR_CURSOR_CTRL2(idx), 0);
+	if (!nv_wait(NV50_PDISPLAY_CURSOR_CURSOR_CTRL2(idx),
+		     NV50_PDISPLAY_CURSOR_CURSOR_CTRL2_STATUS, 0)) {
+		NV_ERROR(dev, "timeout: CURSOR_CTRL2_STATUS == 0\n");
+		NV_ERROR(dev, "CURSOR_CTRL2 = 0x%08x\n",
+			 nv_rd32(dev, NV50_PDISPLAY_CURSOR_CURSOR_CTRL2(idx)));
+	}
+}
+
diff --git a/drivers/gpu/drm/nouveau/nv50_dac.c b/drivers/gpu/drm/nouveau/nv50_dac.c
new file mode 100644
index 000000000000..1fd9537beff6
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv50_dac.c
@@ -0,0 +1,304 @@
+/*
+ * Copyright (C) 2008 Maarten Maathuis.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "drmP.h"
+#include "drm_crtc_helper.h"
+
+#define NOUVEAU_DMA_DEBUG (nouveau_reg_debug & NOUVEAU_REG_DEBUG_EVO)
+#include "nouveau_reg.h"
+#include "nouveau_drv.h"
+#include "nouveau_dma.h"
+#include "nouveau_encoder.h"
+#include "nouveau_connector.h"
+#include "nouveau_crtc.h"
+#include "nv50_display.h"
+
+static void
+nv50_dac_disconnect(struct nouveau_encoder *nv_encoder)
+{
+	struct drm_device *dev = to_drm_encoder(nv_encoder)->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_channel *evo = dev_priv->evo;
+	int ret;
+
+	NV_DEBUG_KMS(dev, "Disconnecting DAC %d\n", nv_encoder->or);
+
+	ret = RING_SPACE(evo, 2);
+	if (ret) {
+		NV_ERROR(dev, "no space while disconnecting DAC\n");
+		return;
+	}
+	BEGIN_RING(evo, 0, NV50_EVO_DAC(nv_encoder->or, MODE_CTRL), 1);
+	OUT_RING(evo, 0);
+}
+
+static enum drm_connector_status
+nv50_dac_detect(struct drm_encoder *encoder, struct drm_connector *connector)
+{
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	struct drm_device *dev = encoder->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	enum drm_connector_status status = connector_status_disconnected;
+	uint32_t dpms_state, load_pattern, load_state;
+	int or = nv_encoder->or;
+
+	nv_wr32(dev, NV50_PDISPLAY_DAC_CLK_CTRL1(or), 0x00000001);
+	dpms_state = nv_rd32(dev, NV50_PDISPLAY_DAC_DPMS_CTRL(or));
+
+	nv_wr32(dev, NV50_PDISPLAY_DAC_DPMS_CTRL(or),
+		0x00150000 | NV50_PDISPLAY_DAC_DPMS_CTRL_PENDING);
+	if (!nv_wait(NV50_PDISPLAY_DAC_DPMS_CTRL(or),
+		     NV50_PDISPLAY_DAC_DPMS_CTRL_PENDING, 0)) {
+		NV_ERROR(dev, "timeout: DAC_DPMS_CTRL_PENDING(%d) == 0\n", or);
+		NV_ERROR(dev, "DAC_DPMS_CTRL(%d) = 0x%08x\n", or,
+			  nv_rd32(dev, NV50_PDISPLAY_DAC_DPMS_CTRL(or)));
+		return status;
+	}
+
+	/* Use bios provided value if possible. */
+	if (dev_priv->vbios.dactestval) {
+		load_pattern = dev_priv->vbios.dactestval;
+		NV_DEBUG_KMS(dev, "Using bios provided load_pattern of %d\n",
+			  load_pattern);
+	} else {
+		load_pattern = 340;
+		NV_DEBUG_KMS(dev, "Using default load_pattern of %d\n",
+			 load_pattern);
+	}
+
+	nv_wr32(dev, NV50_PDISPLAY_DAC_LOAD_CTRL(or),
+		NV50_PDISPLAY_DAC_LOAD_CTRL_ACTIVE | load_pattern);
+	mdelay(45); /* give it some time to process */
+	load_state = nv_rd32(dev, NV50_PDISPLAY_DAC_LOAD_CTRL(or));
+
+	nv_wr32(dev, NV50_PDISPLAY_DAC_LOAD_CTRL(or), 0);
+	nv_wr32(dev, NV50_PDISPLAY_DAC_DPMS_CTRL(or), dpms_state |
+		NV50_PDISPLAY_DAC_DPMS_CTRL_PENDING);
+
+	if ((load_state & NV50_PDISPLAY_DAC_LOAD_CTRL_PRESENT) ==
+			  NV50_PDISPLAY_DAC_LOAD_CTRL_PRESENT)
+		status = connector_status_connected;
+
+	if (status == connector_status_connected)
+		NV_DEBUG_KMS(dev, "Load was detected on output with or %d\n", or);
+	else
+		NV_DEBUG_KMS(dev, "Load was not detected on output with or %d\n", or);
+
+	return status;
+}
+
+static void
+nv50_dac_dpms(struct drm_encoder *encoder, int mode)
+{
+	struct drm_device *dev = encoder->dev;
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	uint32_t val;
+	int or = nv_encoder->or;
+
+	NV_DEBUG_KMS(dev, "or %d mode %d\n", or, mode);
+
+	/* wait for it to be done */
+	if (!nv_wait(NV50_PDISPLAY_DAC_DPMS_CTRL(or),
+		     NV50_PDISPLAY_DAC_DPMS_CTRL_PENDING, 0)) {
+		NV_ERROR(dev, "timeout: DAC_DPMS_CTRL_PENDING(%d) == 0\n", or);
+		NV_ERROR(dev, "DAC_DPMS_CTRL(%d) = 0x%08x\n", or,
+			 nv_rd32(dev, NV50_PDISPLAY_DAC_DPMS_CTRL(or)));
+		return;
+	}
+
+	val = nv_rd32(dev, NV50_PDISPLAY_DAC_DPMS_CTRL(or)) & ~0x7F;
+
+	if (mode != DRM_MODE_DPMS_ON)
+		val |= NV50_PDISPLAY_DAC_DPMS_CTRL_BLANKED;
+
+	switch (mode) {
+	case DRM_MODE_DPMS_STANDBY:
+		val |= NV50_PDISPLAY_DAC_DPMS_CTRL_HSYNC_OFF;
+		break;
+	case DRM_MODE_DPMS_SUSPEND:
+		val |= NV50_PDISPLAY_DAC_DPMS_CTRL_VSYNC_OFF;
+		break;
+	case DRM_MODE_DPMS_OFF:
+		val |= NV50_PDISPLAY_DAC_DPMS_CTRL_OFF;
+		val |= NV50_PDISPLAY_DAC_DPMS_CTRL_HSYNC_OFF;
+		val |= NV50_PDISPLAY_DAC_DPMS_CTRL_VSYNC_OFF;
+		break;
+	default:
+		break;
+	}
+
+	nv_wr32(dev, NV50_PDISPLAY_DAC_DPMS_CTRL(or), val |
+		NV50_PDISPLAY_DAC_DPMS_CTRL_PENDING);
+}
+
+static void
+nv50_dac_save(struct drm_encoder *encoder)
+{
+	NV_ERROR(encoder->dev, "!!\n");
+}
+
+static void
+nv50_dac_restore(struct drm_encoder *encoder)
+{
+	NV_ERROR(encoder->dev, "!!\n");
+}
+
+static bool
+nv50_dac_mode_fixup(struct drm_encoder *encoder, struct drm_display_mode *mode,
+		    struct drm_display_mode *adjusted_mode)
+{
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	struct nouveau_connector *connector;
+
+	NV_DEBUG_KMS(encoder->dev, "or %d\n", nv_encoder->or);
+
+	connector = nouveau_encoder_connector_get(nv_encoder);
+	if (!connector) {
+		NV_ERROR(encoder->dev, "Encoder has no connector\n");
+		return false;
+	}
+
+	if (connector->scaling_mode != DRM_MODE_SCALE_NONE &&
+	     connector->native_mode) {
+		int id = adjusted_mode->base.id;
+		*adjusted_mode = *connector->native_mode;
+		adjusted_mode->base.id = id;
+	}
+
+	return true;
+}
+
+static void
+nv50_dac_prepare(struct drm_encoder *encoder)
+{
+}
+
+static void
+nv50_dac_commit(struct drm_encoder *encoder)
+{
+}
+
+static void
+nv50_dac_mode_set(struct drm_encoder *encoder, struct drm_display_mode *mode,
+		  struct drm_display_mode *adjusted_mode)
+{
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	struct drm_device *dev = encoder->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_channel *evo = dev_priv->evo;
+	struct nouveau_crtc *crtc = nouveau_crtc(encoder->crtc);
+	uint32_t mode_ctl = 0, mode_ctl2 = 0;
+	int ret;
+
+	NV_DEBUG_KMS(dev, "or %d\n", nv_encoder->or);
+
+	nv50_dac_dpms(encoder, DRM_MODE_DPMS_ON);
+
+	if (crtc->index == 1)
+		mode_ctl |= NV50_EVO_DAC_MODE_CTRL_CRTC1;
+	else
+		mode_ctl |= NV50_EVO_DAC_MODE_CTRL_CRTC0;
+
+	/* Lacking a working tv-out, this is not a 100% sure. */
+	if (nv_encoder->dcb->type == OUTPUT_ANALOG)
+		mode_ctl |= 0x40;
+	else
+	if (nv_encoder->dcb->type == OUTPUT_TV)
+		mode_ctl |= 0x100;
+
+	if (adjusted_mode->flags & DRM_MODE_FLAG_NHSYNC)
+		mode_ctl2 |= NV50_EVO_DAC_MODE_CTRL2_NHSYNC;
+
+	if (adjusted_mode->flags & DRM_MODE_FLAG_NVSYNC)
+		mode_ctl2 |= NV50_EVO_DAC_MODE_CTRL2_NVSYNC;
+
+	ret = RING_SPACE(evo, 3);
+	if (ret) {
+		NV_ERROR(dev, "no space while connecting DAC\n");
+		return;
+	}
+	BEGIN_RING(evo, 0, NV50_EVO_DAC(nv_encoder->or, MODE_CTRL), 2);
+	OUT_RING(evo, mode_ctl);
+	OUT_RING(evo, mode_ctl2);
+}
+
+static const struct drm_encoder_helper_funcs nv50_dac_helper_funcs = {
+	.dpms = nv50_dac_dpms,
+	.save = nv50_dac_save,
+	.restore = nv50_dac_restore,
+	.mode_fixup = nv50_dac_mode_fixup,
+	.prepare = nv50_dac_prepare,
+	.commit = nv50_dac_commit,
+	.mode_set = nv50_dac_mode_set,
+	.detect = nv50_dac_detect
+};
+
+static void
+nv50_dac_destroy(struct drm_encoder *encoder)
+{
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+
+	if (!encoder)
+		return;
+
+	NV_DEBUG_KMS(encoder->dev, "\n");
+
+	drm_encoder_cleanup(encoder);
+	kfree(nv_encoder);
+}
+
+static const struct drm_encoder_funcs nv50_dac_encoder_funcs = {
+	.destroy = nv50_dac_destroy,
+};
+
+int
+nv50_dac_create(struct drm_device *dev, struct dcb_entry *entry)
+{
+	struct nouveau_encoder *nv_encoder;
+	struct drm_encoder *encoder;
+
+	NV_DEBUG_KMS(dev, "\n");
+	NV_INFO(dev, "Detected a DAC output\n");
+
+	nv_encoder = kzalloc(sizeof(*nv_encoder), GFP_KERNEL);
+	if (!nv_encoder)
+		return -ENOMEM;
+	encoder = to_drm_encoder(nv_encoder);
+
+	nv_encoder->dcb = entry;
+	nv_encoder->or = ffs(entry->or) - 1;
+
+	nv_encoder->disconnect = nv50_dac_disconnect;
+
+	drm_encoder_init(dev, encoder, &nv50_dac_encoder_funcs,
+			 DRM_MODE_ENCODER_DAC);
+	drm_encoder_helper_add(encoder, &nv50_dac_helper_funcs);
+
+	encoder->possible_crtcs = entry->heads;
+	encoder->possible_clones = 0;
+	return 0;
+}
+
diff --git a/drivers/gpu/drm/nouveau/nv50_display.c b/drivers/gpu/drm/nouveau/nv50_display.c
new file mode 100644
index 000000000000..61a89f2dc553
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv50_display.c
@@ -0,0 +1,998 @@
+/*
+ * Copyright (C) 2008 Maarten Maathuis.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "nv50_display.h"
+#include "nouveau_crtc.h"
+#include "nouveau_encoder.h"
+#include "nouveau_connector.h"
+#include "nouveau_fb.h"
+#include "drm_crtc_helper.h"
+
+static void
+nv50_evo_channel_del(struct nouveau_channel **pchan)
+{
+	struct nouveau_channel *chan = *pchan;
+
+	if (!chan)
+		return;
+	*pchan = NULL;
+
+	nouveau_gpuobj_channel_takedown(chan);
+	nouveau_bo_ref(NULL, &chan->pushbuf_bo);
+
+	if (chan->user)
+		iounmap(chan->user);
+
+	kfree(chan);
+}
+
+static int
+nv50_evo_dmaobj_new(struct nouveau_channel *evo, uint32_t class, uint32_t name,
+		    uint32_t tile_flags, uint32_t magic_flags,
+		    uint32_t offset, uint32_t limit)
+{
+	struct drm_nouveau_private *dev_priv = evo->dev->dev_private;
+	struct drm_device *dev = evo->dev;
+	struct nouveau_gpuobj *obj = NULL;
+	int ret;
+
+	ret = nouveau_gpuobj_new(dev, evo, 6*4, 32, 0, &obj);
+	if (ret)
+		return ret;
+	obj->engine = NVOBJ_ENGINE_DISPLAY;
+
+	ret = nouveau_gpuobj_ref_add(dev, evo, name, obj, NULL);
+	if (ret) {
+		nouveau_gpuobj_del(dev, &obj);
+		return ret;
+	}
+
+	dev_priv->engine.instmem.prepare_access(dev, true);
+	nv_wo32(dev, obj, 0, (tile_flags << 22) | (magic_flags << 16) | class);
+	nv_wo32(dev, obj, 1, limit);
+	nv_wo32(dev, obj, 2, offset);
+	nv_wo32(dev, obj, 3, 0x00000000);
+	nv_wo32(dev, obj, 4, 0x00000000);
+	nv_wo32(dev, obj, 5, 0x00010000);
+	dev_priv->engine.instmem.finish_access(dev);
+
+	return 0;
+}
+
+static int
+nv50_evo_channel_new(struct drm_device *dev, struct nouveau_channel **pchan)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_channel *chan;
+	int ret;
+
+	chan = kzalloc(sizeof(struct nouveau_channel), GFP_KERNEL);
+	if (!chan)
+		return -ENOMEM;
+	*pchan = chan;
+
+	chan->id = -1;
+	chan->dev = dev;
+	chan->user_get = 4;
+	chan->user_put = 0;
+
+	INIT_LIST_HEAD(&chan->ramht_refs);
+
+	ret = nouveau_gpuobj_new_ref(dev, NULL, NULL, 0, 32768, 0x1000,
+				     NVOBJ_FLAG_ZERO_ALLOC, &chan->ramin);
+	if (ret) {
+		NV_ERROR(dev, "Error allocating EVO channel memory: %d\n", ret);
+		nv50_evo_channel_del(pchan);
+		return ret;
+	}
+
+	ret = nouveau_mem_init_heap(&chan->ramin_heap, chan->ramin->gpuobj->
+				    im_pramin->start, 32768);
+	if (ret) {
+		NV_ERROR(dev, "Error initialising EVO PRAMIN heap: %d\n", ret);
+		nv50_evo_channel_del(pchan);
+		return ret;
+	}
+
+	ret = nouveau_gpuobj_new_ref(dev, chan, chan, 0, 4096, 16,
+				     0, &chan->ramht);
+	if (ret) {
+		NV_ERROR(dev, "Unable to allocate EVO RAMHT: %d\n", ret);
+		nv50_evo_channel_del(pchan);
+		return ret;
+	}
+
+	if (dev_priv->chipset != 0x50) {
+		ret = nv50_evo_dmaobj_new(chan, 0x3d, NvEvoFB16, 0x70, 0x19,
+					  0, 0xffffffff);
+		if (ret) {
+			nv50_evo_channel_del(pchan);
+			return ret;
+		}
+
+
+		ret = nv50_evo_dmaobj_new(chan, 0x3d, NvEvoFB32, 0x7a, 0x19,
+					  0, 0xffffffff);
+		if (ret) {
+			nv50_evo_channel_del(pchan);
+			return ret;
+		}
+	}
+
+	ret = nv50_evo_dmaobj_new(chan, 0x3d, NvEvoVRAM, 0, 0x19,
+				  0, nouveau_mem_fb_amount(dev));
+	if (ret) {
+		nv50_evo_channel_del(pchan);
+		return ret;
+	}
+
+	ret = nouveau_bo_new(dev, NULL, 4096, 0, TTM_PL_FLAG_VRAM, 0, 0,
+			     false, true, &chan->pushbuf_bo);
+	if (ret == 0)
+		ret = nouveau_bo_pin(chan->pushbuf_bo, TTM_PL_FLAG_VRAM);
+	if (ret) {
+		NV_ERROR(dev, "Error creating EVO DMA push buffer: %d\n", ret);
+		nv50_evo_channel_del(pchan);
+		return ret;
+	}
+
+	ret = nouveau_bo_map(chan->pushbuf_bo);
+	if (ret) {
+		NV_ERROR(dev, "Error mapping EVO DMA push buffer: %d\n", ret);
+		nv50_evo_channel_del(pchan);
+		return ret;
+	}
+
+	chan->user = ioremap(pci_resource_start(dev->pdev, 0) +
+					NV50_PDISPLAY_USER(0), PAGE_SIZE);
+	if (!chan->user) {
+		NV_ERROR(dev, "Error mapping EVO control regs.\n");
+		nv50_evo_channel_del(pchan);
+		return -ENOMEM;
+	}
+
+	return 0;
+}
+
+int
+nv50_display_init(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_timer_engine *ptimer = &dev_priv->engine.timer;
+	struct nouveau_channel *evo = dev_priv->evo;
+	struct drm_connector *connector;
+	uint32_t val, ram_amount, hpd_en[2];
+	uint64_t start;
+	int ret, i;
+
+	NV_DEBUG_KMS(dev, "\n");
+
+	nv_wr32(dev, 0x00610184, nv_rd32(dev, 0x00614004));
+	/*
+	 * I think the 0x006101XX range is some kind of main control area
+	 * that enables things.
+	 */
+	/* CRTC? */
+	for (i = 0; i < 2; i++) {
+		val = nv_rd32(dev, 0x00616100 + (i * 0x800));
+		nv_wr32(dev, 0x00610190 + (i * 0x10), val);
+		val = nv_rd32(dev, 0x00616104 + (i * 0x800));
+		nv_wr32(dev, 0x00610194 + (i * 0x10), val);
+		val = nv_rd32(dev, 0x00616108 + (i * 0x800));
+		nv_wr32(dev, 0x00610198 + (i * 0x10), val);
+		val = nv_rd32(dev, 0x0061610c + (i * 0x800));
+		nv_wr32(dev, 0x0061019c + (i * 0x10), val);
+	}
+	/* DAC */
+	for (i = 0; i < 3; i++) {
+		val = nv_rd32(dev, 0x0061a000 + (i * 0x800));
+		nv_wr32(dev, 0x006101d0 + (i * 0x04), val);
+	}
+	/* SOR */
+	for (i = 0; i < 4; i++) {
+		val = nv_rd32(dev, 0x0061c000 + (i * 0x800));
+		nv_wr32(dev, 0x006101e0 + (i * 0x04), val);
+	}
+	/* Something not yet in use, tv-out maybe. */
+	for (i = 0; i < 3; i++) {
+		val = nv_rd32(dev, 0x0061e000 + (i * 0x800));
+		nv_wr32(dev, 0x006101f0 + (i * 0x04), val);
+	}
+
+	for (i = 0; i < 3; i++) {
+		nv_wr32(dev, NV50_PDISPLAY_DAC_DPMS_CTRL(i), 0x00550000 |
+			NV50_PDISPLAY_DAC_DPMS_CTRL_PENDING);
+		nv_wr32(dev, NV50_PDISPLAY_DAC_CLK_CTRL1(i), 0x00000001);
+	}
+
+	/* This used to be in crtc unblank, but seems out of place there. */
+	nv_wr32(dev, NV50_PDISPLAY_UNK_380, 0);
+	/* RAM is clamped to 256 MiB. */
+	ram_amount = nouveau_mem_fb_amount(dev);
+	NV_DEBUG_KMS(dev, "ram_amount %d\n", ram_amount);
+	if (ram_amount > 256*1024*1024)
+		ram_amount = 256*1024*1024;
+	nv_wr32(dev, NV50_PDISPLAY_RAM_AMOUNT, ram_amount - 1);
+	nv_wr32(dev, NV50_PDISPLAY_UNK_388, 0x150000);
+	nv_wr32(dev, NV50_PDISPLAY_UNK_38C, 0);
+
+	/* The precise purpose is unknown, i suspect it has something to do
+	 * with text mode.
+	 */
+	if (nv_rd32(dev, NV50_PDISPLAY_INTR_1) & 0x100) {
+		nv_wr32(dev, NV50_PDISPLAY_INTR_1, 0x100);
+		nv_wr32(dev, 0x006194e8, nv_rd32(dev, 0x006194e8) & ~1);
+		if (!nv_wait(0x006194e8, 2, 0)) {
+			NV_ERROR(dev, "timeout: (0x6194e8 & 2) != 0\n");
+			NV_ERROR(dev, "0x6194e8 = 0x%08x\n",
+						nv_rd32(dev, 0x6194e8));
+			return -EBUSY;
+		}
+	}
+
+	/* taken from nv bug #12637, attempts to un-wedge the hw if it's
+	 * stuck in some unspecified state
+	 */
+	start = ptimer->read(dev);
+	nv_wr32(dev, NV50_PDISPLAY_CHANNEL_STAT(0), 0x2b00);
+	while ((val = nv_rd32(dev, NV50_PDISPLAY_CHANNEL_STAT(0))) & 0x1e0000) {
+		if ((val & 0x9f0000) == 0x20000)
+			nv_wr32(dev, NV50_PDISPLAY_CHANNEL_STAT(0),
+							val | 0x800000);
+
+		if ((val & 0x3f0000) == 0x30000)
+			nv_wr32(dev, NV50_PDISPLAY_CHANNEL_STAT(0),
+							val | 0x200000);
+
+		if (ptimer->read(dev) - start > 1000000000ULL) {
+			NV_ERROR(dev, "timeout: (0x610200 & 0x1e0000) != 0\n");
+			NV_ERROR(dev, "0x610200 = 0x%08x\n", val);
+			return -EBUSY;
+		}
+	}
+
+	nv_wr32(dev, NV50_PDISPLAY_CTRL_STATE, NV50_PDISPLAY_CTRL_STATE_ENABLE);
+	nv_wr32(dev, NV50_PDISPLAY_CHANNEL_STAT(0), 0x1000b03);
+	if (!nv_wait(NV50_PDISPLAY_CHANNEL_STAT(0), 0x40000000, 0x40000000)) {
+		NV_ERROR(dev, "timeout: (0x610200 & 0x40000000) == 0x40000000\n");
+		NV_ERROR(dev, "0x610200 = 0x%08x\n",
+			  nv_rd32(dev, NV50_PDISPLAY_CHANNEL_STAT(0)));
+		return -EBUSY;
+	}
+
+	for (i = 0; i < 2; i++) {
+		nv_wr32(dev, NV50_PDISPLAY_CURSOR_CURSOR_CTRL2(i), 0x2000);
+		if (!nv_wait(NV50_PDISPLAY_CURSOR_CURSOR_CTRL2(i),
+			     NV50_PDISPLAY_CURSOR_CURSOR_CTRL2_STATUS, 0)) {
+			NV_ERROR(dev, "timeout: CURSOR_CTRL2_STATUS == 0\n");
+			NV_ERROR(dev, "CURSOR_CTRL2 = 0x%08x\n",
+				 nv_rd32(dev, NV50_PDISPLAY_CURSOR_CURSOR_CTRL2(i)));
+			return -EBUSY;
+		}
+
+		nv_wr32(dev, NV50_PDISPLAY_CURSOR_CURSOR_CTRL2(i),
+			NV50_PDISPLAY_CURSOR_CURSOR_CTRL2_ON);
+		if (!nv_wait(NV50_PDISPLAY_CURSOR_CURSOR_CTRL2(i),
+			     NV50_PDISPLAY_CURSOR_CURSOR_CTRL2_STATUS,
+			     NV50_PDISPLAY_CURSOR_CURSOR_CTRL2_STATUS_ACTIVE)) {
+			NV_ERROR(dev, "timeout: "
+				      "CURSOR_CTRL2_STATUS_ACTIVE(%d)\n", i);
+			NV_ERROR(dev, "CURSOR_CTRL2(%d) = 0x%08x\n", i,
+				 nv_rd32(dev, NV50_PDISPLAY_CURSOR_CURSOR_CTRL2(i)));
+			return -EBUSY;
+		}
+	}
+
+	nv_wr32(dev, NV50_PDISPLAY_OBJECTS, (evo->ramin->instance >> 8) | 9);
+
+	/* initialise fifo */
+	nv_wr32(dev, NV50_PDISPLAY_CHANNEL_DMA_CB(0),
+		((evo->pushbuf_bo->bo.mem.mm_node->start << PAGE_SHIFT) >> 8) |
+		NV50_PDISPLAY_CHANNEL_DMA_CB_LOCATION_VRAM |
+		NV50_PDISPLAY_CHANNEL_DMA_CB_VALID);
+	nv_wr32(dev, NV50_PDISPLAY_CHANNEL_UNK2(0), 0x00010000);
+	nv_wr32(dev, NV50_PDISPLAY_CHANNEL_UNK3(0), 0x00000002);
+	if (!nv_wait(0x610200, 0x80000000, 0x00000000)) {
+		NV_ERROR(dev, "timeout: (0x610200 & 0x80000000) == 0\n");
+		NV_ERROR(dev, "0x610200 = 0x%08x\n", nv_rd32(dev, 0x610200));
+		return -EBUSY;
+	}
+	nv_wr32(dev, NV50_PDISPLAY_CHANNEL_STAT(0),
+		(nv_rd32(dev, NV50_PDISPLAY_CHANNEL_STAT(0)) & ~0x00000003) |
+		 NV50_PDISPLAY_CHANNEL_STAT_DMA_ENABLED);
+	nv_wr32(dev, NV50_PDISPLAY_USER_PUT(0), 0);
+	nv_wr32(dev, NV50_PDISPLAY_CHANNEL_STAT(0), 0x01000003 |
+		NV50_PDISPLAY_CHANNEL_STAT_DMA_ENABLED);
+	nv_wr32(dev, 0x610300, nv_rd32(dev, 0x610300) & ~1);
+
+	evo->dma.max = (4096/4) - 2;
+	evo->dma.put = 0;
+	evo->dma.cur = evo->dma.put;
+	evo->dma.free = evo->dma.max - evo->dma.cur;
+
+	ret = RING_SPACE(evo, NOUVEAU_DMA_SKIPS);
+	if (ret)
+		return ret;
+
+	for (i = 0; i < NOUVEAU_DMA_SKIPS; i++)
+		OUT_RING(evo, 0);
+
+	ret = RING_SPACE(evo, 11);
+	if (ret)
+		return ret;
+	BEGIN_RING(evo, 0, NV50_EVO_UNK84, 2);
+	OUT_RING(evo, NV50_EVO_UNK84_NOTIFY_DISABLED);
+	OUT_RING(evo, NV50_EVO_DMA_NOTIFY_HANDLE_NONE);
+	BEGIN_RING(evo, 0, NV50_EVO_CRTC(0, FB_DMA), 1);
+	OUT_RING(evo, NV50_EVO_CRTC_FB_DMA_HANDLE_NONE);
+	BEGIN_RING(evo, 0, NV50_EVO_CRTC(0, UNK0800), 1);
+	OUT_RING(evo, 0);
+	BEGIN_RING(evo, 0, NV50_EVO_CRTC(0, DISPLAY_START), 1);
+	OUT_RING(evo, 0);
+	BEGIN_RING(evo, 0, NV50_EVO_CRTC(0, UNK082C), 1);
+	OUT_RING(evo, 0);
+	FIRE_RING(evo);
+	if (!nv_wait(0x640004, 0xffffffff, evo->dma.put << 2))
+		NV_ERROR(dev, "evo pushbuf stalled\n");
+
+	/* enable clock change interrupts. */
+	nv_wr32(dev, 0x610028, 0x00010001);
+	nv_wr32(dev, NV50_PDISPLAY_INTR_EN, (NV50_PDISPLAY_INTR_EN_CLK_UNK10 |
+					     NV50_PDISPLAY_INTR_EN_CLK_UNK20 |
+					     NV50_PDISPLAY_INTR_EN_CLK_UNK40));
+
+	/* enable hotplug interrupts */
+	hpd_en[0] = hpd_en[1] = 0;
+	list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
+		struct nouveau_connector *conn = nouveau_connector(connector);
+		struct dcb_gpio_entry *gpio;
+
+		if (conn->dcb->gpio_tag == 0xff)
+			continue;
+
+		gpio = nouveau_bios_gpio_entry(dev, conn->dcb->gpio_tag);
+		if (!gpio)
+			continue;
+
+		hpd_en[gpio->line >> 4] |= (0x00010001 << (gpio->line & 0xf));
+	}
+
+	nv_wr32(dev, 0xe054, 0xffffffff);
+	nv_wr32(dev, 0xe050, hpd_en[0]);
+	if (dev_priv->chipset >= 0x90) {
+		nv_wr32(dev, 0xe074, 0xffffffff);
+		nv_wr32(dev, 0xe070, hpd_en[1]);
+	}
+
+	return 0;
+}
+
+static int nv50_display_disable(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct drm_crtc *drm_crtc;
+	int ret, i;
+
+	NV_DEBUG_KMS(dev, "\n");
+
+	list_for_each_entry(drm_crtc, &dev->mode_config.crtc_list, head) {
+		struct nouveau_crtc *crtc = nouveau_crtc(drm_crtc);
+
+		nv50_crtc_blank(crtc, true);
+	}
+
+	ret = RING_SPACE(dev_priv->evo, 2);
+	if (ret == 0) {
+		BEGIN_RING(dev_priv->evo, 0, NV50_EVO_UPDATE, 1);
+		OUT_RING(dev_priv->evo, 0);
+	}
+	FIRE_RING(dev_priv->evo);
+
+	/* Almost like ack'ing a vblank interrupt, maybe in the spirit of
+	 * cleaning up?
+	 */
+	list_for_each_entry(drm_crtc, &dev->mode_config.crtc_list, head) {
+		struct nouveau_crtc *crtc = nouveau_crtc(drm_crtc);
+		uint32_t mask = NV50_PDISPLAY_INTR_1_VBLANK_CRTC_(crtc->index);
+
+		if (!crtc->base.enabled)
+			continue;
+
+		nv_wr32(dev, NV50_PDISPLAY_INTR_1, mask);
+		if (!nv_wait(NV50_PDISPLAY_INTR_1, mask, mask)) {
+			NV_ERROR(dev, "timeout: (0x610024 & 0x%08x) == "
+				      "0x%08x\n", mask, mask);
+			NV_ERROR(dev, "0x610024 = 0x%08x\n",
+				 nv_rd32(dev, NV50_PDISPLAY_INTR_1));
+		}
+	}
+
+	nv_wr32(dev, NV50_PDISPLAY_CHANNEL_STAT(0), 0);
+	nv_wr32(dev, NV50_PDISPLAY_CTRL_STATE, 0);
+	if (!nv_wait(NV50_PDISPLAY_CHANNEL_STAT(0), 0x1e0000, 0)) {
+		NV_ERROR(dev, "timeout: (0x610200 & 0x1e0000) == 0\n");
+		NV_ERROR(dev, "0x610200 = 0x%08x\n",
+			  nv_rd32(dev, NV50_PDISPLAY_CHANNEL_STAT(0)));
+	}
+
+	for (i = 0; i < 3; i++) {
+		if (!nv_wait(NV50_PDISPLAY_SOR_DPMS_STATE(i),
+			     NV50_PDISPLAY_SOR_DPMS_STATE_WAIT, 0)) {
+			NV_ERROR(dev, "timeout: SOR_DPMS_STATE_WAIT(%d) == 0\n", i);
+			NV_ERROR(dev, "SOR_DPMS_STATE(%d) = 0x%08x\n", i,
+				  nv_rd32(dev, NV50_PDISPLAY_SOR_DPMS_STATE(i)));
+		}
+	}
+
+	/* disable interrupts. */
+	nv_wr32(dev, NV50_PDISPLAY_INTR_EN, 0x00000000);
+
+	/* disable hotplug interrupts */
+	nv_wr32(dev, 0xe054, 0xffffffff);
+	nv_wr32(dev, 0xe050, 0x00000000);
+	if (dev_priv->chipset >= 0x90) {
+		nv_wr32(dev, 0xe074, 0xffffffff);
+		nv_wr32(dev, 0xe070, 0x00000000);
+	}
+	return 0;
+}
+
+int nv50_display_create(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct dcb_table *dcb = &dev_priv->vbios.dcb;
+	int ret, i;
+
+	NV_DEBUG_KMS(dev, "\n");
+
+	/* init basic kernel modesetting */
+	drm_mode_config_init(dev);
+
+	/* Initialise some optional connector properties. */
+	drm_mode_create_scaling_mode_property(dev);
+	drm_mode_create_dithering_property(dev);
+
+	dev->mode_config.min_width = 0;
+	dev->mode_config.min_height = 0;
+
+	dev->mode_config.funcs = (void *)&nouveau_mode_config_funcs;
+
+	dev->mode_config.max_width = 8192;
+	dev->mode_config.max_height = 8192;
+
+	dev->mode_config.fb_base = dev_priv->fb_phys;
+
+	/* Create EVO channel */
+	ret = nv50_evo_channel_new(dev, &dev_priv->evo);
+	if (ret) {
+		NV_ERROR(dev, "Error creating EVO channel: %d\n", ret);
+		return ret;
+	}
+
+	/* Create CRTC objects */
+	for (i = 0; i < 2; i++)
+		nv50_crtc_create(dev, i);
+
+	/* We setup the encoders from the BIOS table */
+	for (i = 0 ; i < dcb->entries; i++) {
+		struct dcb_entry *entry = &dcb->entry[i];
+
+		if (entry->location != DCB_LOC_ON_CHIP) {
+			NV_WARN(dev, "Off-chip encoder %d/%d unsupported\n",
+				entry->type, ffs(entry->or) - 1);
+			continue;
+		}
+
+		switch (entry->type) {
+		case OUTPUT_TMDS:
+		case OUTPUT_LVDS:
+		case OUTPUT_DP:
+			nv50_sor_create(dev, entry);
+			break;
+		case OUTPUT_ANALOG:
+			nv50_dac_create(dev, entry);
+			break;
+		default:
+			NV_WARN(dev, "DCB encoder %d unknown\n", entry->type);
+			continue;
+		}
+	}
+
+	for (i = 0 ; i < dcb->connector.entries; i++) {
+		if (i != 0 && dcb->connector.entry[i].index ==
+			      dcb->connector.entry[i - 1].index)
+			continue;
+		nouveau_connector_create(dev, &dcb->connector.entry[i]);
+	}
+
+	ret = nv50_display_init(dev);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+int nv50_display_destroy(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	NV_DEBUG_KMS(dev, "\n");
+
+	drm_mode_config_cleanup(dev);
+
+	nv50_display_disable(dev);
+	nv50_evo_channel_del(&dev_priv->evo);
+
+	return 0;
+}
+
+static inline uint32_t
+nv50_display_mode_ctrl(struct drm_device *dev, bool sor, int or)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t mc;
+
+	if (sor) {
+		if (dev_priv->chipset < 0x90 ||
+		    dev_priv->chipset == 0x92 || dev_priv->chipset == 0xa0)
+			mc = nv_rd32(dev, NV50_PDISPLAY_SOR_MODE_CTRL_P(or));
+		else
+			mc = nv_rd32(dev, NV90_PDISPLAY_SOR_MODE_CTRL_P(or));
+	} else {
+		mc = nv_rd32(dev, NV50_PDISPLAY_DAC_MODE_CTRL_P(or));
+	}
+
+	return mc;
+}
+
+static int
+nv50_display_irq_head(struct drm_device *dev, int *phead,
+		      struct dcb_entry **pdcbent)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t unk30 = nv_rd32(dev, NV50_PDISPLAY_UNK30_CTRL);
+	uint32_t dac = 0, sor = 0;
+	int head, i, or = 0, type = OUTPUT_ANY;
+
+	/* We're assuming that head 0 *or* head 1 will be active here,
+	 * and not both.  I'm not sure if the hw will even signal both
+	 * ever, but it definitely shouldn't for us as we commit each
+	 * CRTC separately, and submission will be blocked by the GPU
+	 * until we handle each in turn.
+	 */
+	NV_DEBUG_KMS(dev, "0x610030: 0x%08x\n", unk30);
+	head = ffs((unk30 >> 9) & 3) - 1;
+	if (head < 0)
+		return -EINVAL;
+
+	/* This assumes CRTCs are never bound to multiple encoders, which
+	 * should be the case.
+	 */
+	for (i = 0; i < 3 && type == OUTPUT_ANY; i++) {
+		uint32_t mc = nv50_display_mode_ctrl(dev, false, i);
+		if (!(mc & (1 << head)))
+			continue;
+
+		switch ((mc >> 8) & 0xf) {
+		case 0: type = OUTPUT_ANALOG; break;
+		case 1: type = OUTPUT_TV; break;
+		default:
+			NV_ERROR(dev, "unknown dac mode_ctrl: 0x%08x\n", dac);
+			return -1;
+		}
+
+		or = i;
+	}
+
+	for (i = 0; i < 4 && type == OUTPUT_ANY; i++) {
+		uint32_t mc = nv50_display_mode_ctrl(dev, true, i);
+		if (!(mc & (1 << head)))
+			continue;
+
+		switch ((mc >> 8) & 0xf) {
+		case 0: type = OUTPUT_LVDS; break;
+		case 1: type = OUTPUT_TMDS; break;
+		case 2: type = OUTPUT_TMDS; break;
+		case 5: type = OUTPUT_TMDS; break;
+		case 8: type = OUTPUT_DP; break;
+		case 9: type = OUTPUT_DP; break;
+		default:
+			NV_ERROR(dev, "unknown sor mode_ctrl: 0x%08x\n", sor);
+			return -1;
+		}
+
+		or = i;
+	}
+
+	NV_DEBUG_KMS(dev, "type %d, or %d\n", type, or);
+	if (type == OUTPUT_ANY) {
+		NV_ERROR(dev, "unknown encoder!!\n");
+		return -1;
+	}
+
+	for (i = 0; i < dev_priv->vbios.dcb.entries; i++) {
+		struct dcb_entry *dcbent = &dev_priv->vbios.dcb.entry[i];
+
+		if (dcbent->type != type)
+			continue;
+
+		if (!(dcbent->or & (1 << or)))
+			continue;
+
+		*phead = head;
+		*pdcbent = dcbent;
+		return 0;
+	}
+
+	NV_ERROR(dev, "no DCB entry for %d %d\n", dac != 0, or);
+	return 0;
+}
+
+static uint32_t
+nv50_display_script_select(struct drm_device *dev, struct dcb_entry *dcbent,
+			   int pxclk)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_connector *nv_connector = NULL;
+	struct drm_encoder *encoder;
+	struct nvbios *bios = &dev_priv->vbios;
+	uint32_t mc, script = 0, or;
+
+	list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) {
+		struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+
+		if (nv_encoder->dcb != dcbent)
+			continue;
+
+		nv_connector = nouveau_encoder_connector_get(nv_encoder);
+		break;
+	}
+
+	or = ffs(dcbent->or) - 1;
+	mc = nv50_display_mode_ctrl(dev, dcbent->type != OUTPUT_ANALOG, or);
+	switch (dcbent->type) {
+	case OUTPUT_LVDS:
+		script = (mc >> 8) & 0xf;
+		if (bios->fp_no_ddc) {
+			if (bios->fp.dual_link)
+				script |= 0x0100;
+			if (bios->fp.if_is_24bit)
+				script |= 0x0200;
+		} else {
+			if (pxclk >= bios->fp.duallink_transition_clk) {
+				script |= 0x0100;
+				if (bios->fp.strapless_is_24bit & 2)
+					script |= 0x0200;
+			} else
+			if (bios->fp.strapless_is_24bit & 1)
+				script |= 0x0200;
+
+			if (nv_connector && nv_connector->edid &&
+			    (nv_connector->edid->revision >= 4) &&
+			    (nv_connector->edid->input & 0x70) >= 0x20)
+				script |= 0x0200;
+		}
+
+		if (nouveau_uscript_lvds >= 0) {
+			NV_INFO(dev, "override script 0x%04x with 0x%04x "
+				     "for output LVDS-%d\n", script,
+				     nouveau_uscript_lvds, or);
+			script = nouveau_uscript_lvds;
+		}
+		break;
+	case OUTPUT_TMDS:
+		script = (mc >> 8) & 0xf;
+		if (pxclk >= 165000)
+			script |= 0x0100;
+
+		if (nouveau_uscript_tmds >= 0) {
+			NV_INFO(dev, "override script 0x%04x with 0x%04x "
+				     "for output TMDS-%d\n", script,
+				     nouveau_uscript_tmds, or);
+			script = nouveau_uscript_tmds;
+		}
+		break;
+	case OUTPUT_DP:
+		script = (mc >> 8) & 0xf;
+		break;
+	case OUTPUT_ANALOG:
+		script = 0xff;
+		break;
+	default:
+		NV_ERROR(dev, "modeset on unsupported output type!\n");
+		break;
+	}
+
+	return script;
+}
+
+static void
+nv50_display_vblank_crtc_handler(struct drm_device *dev, int crtc)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_channel *chan;
+	struct list_head *entry, *tmp;
+
+	list_for_each_safe(entry, tmp, &dev_priv->vbl_waiting) {
+		chan = list_entry(entry, struct nouveau_channel, nvsw.vbl_wait);
+
+		nouveau_bo_wr32(chan->notifier_bo, chan->nvsw.vblsem_offset,
+						chan->nvsw.vblsem_rval);
+		list_del(&chan->nvsw.vbl_wait);
+	}
+}
+
+static void
+nv50_display_vblank_handler(struct drm_device *dev, uint32_t intr)
+{
+	intr &= NV50_PDISPLAY_INTR_1_VBLANK_CRTC;
+
+	if (intr & NV50_PDISPLAY_INTR_1_VBLANK_CRTC_0)
+		nv50_display_vblank_crtc_handler(dev, 0);
+
+	if (intr & NV50_PDISPLAY_INTR_1_VBLANK_CRTC_1)
+		nv50_display_vblank_crtc_handler(dev, 1);
+
+	nv_wr32(dev, NV50_PDISPLAY_INTR_EN, nv_rd32(dev,
+		     NV50_PDISPLAY_INTR_EN) & ~intr);
+	nv_wr32(dev, NV50_PDISPLAY_INTR_1, intr);
+}
+
+static void
+nv50_display_unk10_handler(struct drm_device *dev)
+{
+	struct dcb_entry *dcbent;
+	int head, ret;
+
+	ret = nv50_display_irq_head(dev, &head, &dcbent);
+	if (ret)
+		goto ack;
+
+	nv_wr32(dev, 0x619494, nv_rd32(dev, 0x619494) & ~8);
+
+	nouveau_bios_run_display_table(dev, dcbent, 0, -1);
+
+ack:
+	nv_wr32(dev, NV50_PDISPLAY_INTR_1, NV50_PDISPLAY_INTR_1_CLK_UNK10);
+	nv_wr32(dev, 0x610030, 0x80000000);
+}
+
+static void
+nv50_display_unk20_handler(struct drm_device *dev)
+{
+	struct dcb_entry *dcbent;
+	uint32_t tmp, pclk, script;
+	int head, or, ret;
+
+	ret = nv50_display_irq_head(dev, &head, &dcbent);
+	if (ret)
+		goto ack;
+	or = ffs(dcbent->or) - 1;
+	pclk = nv_rd32(dev, NV50_PDISPLAY_CRTC_P(head, CLOCK)) & 0x3fffff;
+	script = nv50_display_script_select(dev, dcbent, pclk);
+
+	NV_DEBUG_KMS(dev, "head %d pxclk: %dKHz\n", head, pclk);
+
+	if (dcbent->type != OUTPUT_DP)
+		nouveau_bios_run_display_table(dev, dcbent, 0, -2);
+
+	nv50_crtc_set_clock(dev, head, pclk);
+
+	nouveau_bios_run_display_table(dev, dcbent, script, pclk);
+
+	tmp = nv_rd32(dev, NV50_PDISPLAY_CRTC_CLK_CTRL2(head));
+	tmp &= ~0x000000f;
+	nv_wr32(dev, NV50_PDISPLAY_CRTC_CLK_CTRL2(head), tmp);
+
+	if (dcbent->type != OUTPUT_ANALOG) {
+		tmp = nv_rd32(dev, NV50_PDISPLAY_SOR_CLK_CTRL2(or));
+		tmp &= ~0x00000f0f;
+		if (script & 0x0100)
+			tmp |= 0x00000101;
+		nv_wr32(dev, NV50_PDISPLAY_SOR_CLK_CTRL2(or), tmp);
+	} else {
+		nv_wr32(dev, NV50_PDISPLAY_DAC_CLK_CTRL2(or), 0);
+	}
+
+ack:
+	nv_wr32(dev, NV50_PDISPLAY_INTR_1, NV50_PDISPLAY_INTR_1_CLK_UNK20);
+	nv_wr32(dev, 0x610030, 0x80000000);
+}
+
+static void
+nv50_display_unk40_handler(struct drm_device *dev)
+{
+	struct dcb_entry *dcbent;
+	int head, pclk, script, ret;
+
+	ret = nv50_display_irq_head(dev, &head, &dcbent);
+	if (ret)
+		goto ack;
+	pclk = nv_rd32(dev, NV50_PDISPLAY_CRTC_P(head, CLOCK)) & 0x3fffff;
+	script = nv50_display_script_select(dev, dcbent, pclk);
+
+	nouveau_bios_run_display_table(dev, dcbent, script, -pclk);
+
+ack:
+	nv_wr32(dev, NV50_PDISPLAY_INTR_1, NV50_PDISPLAY_INTR_1_CLK_UNK40);
+	nv_wr32(dev, 0x610030, 0x80000000);
+	nv_wr32(dev, 0x619494, nv_rd32(dev, 0x619494) | 8);
+}
+
+void
+nv50_display_irq_handler_bh(struct work_struct *work)
+{
+	struct drm_nouveau_private *dev_priv =
+		container_of(work, struct drm_nouveau_private, irq_work);
+	struct drm_device *dev = dev_priv->dev;
+
+	for (;;) {
+		uint32_t intr0 = nv_rd32(dev, NV50_PDISPLAY_INTR_0);
+		uint32_t intr1 = nv_rd32(dev, NV50_PDISPLAY_INTR_1);
+
+		NV_DEBUG_KMS(dev, "PDISPLAY_INTR_BH 0x%08x 0x%08x\n", intr0, intr1);
+
+		if (intr1 & NV50_PDISPLAY_INTR_1_CLK_UNK10)
+			nv50_display_unk10_handler(dev);
+		else
+		if (intr1 & NV50_PDISPLAY_INTR_1_CLK_UNK20)
+			nv50_display_unk20_handler(dev);
+		else
+		if (intr1 & NV50_PDISPLAY_INTR_1_CLK_UNK40)
+			nv50_display_unk40_handler(dev);
+		else
+			break;
+	}
+
+	nv_wr32(dev, NV03_PMC_INTR_EN_0, 1);
+}
+
+static void
+nv50_display_error_handler(struct drm_device *dev)
+{
+	uint32_t addr, data;
+
+	nv_wr32(dev, NV50_PDISPLAY_INTR_0, 0x00010000);
+	addr = nv_rd32(dev, NV50_PDISPLAY_TRAPPED_ADDR);
+	data = nv_rd32(dev, NV50_PDISPLAY_TRAPPED_DATA);
+
+	NV_ERROR(dev, "EvoCh %d Mthd 0x%04x Data 0x%08x (0x%04x 0x%02x)\n",
+		 0, addr & 0xffc, data, addr >> 16, (addr >> 12) & 0xf);
+
+	nv_wr32(dev, NV50_PDISPLAY_TRAPPED_ADDR, 0x90000000);
+}
+
+static void
+nv50_display_irq_hotplug(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct drm_connector *connector;
+	const uint32_t gpio_reg[4] = { 0xe104, 0xe108, 0xe280, 0xe284 };
+	uint32_t unplug_mask, plug_mask, change_mask;
+	uint32_t hpd0, hpd1 = 0;
+
+	hpd0 = nv_rd32(dev, 0xe054) & nv_rd32(dev, 0xe050);
+	if (dev_priv->chipset >= 0x90)
+		hpd1 = nv_rd32(dev, 0xe074) & nv_rd32(dev, 0xe070);
+
+	plug_mask   = (hpd0 & 0x0000ffff) | (hpd1 << 16);
+	unplug_mask = (hpd0 >> 16) | (hpd1 & 0xffff0000);
+	change_mask = plug_mask | unplug_mask;
+
+	list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
+		struct drm_encoder_helper_funcs *helper;
+		struct nouveau_connector *nv_connector =
+			nouveau_connector(connector);
+		struct nouveau_encoder *nv_encoder;
+		struct dcb_gpio_entry *gpio;
+		uint32_t reg;
+		bool plugged;
+
+		if (!nv_connector->dcb)
+			continue;
+
+		gpio = nouveau_bios_gpio_entry(dev, nv_connector->dcb->gpio_tag);
+		if (!gpio || !(change_mask & (1 << gpio->line)))
+			continue;
+
+		reg = nv_rd32(dev, gpio_reg[gpio->line >> 3]);
+		plugged = !!(reg & (4 << ((gpio->line & 7) << 2)));
+		NV_INFO(dev, "%splugged %s\n", plugged ? "" : "un",
+			drm_get_connector_name(connector)) ;
+
+		if (!connector->encoder || !connector->encoder->crtc ||
+		    !connector->encoder->crtc->enabled)
+			continue;
+		nv_encoder = nouveau_encoder(connector->encoder);
+		helper = connector->encoder->helper_private;
+
+		if (nv_encoder->dcb->type != OUTPUT_DP)
+			continue;
+
+		if (plugged)
+			helper->dpms(connector->encoder, DRM_MODE_DPMS_ON);
+		else
+			helper->dpms(connector->encoder, DRM_MODE_DPMS_OFF);
+	}
+
+	nv_wr32(dev, 0xe054, nv_rd32(dev, 0xe054));
+	if (dev_priv->chipset >= 0x90)
+		nv_wr32(dev, 0xe074, nv_rd32(dev, 0xe074));
+}
+
+void
+nv50_display_irq_handler(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t delayed = 0;
+
+	while (nv_rd32(dev, NV50_PMC_INTR_0) & NV50_PMC_INTR_0_HOTPLUG)
+		nv50_display_irq_hotplug(dev);
+
+	while (nv_rd32(dev, NV50_PMC_INTR_0) & NV50_PMC_INTR_0_DISPLAY) {
+		uint32_t intr0 = nv_rd32(dev, NV50_PDISPLAY_INTR_0);
+		uint32_t intr1 = nv_rd32(dev, NV50_PDISPLAY_INTR_1);
+		uint32_t clock;
+
+		NV_DEBUG_KMS(dev, "PDISPLAY_INTR 0x%08x 0x%08x\n", intr0, intr1);
+
+		if (!intr0 && !(intr1 & ~delayed))
+			break;
+
+		if (intr0 & 0x00010000) {
+			nv50_display_error_handler(dev);
+			intr0 &= ~0x00010000;
+		}
+
+		if (intr1 & NV50_PDISPLAY_INTR_1_VBLANK_CRTC) {
+			nv50_display_vblank_handler(dev, intr1);
+			intr1 &= ~NV50_PDISPLAY_INTR_1_VBLANK_CRTC;
+		}
+
+		clock = (intr1 & (NV50_PDISPLAY_INTR_1_CLK_UNK10 |
+				  NV50_PDISPLAY_INTR_1_CLK_UNK20 |
+				  NV50_PDISPLAY_INTR_1_CLK_UNK40));
+		if (clock) {
+			nv_wr32(dev, NV03_PMC_INTR_EN_0, 0);
+			if (!work_pending(&dev_priv->irq_work))
+				queue_work(dev_priv->wq, &dev_priv->irq_work);
+			delayed |= clock;
+			intr1 &= ~clock;
+		}
+
+		if (intr0) {
+			NV_ERROR(dev, "unknown PDISPLAY_INTR_0: 0x%08x\n", intr0);
+			nv_wr32(dev, NV50_PDISPLAY_INTR_0, intr0);
+		}
+
+		if (intr1) {
+			NV_ERROR(dev,
+				 "unknown PDISPLAY_INTR_1: 0x%08x\n", intr1);
+			nv_wr32(dev, NV50_PDISPLAY_INTR_1, intr1);
+		}
+	}
+}
+
diff --git a/drivers/gpu/drm/nouveau/nv50_display.h b/drivers/gpu/drm/nouveau/nv50_display.h
new file mode 100644
index 000000000000..3ae8d0725f63
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv50_display.h
@@ -0,0 +1,46 @@
+/*
+ * Copyright (C) 2008 Maarten Maathuis.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __NV50_DISPLAY_H__
+#define __NV50_DISPLAY_H__
+
+#include "drmP.h"
+#include "drm.h"
+#include "nouveau_drv.h"
+#include "nouveau_dma.h"
+#include "nouveau_reg.h"
+#include "nouveau_crtc.h"
+#include "nv50_evo.h"
+
+void nv50_display_irq_handler(struct drm_device *dev);
+void nv50_display_irq_handler_bh(struct work_struct *work);
+int nv50_display_init(struct drm_device *dev);
+int nv50_display_create(struct drm_device *dev);
+int nv50_display_destroy(struct drm_device *dev);
+int nv50_crtc_blank(struct nouveau_crtc *, bool blank);
+int nv50_crtc_set_clock(struct drm_device *, int head, int pclk);
+
+#endif /* __NV50_DISPLAY_H__ */
diff --git a/drivers/gpu/drm/nouveau/nv50_evo.h b/drivers/gpu/drm/nouveau/nv50_evo.h
new file mode 100644
index 000000000000..aae13343bcec
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv50_evo.h
@@ -0,0 +1,113 @@
+/*
+ * Copyright (C) 2008 Maarten Maathuis.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#define NV50_EVO_UPDATE                                              0x00000080
+#define NV50_EVO_UNK84                                               0x00000084
+#define NV50_EVO_UNK84_NOTIFY                                        0x40000000
+#define NV50_EVO_UNK84_NOTIFY_DISABLED                               0x00000000
+#define NV50_EVO_UNK84_NOTIFY_ENABLED                                0x40000000
+#define NV50_EVO_DMA_NOTIFY                                          0x00000088
+#define NV50_EVO_DMA_NOTIFY_HANDLE                                   0xffffffff
+#define NV50_EVO_DMA_NOTIFY_HANDLE_NONE                              0x00000000
+#define NV50_EVO_UNK8C                                               0x0000008C
+
+#define NV50_EVO_DAC(n, r)                       ((n) * 0x80 + NV50_EVO_DAC_##r)
+#define NV50_EVO_DAC_MODE_CTRL                                       0x00000400
+#define NV50_EVO_DAC_MODE_CTRL_CRTC0                                 0x00000001
+#define NV50_EVO_DAC_MODE_CTRL_CRTC1                                 0x00000002
+#define NV50_EVO_DAC_MODE_CTRL2                                      0x00000404
+#define NV50_EVO_DAC_MODE_CTRL2_NHSYNC                               0x00000001
+#define NV50_EVO_DAC_MODE_CTRL2_NVSYNC                               0x00000002
+
+#define NV50_EVO_SOR(n, r)                       ((n) * 0x40 + NV50_EVO_SOR_##r)
+#define NV50_EVO_SOR_MODE_CTRL                                       0x00000600
+#define NV50_EVO_SOR_MODE_CTRL_CRTC0                                 0x00000001
+#define NV50_EVO_SOR_MODE_CTRL_CRTC1                                 0x00000002
+#define NV50_EVO_SOR_MODE_CTRL_TMDS                                  0x00000100
+#define NV50_EVO_SOR_MODE_CTRL_TMDS_DUAL_LINK                        0x00000400
+#define NV50_EVO_SOR_MODE_CTRL_NHSYNC                                0x00001000
+#define NV50_EVO_SOR_MODE_CTRL_NVSYNC                                0x00002000
+
+#define NV50_EVO_CRTC(n, r)                    ((n) * 0x400 + NV50_EVO_CRTC_##r)
+#define NV84_EVO_CRTC(n, r)                    ((n) * 0x400 + NV84_EVO_CRTC_##r)
+#define NV50_EVO_CRTC_UNK0800                                        0x00000800
+#define NV50_EVO_CRTC_CLOCK                                          0x00000804
+#define NV50_EVO_CRTC_INTERLACE                                      0x00000808
+#define NV50_EVO_CRTC_DISPLAY_START                                  0x00000810
+#define NV50_EVO_CRTC_DISPLAY_TOTAL                                  0x00000814
+#define NV50_EVO_CRTC_SYNC_DURATION                                  0x00000818
+#define NV50_EVO_CRTC_SYNC_START_TO_BLANK_END                        0x0000081c
+#define NV50_EVO_CRTC_UNK0820                                        0x00000820
+#define NV50_EVO_CRTC_UNK0824                                        0x00000824
+#define NV50_EVO_CRTC_UNK082C                                        0x0000082c
+#define NV50_EVO_CRTC_CLUT_MODE                                      0x00000840
+/* You can't have a palette in 8 bit mode (=OFF) */
+#define NV50_EVO_CRTC_CLUT_MODE_BLANK                                0x00000000
+#define NV50_EVO_CRTC_CLUT_MODE_OFF                                  0x80000000
+#define NV50_EVO_CRTC_CLUT_MODE_ON                                   0xC0000000
+#define NV50_EVO_CRTC_CLUT_OFFSET                                    0x00000844
+#define NV84_EVO_CRTC_CLUT_DMA                                       0x0000085C
+#define NV84_EVO_CRTC_CLUT_DMA_HANDLE                                0xffffffff
+#define NV84_EVO_CRTC_CLUT_DMA_HANDLE_NONE                           0x00000000
+#define NV50_EVO_CRTC_FB_OFFSET                                      0x00000860
+#define NV50_EVO_CRTC_FB_SIZE                                        0x00000868
+#define NV50_EVO_CRTC_FB_CONFIG                                      0x0000086c
+#define NV50_EVO_CRTC_FB_CONFIG_MODE                                 0x00100000
+#define NV50_EVO_CRTC_FB_CONFIG_MODE_TILE                            0x00000000
+#define NV50_EVO_CRTC_FB_CONFIG_MODE_PITCH                           0x00100000
+#define NV50_EVO_CRTC_FB_DEPTH                                       0x00000870
+#define NV50_EVO_CRTC_FB_DEPTH_8                                     0x00001e00
+#define NV50_EVO_CRTC_FB_DEPTH_15                                    0x0000e900
+#define NV50_EVO_CRTC_FB_DEPTH_16                                    0x0000e800
+#define NV50_EVO_CRTC_FB_DEPTH_24                                    0x0000cf00
+#define NV50_EVO_CRTC_FB_DEPTH_30                                    0x0000d100
+#define NV50_EVO_CRTC_FB_DMA                                         0x00000874
+#define NV50_EVO_CRTC_FB_DMA_HANDLE                                  0xffffffff
+#define NV50_EVO_CRTC_FB_DMA_HANDLE_NONE                             0x00000000
+#define NV50_EVO_CRTC_CURSOR_CTRL                                    0x00000880
+#define NV50_EVO_CRTC_CURSOR_CTRL_HIDE                               0x05000000
+#define NV50_EVO_CRTC_CURSOR_CTRL_SHOW                               0x85000000
+#define NV50_EVO_CRTC_CURSOR_OFFSET                                  0x00000884
+#define NV84_EVO_CRTC_CURSOR_DMA                                     0x0000089c
+#define NV84_EVO_CRTC_CURSOR_DMA_HANDLE                              0xffffffff
+#define NV84_EVO_CRTC_CURSOR_DMA_HANDLE_NONE                         0x00000000
+#define NV50_EVO_CRTC_DITHER_CTRL                                    0x000008a0
+#define NV50_EVO_CRTC_DITHER_CTRL_OFF                                0x00000000
+#define NV50_EVO_CRTC_DITHER_CTRL_ON                                 0x00000011
+#define NV50_EVO_CRTC_SCALE_CTRL                                     0x000008a4
+#define NV50_EVO_CRTC_SCALE_CTRL_INACTIVE                            0x00000000
+#define NV50_EVO_CRTC_SCALE_CTRL_ACTIVE                              0x00000009
+#define NV50_EVO_CRTC_COLOR_CTRL                                     0x000008a8
+#define NV50_EVO_CRTC_COLOR_CTRL_COLOR                               0x00040000
+#define NV50_EVO_CRTC_FB_POS                                         0x000008c0
+#define NV50_EVO_CRTC_REAL_RES                                       0x000008c8
+#define NV50_EVO_CRTC_SCALE_CENTER_OFFSET                            0x000008d4
+#define NV50_EVO_CRTC_SCALE_CENTER_OFFSET_VAL(x, y) \
+	((((unsigned)y << 16) & 0xFFFF0000) | (((unsigned)x) & 0x0000FFFF))
+/* Both of these are needed, otherwise nothing happens. */
+#define NV50_EVO_CRTC_SCALE_RES1                                     0x000008d8
+#define NV50_EVO_CRTC_SCALE_RES2                                     0x000008dc
+
diff --git a/drivers/gpu/drm/nouveau/nv50_fbcon.c b/drivers/gpu/drm/nouveau/nv50_fbcon.c
new file mode 100644
index 000000000000..993c7126fbde
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv50_fbcon.c
@@ -0,0 +1,267 @@
+#include "drmP.h"
+#include "nouveau_drv.h"
+#include "nouveau_dma.h"
+#include "nouveau_fbcon.h"
+
+void
+nv50_fbcon_fillrect(struct fb_info *info, const struct fb_fillrect *rect)
+{
+	struct nouveau_fbcon_par *par = info->par;
+	struct drm_device *dev = par->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_channel *chan = dev_priv->channel;
+
+	if (info->state != FBINFO_STATE_RUNNING)
+		return;
+
+	if (!(info->flags & FBINFO_HWACCEL_DISABLED) &&
+	     RING_SPACE(chan, rect->rop == ROP_COPY ? 7 : 11)) {
+		nouveau_fbcon_gpu_lockup(info);
+	}
+
+	if (info->flags & FBINFO_HWACCEL_DISABLED) {
+		cfb_fillrect(info, rect);
+		return;
+	}
+
+	if (rect->rop != ROP_COPY) {
+		BEGIN_RING(chan, NvSub2D, 0x02ac, 1);
+		OUT_RING(chan, 1);
+	}
+	BEGIN_RING(chan, NvSub2D, 0x0588, 1);
+	if (info->fix.visual == FB_VISUAL_TRUECOLOR ||
+	    info->fix.visual == FB_VISUAL_DIRECTCOLOR)
+		OUT_RING(chan, ((uint32_t *)info->pseudo_palette)[rect->color]);
+	else
+		OUT_RING(chan, rect->color);
+	BEGIN_RING(chan, NvSub2D, 0x0600, 4);
+	OUT_RING(chan, rect->dx);
+	OUT_RING(chan, rect->dy);
+	OUT_RING(chan, rect->dx + rect->width);
+	OUT_RING(chan, rect->dy + rect->height);
+	if (rect->rop != ROP_COPY) {
+		BEGIN_RING(chan, NvSub2D, 0x02ac, 1);
+		OUT_RING(chan, 3);
+	}
+	FIRE_RING(chan);
+}
+
+void
+nv50_fbcon_copyarea(struct fb_info *info, const struct fb_copyarea *region)
+{
+	struct nouveau_fbcon_par *par = info->par;
+	struct drm_device *dev = par->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_channel *chan = dev_priv->channel;
+
+	if (info->state != FBINFO_STATE_RUNNING)
+		return;
+
+	if (!(info->flags & FBINFO_HWACCEL_DISABLED) && RING_SPACE(chan, 12)) {
+		nouveau_fbcon_gpu_lockup(info);
+	}
+
+	if (info->flags & FBINFO_HWACCEL_DISABLED) {
+		cfb_copyarea(info, region);
+		return;
+	}
+
+	BEGIN_RING(chan, NvSub2D, 0x0110, 1);
+	OUT_RING(chan, 0);
+	BEGIN_RING(chan, NvSub2D, 0x08b0, 4);
+	OUT_RING(chan, region->dx);
+	OUT_RING(chan, region->dy);
+	OUT_RING(chan, region->width);
+	OUT_RING(chan, region->height);
+	BEGIN_RING(chan, NvSub2D, 0x08d0, 4);
+	OUT_RING(chan, 0);
+	OUT_RING(chan, region->sx);
+	OUT_RING(chan, 0);
+	OUT_RING(chan, region->sy);
+	FIRE_RING(chan);
+}
+
+void
+nv50_fbcon_imageblit(struct fb_info *info, const struct fb_image *image)
+{
+	struct nouveau_fbcon_par *par = info->par;
+	struct drm_device *dev = par->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_channel *chan = dev_priv->channel;
+	uint32_t width, dwords, *data = (uint32_t *)image->data;
+	uint32_t mask = ~(~0 >> (32 - info->var.bits_per_pixel));
+	uint32_t *palette = info->pseudo_palette;
+
+	if (info->state != FBINFO_STATE_RUNNING)
+		return;
+
+	if (image->depth != 1) {
+		cfb_imageblit(info, image);
+		return;
+	}
+
+	if (!(info->flags & FBINFO_HWACCEL_DISABLED) && RING_SPACE(chan, 11)) {
+		nouveau_fbcon_gpu_lockup(info);
+	}
+
+	if (info->flags & FBINFO_HWACCEL_DISABLED) {
+		cfb_imageblit(info, image);
+		return;
+	}
+
+	width = ALIGN(image->width, 32);
+	dwords = (width * image->height) >> 5;
+
+	BEGIN_RING(chan, NvSub2D, 0x0814, 2);
+	if (info->fix.visual == FB_VISUAL_TRUECOLOR ||
+	    info->fix.visual == FB_VISUAL_DIRECTCOLOR) {
+		OUT_RING(chan, palette[image->bg_color] | mask);
+		OUT_RING(chan, palette[image->fg_color] | mask);
+	} else {
+		OUT_RING(chan, image->bg_color);
+		OUT_RING(chan, image->fg_color);
+	}
+	BEGIN_RING(chan, NvSub2D, 0x0838, 2);
+	OUT_RING(chan, image->width);
+	OUT_RING(chan, image->height);
+	BEGIN_RING(chan, NvSub2D, 0x0850, 4);
+	OUT_RING(chan, 0);
+	OUT_RING(chan, image->dx);
+	OUT_RING(chan, 0);
+	OUT_RING(chan, image->dy);
+
+	while (dwords) {
+		int push = dwords > 2047 ? 2047 : dwords;
+
+		if (RING_SPACE(chan, push + 1)) {
+			nouveau_fbcon_gpu_lockup(info);
+			cfb_imageblit(info, image);
+			return;
+		}
+
+		dwords -= push;
+
+		BEGIN_RING(chan, NvSub2D, 0x40000860, push);
+		OUT_RINGp(chan, data, push);
+		data += push;
+	}
+
+	FIRE_RING(chan);
+}
+
+int
+nv50_fbcon_accel_init(struct fb_info *info)
+{
+	struct nouveau_fbcon_par *par = info->par;
+	struct drm_device *dev = par->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_channel *chan = dev_priv->channel;
+	struct nouveau_gpuobj *eng2d = NULL;
+	int ret, format;
+
+	switch (info->var.bits_per_pixel) {
+	case 8:
+		format = 0xf3;
+		break;
+	case 15:
+		format = 0xf8;
+		break;
+	case 16:
+		format = 0xe8;
+		break;
+	case 32:
+		switch (info->var.transp.length) {
+		case 0: /* depth 24 */
+		case 8: /* depth 32, just use 24.. */
+			format = 0xe6;
+			break;
+		case 2: /* depth 30 */
+			format = 0xd1;
+			break;
+		default:
+			return -EINVAL;
+		}
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	ret = nouveau_gpuobj_gr_new(dev_priv->channel, 0x502d, &eng2d);
+	if (ret)
+		return ret;
+
+	ret = nouveau_gpuobj_ref_add(dev, dev_priv->channel, Nv2D, eng2d, NULL);
+	if (ret)
+		return ret;
+
+	ret = RING_SPACE(chan, 59);
+	if (ret) {
+		nouveau_fbcon_gpu_lockup(info);
+		return ret;
+	}
+
+	BEGIN_RING(chan, NvSub2D, 0x0000, 1);
+	OUT_RING(chan, Nv2D);
+	BEGIN_RING(chan, NvSub2D, 0x0180, 4);
+	OUT_RING(chan, NvNotify0);
+	OUT_RING(chan, chan->vram_handle);
+	OUT_RING(chan, chan->vram_handle);
+	OUT_RING(chan, chan->vram_handle);
+	BEGIN_RING(chan, NvSub2D, 0x0290, 1);
+	OUT_RING(chan, 0);
+	BEGIN_RING(chan, NvSub2D, 0x0888, 1);
+	OUT_RING(chan, 1);
+	BEGIN_RING(chan, NvSub2D, 0x02ac, 1);
+	OUT_RING(chan, 3);
+	BEGIN_RING(chan, NvSub2D, 0x02a0, 1);
+	OUT_RING(chan, 0x55);
+	BEGIN_RING(chan, NvSub2D, 0x08c0, 4);
+	OUT_RING(chan, 0);
+	OUT_RING(chan, 1);
+	OUT_RING(chan, 0);
+	OUT_RING(chan, 1);
+	BEGIN_RING(chan, NvSub2D, 0x0580, 2);
+	OUT_RING(chan, 4);
+	OUT_RING(chan, format);
+	BEGIN_RING(chan, NvSub2D, 0x02e8, 2);
+	OUT_RING(chan, 2);
+	OUT_RING(chan, 1);
+	BEGIN_RING(chan, NvSub2D, 0x0804, 1);
+	OUT_RING(chan, format);
+	BEGIN_RING(chan, NvSub2D, 0x0800, 1);
+	OUT_RING(chan, 1);
+	BEGIN_RING(chan, NvSub2D, 0x0808, 3);
+	OUT_RING(chan, 0);
+	OUT_RING(chan, 0);
+	OUT_RING(chan, 0);
+	BEGIN_RING(chan, NvSub2D, 0x081c, 1);
+	OUT_RING(chan, 1);
+	BEGIN_RING(chan, NvSub2D, 0x0840, 4);
+	OUT_RING(chan, 0);
+	OUT_RING(chan, 1);
+	OUT_RING(chan, 0);
+	OUT_RING(chan, 1);
+	BEGIN_RING(chan, NvSub2D, 0x0200, 2);
+	OUT_RING(chan, format);
+	OUT_RING(chan, 1);
+	BEGIN_RING(chan, NvSub2D, 0x0214, 5);
+	OUT_RING(chan, info->fix.line_length);
+	OUT_RING(chan, info->var.xres_virtual);
+	OUT_RING(chan, info->var.yres_virtual);
+	OUT_RING(chan, 0);
+	OUT_RING(chan, info->fix.smem_start - dev_priv->fb_phys +
+			 dev_priv->vm_vram_base);
+	BEGIN_RING(chan, NvSub2D, 0x0230, 2);
+	OUT_RING(chan, format);
+	OUT_RING(chan, 1);
+	BEGIN_RING(chan, NvSub2D, 0x0244, 5);
+	OUT_RING(chan, info->fix.line_length);
+	OUT_RING(chan, info->var.xres_virtual);
+	OUT_RING(chan, info->var.yres_virtual);
+	OUT_RING(chan, 0);
+	OUT_RING(chan, info->fix.smem_start - dev_priv->fb_phys +
+			 dev_priv->vm_vram_base);
+
+	return 0;
+}
+
diff --git a/drivers/gpu/drm/nouveau/nv50_fifo.c b/drivers/gpu/drm/nouveau/nv50_fifo.c
new file mode 100644
index 000000000000..e20c0e2474f3
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv50_fifo.c
@@ -0,0 +1,500 @@
+/*
+ * Copyright (C) 2007 Ben Skeggs.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "drmP.h"
+#include "drm.h"
+#include "nouveau_drv.h"
+
+struct nv50_fifo_priv {
+	struct nouveau_gpuobj_ref *thingo[2];
+	int cur_thingo;
+};
+
+#define IS_G80 ((dev_priv->chipset & 0xf0) == 0x50)
+
+static void
+nv50_fifo_init_thingo(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nv50_fifo_priv *priv = dev_priv->engine.fifo.priv;
+	struct nouveau_gpuobj_ref *cur;
+	int i, nr;
+
+	NV_DEBUG(dev, "\n");
+
+	cur = priv->thingo[priv->cur_thingo];
+	priv->cur_thingo = !priv->cur_thingo;
+
+	/* We never schedule channel 0 or 127 */
+	dev_priv->engine.instmem.prepare_access(dev, true);
+	for (i = 1, nr = 0; i < 127; i++) {
+		if (dev_priv->fifos[i] && dev_priv->fifos[i]->ramfc)
+			nv_wo32(dev, cur->gpuobj, nr++, i);
+	}
+	dev_priv->engine.instmem.finish_access(dev);
+
+	nv_wr32(dev, 0x32f4, cur->instance >> 12);
+	nv_wr32(dev, 0x32ec, nr);
+	nv_wr32(dev, 0x2500, 0x101);
+}
+
+static int
+nv50_fifo_channel_enable(struct drm_device *dev, int channel, bool nt)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_channel *chan = dev_priv->fifos[channel];
+	uint32_t inst;
+
+	NV_DEBUG(dev, "ch%d\n", channel);
+
+	if (!chan->ramfc)
+		return -EINVAL;
+
+	if (IS_G80)
+		inst = chan->ramfc->instance >> 12;
+	else
+		inst = chan->ramfc->instance >> 8;
+	nv_wr32(dev, NV50_PFIFO_CTX_TABLE(channel),
+		 inst | NV50_PFIFO_CTX_TABLE_CHANNEL_ENABLED);
+
+	if (!nt)
+		nv50_fifo_init_thingo(dev);
+	return 0;
+}
+
+static void
+nv50_fifo_channel_disable(struct drm_device *dev, int channel, bool nt)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t inst;
+
+	NV_DEBUG(dev, "ch%d, nt=%d\n", channel, nt);
+
+	if (IS_G80)
+		inst = NV50_PFIFO_CTX_TABLE_INSTANCE_MASK_G80;
+	else
+		inst = NV50_PFIFO_CTX_TABLE_INSTANCE_MASK_G84;
+	nv_wr32(dev, NV50_PFIFO_CTX_TABLE(channel), inst);
+
+	if (!nt)
+		nv50_fifo_init_thingo(dev);
+}
+
+static void
+nv50_fifo_init_reset(struct drm_device *dev)
+{
+	uint32_t pmc_e = NV_PMC_ENABLE_PFIFO;
+
+	NV_DEBUG(dev, "\n");
+
+	nv_wr32(dev, NV03_PMC_ENABLE, nv_rd32(dev, NV03_PMC_ENABLE) & ~pmc_e);
+	nv_wr32(dev, NV03_PMC_ENABLE, nv_rd32(dev, NV03_PMC_ENABLE) |  pmc_e);
+}
+
+static void
+nv50_fifo_init_intr(struct drm_device *dev)
+{
+	NV_DEBUG(dev, "\n");
+
+	nv_wr32(dev, NV03_PFIFO_INTR_0, 0xFFFFFFFF);
+	nv_wr32(dev, NV03_PFIFO_INTR_EN_0, 0xFFFFFFFF);
+}
+
+static void
+nv50_fifo_init_context_table(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	int i;
+
+	NV_DEBUG(dev, "\n");
+
+	for (i = 0; i < NV50_PFIFO_CTX_TABLE__SIZE; i++) {
+		if (dev_priv->fifos[i])
+			nv50_fifo_channel_enable(dev, i, true);
+		else
+			nv50_fifo_channel_disable(dev, i, true);
+	}
+
+	nv50_fifo_init_thingo(dev);
+}
+
+static void
+nv50_fifo_init_regs__nv(struct drm_device *dev)
+{
+	NV_DEBUG(dev, "\n");
+
+	nv_wr32(dev, 0x250c, 0x6f3cfc34);
+}
+
+static void
+nv50_fifo_init_regs(struct drm_device *dev)
+{
+	NV_DEBUG(dev, "\n");
+
+	nv_wr32(dev, 0x2500, 0);
+	nv_wr32(dev, 0x3250, 0);
+	nv_wr32(dev, 0x3220, 0);
+	nv_wr32(dev, 0x3204, 0);
+	nv_wr32(dev, 0x3210, 0);
+	nv_wr32(dev, 0x3270, 0);
+
+	/* Enable dummy channels setup by nv50_instmem.c */
+	nv50_fifo_channel_enable(dev, 0, true);
+	nv50_fifo_channel_enable(dev, 127, true);
+}
+
+int
+nv50_fifo_init(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nv50_fifo_priv *priv;
+	int ret;
+
+	NV_DEBUG(dev, "\n");
+
+	priv = dev_priv->engine.fifo.priv;
+	if (priv) {
+		priv->cur_thingo = !priv->cur_thingo;
+		goto just_reset;
+	}
+
+	priv = kzalloc(sizeof(*priv), GFP_KERNEL);
+	if (!priv)
+		return -ENOMEM;
+	dev_priv->engine.fifo.priv = priv;
+
+	ret = nouveau_gpuobj_new_ref(dev, NULL, NULL, 0, 128*4, 0x1000,
+				     NVOBJ_FLAG_ZERO_ALLOC, &priv->thingo[0]);
+	if (ret) {
+		NV_ERROR(dev, "error creating thingo0: %d\n", ret);
+		return ret;
+	}
+
+	ret = nouveau_gpuobj_new_ref(dev, NULL, NULL, 0, 128*4, 0x1000,
+				     NVOBJ_FLAG_ZERO_ALLOC, &priv->thingo[1]);
+	if (ret) {
+		NV_ERROR(dev, "error creating thingo1: %d\n", ret);
+		return ret;
+	}
+
+just_reset:
+	nv50_fifo_init_reset(dev);
+	nv50_fifo_init_intr(dev);
+	nv50_fifo_init_context_table(dev);
+	nv50_fifo_init_regs__nv(dev);
+	nv50_fifo_init_regs(dev);
+	dev_priv->engine.fifo.enable(dev);
+	dev_priv->engine.fifo.reassign(dev, true);
+
+	return 0;
+}
+
+void
+nv50_fifo_takedown(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nv50_fifo_priv *priv = dev_priv->engine.fifo.priv;
+
+	NV_DEBUG(dev, "\n");
+
+	if (!priv)
+		return;
+
+	nouveau_gpuobj_ref_del(dev, &priv->thingo[0]);
+	nouveau_gpuobj_ref_del(dev, &priv->thingo[1]);
+
+	dev_priv->engine.fifo.priv = NULL;
+	kfree(priv);
+}
+
+int
+nv50_fifo_channel_id(struct drm_device *dev)
+{
+	return nv_rd32(dev, NV03_PFIFO_CACHE1_PUSH1) &
+			NV50_PFIFO_CACHE1_PUSH1_CHID_MASK;
+}
+
+int
+nv50_fifo_create_context(struct nouveau_channel *chan)
+{
+	struct drm_device *dev = chan->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_gpuobj *ramfc = NULL;
+	unsigned long flags;
+	int ret;
+
+	NV_DEBUG(dev, "ch%d\n", chan->id);
+
+	if (IS_G80) {
+		uint32_t ramin_poffset = chan->ramin->gpuobj->im_pramin->start;
+		uint32_t ramin_voffset = chan->ramin->gpuobj->im_backing_start;
+
+		ret = nouveau_gpuobj_new_fake(dev, ramin_poffset, ramin_voffset,
+					      0x100, NVOBJ_FLAG_ZERO_ALLOC |
+					      NVOBJ_FLAG_ZERO_FREE, &ramfc,
+					      &chan->ramfc);
+		if (ret)
+			return ret;
+
+		ret = nouveau_gpuobj_new_fake(dev, ramin_poffset + 0x0400,
+					      ramin_voffset + 0x0400, 4096,
+					      0, NULL, &chan->cache);
+		if (ret)
+			return ret;
+	} else {
+		ret = nouveau_gpuobj_new_ref(dev, chan, NULL, 0, 0x100, 256,
+					     NVOBJ_FLAG_ZERO_ALLOC |
+					     NVOBJ_FLAG_ZERO_FREE,
+					     &chan->ramfc);
+		if (ret)
+			return ret;
+		ramfc = chan->ramfc->gpuobj;
+
+		ret = nouveau_gpuobj_new_ref(dev, chan, NULL, 0, 4096, 1024,
+					     0, &chan->cache);
+		if (ret)
+			return ret;
+	}
+
+	spin_lock_irqsave(&dev_priv->context_switch_lock, flags);
+
+	dev_priv->engine.instmem.prepare_access(dev, true);
+
+	nv_wo32(dev, ramfc, 0x48/4, chan->pushbuf->instance >> 4);
+	nv_wo32(dev, ramfc, 0x80/4, (0xc << 24) | (chan->ramht->instance >> 4));
+	nv_wo32(dev, ramfc, 0x44/4, 0x2101ffff);
+	nv_wo32(dev, ramfc, 0x60/4, 0x7fffffff);
+	nv_wo32(dev, ramfc, 0x40/4, 0x00000000);
+	nv_wo32(dev, ramfc, 0x7c/4, 0x30000001);
+	nv_wo32(dev, ramfc, 0x78/4, 0x00000000);
+	nv_wo32(dev, ramfc, 0x3c/4, 0x403f6078);
+	nv_wo32(dev, ramfc, 0x50/4, chan->pushbuf_base +
+				    chan->dma.ib_base * 4);
+	nv_wo32(dev, ramfc, 0x54/4, drm_order(chan->dma.ib_max + 1) << 16);
+
+	if (!IS_G80) {
+		nv_wo32(dev, chan->ramin->gpuobj, 0, chan->id);
+		nv_wo32(dev, chan->ramin->gpuobj, 1,
+						chan->ramfc->instance >> 8);
+
+		nv_wo32(dev, ramfc, 0x88/4, chan->cache->instance >> 10);
+		nv_wo32(dev, ramfc, 0x98/4, chan->ramin->instance >> 12);
+	}
+
+	dev_priv->engine.instmem.finish_access(dev);
+
+	ret = nv50_fifo_channel_enable(dev, chan->id, false);
+	if (ret) {
+		NV_ERROR(dev, "error enabling ch%d: %d\n", chan->id, ret);
+		spin_unlock_irqrestore(&dev_priv->context_switch_lock, flags);
+		nouveau_gpuobj_ref_del(dev, &chan->ramfc);
+		return ret;
+	}
+
+	spin_unlock_irqrestore(&dev_priv->context_switch_lock, flags);
+	return 0;
+}
+
+void
+nv50_fifo_destroy_context(struct nouveau_channel *chan)
+{
+	struct drm_device *dev = chan->dev;
+	struct nouveau_gpuobj_ref *ramfc = chan->ramfc;
+
+	NV_DEBUG(dev, "ch%d\n", chan->id);
+
+	/* This will ensure the channel is seen as disabled. */
+	chan->ramfc = NULL;
+	nv50_fifo_channel_disable(dev, chan->id, false);
+
+	/* Dummy channel, also used on ch 127 */
+	if (chan->id == 0)
+		nv50_fifo_channel_disable(dev, 127, false);
+
+	nouveau_gpuobj_ref_del(dev, &ramfc);
+	nouveau_gpuobj_ref_del(dev, &chan->cache);
+}
+
+int
+nv50_fifo_load_context(struct nouveau_channel *chan)
+{
+	struct drm_device *dev = chan->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_gpuobj *ramfc = chan->ramfc->gpuobj;
+	struct nouveau_gpuobj *cache = chan->cache->gpuobj;
+	int ptr, cnt;
+
+	NV_DEBUG(dev, "ch%d\n", chan->id);
+
+	dev_priv->engine.instmem.prepare_access(dev, false);
+
+	nv_wr32(dev, 0x3330, nv_ro32(dev, ramfc, 0x00/4));
+	nv_wr32(dev, 0x3334, nv_ro32(dev, ramfc, 0x04/4));
+	nv_wr32(dev, 0x3240, nv_ro32(dev, ramfc, 0x08/4));
+	nv_wr32(dev, 0x3320, nv_ro32(dev, ramfc, 0x0c/4));
+	nv_wr32(dev, 0x3244, nv_ro32(dev, ramfc, 0x10/4));
+	nv_wr32(dev, 0x3328, nv_ro32(dev, ramfc, 0x14/4));
+	nv_wr32(dev, 0x3368, nv_ro32(dev, ramfc, 0x18/4));
+	nv_wr32(dev, 0x336c, nv_ro32(dev, ramfc, 0x1c/4));
+	nv_wr32(dev, 0x3370, nv_ro32(dev, ramfc, 0x20/4));
+	nv_wr32(dev, 0x3374, nv_ro32(dev, ramfc, 0x24/4));
+	nv_wr32(dev, 0x3378, nv_ro32(dev, ramfc, 0x28/4));
+	nv_wr32(dev, 0x337c, nv_ro32(dev, ramfc, 0x2c/4));
+	nv_wr32(dev, 0x3228, nv_ro32(dev, ramfc, 0x30/4));
+	nv_wr32(dev, 0x3364, nv_ro32(dev, ramfc, 0x34/4));
+	nv_wr32(dev, 0x32a0, nv_ro32(dev, ramfc, 0x38/4));
+	nv_wr32(dev, 0x3224, nv_ro32(dev, ramfc, 0x3c/4));
+	nv_wr32(dev, 0x324c, nv_ro32(dev, ramfc, 0x40/4));
+	nv_wr32(dev, 0x2044, nv_ro32(dev, ramfc, 0x44/4));
+	nv_wr32(dev, 0x322c, nv_ro32(dev, ramfc, 0x48/4));
+	nv_wr32(dev, 0x3234, nv_ro32(dev, ramfc, 0x4c/4));
+	nv_wr32(dev, 0x3340, nv_ro32(dev, ramfc, 0x50/4));
+	nv_wr32(dev, 0x3344, nv_ro32(dev, ramfc, 0x54/4));
+	nv_wr32(dev, 0x3280, nv_ro32(dev, ramfc, 0x58/4));
+	nv_wr32(dev, 0x3254, nv_ro32(dev, ramfc, 0x5c/4));
+	nv_wr32(dev, 0x3260, nv_ro32(dev, ramfc, 0x60/4));
+	nv_wr32(dev, 0x3264, nv_ro32(dev, ramfc, 0x64/4));
+	nv_wr32(dev, 0x3268, nv_ro32(dev, ramfc, 0x68/4));
+	nv_wr32(dev, 0x326c, nv_ro32(dev, ramfc, 0x6c/4));
+	nv_wr32(dev, 0x32e4, nv_ro32(dev, ramfc, 0x70/4));
+	nv_wr32(dev, 0x3248, nv_ro32(dev, ramfc, 0x74/4));
+	nv_wr32(dev, 0x2088, nv_ro32(dev, ramfc, 0x78/4));
+	nv_wr32(dev, 0x2058, nv_ro32(dev, ramfc, 0x7c/4));
+	nv_wr32(dev, 0x2210, nv_ro32(dev, ramfc, 0x80/4));
+
+	cnt = nv_ro32(dev, ramfc, 0x84/4);
+	for (ptr = 0; ptr < cnt; ptr++) {
+		nv_wr32(dev, NV40_PFIFO_CACHE1_METHOD(ptr),
+			nv_ro32(dev, cache, (ptr * 2) + 0));
+		nv_wr32(dev, NV40_PFIFO_CACHE1_DATA(ptr),
+			nv_ro32(dev, cache, (ptr * 2) + 1));
+	}
+	nv_wr32(dev, NV03_PFIFO_CACHE1_PUT, cnt << 2);
+	nv_wr32(dev, NV03_PFIFO_CACHE1_GET, 0);
+
+	/* guessing that all the 0x34xx regs aren't on NV50 */
+	if (!IS_G80) {
+		nv_wr32(dev, 0x340c, nv_ro32(dev, ramfc, 0x88/4));
+		nv_wr32(dev, 0x3400, nv_ro32(dev, ramfc, 0x8c/4));
+		nv_wr32(dev, 0x3404, nv_ro32(dev, ramfc, 0x90/4));
+		nv_wr32(dev, 0x3408, nv_ro32(dev, ramfc, 0x94/4));
+		nv_wr32(dev, 0x3410, nv_ro32(dev, ramfc, 0x98/4));
+	}
+
+	dev_priv->engine.instmem.finish_access(dev);
+
+	nv_wr32(dev, NV03_PFIFO_CACHE1_PUSH1, chan->id | (1<<16));
+	return 0;
+}
+
+int
+nv50_fifo_unload_context(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_fifo_engine *pfifo = &dev_priv->engine.fifo;
+	struct nouveau_gpuobj *ramfc, *cache;
+	struct nouveau_channel *chan = NULL;
+	int chid, get, put, ptr;
+
+	NV_DEBUG(dev, "\n");
+
+	chid = pfifo->channel_id(dev);
+	if (chid < 1 || chid >= dev_priv->engine.fifo.channels - 1)
+		return 0;
+
+	chan = dev_priv->fifos[chid];
+	if (!chan) {
+		NV_ERROR(dev, "Inactive channel on PFIFO: %d\n", chid);
+		return -EINVAL;
+	}
+	NV_DEBUG(dev, "ch%d\n", chan->id);
+	ramfc = chan->ramfc->gpuobj;
+	cache = chan->cache->gpuobj;
+
+	dev_priv->engine.instmem.prepare_access(dev, true);
+
+	nv_wo32(dev, ramfc, 0x00/4, nv_rd32(dev, 0x3330));
+	nv_wo32(dev, ramfc, 0x04/4, nv_rd32(dev, 0x3334));
+	nv_wo32(dev, ramfc, 0x08/4, nv_rd32(dev, 0x3240));
+	nv_wo32(dev, ramfc, 0x0c/4, nv_rd32(dev, 0x3320));
+	nv_wo32(dev, ramfc, 0x10/4, nv_rd32(dev, 0x3244));
+	nv_wo32(dev, ramfc, 0x14/4, nv_rd32(dev, 0x3328));
+	nv_wo32(dev, ramfc, 0x18/4, nv_rd32(dev, 0x3368));
+	nv_wo32(dev, ramfc, 0x1c/4, nv_rd32(dev, 0x336c));
+	nv_wo32(dev, ramfc, 0x20/4, nv_rd32(dev, 0x3370));
+	nv_wo32(dev, ramfc, 0x24/4, nv_rd32(dev, 0x3374));
+	nv_wo32(dev, ramfc, 0x28/4, nv_rd32(dev, 0x3378));
+	nv_wo32(dev, ramfc, 0x2c/4, nv_rd32(dev, 0x337c));
+	nv_wo32(dev, ramfc, 0x30/4, nv_rd32(dev, 0x3228));
+	nv_wo32(dev, ramfc, 0x34/4, nv_rd32(dev, 0x3364));
+	nv_wo32(dev, ramfc, 0x38/4, nv_rd32(dev, 0x32a0));
+	nv_wo32(dev, ramfc, 0x3c/4, nv_rd32(dev, 0x3224));
+	nv_wo32(dev, ramfc, 0x40/4, nv_rd32(dev, 0x324c));
+	nv_wo32(dev, ramfc, 0x44/4, nv_rd32(dev, 0x2044));
+	nv_wo32(dev, ramfc, 0x48/4, nv_rd32(dev, 0x322c));
+	nv_wo32(dev, ramfc, 0x4c/4, nv_rd32(dev, 0x3234));
+	nv_wo32(dev, ramfc, 0x50/4, nv_rd32(dev, 0x3340));
+	nv_wo32(dev, ramfc, 0x54/4, nv_rd32(dev, 0x3344));
+	nv_wo32(dev, ramfc, 0x58/4, nv_rd32(dev, 0x3280));
+	nv_wo32(dev, ramfc, 0x5c/4, nv_rd32(dev, 0x3254));
+	nv_wo32(dev, ramfc, 0x60/4, nv_rd32(dev, 0x3260));
+	nv_wo32(dev, ramfc, 0x64/4, nv_rd32(dev, 0x3264));
+	nv_wo32(dev, ramfc, 0x68/4, nv_rd32(dev, 0x3268));
+	nv_wo32(dev, ramfc, 0x6c/4, nv_rd32(dev, 0x326c));
+	nv_wo32(dev, ramfc, 0x70/4, nv_rd32(dev, 0x32e4));
+	nv_wo32(dev, ramfc, 0x74/4, nv_rd32(dev, 0x3248));
+	nv_wo32(dev, ramfc, 0x78/4, nv_rd32(dev, 0x2088));
+	nv_wo32(dev, ramfc, 0x7c/4, nv_rd32(dev, 0x2058));
+	nv_wo32(dev, ramfc, 0x80/4, nv_rd32(dev, 0x2210));
+
+	put = (nv_rd32(dev, NV03_PFIFO_CACHE1_PUT) & 0x7ff) >> 2;
+	get = (nv_rd32(dev, NV03_PFIFO_CACHE1_GET) & 0x7ff) >> 2;
+	ptr = 0;
+	while (put != get) {
+		nv_wo32(dev, cache, ptr++,
+			    nv_rd32(dev, NV40_PFIFO_CACHE1_METHOD(get)));
+		nv_wo32(dev, cache, ptr++,
+			    nv_rd32(dev, NV40_PFIFO_CACHE1_DATA(get)));
+		get = (get + 1) & 0x1ff;
+	}
+
+	/* guessing that all the 0x34xx regs aren't on NV50 */
+	if (!IS_G80) {
+		nv_wo32(dev, ramfc, 0x84/4, ptr >> 1);
+		nv_wo32(dev, ramfc, 0x88/4, nv_rd32(dev, 0x340c));
+		nv_wo32(dev, ramfc, 0x8c/4, nv_rd32(dev, 0x3400));
+		nv_wo32(dev, ramfc, 0x90/4, nv_rd32(dev, 0x3404));
+		nv_wo32(dev, ramfc, 0x94/4, nv_rd32(dev, 0x3408));
+		nv_wo32(dev, ramfc, 0x98/4, nv_rd32(dev, 0x3410));
+	}
+
+	dev_priv->engine.instmem.finish_access(dev);
+
+	/*XXX: probably reload ch127 (NULL) state back too */
+	nv_wr32(dev, NV03_PFIFO_CACHE1_PUSH1, 127);
+	return 0;
+}
+
diff --git a/drivers/gpu/drm/nouveau/nv50_graph.c b/drivers/gpu/drm/nouveau/nv50_graph.c
new file mode 100644
index 000000000000..857a09671a39
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv50_graph.c
@@ -0,0 +1,404 @@
+/*
+ * Copyright (C) 2007 Ben Skeggs.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "drmP.h"
+#include "drm.h"
+#include "nouveau_drv.h"
+
+#include "nouveau_grctx.h"
+
+#define IS_G80 ((dev_priv->chipset & 0xf0) == 0x50)
+
+static void
+nv50_graph_init_reset(struct drm_device *dev)
+{
+	uint32_t pmc_e = NV_PMC_ENABLE_PGRAPH | (1 << 21);
+
+	NV_DEBUG(dev, "\n");
+
+	nv_wr32(dev, NV03_PMC_ENABLE, nv_rd32(dev, NV03_PMC_ENABLE) & ~pmc_e);
+	nv_wr32(dev, NV03_PMC_ENABLE, nv_rd32(dev, NV03_PMC_ENABLE) |  pmc_e);
+}
+
+static void
+nv50_graph_init_intr(struct drm_device *dev)
+{
+	NV_DEBUG(dev, "\n");
+
+	nv_wr32(dev, NV03_PGRAPH_INTR, 0xffffffff);
+	nv_wr32(dev, 0x400138, 0xffffffff);
+	nv_wr32(dev, NV40_PGRAPH_INTR_EN, 0xffffffff);
+}
+
+static void
+nv50_graph_init_regs__nv(struct drm_device *dev)
+{
+	NV_DEBUG(dev, "\n");
+
+	nv_wr32(dev, 0x400804, 0xc0000000);
+	nv_wr32(dev, 0x406800, 0xc0000000);
+	nv_wr32(dev, 0x400c04, 0xc0000000);
+	nv_wr32(dev, 0x401800, 0xc0000000);
+	nv_wr32(dev, 0x405018, 0xc0000000);
+	nv_wr32(dev, 0x402000, 0xc0000000);
+
+	nv_wr32(dev, 0x400108, 0xffffffff);
+
+	nv_wr32(dev, 0x400824, 0x00004000);
+	nv_wr32(dev, 0x400500, 0x00010001);
+}
+
+static void
+nv50_graph_init_regs(struct drm_device *dev)
+{
+	NV_DEBUG(dev, "\n");
+
+	nv_wr32(dev, NV04_PGRAPH_DEBUG_3,
+				(1 << 2) /* HW_CONTEXT_SWITCH_ENABLED */);
+	nv_wr32(dev, 0x402ca8, 0x800);
+}
+
+static int
+nv50_graph_init_ctxctl(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	NV_DEBUG(dev, "\n");
+
+	if (nouveau_ctxfw) {
+		nouveau_grctx_prog_load(dev);
+		dev_priv->engine.graph.grctx_size = 0x70000;
+	}
+	if (!dev_priv->engine.graph.ctxprog) {
+		struct nouveau_grctx ctx = {};
+		uint32_t *cp = kmalloc(512 * 4, GFP_KERNEL);
+		int i;
+		if (!cp) {
+			NV_ERROR(dev, "Couldn't alloc ctxprog! Disabling acceleration.\n");
+			dev_priv->engine.graph.accel_blocked = true;
+			return 0;
+		}
+		ctx.dev = dev;
+		ctx.mode = NOUVEAU_GRCTX_PROG;
+		ctx.data = cp;
+		ctx.ctxprog_max = 512;
+		if (!nv50_grctx_init(&ctx)) {
+			dev_priv->engine.graph.grctx_size = ctx.ctxvals_pos * 4;
+
+			nv_wr32(dev, NV40_PGRAPH_CTXCTL_UCODE_INDEX, 0);
+			for (i = 0; i < ctx.ctxprog_len; i++)
+				nv_wr32(dev, NV40_PGRAPH_CTXCTL_UCODE_DATA, cp[i]);
+		} else {
+			dev_priv->engine.graph.accel_blocked = true;
+		}
+		kfree(cp);
+	}
+
+	nv_wr32(dev, 0x400320, 4);
+	nv_wr32(dev, NV40_PGRAPH_CTXCTL_CUR, 0);
+	nv_wr32(dev, NV20_PGRAPH_CHANNEL_CTX_POINTER, 0);
+	return 0;
+}
+
+int
+nv50_graph_init(struct drm_device *dev)
+{
+	int ret;
+
+	NV_DEBUG(dev, "\n");
+
+	nv50_graph_init_reset(dev);
+	nv50_graph_init_regs__nv(dev);
+	nv50_graph_init_regs(dev);
+	nv50_graph_init_intr(dev);
+
+	ret = nv50_graph_init_ctxctl(dev);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+void
+nv50_graph_takedown(struct drm_device *dev)
+{
+	NV_DEBUG(dev, "\n");
+	nouveau_grctx_fini(dev);
+}
+
+void
+nv50_graph_fifo_access(struct drm_device *dev, bool enabled)
+{
+	const uint32_t mask = 0x00010001;
+
+	if (enabled)
+		nv_wr32(dev, 0x400500, nv_rd32(dev, 0x400500) | mask);
+	else
+		nv_wr32(dev, 0x400500, nv_rd32(dev, 0x400500) & ~mask);
+}
+
+struct nouveau_channel *
+nv50_graph_channel(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	uint32_t inst;
+	int i;
+
+	/* Be sure we're not in the middle of a context switch or bad things
+	 * will happen, such as unloading the wrong pgraph context.
+	 */
+	if (!nv_wait(0x400300, 0x00000001, 0x00000000))
+		NV_ERROR(dev, "Ctxprog is still running\n");
+
+	inst = nv_rd32(dev, NV50_PGRAPH_CTXCTL_CUR);
+	if (!(inst & NV50_PGRAPH_CTXCTL_CUR_LOADED))
+		return NULL;
+	inst = (inst & NV50_PGRAPH_CTXCTL_CUR_INSTANCE) << 12;
+
+	for (i = 0; i < dev_priv->engine.fifo.channels; i++) {
+		struct nouveau_channel *chan = dev_priv->fifos[i];
+
+		if (chan && chan->ramin && chan->ramin->instance == inst)
+			return chan;
+	}
+
+	return NULL;
+}
+
+int
+nv50_graph_create_context(struct nouveau_channel *chan)
+{
+	struct drm_device *dev = chan->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_gpuobj *ramin = chan->ramin->gpuobj;
+	struct nouveau_gpuobj *ctx;
+	struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
+	int hdr, ret;
+
+	NV_DEBUG(dev, "ch%d\n", chan->id);
+
+	ret = nouveau_gpuobj_new_ref(dev, chan, NULL, 0, pgraph->grctx_size,
+				     0x1000, NVOBJ_FLAG_ZERO_ALLOC |
+				     NVOBJ_FLAG_ZERO_FREE, &chan->ramin_grctx);
+	if (ret)
+		return ret;
+	ctx = chan->ramin_grctx->gpuobj;
+
+	hdr = IS_G80 ? 0x200 : 0x20;
+	dev_priv->engine.instmem.prepare_access(dev, true);
+	nv_wo32(dev, ramin, (hdr + 0x00)/4, 0x00190002);
+	nv_wo32(dev, ramin, (hdr + 0x04)/4, chan->ramin_grctx->instance +
+					   pgraph->grctx_size - 1);
+	nv_wo32(dev, ramin, (hdr + 0x08)/4, chan->ramin_grctx->instance);
+	nv_wo32(dev, ramin, (hdr + 0x0c)/4, 0);
+	nv_wo32(dev, ramin, (hdr + 0x10)/4, 0);
+	nv_wo32(dev, ramin, (hdr + 0x14)/4, 0x00010000);
+	dev_priv->engine.instmem.finish_access(dev);
+
+	dev_priv->engine.instmem.prepare_access(dev, true);
+	if (!pgraph->ctxprog) {
+		struct nouveau_grctx ctx = {};
+		ctx.dev = chan->dev;
+		ctx.mode = NOUVEAU_GRCTX_VALS;
+		ctx.data = chan->ramin_grctx->gpuobj;
+		nv50_grctx_init(&ctx);
+	} else {
+		nouveau_grctx_vals_load(dev, ctx);
+	}
+	nv_wo32(dev, ctx, 0x00000/4, chan->ramin->instance >> 12);
+	if ((dev_priv->chipset & 0xf0) == 0xa0)
+		nv_wo32(dev, ctx, 0x00004/4, 0x00000000);
+	else
+		nv_wo32(dev, ctx, 0x0011c/4, 0x00000000);
+	dev_priv->engine.instmem.finish_access(dev);
+
+	return 0;
+}
+
+void
+nv50_graph_destroy_context(struct nouveau_channel *chan)
+{
+	struct drm_device *dev = chan->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	int i, hdr = IS_G80 ? 0x200 : 0x20;
+
+	NV_DEBUG(dev, "ch%d\n", chan->id);
+
+	if (!chan->ramin || !chan->ramin->gpuobj)
+		return;
+
+	dev_priv->engine.instmem.prepare_access(dev, true);
+	for (i = hdr; i < hdr + 24; i += 4)
+		nv_wo32(dev, chan->ramin->gpuobj, i/4, 0);
+	dev_priv->engine.instmem.finish_access(dev);
+
+	nouveau_gpuobj_ref_del(dev, &chan->ramin_grctx);
+}
+
+static int
+nv50_graph_do_load_context(struct drm_device *dev, uint32_t inst)
+{
+	uint32_t fifo = nv_rd32(dev, 0x400500);
+
+	nv_wr32(dev, 0x400500, fifo & ~1);
+	nv_wr32(dev, 0x400784, inst);
+	nv_wr32(dev, 0x400824, nv_rd32(dev, 0x400824) | 0x40);
+	nv_wr32(dev, 0x400320, nv_rd32(dev, 0x400320) | 0x11);
+	nv_wr32(dev, 0x400040, 0xffffffff);
+	(void)nv_rd32(dev, 0x400040);
+	nv_wr32(dev, 0x400040, 0x00000000);
+	nv_wr32(dev, 0x400304, nv_rd32(dev, 0x400304) | 1);
+
+	if (nouveau_wait_for_idle(dev))
+		nv_wr32(dev, 0x40032c, inst | (1<<31));
+	nv_wr32(dev, 0x400500, fifo);
+
+	return 0;
+}
+
+int
+nv50_graph_load_context(struct nouveau_channel *chan)
+{
+	uint32_t inst = chan->ramin->instance >> 12;
+
+	NV_DEBUG(chan->dev, "ch%d\n", chan->id);
+	return nv50_graph_do_load_context(chan->dev, inst);
+}
+
+int
+nv50_graph_unload_context(struct drm_device *dev)
+{
+	uint32_t inst;
+
+	inst  = nv_rd32(dev, NV50_PGRAPH_CTXCTL_CUR);
+	if (!(inst & NV50_PGRAPH_CTXCTL_CUR_LOADED))
+		return 0;
+	inst &= NV50_PGRAPH_CTXCTL_CUR_INSTANCE;
+
+	nouveau_wait_for_idle(dev);
+	nv_wr32(dev, 0x400784, inst);
+	nv_wr32(dev, 0x400824, nv_rd32(dev, 0x400824) | 0x20);
+	nv_wr32(dev, 0x400304, nv_rd32(dev, 0x400304) | 0x01);
+	nouveau_wait_for_idle(dev);
+
+	nv_wr32(dev, NV50_PGRAPH_CTXCTL_CUR, inst);
+	return 0;
+}
+
+void
+nv50_graph_context_switch(struct drm_device *dev)
+{
+	uint32_t inst;
+
+	nv50_graph_unload_context(dev);
+
+	inst  = nv_rd32(dev, NV50_PGRAPH_CTXCTL_NEXT);
+	inst &= NV50_PGRAPH_CTXCTL_NEXT_INSTANCE;
+	nv50_graph_do_load_context(dev, inst);
+
+	nv_wr32(dev, NV40_PGRAPH_INTR_EN, nv_rd32(dev,
+		NV40_PGRAPH_INTR_EN) | NV_PGRAPH_INTR_CONTEXT_SWITCH);
+}
+
+static int
+nv50_graph_nvsw_dma_vblsem(struct nouveau_channel *chan, int grclass,
+			   int mthd, uint32_t data)
+{
+	struct nouveau_gpuobj_ref *ref = NULL;
+
+	if (nouveau_gpuobj_ref_find(chan, data, &ref))
+		return -ENOENT;
+
+	if (nouveau_notifier_offset(ref->gpuobj, NULL))
+		return -EINVAL;
+
+	chan->nvsw.vblsem = ref->gpuobj;
+	chan->nvsw.vblsem_offset = ~0;
+	return 0;
+}
+
+static int
+nv50_graph_nvsw_vblsem_offset(struct nouveau_channel *chan, int grclass,
+			      int mthd, uint32_t data)
+{
+	if (nouveau_notifier_offset(chan->nvsw.vblsem, &data))
+		return -ERANGE;
+
+	chan->nvsw.vblsem_offset = data >> 2;
+	return 0;
+}
+
+static int
+nv50_graph_nvsw_vblsem_release_val(struct nouveau_channel *chan, int grclass,
+				   int mthd, uint32_t data)
+{
+	chan->nvsw.vblsem_rval = data;
+	return 0;
+}
+
+static int
+nv50_graph_nvsw_vblsem_release(struct nouveau_channel *chan, int grclass,
+			       int mthd, uint32_t data)
+{
+	struct drm_device *dev = chan->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	if (!chan->nvsw.vblsem || chan->nvsw.vblsem_offset == ~0 || data > 1)
+		return -EINVAL;
+
+	if (!(nv_rd32(dev, NV50_PDISPLAY_INTR_EN) &
+		      NV50_PDISPLAY_INTR_EN_VBLANK_CRTC_(data))) {
+		nv_wr32(dev, NV50_PDISPLAY_INTR_1,
+			NV50_PDISPLAY_INTR_1_VBLANK_CRTC_(data));
+		nv_wr32(dev, NV50_PDISPLAY_INTR_EN, nv_rd32(dev,
+			NV50_PDISPLAY_INTR_EN) |
+			NV50_PDISPLAY_INTR_EN_VBLANK_CRTC_(data));
+	}
+
+	list_add(&chan->nvsw.vbl_wait, &dev_priv->vbl_waiting);
+	return 0;
+}
+
+static struct nouveau_pgraph_object_method nv50_graph_nvsw_methods[] = {
+	{ 0x018c, nv50_graph_nvsw_dma_vblsem },
+	{ 0x0400, nv50_graph_nvsw_vblsem_offset },
+	{ 0x0404, nv50_graph_nvsw_vblsem_release_val },
+	{ 0x0408, nv50_graph_nvsw_vblsem_release },
+	{}
+};
+
+struct nouveau_pgraph_object_class nv50_graph_grclass[] = {
+	{ 0x506e, true, nv50_graph_nvsw_methods }, /* nvsw */
+	{ 0x0030, false, NULL }, /* null */
+	{ 0x5039, false, NULL }, /* m2mf */
+	{ 0x502d, false, NULL }, /* 2d */
+	{ 0x50c0, false, NULL }, /* compute */
+	{ 0x5097, false, NULL }, /* tesla (nv50) */
+	{ 0x8297, false, NULL }, /* tesla (nv80/nv90) */
+	{ 0x8397, false, NULL }, /* tesla (nva0) */
+	{ 0x8597, false, NULL }, /* tesla (nva8) */
+	{}
+};
diff --git a/drivers/gpu/drm/nouveau/nv50_grctx.c b/drivers/gpu/drm/nouveau/nv50_grctx.c
new file mode 100644
index 000000000000..d105fcd42ca0
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv50_grctx.c
@@ -0,0 +1,2367 @@
+/*
+ * Copyright 2009 Marcin Kościelnicki
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#define CP_FLAG_CLEAR                 0
+#define CP_FLAG_SET                   1
+#define CP_FLAG_SWAP_DIRECTION        ((0 * 32) + 0)
+#define CP_FLAG_SWAP_DIRECTION_LOAD   0
+#define CP_FLAG_SWAP_DIRECTION_SAVE   1
+#define CP_FLAG_UNK01                 ((0 * 32) + 1)
+#define CP_FLAG_UNK01_CLEAR           0
+#define CP_FLAG_UNK01_SET             1
+#define CP_FLAG_UNK03                 ((0 * 32) + 3)
+#define CP_FLAG_UNK03_CLEAR           0
+#define CP_FLAG_UNK03_SET             1
+#define CP_FLAG_USER_SAVE             ((0 * 32) + 5)
+#define CP_FLAG_USER_SAVE_NOT_PENDING 0
+#define CP_FLAG_USER_SAVE_PENDING     1
+#define CP_FLAG_USER_LOAD             ((0 * 32) + 6)
+#define CP_FLAG_USER_LOAD_NOT_PENDING 0
+#define CP_FLAG_USER_LOAD_PENDING     1
+#define CP_FLAG_UNK0B                 ((0 * 32) + 0xb)
+#define CP_FLAG_UNK0B_CLEAR           0
+#define CP_FLAG_UNK0B_SET             1
+#define CP_FLAG_UNK1D                 ((0 * 32) + 0x1d)
+#define CP_FLAG_UNK1D_CLEAR           0
+#define CP_FLAG_UNK1D_SET             1
+#define CP_FLAG_UNK20                 ((1 * 32) + 0)
+#define CP_FLAG_UNK20_CLEAR           0
+#define CP_FLAG_UNK20_SET             1
+#define CP_FLAG_STATUS                ((2 * 32) + 0)
+#define CP_FLAG_STATUS_BUSY           0
+#define CP_FLAG_STATUS_IDLE           1
+#define CP_FLAG_AUTO_SAVE             ((2 * 32) + 4)
+#define CP_FLAG_AUTO_SAVE_NOT_PENDING 0
+#define CP_FLAG_AUTO_SAVE_PENDING     1
+#define CP_FLAG_AUTO_LOAD             ((2 * 32) + 5)
+#define CP_FLAG_AUTO_LOAD_NOT_PENDING 0
+#define CP_FLAG_AUTO_LOAD_PENDING     1
+#define CP_FLAG_XFER                  ((2 * 32) + 11)
+#define CP_FLAG_XFER_IDLE             0
+#define CP_FLAG_XFER_BUSY             1
+#define CP_FLAG_NEWCTX                ((2 * 32) + 12)
+#define CP_FLAG_NEWCTX_BUSY           0
+#define CP_FLAG_NEWCTX_DONE           1
+#define CP_FLAG_ALWAYS                ((2 * 32) + 13)
+#define CP_FLAG_ALWAYS_FALSE          0
+#define CP_FLAG_ALWAYS_TRUE           1
+
+#define CP_CTX                   0x00100000
+#define CP_CTX_COUNT             0x000f0000
+#define CP_CTX_COUNT_SHIFT               16
+#define CP_CTX_REG               0x00003fff
+#define CP_LOAD_SR               0x00200000
+#define CP_LOAD_SR_VALUE         0x000fffff
+#define CP_BRA                   0x00400000
+#define CP_BRA_IP                0x0001ff00
+#define CP_BRA_IP_SHIFT                   8
+#define CP_BRA_IF_CLEAR          0x00000080
+#define CP_BRA_FLAG              0x0000007f
+#define CP_WAIT                  0x00500000
+#define CP_WAIT_SET              0x00000080
+#define CP_WAIT_FLAG             0x0000007f
+#define CP_SET                   0x00700000
+#define CP_SET_1                 0x00000080
+#define CP_SET_FLAG              0x0000007f
+#define CP_NEWCTX                0x00600004
+#define CP_NEXT_TO_SWAP          0x00600005
+#define CP_SET_CONTEXT_POINTER   0x00600006
+#define CP_SET_XFER_POINTER      0x00600007
+#define CP_ENABLE                0x00600009
+#define CP_END                   0x0060000c
+#define CP_NEXT_TO_CURRENT       0x0060000d
+#define CP_DISABLE1              0x0090ffff
+#define CP_DISABLE2              0x0091ffff
+#define CP_XFER_1      0x008000ff
+#define CP_XFER_2      0x008800ff
+#define CP_SEEK_1      0x00c000ff
+#define CP_SEEK_2      0x00c800ff
+
+#include "drmP.h"
+#include "nouveau_drv.h"
+#include "nouveau_grctx.h"
+
+/*
+ * This code deals with PGRAPH contexts on NV50 family cards. Like NV40, it's
+ * the GPU itself that does context-switching, but it needs a special
+ * microcode to do it. And it's the driver's task to supply this microcode,
+ * further known as ctxprog, as well as the initial context values, known
+ * as ctxvals.
+ *
+ * Without ctxprog, you cannot switch contexts. Not even in software, since
+ * the majority of context [xfer strands] isn't accessible directly. You're
+ * stuck with a single channel, and you also suffer all the problems resulting
+ * from missing ctxvals, since you cannot load them.
+ *
+ * Without ctxvals, you're stuck with PGRAPH's default context. It's enough to
+ * run 2d operations, but trying to utilise 3d or CUDA will just lock you up,
+ * since you don't have... some sort of needed setup.
+ *
+ * Nouveau will just disable acceleration if not given ctxprog + ctxvals, since
+ * it's too much hassle to handle no-ctxprog as a special case.
+ */
+
+/*
+ * How ctxprogs work.
+ *
+ * The ctxprog is written in its own kind of microcode, with very small and
+ * crappy set of available commands. You upload it to a small [512 insns]
+ * area of memory on PGRAPH, and it'll be run when PFIFO wants PGRAPH to
+ * switch channel. or when the driver explicitely requests it. Stuff visible
+ * to ctxprog consists of: PGRAPH MMIO registers, PGRAPH context strands,
+ * the per-channel context save area in VRAM [known as ctxvals or grctx],
+ * 4 flags registers, a scratch register, two grctx pointers, plus many
+ * random poorly-understood details.
+ *
+ * When ctxprog runs, it's supposed to check what operations are asked of it,
+ * save old context if requested, optionally reset PGRAPH and switch to the
+ * new channel, and load the new context. Context consists of three major
+ * parts: subset of MMIO registers and two "xfer areas".
+ */
+
+/* TODO:
+ *  - document unimplemented bits compared to nvidia
+ *  - NVAx: make a TP subroutine, use it.
+ *  - use 0x4008fc instead of 0x1540?
+ */
+
+enum cp_label {
+	cp_check_load = 1,
+	cp_setup_auto_load,
+	cp_setup_load,
+	cp_setup_save,
+	cp_swap_state,
+	cp_prepare_exit,
+	cp_exit,
+};
+
+static void nv50_graph_construct_mmio(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_xfer1(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_xfer2(struct nouveau_grctx *ctx);
+
+/* Main function: construct the ctxprog skeleton, call the other functions. */
+
+int
+nv50_grctx_init(struct nouveau_grctx *ctx)
+{
+	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
+
+	switch (dev_priv->chipset) {
+	case 0x50:
+	case 0x84:
+	case 0x86:
+	case 0x92:
+	case 0x94:
+	case 0x96:
+	case 0x98:
+	case 0xa0:
+	case 0xa5:
+	case 0xa8:
+	case 0xaa:
+	case 0xac:
+		break;
+	default:
+		NV_ERROR(ctx->dev, "I don't know how to make a ctxprog for "
+				   "your NV%x card.\n", dev_priv->chipset);
+		NV_ERROR(ctx->dev, "Disabling acceleration. Please contact "
+				   "the devs.\n");
+		return -ENOSYS;
+	}
+	/* decide whether we're loading/unloading the context */
+	cp_bra (ctx, AUTO_SAVE, PENDING, cp_setup_save);
+	cp_bra (ctx, USER_SAVE, PENDING, cp_setup_save);
+
+	cp_name(ctx, cp_check_load);
+	cp_bra (ctx, AUTO_LOAD, PENDING, cp_setup_auto_load);
+	cp_bra (ctx, USER_LOAD, PENDING, cp_setup_load);
+	cp_bra (ctx, ALWAYS, TRUE, cp_exit);
+
+	/* setup for context load */
+	cp_name(ctx, cp_setup_auto_load);
+	cp_out (ctx, CP_DISABLE1);
+	cp_out (ctx, CP_DISABLE2);
+	cp_out (ctx, CP_ENABLE);
+	cp_out (ctx, CP_NEXT_TO_SWAP);
+	cp_set (ctx, UNK01, SET);
+	cp_name(ctx, cp_setup_load);
+	cp_out (ctx, CP_NEWCTX);
+	cp_wait(ctx, NEWCTX, BUSY);
+	cp_set (ctx, UNK1D, CLEAR);
+	cp_set (ctx, SWAP_DIRECTION, LOAD);
+	cp_bra (ctx, UNK0B, SET, cp_prepare_exit);
+	cp_bra (ctx, ALWAYS, TRUE, cp_swap_state);
+
+	/* setup for context save */
+	cp_name(ctx, cp_setup_save);
+	cp_set (ctx, UNK1D, SET);
+	cp_wait(ctx, STATUS, BUSY);
+	cp_set (ctx, UNK01, SET);
+	cp_set (ctx, SWAP_DIRECTION, SAVE);
+
+	/* general PGRAPH state */
+	cp_name(ctx, cp_swap_state);
+	cp_set (ctx, UNK03, SET);
+	cp_pos (ctx, 0x00004/4);
+	cp_ctx (ctx, 0x400828, 1); /* needed. otherwise, flickering happens. */
+	cp_pos (ctx, 0x00100/4);
+	nv50_graph_construct_mmio(ctx);
+	nv50_graph_construct_xfer1(ctx);
+	nv50_graph_construct_xfer2(ctx);
+
+	cp_bra (ctx, SWAP_DIRECTION, SAVE, cp_check_load);
+
+	cp_set (ctx, UNK20, SET);
+	cp_set (ctx, SWAP_DIRECTION, SAVE); /* no idea why this is needed, but fixes at least one lockup. */
+	cp_lsr (ctx, ctx->ctxvals_base);
+	cp_out (ctx, CP_SET_XFER_POINTER);
+	cp_lsr (ctx, 4);
+	cp_out (ctx, CP_SEEK_1);
+	cp_out (ctx, CP_XFER_1);
+	cp_wait(ctx, XFER, BUSY);
+
+	/* pre-exit state updates */
+	cp_name(ctx, cp_prepare_exit);
+	cp_set (ctx, UNK01, CLEAR);
+	cp_set (ctx, UNK03, CLEAR);
+	cp_set (ctx, UNK1D, CLEAR);
+
+	cp_bra (ctx, USER_SAVE, PENDING, cp_exit);
+	cp_out (ctx, CP_NEXT_TO_CURRENT);
+
+	cp_name(ctx, cp_exit);
+	cp_set (ctx, USER_SAVE, NOT_PENDING);
+	cp_set (ctx, USER_LOAD, NOT_PENDING);
+	cp_out (ctx, CP_END);
+	ctx->ctxvals_pos += 0x400; /* padding... no idea why you need it */
+
+	return 0;
+}
+
+/*
+ * Constructs MMIO part of ctxprog and ctxvals. Just a matter of knowing which
+ * registers to save/restore and the default values for them.
+ */
+
+static void
+nv50_graph_construct_mmio(struct nouveau_grctx *ctx)
+{
+	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
+	int i, j;
+	int offset, base;
+	uint32_t units = nv_rd32 (ctx->dev, 0x1540);
+
+	/* 0800 */
+	cp_ctx(ctx, 0x400808, 7);
+	gr_def(ctx, 0x400814, 0x00000030);
+	cp_ctx(ctx, 0x400834, 0x32);
+	if (dev_priv->chipset == 0x50) {
+		gr_def(ctx, 0x400834, 0xff400040);
+		gr_def(ctx, 0x400838, 0xfff00080);
+		gr_def(ctx, 0x40083c, 0xfff70090);
+		gr_def(ctx, 0x400840, 0xffe806a8);
+	}
+	gr_def(ctx, 0x400844, 0x00000002);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+		gr_def(ctx, 0x400894, 0x00001000);
+	gr_def(ctx, 0x4008e8, 0x00000003);
+	gr_def(ctx, 0x4008ec, 0x00001000);
+	if (dev_priv->chipset == 0x50)
+		cp_ctx(ctx, 0x400908, 0xb);
+	else if (dev_priv->chipset < 0xa0)
+		cp_ctx(ctx, 0x400908, 0xc);
+	else
+		cp_ctx(ctx, 0x400908, 0xe);
+
+	if (dev_priv->chipset >= 0xa0)
+		cp_ctx(ctx, 0x400b00, 0x1);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) {
+		cp_ctx(ctx, 0x400b10, 0x1);
+		gr_def(ctx, 0x400b10, 0x0001629d);
+		cp_ctx(ctx, 0x400b20, 0x1);
+		gr_def(ctx, 0x400b20, 0x0001629d);
+	}
+
+	/* 0C00 */
+	cp_ctx(ctx, 0x400c08, 0x2);
+	gr_def(ctx, 0x400c08, 0x0000fe0c);
+
+	/* 1000 */
+	if (dev_priv->chipset < 0xa0) {
+		cp_ctx(ctx, 0x401008, 0x4);
+		gr_def(ctx, 0x401014, 0x00001000);
+	} else if (dev_priv->chipset == 0xa0 || dev_priv->chipset >= 0xaa) {
+		cp_ctx(ctx, 0x401008, 0x5);
+		gr_def(ctx, 0x401018, 0x00001000);
+	} else {
+		cp_ctx(ctx, 0x401008, 0x5);
+		gr_def(ctx, 0x401018, 0x00004000);
+	}
+
+	/* 1400 */
+	cp_ctx(ctx, 0x401400, 0x8);
+	cp_ctx(ctx, 0x401424, 0x3);
+	if (dev_priv->chipset == 0x50)
+		gr_def(ctx, 0x40142c, 0x0001fd87);
+	else
+		gr_def(ctx, 0x40142c, 0x00000187);
+	cp_ctx(ctx, 0x401540, 0x5);
+	gr_def(ctx, 0x401550, 0x00001018);
+
+	/* 1800 */
+	cp_ctx(ctx, 0x401814, 0x1);
+	gr_def(ctx, 0x401814, 0x000000ff);
+	if (dev_priv->chipset == 0x50) {
+		cp_ctx(ctx, 0x40181c, 0xe);
+		gr_def(ctx, 0x401850, 0x00000004);
+	} else if (dev_priv->chipset < 0xa0) {
+		cp_ctx(ctx, 0x40181c, 0xf);
+		gr_def(ctx, 0x401854, 0x00000004);
+	} else {
+		cp_ctx(ctx, 0x40181c, 0x13);
+		gr_def(ctx, 0x401864, 0x00000004);
+	}
+
+	/* 1C00 */
+	cp_ctx(ctx, 0x401c00, 0x1);
+	switch (dev_priv->chipset) {
+	case 0x50:
+		gr_def(ctx, 0x401c00, 0x0001005f);
+		break;
+	case 0x84:
+	case 0x86:
+	case 0x94:
+		gr_def(ctx, 0x401c00, 0x044d00df);
+		break;
+	case 0x92:
+	case 0x96:
+	case 0x98:
+	case 0xa0:
+	case 0xaa:
+	case 0xac:
+		gr_def(ctx, 0x401c00, 0x042500df);
+		break;
+	case 0xa5:
+	case 0xa8:
+		gr_def(ctx, 0x401c00, 0x142500df);
+		break;
+	}
+
+	/* 2400 */
+	cp_ctx(ctx, 0x402400, 0x1);
+	if (dev_priv->chipset == 0x50)
+		cp_ctx(ctx, 0x402408, 0x1);
+	else
+		cp_ctx(ctx, 0x402408, 0x2);
+	gr_def(ctx, 0x402408, 0x00000600);
+
+	/* 2800 */
+	cp_ctx(ctx, 0x402800, 0x1);
+	if (dev_priv->chipset == 0x50)
+		gr_def(ctx, 0x402800, 0x00000006);
+
+	/* 2C00 */
+	cp_ctx(ctx, 0x402c08, 0x6);
+	if (dev_priv->chipset != 0x50)
+		gr_def(ctx, 0x402c14, 0x01000000);
+	gr_def(ctx, 0x402c18, 0x000000ff);
+	if (dev_priv->chipset == 0x50)
+		cp_ctx(ctx, 0x402ca0, 0x1);
+	else
+		cp_ctx(ctx, 0x402ca0, 0x2);
+	if (dev_priv->chipset < 0xa0)
+		gr_def(ctx, 0x402ca0, 0x00000400);
+	else if (dev_priv->chipset == 0xa0 || dev_priv->chipset >= 0xaa)
+		gr_def(ctx, 0x402ca0, 0x00000800);
+	else
+		gr_def(ctx, 0x402ca0, 0x00000400);
+	cp_ctx(ctx, 0x402cac, 0x4);
+
+	/* 3000 */
+	cp_ctx(ctx, 0x403004, 0x1);
+	gr_def(ctx, 0x403004, 0x00000001);
+
+	/* 3404 */
+	if (dev_priv->chipset >= 0xa0) {
+		cp_ctx(ctx, 0x403404, 0x1);
+		gr_def(ctx, 0x403404, 0x00000001);
+	}
+
+	/* 5000 */
+	cp_ctx(ctx, 0x405000, 0x1);
+	switch (dev_priv->chipset) {
+	case 0x50:
+		gr_def(ctx, 0x405000, 0x00300080);
+		break;
+	case 0x84:
+	case 0xa0:
+	case 0xa5:
+	case 0xa8:
+	case 0xaa:
+	case 0xac:
+		gr_def(ctx, 0x405000, 0x000e0080);
+		break;
+	case 0x86:
+	case 0x92:
+	case 0x94:
+	case 0x96:
+	case 0x98:
+		gr_def(ctx, 0x405000, 0x00000080);
+		break;
+	}
+	cp_ctx(ctx, 0x405014, 0x1);
+	gr_def(ctx, 0x405014, 0x00000004);
+	cp_ctx(ctx, 0x40501c, 0x1);
+	cp_ctx(ctx, 0x405024, 0x1);
+	cp_ctx(ctx, 0x40502c, 0x1);
+
+	/* 5400 or maybe 4800 */
+	if (dev_priv->chipset == 0x50) {
+		offset = 0x405400;
+		cp_ctx(ctx, 0x405400, 0xea);
+	} else if (dev_priv->chipset < 0x94) {
+		offset = 0x405400;
+		cp_ctx(ctx, 0x405400, 0xcb);
+	} else if (dev_priv->chipset < 0xa0) {
+		offset = 0x405400;
+		cp_ctx(ctx, 0x405400, 0xcc);
+	} else if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) {
+		offset = 0x404800;
+		cp_ctx(ctx, 0x404800, 0xda);
+	} else {
+		offset = 0x405400;
+		cp_ctx(ctx, 0x405400, 0xd4);
+	}
+	gr_def(ctx, offset + 0x0c, 0x00000002);
+	gr_def(ctx, offset + 0x10, 0x00000001);
+	if (dev_priv->chipset >= 0x94)
+		offset += 4;
+	gr_def(ctx, offset + 0x1c, 0x00000001);
+	gr_def(ctx, offset + 0x20, 0x00000100);
+	gr_def(ctx, offset + 0x38, 0x00000002);
+	gr_def(ctx, offset + 0x3c, 0x00000001);
+	gr_def(ctx, offset + 0x40, 0x00000001);
+	gr_def(ctx, offset + 0x50, 0x00000001);
+	gr_def(ctx, offset + 0x54, 0x003fffff);
+	gr_def(ctx, offset + 0x58, 0x00001fff);
+	gr_def(ctx, offset + 0x60, 0x00000001);
+	gr_def(ctx, offset + 0x64, 0x00000001);
+	gr_def(ctx, offset + 0x6c, 0x00000001);
+	gr_def(ctx, offset + 0x70, 0x00000001);
+	gr_def(ctx, offset + 0x74, 0x00000001);
+	gr_def(ctx, offset + 0x78, 0x00000004);
+	gr_def(ctx, offset + 0x7c, 0x00000001);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+		offset += 4;
+	gr_def(ctx, offset + 0x80, 0x00000001);
+	gr_def(ctx, offset + 0x84, 0x00000001);
+	gr_def(ctx, offset + 0x88, 0x00000007);
+	gr_def(ctx, offset + 0x8c, 0x00000001);
+	gr_def(ctx, offset + 0x90, 0x00000007);
+	gr_def(ctx, offset + 0x94, 0x00000001);
+	gr_def(ctx, offset + 0x98, 0x00000001);
+	gr_def(ctx, offset + 0x9c, 0x00000001);
+	if (dev_priv->chipset == 0x50) {
+		 gr_def(ctx, offset + 0xb0, 0x00000001);
+		 gr_def(ctx, offset + 0xb4, 0x00000001);
+		 gr_def(ctx, offset + 0xbc, 0x00000001);
+		 gr_def(ctx, offset + 0xc0, 0x0000000a);
+		 gr_def(ctx, offset + 0xd0, 0x00000040);
+		 gr_def(ctx, offset + 0xd8, 0x00000002);
+		 gr_def(ctx, offset + 0xdc, 0x00000100);
+		 gr_def(ctx, offset + 0xe0, 0x00000001);
+		 gr_def(ctx, offset + 0xe4, 0x00000100);
+		 gr_def(ctx, offset + 0x100, 0x00000001);
+		 gr_def(ctx, offset + 0x124, 0x00000004);
+		 gr_def(ctx, offset + 0x13c, 0x00000001);
+		 gr_def(ctx, offset + 0x140, 0x00000100);
+		 gr_def(ctx, offset + 0x148, 0x00000001);
+		 gr_def(ctx, offset + 0x154, 0x00000100);
+		 gr_def(ctx, offset + 0x158, 0x00000001);
+		 gr_def(ctx, offset + 0x15c, 0x00000100);
+		 gr_def(ctx, offset + 0x164, 0x00000001);
+		 gr_def(ctx, offset + 0x170, 0x00000100);
+		 gr_def(ctx, offset + 0x174, 0x00000001);
+		 gr_def(ctx, offset + 0x17c, 0x00000001);
+		 gr_def(ctx, offset + 0x188, 0x00000002);
+		 gr_def(ctx, offset + 0x190, 0x00000001);
+		 gr_def(ctx, offset + 0x198, 0x00000001);
+		 gr_def(ctx, offset + 0x1ac, 0x00000003);
+		 offset += 0xd0;
+	} else {
+		gr_def(ctx, offset + 0xb0, 0x00000001);
+		gr_def(ctx, offset + 0xb4, 0x00000100);
+		gr_def(ctx, offset + 0xbc, 0x00000001);
+		gr_def(ctx, offset + 0xc8, 0x00000100);
+		gr_def(ctx, offset + 0xcc, 0x00000001);
+		gr_def(ctx, offset + 0xd0, 0x00000100);
+		gr_def(ctx, offset + 0xd8, 0x00000001);
+		gr_def(ctx, offset + 0xe4, 0x00000100);
+	}
+	gr_def(ctx, offset + 0xf8, 0x00000004);
+	gr_def(ctx, offset + 0xfc, 0x00000070);
+	gr_def(ctx, offset + 0x100, 0x00000080);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+		offset += 4;
+	gr_def(ctx, offset + 0x114, 0x0000000c);
+	if (dev_priv->chipset == 0x50)
+		offset -= 4;
+	gr_def(ctx, offset + 0x11c, 0x00000008);
+	gr_def(ctx, offset + 0x120, 0x00000014);
+	if (dev_priv->chipset == 0x50) {
+		gr_def(ctx, offset + 0x124, 0x00000026);
+		offset -= 0x18;
+	} else {
+		gr_def(ctx, offset + 0x128, 0x00000029);
+		gr_def(ctx, offset + 0x12c, 0x00000027);
+		gr_def(ctx, offset + 0x130, 0x00000026);
+		gr_def(ctx, offset + 0x134, 0x00000008);
+		gr_def(ctx, offset + 0x138, 0x00000004);
+		gr_def(ctx, offset + 0x13c, 0x00000027);
+	}
+	gr_def(ctx, offset + 0x148, 0x00000001);
+	gr_def(ctx, offset + 0x14c, 0x00000002);
+	gr_def(ctx, offset + 0x150, 0x00000003);
+	gr_def(ctx, offset + 0x154, 0x00000004);
+	gr_def(ctx, offset + 0x158, 0x00000005);
+	gr_def(ctx, offset + 0x15c, 0x00000006);
+	gr_def(ctx, offset + 0x160, 0x00000007);
+	gr_def(ctx, offset + 0x164, 0x00000001);
+	gr_def(ctx, offset + 0x1a8, 0x000000cf);
+	if (dev_priv->chipset == 0x50)
+		offset -= 4;
+	gr_def(ctx, offset + 0x1d8, 0x00000080);
+	gr_def(ctx, offset + 0x1dc, 0x00000004);
+	gr_def(ctx, offset + 0x1e0, 0x00000004);
+	if (dev_priv->chipset == 0x50)
+		offset -= 4;
+	else
+		gr_def(ctx, offset + 0x1e4, 0x00000003);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) {
+		gr_def(ctx, offset + 0x1ec, 0x00000003);
+		offset += 8;
+	}
+	gr_def(ctx, offset + 0x1e8, 0x00000001);
+	if (dev_priv->chipset == 0x50)
+		offset -= 4;
+	gr_def(ctx, offset + 0x1f4, 0x00000012);
+	gr_def(ctx, offset + 0x1f8, 0x00000010);
+	gr_def(ctx, offset + 0x1fc, 0x0000000c);
+	gr_def(ctx, offset + 0x200, 0x00000001);
+	gr_def(ctx, offset + 0x210, 0x00000004);
+	gr_def(ctx, offset + 0x214, 0x00000002);
+	gr_def(ctx, offset + 0x218, 0x00000004);
+	if (dev_priv->chipset >= 0xa0)
+		offset += 4;
+	gr_def(ctx, offset + 0x224, 0x003fffff);
+	gr_def(ctx, offset + 0x228, 0x00001fff);
+	if (dev_priv->chipset == 0x50)
+		offset -= 0x20;
+	else if (dev_priv->chipset >= 0xa0) {
+		gr_def(ctx, offset + 0x250, 0x00000001);
+		gr_def(ctx, offset + 0x254, 0x00000001);
+		gr_def(ctx, offset + 0x258, 0x00000002);
+		offset += 0x10;
+	}
+	gr_def(ctx, offset + 0x250, 0x00000004);
+	gr_def(ctx, offset + 0x254, 0x00000014);
+	gr_def(ctx, offset + 0x258, 0x00000001);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+		offset += 4;
+	gr_def(ctx, offset + 0x264, 0x00000002);
+	if (dev_priv->chipset >= 0xa0)
+		offset += 8;
+	gr_def(ctx, offset + 0x270, 0x00000001);
+	gr_def(ctx, offset + 0x278, 0x00000002);
+	gr_def(ctx, offset + 0x27c, 0x00001000);
+	if (dev_priv->chipset == 0x50)
+		offset -= 0xc;
+	else {
+		gr_def(ctx, offset + 0x280, 0x00000e00);
+		gr_def(ctx, offset + 0x284, 0x00001000);
+		gr_def(ctx, offset + 0x288, 0x00001e00);
+	}
+	gr_def(ctx, offset + 0x290, 0x00000001);
+	gr_def(ctx, offset + 0x294, 0x00000001);
+	gr_def(ctx, offset + 0x298, 0x00000001);
+	gr_def(ctx, offset + 0x29c, 0x00000001);
+	gr_def(ctx, offset + 0x2a0, 0x00000001);
+	gr_def(ctx, offset + 0x2b0, 0x00000200);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) {
+		gr_def(ctx, offset + 0x2b4, 0x00000200);
+		offset += 4;
+	}
+	if (dev_priv->chipset < 0xa0) {
+		gr_def(ctx, offset + 0x2b8, 0x00000001);
+		gr_def(ctx, offset + 0x2bc, 0x00000070);
+		gr_def(ctx, offset + 0x2c0, 0x00000080);
+		gr_def(ctx, offset + 0x2cc, 0x00000001);
+		gr_def(ctx, offset + 0x2d0, 0x00000070);
+		gr_def(ctx, offset + 0x2d4, 0x00000080);
+	} else {
+		gr_def(ctx, offset + 0x2b8, 0x00000001);
+		gr_def(ctx, offset + 0x2bc, 0x000000f0);
+		gr_def(ctx, offset + 0x2c0, 0x000000ff);
+		gr_def(ctx, offset + 0x2cc, 0x00000001);
+		gr_def(ctx, offset + 0x2d0, 0x000000f0);
+		gr_def(ctx, offset + 0x2d4, 0x000000ff);
+		gr_def(ctx, offset + 0x2dc, 0x00000009);
+		offset += 4;
+	}
+	gr_def(ctx, offset + 0x2e4, 0x00000001);
+	gr_def(ctx, offset + 0x2e8, 0x000000cf);
+	gr_def(ctx, offset + 0x2f0, 0x00000001);
+	gr_def(ctx, offset + 0x300, 0x000000cf);
+	gr_def(ctx, offset + 0x308, 0x00000002);
+	gr_def(ctx, offset + 0x310, 0x00000001);
+	gr_def(ctx, offset + 0x318, 0x00000001);
+	gr_def(ctx, offset + 0x320, 0x000000cf);
+	gr_def(ctx, offset + 0x324, 0x000000cf);
+	gr_def(ctx, offset + 0x328, 0x00000001);
+
+	/* 6000? */
+	if (dev_priv->chipset == 0x50)
+		cp_ctx(ctx, 0x4063e0, 0x1);
+
+	/* 6800 */
+	if (dev_priv->chipset < 0x90) {
+		cp_ctx(ctx, 0x406814, 0x2b);
+		gr_def(ctx, 0x406818, 0x00000f80);
+		gr_def(ctx, 0x406860, 0x007f0080);
+		gr_def(ctx, 0x40689c, 0x007f0080);
+	} else {
+		cp_ctx(ctx, 0x406814, 0x4);
+		if (dev_priv->chipset == 0x98)
+			gr_def(ctx, 0x406818, 0x00000f80);
+		else
+			gr_def(ctx, 0x406818, 0x00001f80);
+		if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+			gr_def(ctx, 0x40681c, 0x00000030);
+		cp_ctx(ctx, 0x406830, 0x3);
+	}
+
+	/* 7000: per-ROP group state */
+	for (i = 0; i < 8; i++) {
+		if (units & (1<<(i+16))) {
+			cp_ctx(ctx, 0x407000 + (i<<8), 3);
+			if (dev_priv->chipset == 0x50)
+				gr_def(ctx, 0x407000 + (i<<8), 0x1b74f820);
+			else if (dev_priv->chipset != 0xa5)
+				gr_def(ctx, 0x407000 + (i<<8), 0x3b74f821);
+			else
+				gr_def(ctx, 0x407000 + (i<<8), 0x7b74f821);
+			gr_def(ctx, 0x407004 + (i<<8), 0x89058001);
+
+			if (dev_priv->chipset == 0x50) {
+				cp_ctx(ctx, 0x407010 + (i<<8), 1);
+			} else if (dev_priv->chipset < 0xa0) {
+				cp_ctx(ctx, 0x407010 + (i<<8), 2);
+				gr_def(ctx, 0x407010 + (i<<8), 0x00001000);
+				gr_def(ctx, 0x407014 + (i<<8), 0x0000001f);
+			} else {
+				cp_ctx(ctx, 0x407010 + (i<<8), 3);
+				gr_def(ctx, 0x407010 + (i<<8), 0x00001000);
+				if (dev_priv->chipset != 0xa5)
+					gr_def(ctx, 0x407014 + (i<<8), 0x000000ff);
+				else
+					gr_def(ctx, 0x407014 + (i<<8), 0x000001ff);
+			}
+
+			cp_ctx(ctx, 0x407080 + (i<<8), 4);
+			if (dev_priv->chipset != 0xa5)
+				gr_def(ctx, 0x407080 + (i<<8), 0x027c10fa);
+			else
+				gr_def(ctx, 0x407080 + (i<<8), 0x827c10fa);
+			if (dev_priv->chipset == 0x50)
+				gr_def(ctx, 0x407084 + (i<<8), 0x000000c0);
+			else
+				gr_def(ctx, 0x407084 + (i<<8), 0x400000c0);
+			gr_def(ctx, 0x407088 + (i<<8), 0xb7892080);
+
+			if (dev_priv->chipset < 0xa0)
+				cp_ctx(ctx, 0x407094 + (i<<8), 1);
+			else if (dev_priv->chipset <= 0xa0 || dev_priv->chipset >= 0xaa)
+				cp_ctx(ctx, 0x407094 + (i<<8), 3);
+			else {
+				cp_ctx(ctx, 0x407094 + (i<<8), 4);
+				gr_def(ctx, 0x4070a0 + (i<<8), 1);
+			}
+		}
+	}
+
+	cp_ctx(ctx, 0x407c00, 0x3);
+	if (dev_priv->chipset < 0x90)
+		gr_def(ctx, 0x407c00, 0x00010040);
+	else if (dev_priv->chipset < 0xa0)
+		gr_def(ctx, 0x407c00, 0x00390040);
+	else
+		gr_def(ctx, 0x407c00, 0x003d0040);
+	gr_def(ctx, 0x407c08, 0x00000022);
+	if (dev_priv->chipset >= 0xa0) {
+		cp_ctx(ctx, 0x407c10, 0x3);
+		cp_ctx(ctx, 0x407c20, 0x1);
+		cp_ctx(ctx, 0x407c2c, 0x1);
+	}
+
+	if (dev_priv->chipset < 0xa0) {
+		cp_ctx(ctx, 0x407d00, 0x9);
+	} else {
+		cp_ctx(ctx, 0x407d00, 0x15);
+	}
+	if (dev_priv->chipset == 0x98)
+		gr_def(ctx, 0x407d08, 0x00380040);
+	else {
+		if (dev_priv->chipset < 0x90)
+			gr_def(ctx, 0x407d08, 0x00010040);
+		else if (dev_priv->chipset < 0xa0)
+			gr_def(ctx, 0x407d08, 0x00390040);
+		else
+			gr_def(ctx, 0x407d08, 0x003d0040);
+		gr_def(ctx, 0x407d0c, 0x00000022);
+	}
+
+	/* 8000+: per-TP state */
+	for (i = 0; i < 10; i++) {
+		if (units & (1<<i)) {
+			if (dev_priv->chipset < 0xa0)
+				base = 0x408000 + (i<<12);
+			else
+				base = 0x408000 + (i<<11);
+			if (dev_priv->chipset < 0xa0)
+				offset = base + 0xc00;
+			else
+				offset = base + 0x80;
+			cp_ctx(ctx, offset + 0x00, 1);
+			gr_def(ctx, offset + 0x00, 0x0000ff0a);
+			cp_ctx(ctx, offset + 0x08, 1);
+
+			/* per-MP state */
+			for (j = 0; j < (dev_priv->chipset < 0xa0 ? 2 : 4); j++) {
+				if (!(units & (1 << (j+24)))) continue;
+				if (dev_priv->chipset < 0xa0)
+					offset = base + 0x200 + (j<<7);
+				else
+					offset = base + 0x100 + (j<<7);
+				cp_ctx(ctx, offset, 0x20);
+				gr_def(ctx, offset + 0x00, 0x01800000);
+				gr_def(ctx, offset + 0x04, 0x00160000);
+				gr_def(ctx, offset + 0x08, 0x01800000);
+				gr_def(ctx, offset + 0x18, 0x0003ffff);
+				switch (dev_priv->chipset) {
+				case 0x50:
+					gr_def(ctx, offset + 0x1c, 0x00080000);
+					break;
+				case 0x84:
+					gr_def(ctx, offset + 0x1c, 0x00880000);
+					break;
+				case 0x86:
+					gr_def(ctx, offset + 0x1c, 0x008c0000);
+					break;
+				case 0x92:
+				case 0x96:
+				case 0x98:
+					gr_def(ctx, offset + 0x1c, 0x118c0000);
+					break;
+				case 0x94:
+					gr_def(ctx, offset + 0x1c, 0x10880000);
+					break;
+				case 0xa0:
+				case 0xa5:
+					gr_def(ctx, offset + 0x1c, 0x310c0000);
+					break;
+				case 0xa8:
+				case 0xaa:
+				case 0xac:
+					gr_def(ctx, offset + 0x1c, 0x300c0000);
+					break;
+				}
+				gr_def(ctx, offset + 0x40, 0x00010401);
+				if (dev_priv->chipset == 0x50)
+					gr_def(ctx, offset + 0x48, 0x00000040);
+				else
+					gr_def(ctx, offset + 0x48, 0x00000078);
+				gr_def(ctx, offset + 0x50, 0x000000bf);
+				gr_def(ctx, offset + 0x58, 0x00001210);
+				if (dev_priv->chipset == 0x50)
+					gr_def(ctx, offset + 0x5c, 0x00000080);
+				else
+					gr_def(ctx, offset + 0x5c, 0x08000080);
+				if (dev_priv->chipset >= 0xa0)
+					gr_def(ctx, offset + 0x68, 0x0000003e);
+			}
+
+			if (dev_priv->chipset < 0xa0)
+				cp_ctx(ctx, base + 0x300, 0x4);
+			else
+				cp_ctx(ctx, base + 0x300, 0x5);
+			if (dev_priv->chipset == 0x50)
+				gr_def(ctx, base + 0x304, 0x00007070);
+			else if (dev_priv->chipset < 0xa0)
+				gr_def(ctx, base + 0x304, 0x00027070);
+			else if (dev_priv->chipset <= 0xa0 || dev_priv->chipset >= 0xaa)
+				gr_def(ctx, base + 0x304, 0x01127070);
+			else
+				gr_def(ctx, base + 0x304, 0x05127070);
+
+			if (dev_priv->chipset < 0xa0)
+				cp_ctx(ctx, base + 0x318, 1);
+			else
+				cp_ctx(ctx, base + 0x320, 1);
+			if (dev_priv->chipset == 0x50)
+				gr_def(ctx, base + 0x318, 0x0003ffff);
+			else if (dev_priv->chipset < 0xa0)
+				gr_def(ctx, base + 0x318, 0x03ffffff);
+			else
+				gr_def(ctx, base + 0x320, 0x07ffffff);
+
+			if (dev_priv->chipset < 0xa0)
+				cp_ctx(ctx, base + 0x324, 5);
+			else
+				cp_ctx(ctx, base + 0x328, 4);
+
+			if (dev_priv->chipset < 0xa0) {
+				cp_ctx(ctx, base + 0x340, 9);
+				offset = base + 0x340;
+			} else if (dev_priv->chipset <= 0xa0 || dev_priv->chipset >= 0xaa) {
+				cp_ctx(ctx, base + 0x33c, 0xb);
+				offset = base + 0x344;
+			} else {
+				cp_ctx(ctx, base + 0x33c, 0xd);
+				offset = base + 0x344;
+			}
+			gr_def(ctx, offset + 0x0, 0x00120407);
+			gr_def(ctx, offset + 0x4, 0x05091507);
+			if (dev_priv->chipset == 0x84)
+				gr_def(ctx, offset + 0x8, 0x05100202);
+			else
+				gr_def(ctx, offset + 0x8, 0x05010202);
+			gr_def(ctx, offset + 0xc, 0x00030201);
+
+			cp_ctx(ctx, base + 0x400, 2);
+			gr_def(ctx, base + 0x404, 0x00000040);
+			cp_ctx(ctx, base + 0x40c, 2);
+			gr_def(ctx, base + 0x40c, 0x0d0c0b0a);
+			gr_def(ctx, base + 0x410, 0x00141210);
+
+			if (dev_priv->chipset < 0xa0)
+				offset = base + 0x800;
+			else
+				offset = base + 0x500;
+			cp_ctx(ctx, offset, 6);
+			gr_def(ctx, offset + 0x0, 0x000001f0);
+			gr_def(ctx, offset + 0x4, 0x00000001);
+			gr_def(ctx, offset + 0x8, 0x00000003);
+			if (dev_priv->chipset == 0x50 || dev_priv->chipset >= 0xaa)
+				gr_def(ctx, offset + 0xc, 0x00008000);
+			gr_def(ctx, offset + 0x14, 0x00039e00);
+			cp_ctx(ctx, offset + 0x1c, 2);
+			if (dev_priv->chipset == 0x50)
+				gr_def(ctx, offset + 0x1c, 0x00000040);
+			else
+				gr_def(ctx, offset + 0x1c, 0x00000100);
+			gr_def(ctx, offset + 0x20, 0x00003800);
+
+			if (dev_priv->chipset >= 0xa0) {
+				cp_ctx(ctx, base + 0x54c, 2);
+				if (dev_priv->chipset <= 0xa0 || dev_priv->chipset >= 0xaa)
+					gr_def(ctx, base + 0x54c, 0x003fe006);
+				else
+					gr_def(ctx, base + 0x54c, 0x003fe007);
+				gr_def(ctx, base + 0x550, 0x003fe000);
+			}
+
+			if (dev_priv->chipset < 0xa0)
+				offset = base + 0xa00;
+			else
+				offset = base + 0x680;
+			cp_ctx(ctx, offset, 1);
+			gr_def(ctx, offset, 0x00404040);
+
+			if (dev_priv->chipset < 0xa0)
+				offset = base + 0xe00;
+			else
+				offset = base + 0x700;
+			cp_ctx(ctx, offset, 2);
+			if (dev_priv->chipset < 0xa0)
+				gr_def(ctx, offset, 0x0077f005);
+			else if (dev_priv->chipset == 0xa5)
+				gr_def(ctx, offset, 0x6cf7f007);
+			else if (dev_priv->chipset == 0xa8)
+				gr_def(ctx, offset, 0x6cfff007);
+			else if (dev_priv->chipset == 0xac)
+				gr_def(ctx, offset, 0x0cfff007);
+			else
+				gr_def(ctx, offset, 0x0cf7f007);
+			if (dev_priv->chipset == 0x50)
+				gr_def(ctx, offset + 0x4, 0x00007fff);
+			else if (dev_priv->chipset < 0xa0)
+				gr_def(ctx, offset + 0x4, 0x003f7fff);
+			else
+				gr_def(ctx, offset + 0x4, 0x02bf7fff);
+			cp_ctx(ctx, offset + 0x2c, 1);
+			if (dev_priv->chipset == 0x50) {
+				cp_ctx(ctx, offset + 0x50, 9);
+				gr_def(ctx, offset + 0x54, 0x000003ff);
+				gr_def(ctx, offset + 0x58, 0x00000003);
+				gr_def(ctx, offset + 0x5c, 0x00000003);
+				gr_def(ctx, offset + 0x60, 0x000001ff);
+				gr_def(ctx, offset + 0x64, 0x0000001f);
+				gr_def(ctx, offset + 0x68, 0x0000000f);
+				gr_def(ctx, offset + 0x6c, 0x0000000f);
+			} else if(dev_priv->chipset < 0xa0) {
+				cp_ctx(ctx, offset + 0x50, 1);
+				cp_ctx(ctx, offset + 0x70, 1);
+			} else {
+				cp_ctx(ctx, offset + 0x50, 1);
+				cp_ctx(ctx, offset + 0x60, 5);
+			}
+		}
+	}
+}
+
+/*
+ * xfer areas. These are a pain.
+ *
+ * There are 2 xfer areas: the first one is big and contains all sorts of
+ * stuff, the second is small and contains some per-TP context.
+ *
+ * Each area is split into 8 "strands". The areas, when saved to grctx,
+ * are made of 8-word blocks. Each block contains a single word from
+ * each strand. The strands are independent of each other, their
+ * addresses are unrelated to each other, and data in them is closely
+ * packed together. The strand layout varies a bit between cards: here
+ * and there, a single word is thrown out in the middle and the whole
+ * strand is offset by a bit from corresponding one on another chipset.
+ * For this reason, addresses of stuff in strands are almost useless.
+ * Knowing sequence of stuff and size of gaps between them is much more
+ * useful, and that's how we build the strands in our generator.
+ *
+ * NVA0 takes this mess to a whole new level by cutting the old strands
+ * into a few dozen pieces [known as genes], rearranging them randomly,
+ * and putting them back together to make new strands. Hopefully these
+ * genes correspond more or less directly to the same PGRAPH subunits
+ * as in 400040 register.
+ *
+ * The most common value in default context is 0, and when the genes
+ * are separated by 0's, gene bounduaries are quite speculative...
+ * some of them can be clearly deduced, others can be guessed, and yet
+ * others won't be resolved without figuring out the real meaning of
+ * given ctxval. For the same reason, ending point of each strand
+ * is unknown. Except for strand 0, which is the longest strand and
+ * its end corresponds to end of the whole xfer.
+ *
+ * An unsolved mystery is the seek instruction: it takes an argument
+ * in bits 8-18, and that argument is clearly the place in strands to
+ * seek to... but the offsets don't seem to correspond to offsets as
+ * seen in grctx. Perhaps there's another, real, not randomly-changing
+ * addressing in strands, and the xfer insn just happens to skip over
+ * the unused bits? NV10-NV30 PIPE comes to mind...
+ *
+ * As far as I know, there's no way to access the xfer areas directly
+ * without the help of ctxprog.
+ */
+
+static inline void
+xf_emit(struct nouveau_grctx *ctx, int num, uint32_t val) {
+	int i;
+	if (val && ctx->mode == NOUVEAU_GRCTX_VALS)
+		for (i = 0; i < num; i++)
+			nv_wo32(ctx->dev, ctx->data, ctx->ctxvals_pos + (i << 3), val);
+	ctx->ctxvals_pos += num << 3;
+}
+
+/* Gene declarations... */
+
+static void nv50_graph_construct_gene_m2mf(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_gene_unk1(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_gene_unk2(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_gene_unk3(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_gene_unk4(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_gene_unk5(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_gene_unk6(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_gene_unk7(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_gene_unk8(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_gene_unk9(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_gene_unk10(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_gene_ropc(struct nouveau_grctx *ctx);
+static void nv50_graph_construct_xfer_tp(struct nouveau_grctx *ctx);
+
+static void
+nv50_graph_construct_xfer1(struct nouveau_grctx *ctx)
+{
+	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
+	int i;
+	int offset;
+	int size = 0;
+	uint32_t units = nv_rd32 (ctx->dev, 0x1540);
+
+	offset = (ctx->ctxvals_pos+0x3f)&~0x3f;
+	ctx->ctxvals_base = offset;
+
+	if (dev_priv->chipset < 0xa0) {
+		/* Strand 0 */
+		ctx->ctxvals_pos = offset;
+		switch (dev_priv->chipset) {
+		case 0x50:
+			xf_emit(ctx, 0x99, 0);
+			break;
+		case 0x84:
+		case 0x86:
+			xf_emit(ctx, 0x384, 0);
+			break;
+		case 0x92:
+		case 0x94:
+		case 0x96:
+		case 0x98:
+			xf_emit(ctx, 0x380, 0);
+			break;
+		}
+		nv50_graph_construct_gene_m2mf (ctx);
+		switch (dev_priv->chipset) {
+		case 0x50:
+		case 0x84:
+		case 0x86:
+		case 0x98:
+			xf_emit(ctx, 0x4c4, 0);
+			break;
+		case 0x92:
+		case 0x94:
+		case 0x96:
+			xf_emit(ctx, 0x984, 0);
+			break;
+		}
+		nv50_graph_construct_gene_unk5(ctx);
+		if (dev_priv->chipset == 0x50)
+			xf_emit(ctx, 0xa, 0);
+		else
+			xf_emit(ctx, 0xb, 0);
+		nv50_graph_construct_gene_unk4(ctx);
+		nv50_graph_construct_gene_unk3(ctx);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+
+		/* Strand 1 */
+		ctx->ctxvals_pos = offset + 0x1;
+		nv50_graph_construct_gene_unk6(ctx);
+		nv50_graph_construct_gene_unk7(ctx);
+		nv50_graph_construct_gene_unk8(ctx);
+		switch (dev_priv->chipset) {
+		case 0x50:
+		case 0x92:
+			xf_emit(ctx, 0xfb, 0);
+			break;
+		case 0x84:
+			xf_emit(ctx, 0xd3, 0);
+			break;
+		case 0x94:
+		case 0x96:
+			xf_emit(ctx, 0xab, 0);
+			break;
+		case 0x86:
+		case 0x98:
+			xf_emit(ctx, 0x6b, 0);
+			break;
+		}
+		xf_emit(ctx, 2, 0x4e3bfdf);
+		xf_emit(ctx, 4, 0);
+		xf_emit(ctx, 1, 0x0fac6881);
+		xf_emit(ctx, 0xb, 0);
+		xf_emit(ctx, 2, 0x4e3bfdf);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+
+		/* Strand 2 */
+		ctx->ctxvals_pos = offset + 0x2;
+		switch (dev_priv->chipset) {
+		case 0x50:
+		case 0x92:
+			xf_emit(ctx, 0xa80, 0);
+			break;
+		case 0x84:
+			xf_emit(ctx, 0xa7e, 0);
+			break;
+		case 0x94:
+		case 0x96:
+			xf_emit(ctx, 0xa7c, 0);
+			break;
+		case 0x86:
+		case 0x98:
+			xf_emit(ctx, 0xa7a, 0);
+			break;
+		}
+		xf_emit(ctx, 1, 0x3fffff);
+		xf_emit(ctx, 2, 0);
+		xf_emit(ctx, 1, 0x1fff);
+		xf_emit(ctx, 0xe, 0);
+		nv50_graph_construct_gene_unk9(ctx);
+		nv50_graph_construct_gene_unk2(ctx);
+		nv50_graph_construct_gene_unk1(ctx);
+		nv50_graph_construct_gene_unk10(ctx);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+
+		/* Strand 3: per-ROP group state */
+		ctx->ctxvals_pos = offset + 3;
+		for (i = 0; i < 6; i++)
+			if (units & (1 << (i + 16)))
+				nv50_graph_construct_gene_ropc(ctx);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+
+		/* Strands 4-7: per-TP state */
+		for (i = 0; i < 4; i++) {
+			ctx->ctxvals_pos = offset + 4 + i;
+			if (units & (1 << (2 * i)))
+				nv50_graph_construct_xfer_tp(ctx);
+			if (units & (1 << (2 * i + 1)))
+				nv50_graph_construct_xfer_tp(ctx);
+			if ((ctx->ctxvals_pos-offset)/8 > size)
+				size = (ctx->ctxvals_pos-offset)/8;
+		}
+	} else {
+		/* Strand 0 */
+		ctx->ctxvals_pos = offset;
+		if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+			xf_emit(ctx, 0x385, 0);
+		else
+			xf_emit(ctx, 0x384, 0);
+		nv50_graph_construct_gene_m2mf(ctx);
+		xf_emit(ctx, 0x950, 0);
+		nv50_graph_construct_gene_unk10(ctx);
+		xf_emit(ctx, 1, 0x0fac6881);
+		if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) {
+			xf_emit(ctx, 1, 1);
+			xf_emit(ctx, 3, 0);
+		}
+		nv50_graph_construct_gene_unk8(ctx);
+		if (dev_priv->chipset == 0xa0)
+			xf_emit(ctx, 0x189, 0);
+		else if (dev_priv->chipset < 0xa8)
+			xf_emit(ctx, 0x99, 0);
+		else if (dev_priv->chipset == 0xaa)
+			xf_emit(ctx, 0x65, 0);
+		else
+			xf_emit(ctx, 0x6d, 0);
+		nv50_graph_construct_gene_unk9(ctx);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+
+		/* Strand 1 */
+		ctx->ctxvals_pos = offset + 1;
+		nv50_graph_construct_gene_unk1(ctx);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+
+		/* Strand 2 */
+		ctx->ctxvals_pos = offset + 2;
+		if (dev_priv->chipset == 0xa0) {
+			nv50_graph_construct_gene_unk2(ctx);
+		}
+		xf_emit(ctx, 0x36, 0);
+		nv50_graph_construct_gene_unk5(ctx);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+
+		/* Strand 3 */
+		ctx->ctxvals_pos = offset + 3;
+		xf_emit(ctx, 1, 0);
+		xf_emit(ctx, 1, 1);
+		nv50_graph_construct_gene_unk6(ctx);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+
+		/* Strand 4 */
+		ctx->ctxvals_pos = offset + 4;
+		if (dev_priv->chipset == 0xa0)
+			xf_emit(ctx, 0xa80, 0);
+		else
+			xf_emit(ctx, 0xa7a, 0);
+		xf_emit(ctx, 1, 0x3fffff);
+		xf_emit(ctx, 2, 0);
+		xf_emit(ctx, 1, 0x1fff);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+
+		/* Strand 5 */
+		ctx->ctxvals_pos = offset + 5;
+		xf_emit(ctx, 1, 0);
+		xf_emit(ctx, 1, 0x0fac6881);
+		xf_emit(ctx, 0xb, 0);
+		xf_emit(ctx, 2, 0x4e3bfdf);
+		xf_emit(ctx, 3, 0);
+		if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+			xf_emit(ctx, 1, 0x11);
+		xf_emit(ctx, 1, 0);
+		xf_emit(ctx, 2, 0x4e3bfdf);
+		xf_emit(ctx, 2, 0);
+		if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+			xf_emit(ctx, 1, 0x11);
+		xf_emit(ctx, 1, 0);
+		for (i = 0; i < 8; i++)
+			if (units & (1<<(i+16)))
+				nv50_graph_construct_gene_ropc(ctx);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+
+		/* Strand 6 */
+		ctx->ctxvals_pos = offset + 6;
+		nv50_graph_construct_gene_unk3(ctx);
+		xf_emit(ctx, 0xb, 0);
+		nv50_graph_construct_gene_unk4(ctx);
+		nv50_graph_construct_gene_unk7(ctx);
+		if (units & (1 << 0))
+			nv50_graph_construct_xfer_tp(ctx);
+		if (units & (1 << 1))
+			nv50_graph_construct_xfer_tp(ctx);
+		if (units & (1 << 2))
+			nv50_graph_construct_xfer_tp(ctx);
+		if (units & (1 << 3))
+			nv50_graph_construct_xfer_tp(ctx);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+
+		/* Strand 7 */
+		ctx->ctxvals_pos = offset + 7;
+		if (dev_priv->chipset == 0xa0) {
+			if (units & (1 << 4))
+				nv50_graph_construct_xfer_tp(ctx);
+			if (units & (1 << 5))
+				nv50_graph_construct_xfer_tp(ctx);
+			if (units & (1 << 6))
+				nv50_graph_construct_xfer_tp(ctx);
+			if (units & (1 << 7))
+				nv50_graph_construct_xfer_tp(ctx);
+			if (units & (1 << 8))
+				nv50_graph_construct_xfer_tp(ctx);
+			if (units & (1 << 9))
+				nv50_graph_construct_xfer_tp(ctx);
+		} else {
+			nv50_graph_construct_gene_unk2(ctx);
+		}
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+	}
+
+	ctx->ctxvals_pos = offset + size * 8;
+	ctx->ctxvals_pos = (ctx->ctxvals_pos+0x3f)&~0x3f;
+	cp_lsr (ctx, offset);
+	cp_out (ctx, CP_SET_XFER_POINTER);
+	cp_lsr (ctx, size);
+	cp_out (ctx, CP_SEEK_1);
+	cp_out (ctx, CP_XFER_1);
+	cp_wait(ctx, XFER, BUSY);
+}
+
+/*
+ * non-trivial demagiced parts of ctx init go here
+ */
+
+static void
+nv50_graph_construct_gene_m2mf(struct nouveau_grctx *ctx)
+{
+	/* m2mf state */
+	xf_emit (ctx, 1, 0);		/* DMA_NOTIFY instance >> 4 */
+	xf_emit (ctx, 1, 0);		/* DMA_BUFFER_IN instance >> 4 */
+	xf_emit (ctx, 1, 0);		/* DMA_BUFFER_OUT instance >> 4 */
+	xf_emit (ctx, 1, 0);		/* OFFSET_IN */
+	xf_emit (ctx, 1, 0);		/* OFFSET_OUT */
+	xf_emit (ctx, 1, 0);		/* PITCH_IN */
+	xf_emit (ctx, 1, 0);		/* PITCH_OUT */
+	xf_emit (ctx, 1, 0);		/* LINE_LENGTH */
+	xf_emit (ctx, 1, 0);		/* LINE_COUNT */
+	xf_emit (ctx, 1, 0x21);		/* FORMAT: bits 0-4 INPUT_INC, bits 5-9 OUTPUT_INC */
+	xf_emit (ctx, 1, 1);		/* LINEAR_IN */
+	xf_emit (ctx, 1, 0x2);		/* TILING_MODE_IN: bits 0-2 y tiling, bits 3-5 z tiling */
+	xf_emit (ctx, 1, 0x100);	/* TILING_PITCH_IN */
+	xf_emit (ctx, 1, 0x100);	/* TILING_HEIGHT_IN */
+	xf_emit (ctx, 1, 1);		/* TILING_DEPTH_IN */
+	xf_emit (ctx, 1, 0);		/* TILING_POSITION_IN_Z */
+	xf_emit (ctx, 1, 0);		/* TILING_POSITION_IN */
+	xf_emit (ctx, 1, 1);		/* LINEAR_OUT */
+	xf_emit (ctx, 1, 0x2);		/* TILING_MODE_OUT: bits 0-2 y tiling, bits 3-5 z tiling */
+	xf_emit (ctx, 1, 0x100);	/* TILING_PITCH_OUT */
+	xf_emit (ctx, 1, 0x100);	/* TILING_HEIGHT_OUT */
+	xf_emit (ctx, 1, 1);		/* TILING_DEPTH_OUT */
+	xf_emit (ctx, 1, 0);		/* TILING_POSITION_OUT_Z */
+	xf_emit (ctx, 1, 0);		/* TILING_POSITION_OUT */
+	xf_emit (ctx, 1, 0);		/* OFFSET_IN_HIGH */
+	xf_emit (ctx, 1, 0);		/* OFFSET_OUT_HIGH */
+}
+
+static void
+nv50_graph_construct_gene_unk1(struct nouveau_grctx *ctx)
+{
+	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
+	/* end of area 2 on pre-NVA0, area 1 on NVAx */
+	xf_emit(ctx, 2, 4);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 0x80);
+	xf_emit(ctx, 1, 4);
+	xf_emit(ctx, 1, 0x80c14);
+	xf_emit(ctx, 1, 0);
+	if (dev_priv->chipset == 0x50)
+		xf_emit(ctx, 1, 0x3ff);
+	else
+		xf_emit(ctx, 1, 0x7ff);
+	switch (dev_priv->chipset) {
+	case 0x50:
+	case 0x86:
+	case 0x98:
+	case 0xaa:
+	case 0xac:
+		xf_emit(ctx, 0x542, 0);
+		break;
+	case 0x84:
+	case 0x92:
+	case 0x94:
+	case 0x96:
+		xf_emit(ctx, 0x942, 0);
+		break;
+	case 0xa0:
+		xf_emit(ctx, 0x2042, 0);
+		break;
+	case 0xa5:
+	case 0xa8:
+		xf_emit(ctx, 0x842, 0);
+		break;
+	}
+	xf_emit(ctx, 2, 4);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 0x80);
+	xf_emit(ctx, 1, 4);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 0x27);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 0x26);
+	xf_emit(ctx, 3, 0);
+}
+
+static void
+nv50_graph_construct_gene_unk10(struct nouveau_grctx *ctx)
+{
+	/* end of area 2 on pre-NVA0, area 1 on NVAx */
+	xf_emit(ctx, 0x10, 0x04000000);
+	xf_emit(ctx, 0x24, 0);
+	xf_emit(ctx, 2, 0x04e3bfdf);
+	xf_emit(ctx, 2, 0);
+	xf_emit(ctx, 1, 0x1fe21);
+}
+
+static void
+nv50_graph_construct_gene_unk2(struct nouveau_grctx *ctx)
+{
+	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
+	/* middle of area 2 on pre-NVA0, beginning of area 2 on NVA0, area 7 on >NVA0 */
+	if (dev_priv->chipset != 0x50) {
+		xf_emit(ctx, 5, 0);
+		xf_emit(ctx, 1, 0x80c14);
+		xf_emit(ctx, 2, 0);
+		xf_emit(ctx, 1, 0x804);
+		xf_emit(ctx, 1, 0);
+		xf_emit(ctx, 2, 4);
+		xf_emit(ctx, 1, 0x8100c12);
+	}
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 2, 4);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 0x10);
+	if (dev_priv->chipset == 0x50)
+		xf_emit(ctx, 3, 0);
+	else
+		xf_emit(ctx, 4, 0);
+	xf_emit(ctx, 1, 0x804);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 0x1a);
+	if (dev_priv->chipset != 0x50)
+		xf_emit(ctx, 1, 0x7f);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 0x80c14);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 0x8100c12);
+	xf_emit(ctx, 2, 4);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 0x10);
+	xf_emit(ctx, 3, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 0x8100c12);
+	xf_emit(ctx, 6, 0);
+	if (dev_priv->chipset == 0x50)
+		xf_emit(ctx, 1, 0x3ff);
+	else
+		xf_emit(ctx, 1, 0x7ff);
+	xf_emit(ctx, 1, 0x80c14);
+	xf_emit(ctx, 0x38, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 2, 0);
+	xf_emit(ctx, 1, 0x10);
+	xf_emit(ctx, 0x38, 0);
+	xf_emit(ctx, 2, 0x88);
+	xf_emit(ctx, 2, 0);
+	xf_emit(ctx, 1, 4);
+	xf_emit(ctx, 0x16, 0);
+	xf_emit(ctx, 1, 0x26);
+	xf_emit(ctx, 2, 0);
+	xf_emit(ctx, 1, 0x3f800000);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+		xf_emit(ctx, 4, 0);
+	else
+		xf_emit(ctx, 3, 0);
+	xf_emit(ctx, 1, 0x1a);
+	xf_emit(ctx, 1, 0x10);
+	if (dev_priv->chipset != 0x50)
+		xf_emit(ctx, 0x28, 0);
+	else
+		xf_emit(ctx, 0x25, 0);
+	xf_emit(ctx, 1, 0x52);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 0x26);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 2, 4);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 0x1a);
+	xf_emit(ctx, 2, 0);
+	xf_emit(ctx, 1, 0x00ffff00);
+	xf_emit(ctx, 1, 0);
+}
+
+static void
+nv50_graph_construct_gene_unk3(struct nouveau_grctx *ctx)
+{
+	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
+	/* end of area 0 on pre-NVA0, beginning of area 6 on NVAx */
+	xf_emit(ctx, 1, 0x3f);
+	xf_emit(ctx, 0xa, 0);
+	xf_emit(ctx, 1, 2);
+	xf_emit(ctx, 2, 0x04000000);
+	xf_emit(ctx, 8, 0);
+	xf_emit(ctx, 1, 4);
+	xf_emit(ctx, 3, 0);
+	xf_emit(ctx, 1, 4);
+	if (dev_priv->chipset == 0x50)
+		xf_emit(ctx, 0x10, 0);
+	else
+		xf_emit(ctx, 0x11, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 0x1001);
+	xf_emit(ctx, 4, 0xffff);
+	xf_emit(ctx, 0x20, 0);
+	xf_emit(ctx, 0x10, 0x3f800000);
+	xf_emit(ctx, 1, 0x10);
+	if (dev_priv->chipset == 0x50)
+		xf_emit(ctx, 1, 0);
+	else
+		xf_emit(ctx, 2, 0);
+	xf_emit(ctx, 1, 3);
+	xf_emit(ctx, 2, 0);
+}
+
+static void
+nv50_graph_construct_gene_unk4(struct nouveau_grctx *ctx)
+{
+	/* middle of area 0 on pre-NVA0, middle of area 6 on NVAx */
+	xf_emit(ctx, 2, 0x04000000);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 0x80);
+	xf_emit(ctx, 3, 0);
+	xf_emit(ctx, 1, 0x80);
+	xf_emit(ctx, 1, 0);
+}
+
+static void
+nv50_graph_construct_gene_unk5(struct nouveau_grctx *ctx)
+{
+	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
+	/* middle of area 0 on pre-NVA0 [after m2mf], end of area 2 on NVAx */
+	xf_emit(ctx, 2, 4);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+		xf_emit(ctx, 0x1c4d, 0);
+	else
+		xf_emit(ctx, 0x1c4b, 0);
+	xf_emit(ctx, 2, 4);
+	xf_emit(ctx, 1, 0x8100c12);
+	if (dev_priv->chipset != 0x50)
+		xf_emit(ctx, 1, 3);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 0x8100c12);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 0x80c14);
+	xf_emit(ctx, 1, 1);
+	if (dev_priv->chipset >= 0xa0)
+		xf_emit(ctx, 2, 4);
+	xf_emit(ctx, 1, 0x80c14);
+	xf_emit(ctx, 2, 0);
+	xf_emit(ctx, 1, 0x8100c12);
+	xf_emit(ctx, 1, 0x27);
+	xf_emit(ctx, 2, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 0x3c1, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 0x16, 0);
+	xf_emit(ctx, 1, 0x8100c12);
+	xf_emit(ctx, 1, 0);
+}
+
+static void
+nv50_graph_construct_gene_unk6(struct nouveau_grctx *ctx)
+{
+	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
+	/* beginning of area 1 on pre-NVA0 [after m2mf], area 3 on NVAx */
+	xf_emit(ctx, 4, 0);
+	xf_emit(ctx, 1, 0xf);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+		xf_emit(ctx, 8, 0);
+	else
+		xf_emit(ctx, 4, 0);
+	xf_emit(ctx, 1, 0x20);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+		xf_emit(ctx, 0x11, 0);
+	else if (dev_priv->chipset >= 0xa0)
+		xf_emit(ctx, 0xf, 0);
+	else
+		xf_emit(ctx, 0xe, 0);
+	xf_emit(ctx, 1, 0x1a);
+	xf_emit(ctx, 0xd, 0);
+	xf_emit(ctx, 2, 4);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 4);
+	xf_emit(ctx, 1, 8);
+	xf_emit(ctx, 1, 0);
+	if (dev_priv->chipset == 0x50)
+		xf_emit(ctx, 1, 0x3ff);
+	else
+		xf_emit(ctx, 1, 0x7ff);
+	if (dev_priv->chipset == 0xa8)
+		xf_emit(ctx, 1, 0x1e00);
+	xf_emit(ctx, 0xc, 0);
+	xf_emit(ctx, 1, 0xf);
+	if (dev_priv->chipset == 0x50)
+		xf_emit(ctx, 0x125, 0);
+	else if (dev_priv->chipset < 0xa0)
+		xf_emit(ctx, 0x126, 0);
+	else if (dev_priv->chipset == 0xa0 || dev_priv->chipset >= 0xaa)
+		xf_emit(ctx, 0x124, 0);
+	else
+		xf_emit(ctx, 0x1f7, 0);
+	xf_emit(ctx, 1, 0xf);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+		xf_emit(ctx, 3, 0);
+	else
+		xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 1);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+		xf_emit(ctx, 0xa1, 0);
+	else
+		xf_emit(ctx, 0x5a, 0);
+	xf_emit(ctx, 1, 0xf);
+	if (dev_priv->chipset < 0xa0)
+		xf_emit(ctx, 0x834, 0);
+	else if (dev_priv->chipset == 0xa0)
+		xf_emit(ctx, 0x1873, 0);
+	else if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+		xf_emit(ctx, 0x8ba, 0);
+	else
+		xf_emit(ctx, 0x833, 0);
+	xf_emit(ctx, 1, 0xf);
+	xf_emit(ctx, 0xf, 0);
+}
+
+static void
+nv50_graph_construct_gene_unk7(struct nouveau_grctx *ctx)
+{
+	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
+	/* middle of area 1 on pre-NVA0 [after m2mf], middle of area 6 on NVAx */
+	xf_emit(ctx, 2, 0);
+	if (dev_priv->chipset == 0x50)
+		xf_emit(ctx, 2, 1);
+	else
+		xf_emit(ctx, 2, 0);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 2, 0x100);
+	xf_emit(ctx, 1, 0x11);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 8);
+	xf_emit(ctx, 5, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 3, 1);
+	xf_emit(ctx, 1, 0xcf);
+	xf_emit(ctx, 1, 2);
+	xf_emit(ctx, 6, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 3, 1);
+	xf_emit(ctx, 4, 0);
+	xf_emit(ctx, 1, 4);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 0x15);
+	xf_emit(ctx, 3, 0);
+	xf_emit(ctx, 1, 0x4444480);
+	xf_emit(ctx, 0x37, 0);
+}
+
+static void
+nv50_graph_construct_gene_unk8(struct nouveau_grctx *ctx)
+{
+	/* middle of area 1 on pre-NVA0 [after m2mf], middle of area 0 on NVAx */
+	xf_emit(ctx, 4, 0);
+	xf_emit(ctx, 1, 0x8100c12);
+	xf_emit(ctx, 4, 0);
+	xf_emit(ctx, 1, 0x100);
+	xf_emit(ctx, 2, 0);
+	xf_emit(ctx, 1, 0x10001);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 0x10001);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 0x10001);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 4);
+	xf_emit(ctx, 1, 2);
+}
+
+static void
+nv50_graph_construct_gene_unk9(struct nouveau_grctx *ctx)
+{
+	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
+	/* middle of area 2 on pre-NVA0 [after m2mf], end of area 0 on NVAx */
+	xf_emit(ctx, 1, 0x3f800000);
+	xf_emit(ctx, 6, 0);
+	xf_emit(ctx, 1, 4);
+	xf_emit(ctx, 1, 0x1a);
+	xf_emit(ctx, 2, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 0x12, 0);
+	xf_emit(ctx, 1, 0x00ffff00);
+	xf_emit(ctx, 6, 0);
+	xf_emit(ctx, 1, 0xf);
+	xf_emit(ctx, 7, 0);
+	xf_emit(ctx, 1, 0x0fac6881);
+	xf_emit(ctx, 1, 0x11);
+	xf_emit(ctx, 0xf, 0);
+	xf_emit(ctx, 1, 4);
+	xf_emit(ctx, 2, 0);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+		xf_emit(ctx, 1, 3);
+	else if (dev_priv->chipset >= 0xa0)
+		xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 2, 0);
+	xf_emit(ctx, 1, 2);
+	xf_emit(ctx, 2, 0x04000000);
+	xf_emit(ctx, 3, 0);
+	xf_emit(ctx, 1, 5);
+	xf_emit(ctx, 1, 0x52);
+	if (dev_priv->chipset == 0x50) {
+		xf_emit(ctx, 0x13, 0);
+	} else {
+		xf_emit(ctx, 4, 0);
+		xf_emit(ctx, 1, 1);
+		if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+			xf_emit(ctx, 0x11, 0);
+		else
+			xf_emit(ctx, 0x10, 0);
+	}
+	xf_emit(ctx, 0x10, 0x3f800000);
+	xf_emit(ctx, 1, 0x10);
+	xf_emit(ctx, 0x26, 0);
+	xf_emit(ctx, 1, 0x8100c12);
+	xf_emit(ctx, 1, 5);
+	xf_emit(ctx, 2, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 4, 0xffff);
+	if (dev_priv->chipset != 0x50)
+		xf_emit(ctx, 1, 3);
+	if (dev_priv->chipset < 0xa0)
+		xf_emit(ctx, 0x1f, 0);
+	else if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+		xf_emit(ctx, 0xc, 0);
+	else
+		xf_emit(ctx, 3, 0);
+	xf_emit(ctx, 1, 0x00ffff00);
+	xf_emit(ctx, 1, 0x1a);
+	if (dev_priv->chipset != 0x50) {
+		xf_emit(ctx, 1, 0);
+		xf_emit(ctx, 1, 3);
+	}
+	if (dev_priv->chipset < 0xa0)
+		xf_emit(ctx, 0x26, 0);
+	else
+		xf_emit(ctx, 0x3c, 0);
+	xf_emit(ctx, 1, 0x102);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 4, 4);
+	if (dev_priv->chipset >= 0xa0)
+		xf_emit(ctx, 8, 0);
+	xf_emit(ctx, 2, 4);
+	xf_emit(ctx, 1, 0);
+	if (dev_priv->chipset == 0x50)
+		xf_emit(ctx, 1, 0x3ff);
+	else
+		xf_emit(ctx, 1, 0x7ff);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 0x102);
+	xf_emit(ctx, 9, 0);
+	xf_emit(ctx, 4, 4);
+	xf_emit(ctx, 0x2c, 0);
+}
+
+static void
+nv50_graph_construct_gene_ropc(struct nouveau_grctx *ctx)
+{
+	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
+	int magic2;
+	if (dev_priv->chipset == 0x50) {
+		magic2 = 0x00003e60;
+	} else if (dev_priv->chipset <= 0xa0 || dev_priv->chipset >= 0xaa) {
+		magic2 = 0x001ffe67;
+	} else {
+		magic2 = 0x00087e67;
+	}
+	xf_emit(ctx, 8, 0);
+	xf_emit(ctx, 1, 2);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, magic2);
+	xf_emit(ctx, 4, 0);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+		xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 7, 0);
+	if (dev_priv->chipset >= 0xa0 && dev_priv->chipset < 0xaa)
+		xf_emit(ctx, 1, 0x15);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 0x10);
+	xf_emit(ctx, 2, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 4, 0);
+	if (dev_priv->chipset == 0x86 || dev_priv->chipset == 0x92 || dev_priv->chipset == 0x98 || dev_priv->chipset >= 0xa0) {
+		xf_emit(ctx, 1, 4);
+		xf_emit(ctx, 1, 0x400);
+		xf_emit(ctx, 1, 0x300);
+		xf_emit(ctx, 1, 0x1001);
+		if (dev_priv->chipset != 0xa0) {
+			if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+				xf_emit(ctx, 1, 0);
+			else
+				xf_emit(ctx, 1, 0x15);
+		}
+		xf_emit(ctx, 3, 0);
+	}
+	xf_emit(ctx, 2, 0);
+	xf_emit(ctx, 1, 2);
+	xf_emit(ctx, 8, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 0x10);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 0x13, 0);
+	xf_emit(ctx, 1, 0x10);
+	xf_emit(ctx, 0x10, 0);
+	xf_emit(ctx, 0x10, 0x3f800000);
+	xf_emit(ctx, 0x19, 0);
+	xf_emit(ctx, 1, 0x10);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 0x3f);
+	xf_emit(ctx, 6, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 1);
+	if (dev_priv->chipset >= 0xa0) {
+		xf_emit(ctx, 2, 0);
+		xf_emit(ctx, 1, 0x1001);
+		xf_emit(ctx, 0xb, 0);
+	} else {
+		xf_emit(ctx, 0xc, 0);
+	}
+	xf_emit(ctx, 1, 0x11);
+	xf_emit(ctx, 7, 0);
+	xf_emit(ctx, 1, 0xf);
+	xf_emit(ctx, 7, 0);
+	xf_emit(ctx, 1, 0x11);
+	if (dev_priv->chipset == 0x50)
+		xf_emit(ctx, 4, 0);
+	else
+		xf_emit(ctx, 6, 0);
+	xf_emit(ctx, 3, 1);
+	xf_emit(ctx, 1, 2);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 2);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, magic2);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 0x0fac6881);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) {
+		xf_emit(ctx, 1, 0);
+		xf_emit(ctx, 0x18, 1);
+		xf_emit(ctx, 8, 2);
+		xf_emit(ctx, 8, 1);
+		xf_emit(ctx, 8, 2);
+		xf_emit(ctx, 8, 1);
+		xf_emit(ctx, 3, 0);
+		xf_emit(ctx, 1, 1);
+		xf_emit(ctx, 5, 0);
+		xf_emit(ctx, 1, 1);
+		xf_emit(ctx, 0x16, 0);
+	} else {
+		if (dev_priv->chipset >= 0xa0)
+			xf_emit(ctx, 0x1b, 0);
+		else
+			xf_emit(ctx, 0x15, 0);
+	}
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 2);
+	xf_emit(ctx, 2, 1);
+	xf_emit(ctx, 1, 2);
+	xf_emit(ctx, 2, 1);
+	if (dev_priv->chipset >= 0xa0)
+		xf_emit(ctx, 4, 0);
+	else
+		xf_emit(ctx, 3, 0);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) {
+		xf_emit(ctx, 0x10, 1);
+		xf_emit(ctx, 8, 2);
+		xf_emit(ctx, 0x10, 1);
+		xf_emit(ctx, 8, 2);
+		xf_emit(ctx, 8, 1);
+		xf_emit(ctx, 3, 0);
+	}
+	xf_emit(ctx, 1, 0x11);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 0x5b, 0);
+}
+
+static void
+nv50_graph_construct_xfer_tp_x1(struct nouveau_grctx *ctx)
+{
+	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
+	int magic3;
+	if (dev_priv->chipset == 0x50)
+		magic3 = 0x1000;
+	else if (dev_priv->chipset == 0x86 || dev_priv->chipset == 0x98 || dev_priv->chipset >= 0xa8)
+		magic3 = 0x1e00;
+	else
+		magic3 = 0;
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 4);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+		xf_emit(ctx, 0x24, 0);
+	else if (dev_priv->chipset >= 0xa0)
+		xf_emit(ctx, 0x14, 0);
+	else
+		xf_emit(ctx, 0x15, 0);
+	xf_emit(ctx, 2, 4);
+	if (dev_priv->chipset >= 0xa0)
+		xf_emit(ctx, 1, 0x03020100);
+	else
+		xf_emit(ctx, 1, 0x00608080);
+	xf_emit(ctx, 4, 0);
+	xf_emit(ctx, 1, 4);
+	xf_emit(ctx, 2, 0);
+	xf_emit(ctx, 2, 4);
+	xf_emit(ctx, 1, 0x80);
+	if (magic3)
+		xf_emit(ctx, 1, magic3);
+	xf_emit(ctx, 1, 4);
+	xf_emit(ctx, 0x24, 0);
+	xf_emit(ctx, 1, 4);
+	xf_emit(ctx, 1, 0x80);
+	xf_emit(ctx, 1, 4);
+	xf_emit(ctx, 1, 0x03020100);
+	xf_emit(ctx, 1, 3);
+	if (magic3)
+		xf_emit(ctx, 1, magic3);
+	xf_emit(ctx, 1, 4);
+	xf_emit(ctx, 4, 0);
+	xf_emit(ctx, 1, 4);
+	xf_emit(ctx, 1, 3);
+	xf_emit(ctx, 3, 0);
+	xf_emit(ctx, 1, 4);
+	if (dev_priv->chipset == 0x94 || dev_priv->chipset == 0x96)
+		xf_emit(ctx, 0x1024, 0);
+	else if (dev_priv->chipset < 0xa0)
+		xf_emit(ctx, 0xa24, 0);
+	else if (dev_priv->chipset == 0xa0 || dev_priv->chipset >= 0xaa)
+		xf_emit(ctx, 0x214, 0);
+	else
+		xf_emit(ctx, 0x414, 0);
+	xf_emit(ctx, 1, 4);
+	xf_emit(ctx, 1, 3);
+	xf_emit(ctx, 2, 0);
+}
+
+static void
+nv50_graph_construct_xfer_tp_x2(struct nouveau_grctx *ctx)
+{
+	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
+	int magic1, magic2;
+	if (dev_priv->chipset == 0x50) {
+		magic1 = 0x3ff;
+		magic2 = 0x00003e60;
+	} else if (dev_priv->chipset <= 0xa0 || dev_priv->chipset >= 0xaa) {
+		magic1 = 0x7ff;
+		magic2 = 0x001ffe67;
+	} else {
+		magic1 = 0x7ff;
+		magic2 = 0x00087e67;
+	}
+	xf_emit(ctx, 3, 0);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+		xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 0xc, 0);
+	xf_emit(ctx, 1, 0xf);
+	xf_emit(ctx, 0xb, 0);
+	xf_emit(ctx, 1, 4);
+	xf_emit(ctx, 4, 0xffff);
+	xf_emit(ctx, 8, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 3, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 5, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 2, 0);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) {
+		xf_emit(ctx, 1, 3);
+		xf_emit(ctx, 1, 0);
+	} else if (dev_priv->chipset >= 0xa0)
+		xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 0xa, 0);
+	xf_emit(ctx, 2, 1);
+	xf_emit(ctx, 1, 2);
+	xf_emit(ctx, 2, 1);
+	xf_emit(ctx, 1, 2);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) {
+		xf_emit(ctx, 1, 0);
+		xf_emit(ctx, 0x18, 1);
+		xf_emit(ctx, 8, 2);
+		xf_emit(ctx, 8, 1);
+		xf_emit(ctx, 8, 2);
+		xf_emit(ctx, 8, 1);
+		xf_emit(ctx, 1, 0);
+	}
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 0x11);
+	xf_emit(ctx, 7, 0);
+	xf_emit(ctx, 1, 0x0fac6881);
+	xf_emit(ctx, 2, 0);
+	xf_emit(ctx, 1, 4);
+	xf_emit(ctx, 3, 0);
+	xf_emit(ctx, 1, 0x11);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 3, 0xcf);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+		xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 0xa, 0);
+	xf_emit(ctx, 2, 1);
+	xf_emit(ctx, 1, 2);
+	xf_emit(ctx, 2, 1);
+	xf_emit(ctx, 1, 2);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 8, 1);
+	xf_emit(ctx, 1, 0x11);
+	xf_emit(ctx, 7, 0);
+	xf_emit(ctx, 1, 0x0fac6881);
+	xf_emit(ctx, 1, 0xf);
+	xf_emit(ctx, 7, 0);
+	xf_emit(ctx, 1, magic2);
+	xf_emit(ctx, 2, 0);
+	xf_emit(ctx, 1, 0x11);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+		xf_emit(ctx, 2, 1);
+	else
+		xf_emit(ctx, 1, 1);
+	if(dev_priv->chipset == 0x50)
+		xf_emit(ctx, 1, 0);
+	else
+		xf_emit(ctx, 3, 0);
+	xf_emit(ctx, 1, 4);
+	xf_emit(ctx, 5, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 4, 0);
+	xf_emit(ctx, 1, 0x11);
+	xf_emit(ctx, 7, 0);
+	xf_emit(ctx, 1, 0x0fac6881);
+	xf_emit(ctx, 3, 0);
+	xf_emit(ctx, 1, 0x11);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, magic1);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 2, 0);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+		xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 0x28, 0);
+	xf_emit(ctx, 8, 8);
+	xf_emit(ctx, 1, 0x11);
+	xf_emit(ctx, 7, 0);
+	xf_emit(ctx, 1, 0x0fac6881);
+	xf_emit(ctx, 8, 0x400);
+	xf_emit(ctx, 8, 0x300);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 0xf);
+	xf_emit(ctx, 7, 0);
+	xf_emit(ctx, 1, 0x20);
+	xf_emit(ctx, 1, 0x11);
+	xf_emit(ctx, 1, 0x100);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 2, 0);
+	xf_emit(ctx, 1, 0x40);
+	xf_emit(ctx, 1, 0x100);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 3);
+	xf_emit(ctx, 4, 0);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+		xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, magic2);
+	xf_emit(ctx, 3, 0);
+	xf_emit(ctx, 1, 2);
+	xf_emit(ctx, 1, 0x0fac6881);
+	xf_emit(ctx, 9, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 4, 0);
+	xf_emit(ctx, 1, 4);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 0x400);
+	xf_emit(ctx, 1, 0x300);
+	xf_emit(ctx, 1, 0x1001);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+		xf_emit(ctx, 4, 0);
+	else
+		xf_emit(ctx, 3, 0);
+	xf_emit(ctx, 1, 0x11);
+	xf_emit(ctx, 7, 0);
+	xf_emit(ctx, 1, 0x0fac6881);
+	xf_emit(ctx, 1, 0xf);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) {
+		xf_emit(ctx, 0x15, 0);
+		xf_emit(ctx, 1, 1);
+		xf_emit(ctx, 3, 0);
+	} else
+		xf_emit(ctx, 0x17, 0);
+	if (dev_priv->chipset >= 0xa0)
+		xf_emit(ctx, 1, 0x0fac6881);
+	xf_emit(ctx, 1, magic2);
+	xf_emit(ctx, 3, 0);
+	xf_emit(ctx, 1, 0x11);
+	xf_emit(ctx, 2, 0);
+	xf_emit(ctx, 1, 4);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 2, 1);
+	xf_emit(ctx, 3, 0);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+		xf_emit(ctx, 2, 1);
+	else
+		xf_emit(ctx, 1, 1);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+		xf_emit(ctx, 2, 0);
+	else if (dev_priv->chipset != 0x50)
+		xf_emit(ctx, 1, 0);
+}
+
+static void
+nv50_graph_construct_xfer_tp_x3(struct nouveau_grctx *ctx)
+{
+	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
+	xf_emit(ctx, 3, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 1);
+	if (dev_priv->chipset == 0x50)
+		xf_emit(ctx, 2, 0);
+	else
+		xf_emit(ctx, 3, 0);
+	xf_emit(ctx, 1, 0x2a712488);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 0x4085c000);
+	xf_emit(ctx, 1, 0x40);
+	xf_emit(ctx, 1, 0x100);
+	xf_emit(ctx, 1, 0x10100);
+	xf_emit(ctx, 1, 0x02800000);
+}
+
+static void
+nv50_graph_construct_xfer_tp_x4(struct nouveau_grctx *ctx)
+{
+	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
+	xf_emit(ctx, 2, 0x04e3bfdf);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 0x00ffff00);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+		xf_emit(ctx, 2, 1);
+	else
+		xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 2, 0);
+	xf_emit(ctx, 1, 0x00ffff00);
+	xf_emit(ctx, 8, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 0x30201000);
+	xf_emit(ctx, 1, 0x70605040);
+	xf_emit(ctx, 1, 0xb8a89888);
+	xf_emit(ctx, 1, 0xf8e8d8c8);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 0x1a);
+}
+
+static void
+nv50_graph_construct_xfer_tp_x5(struct nouveau_grctx *ctx)
+{
+	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
+	xf_emit(ctx, 3, 0);
+	xf_emit(ctx, 1, 0xfac6881);
+	xf_emit(ctx, 4, 0);
+	xf_emit(ctx, 1, 4);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 2, 1);
+	xf_emit(ctx, 2, 0);
+	xf_emit(ctx, 1, 1);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+		xf_emit(ctx, 0xb, 0);
+	else
+		xf_emit(ctx, 0xa, 0);
+	xf_emit(ctx, 8, 1);
+	xf_emit(ctx, 1, 0x11);
+	xf_emit(ctx, 7, 0);
+	xf_emit(ctx, 1, 0xfac6881);
+	xf_emit(ctx, 1, 0xf);
+	xf_emit(ctx, 7, 0);
+	xf_emit(ctx, 1, 0x11);
+	xf_emit(ctx, 1, 1);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) {
+		xf_emit(ctx, 6, 0);
+		xf_emit(ctx, 1, 1);
+		xf_emit(ctx, 6, 0);
+	} else {
+		xf_emit(ctx, 0xb, 0);
+	}
+}
+
+static void
+nv50_graph_construct_xfer_tp(struct nouveau_grctx *ctx)
+{
+	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
+	if (dev_priv->chipset < 0xa0) {
+		nv50_graph_construct_xfer_tp_x1(ctx);
+		nv50_graph_construct_xfer_tp_x2(ctx);
+		nv50_graph_construct_xfer_tp_x3(ctx);
+		if (dev_priv->chipset == 0x50)
+			xf_emit(ctx, 0xf, 0);
+		else
+			xf_emit(ctx, 0x12, 0);
+		nv50_graph_construct_xfer_tp_x4(ctx);
+	} else {
+		nv50_graph_construct_xfer_tp_x3(ctx);
+		if (dev_priv->chipset < 0xaa)
+			xf_emit(ctx, 0xc, 0);
+		else
+			xf_emit(ctx, 0xa, 0);
+		nv50_graph_construct_xfer_tp_x2(ctx);
+		nv50_graph_construct_xfer_tp_x5(ctx);
+		nv50_graph_construct_xfer_tp_x4(ctx);
+		nv50_graph_construct_xfer_tp_x1(ctx);
+	}
+}
+
+static void
+nv50_graph_construct_xfer_tp2(struct nouveau_grctx *ctx)
+{
+	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
+	int i, mpcnt;
+	if (dev_priv->chipset == 0x98 || dev_priv->chipset == 0xaa)
+		mpcnt = 1;
+	else if (dev_priv->chipset < 0xa0 || dev_priv->chipset >= 0xa8)
+		mpcnt = 2;
+	else
+		mpcnt = 3;
+	for (i = 0; i < mpcnt; i++) {
+		xf_emit(ctx, 1, 0);
+		xf_emit(ctx, 1, 0x80);
+		xf_emit(ctx, 1, 0x80007004);
+		xf_emit(ctx, 1, 0x04000400);
+		if (dev_priv->chipset >= 0xa0)
+			xf_emit(ctx, 1, 0xc0);
+		xf_emit(ctx, 1, 0x1000);
+		xf_emit(ctx, 2, 0);
+		if (dev_priv->chipset == 0x86 || dev_priv->chipset == 0x98 || dev_priv->chipset >= 0xa8) {
+			xf_emit(ctx, 1, 0xe00);
+			xf_emit(ctx, 1, 0x1e00);
+		}
+		xf_emit(ctx, 1, 1);
+		xf_emit(ctx, 2, 0);
+		if (dev_priv->chipset == 0x50)
+			xf_emit(ctx, 2, 0x1000);
+		xf_emit(ctx, 1, 1);
+		xf_emit(ctx, 1, 0);
+		xf_emit(ctx, 1, 4);
+		xf_emit(ctx, 1, 2);
+		if (dev_priv->chipset >= 0xaa)
+			xf_emit(ctx, 0xb, 0);
+		else if (dev_priv->chipset >= 0xa0)
+			xf_emit(ctx, 0xc, 0);
+		else
+			xf_emit(ctx, 0xa, 0);
+	}
+	xf_emit(ctx, 1, 0x08100c12);
+	xf_emit(ctx, 1, 0);
+	if (dev_priv->chipset >= 0xa0) {
+		xf_emit(ctx, 1, 0x1fe21);
+	}
+	xf_emit(ctx, 5, 0);
+	xf_emit(ctx, 4, 0xffff);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 2, 0x10001);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 0x1fe21);
+	xf_emit(ctx, 1, 0);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+		xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 4, 0);
+	xf_emit(ctx, 1, 0x08100c12);
+	xf_emit(ctx, 1, 4);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 2);
+	xf_emit(ctx, 1, 0x11);
+	xf_emit(ctx, 8, 0);
+	xf_emit(ctx, 1, 0xfac6881);
+	xf_emit(ctx, 1, 0);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa)
+		xf_emit(ctx, 1, 3);
+	xf_emit(ctx, 3, 0);
+	xf_emit(ctx, 1, 4);
+	xf_emit(ctx, 9, 0);
+	xf_emit(ctx, 1, 2);
+	xf_emit(ctx, 2, 1);
+	xf_emit(ctx, 1, 2);
+	xf_emit(ctx, 3, 1);
+	xf_emit(ctx, 1, 0);
+	if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) {
+		xf_emit(ctx, 8, 2);
+		xf_emit(ctx, 0x10, 1);
+		xf_emit(ctx, 8, 2);
+		xf_emit(ctx, 0x18, 1);
+		xf_emit(ctx, 3, 0);
+	}
+	xf_emit(ctx, 1, 4);
+	if (dev_priv->chipset == 0x50)
+		xf_emit(ctx, 0x3a0, 0);
+	else if (dev_priv->chipset < 0x94)
+		xf_emit(ctx, 0x3a2, 0);
+	else if (dev_priv->chipset == 0x98 || dev_priv->chipset == 0xaa)
+		xf_emit(ctx, 0x39f, 0);
+	else
+		xf_emit(ctx, 0x3a3, 0);
+	xf_emit(ctx, 1, 0x11);
+	xf_emit(ctx, 1, 0);
+	xf_emit(ctx, 1, 1);
+	xf_emit(ctx, 0x2d, 0);
+}
+
+static void
+nv50_graph_construct_xfer2(struct nouveau_grctx *ctx)
+{
+	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
+	int i;
+	uint32_t offset;
+	uint32_t units = nv_rd32 (ctx->dev, 0x1540);
+	int size = 0;
+
+	offset = (ctx->ctxvals_pos+0x3f)&~0x3f;
+
+	if (dev_priv->chipset < 0xa0) {
+		for (i = 0; i < 8; i++) {
+			ctx->ctxvals_pos = offset + i;
+			if (i == 0)
+				xf_emit(ctx, 1, 0x08100c12);
+			if (units & (1 << i))
+				nv50_graph_construct_xfer_tp2(ctx);
+			if ((ctx->ctxvals_pos-offset)/8 > size)
+				size = (ctx->ctxvals_pos-offset)/8;
+		}
+	} else {
+		/* Strand 0: TPs 0, 1 */
+		ctx->ctxvals_pos = offset;
+		xf_emit(ctx, 1, 0x08100c12);
+		if (units & (1 << 0))
+			nv50_graph_construct_xfer_tp2(ctx);
+		if (units & (1 << 1))
+			nv50_graph_construct_xfer_tp2(ctx);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+
+		/* Strand 0: TPs 2, 3 */
+		ctx->ctxvals_pos = offset + 1;
+		if (units & (1 << 2))
+			nv50_graph_construct_xfer_tp2(ctx);
+		if (units & (1 << 3))
+			nv50_graph_construct_xfer_tp2(ctx);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+
+		/* Strand 0: TPs 4, 5, 6 */
+		ctx->ctxvals_pos = offset + 2;
+		if (units & (1 << 4))
+			nv50_graph_construct_xfer_tp2(ctx);
+		if (units & (1 << 5))
+			nv50_graph_construct_xfer_tp2(ctx);
+		if (units & (1 << 6))
+			nv50_graph_construct_xfer_tp2(ctx);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+
+		/* Strand 0: TPs 7, 8, 9 */
+		ctx->ctxvals_pos = offset + 3;
+		if (units & (1 << 7))
+			nv50_graph_construct_xfer_tp2(ctx);
+		if (units & (1 << 8))
+			nv50_graph_construct_xfer_tp2(ctx);
+		if (units & (1 << 9))
+			nv50_graph_construct_xfer_tp2(ctx);
+		if ((ctx->ctxvals_pos-offset)/8 > size)
+			size = (ctx->ctxvals_pos-offset)/8;
+	}
+	ctx->ctxvals_pos = offset + size * 8;
+	ctx->ctxvals_pos = (ctx->ctxvals_pos+0x3f)&~0x3f;
+	cp_lsr (ctx, offset);
+	cp_out (ctx, CP_SET_XFER_POINTER);
+	cp_lsr (ctx, size);
+	cp_out (ctx, CP_SEEK_2);
+	cp_out (ctx, CP_XFER_2);
+	cp_wait(ctx, XFER, BUSY);
+}
diff --git a/drivers/gpu/drm/nouveau/nv50_instmem.c b/drivers/gpu/drm/nouveau/nv50_instmem.c
new file mode 100644
index 000000000000..de1f5b0062c5
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv50_instmem.c
@@ -0,0 +1,531 @@
+/*
+ * Copyright (C) 2007 Ben Skeggs.
+ *
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "drmP.h"
+#include "drm.h"
+#include "nouveau_drv.h"
+
+struct nv50_instmem_priv {
+	uint32_t save1700[5]; /* 0x1700->0x1710 */
+
+	struct nouveau_gpuobj_ref *pramin_pt;
+	struct nouveau_gpuobj_ref *pramin_bar;
+	struct nouveau_gpuobj_ref *fb_bar;
+
+	bool last_access_wr;
+};
+
+#define NV50_INSTMEM_PAGE_SHIFT 12
+#define NV50_INSTMEM_PAGE_SIZE  (1 << NV50_INSTMEM_PAGE_SHIFT)
+#define NV50_INSTMEM_PT_SIZE(a)	(((a) >> 12) << 3)
+
+/*NOTE: - Assumes 0x1700 already covers the correct MiB of PRAMIN
+ */
+#define BAR0_WI32(g, o, v) do {                                   \
+	uint32_t offset;                                          \
+	if ((g)->im_backing) {                                    \
+		offset = (g)->im_backing_start;                   \
+	} else {                                                  \
+		offset  = chan->ramin->gpuobj->im_backing_start;  \
+		offset += (g)->im_pramin->start;                  \
+	}                                                         \
+	offset += (o);                                            \
+	nv_wr32(dev, NV_RAMIN + (offset & 0xfffff), (v));              \
+} while (0)
+
+int
+nv50_instmem_init(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_channel *chan;
+	uint32_t c_offset, c_size, c_ramfc, c_vmpd, c_base, pt_size;
+	struct nv50_instmem_priv *priv;
+	int ret, i;
+	uint32_t v, save_nv001700;
+
+	priv = kzalloc(sizeof(*priv), GFP_KERNEL);
+	if (!priv)
+		return -ENOMEM;
+	dev_priv->engine.instmem.priv = priv;
+
+	/* Save state, will restore at takedown. */
+	for (i = 0x1700; i <= 0x1710; i += 4)
+		priv->save1700[(i-0x1700)/4] = nv_rd32(dev, i);
+
+	if (dev_priv->chipset == 0xaa || dev_priv->chipset == 0xac)
+		dev_priv->vram_sys_base = nv_rd32(dev, 0x100e10) << 12;
+	else
+		dev_priv->vram_sys_base = 0;
+
+	/* Reserve the last MiB of VRAM, we should probably try to avoid
+	 * setting up the below tables over the top of the VBIOS image at
+	 * some point.
+	 */
+	dev_priv->ramin_rsvd_vram = 1 << 20;
+	c_offset = nouveau_mem_fb_amount(dev) - dev_priv->ramin_rsvd_vram;
+	c_size   = 128 << 10;
+	c_vmpd   = ((dev_priv->chipset & 0xf0) == 0x50) ? 0x1400 : 0x200;
+	c_ramfc  = ((dev_priv->chipset & 0xf0) == 0x50) ? 0x0 : 0x20;
+	c_base   = c_vmpd + 0x4000;
+	pt_size  = NV50_INSTMEM_PT_SIZE(dev_priv->ramin_size);
+
+	NV_DEBUG(dev, " Rsvd VRAM base: 0x%08x\n", c_offset);
+	NV_DEBUG(dev, "    VBIOS image: 0x%08x\n",
+				(nv_rd32(dev, 0x619f04) & ~0xff) << 8);
+	NV_DEBUG(dev, "  Aperture size: %d MiB\n", dev_priv->ramin_size >> 20);
+	NV_DEBUG(dev, "        PT size: %d KiB\n", pt_size >> 10);
+
+	/* Determine VM layout, we need to do this first to make sure
+	 * we allocate enough memory for all the page tables.
+	 */
+	dev_priv->vm_gart_base = roundup(NV50_VM_BLOCK, NV50_VM_BLOCK);
+	dev_priv->vm_gart_size = NV50_VM_BLOCK;
+
+	dev_priv->vm_vram_base = dev_priv->vm_gart_base + dev_priv->vm_gart_size;
+	dev_priv->vm_vram_size = nouveau_mem_fb_amount(dev);
+	if (dev_priv->vm_vram_size > NV50_VM_MAX_VRAM)
+		dev_priv->vm_vram_size = NV50_VM_MAX_VRAM;
+	dev_priv->vm_vram_size = roundup(dev_priv->vm_vram_size, NV50_VM_BLOCK);
+	dev_priv->vm_vram_pt_nr = dev_priv->vm_vram_size / NV50_VM_BLOCK;
+
+	dev_priv->vm_end = dev_priv->vm_vram_base + dev_priv->vm_vram_size;
+
+	NV_DEBUG(dev, "NV50VM: GART 0x%016llx-0x%016llx\n",
+		 dev_priv->vm_gart_base,
+		 dev_priv->vm_gart_base + dev_priv->vm_gart_size - 1);
+	NV_DEBUG(dev, "NV50VM: VRAM 0x%016llx-0x%016llx\n",
+		 dev_priv->vm_vram_base,
+		 dev_priv->vm_vram_base + dev_priv->vm_vram_size - 1);
+
+	c_size += dev_priv->vm_vram_pt_nr * (NV50_VM_BLOCK / 65536 * 8);
+
+	/* Map BAR0 PRAMIN aperture over the memory we want to use */
+	save_nv001700 = nv_rd32(dev, NV50_PUNK_BAR0_PRAMIN);
+	nv_wr32(dev, NV50_PUNK_BAR0_PRAMIN, (c_offset >> 16));
+
+	/* Create a fake channel, and use it as our "dummy" channels 0/127.
+	 * The main reason for creating a channel is so we can use the gpuobj
+	 * code.  However, it's probably worth noting that NVIDIA also setup
+	 * their channels 0/127 with the same values they configure here.
+	 * So, there may be some other reason for doing this.
+	 *
+	 * Have to create the entire channel manually, as the real channel
+	 * creation code assumes we have PRAMIN access, and we don't until
+	 * we're done here.
+	 */
+	chan = kzalloc(sizeof(*chan), GFP_KERNEL);
+	if (!chan)
+		return -ENOMEM;
+	chan->id = 0;
+	chan->dev = dev;
+	chan->file_priv = (struct drm_file *)-2;
+	dev_priv->fifos[0] = dev_priv->fifos[127] = chan;
+
+	/* Channel's PRAMIN object + heap */
+	ret = nouveau_gpuobj_new_fake(dev, 0, c_offset, c_size, 0,
+							NULL, &chan->ramin);
+	if (ret)
+		return ret;
+
+	if (nouveau_mem_init_heap(&chan->ramin_heap, c_base, c_size - c_base))
+		return -ENOMEM;
+
+	/* RAMFC + zero channel's PRAMIN up to start of VM pagedir */
+	ret = nouveau_gpuobj_new_fake(dev, c_ramfc, c_offset + c_ramfc,
+						0x4000, 0, NULL, &chan->ramfc);
+	if (ret)
+		return ret;
+
+	for (i = 0; i < c_vmpd; i += 4)
+		BAR0_WI32(chan->ramin->gpuobj, i, 0);
+
+	/* VM page directory */
+	ret = nouveau_gpuobj_new_fake(dev, c_vmpd, c_offset + c_vmpd,
+					   0x4000, 0, &chan->vm_pd, NULL);
+	if (ret)
+		return ret;
+	for (i = 0; i < 0x4000; i += 8) {
+		BAR0_WI32(chan->vm_pd, i + 0x00, 0x00000000);
+		BAR0_WI32(chan->vm_pd, i + 0x04, 0x00000000);
+	}
+
+	/* PRAMIN page table, cheat and map into VM at 0x0000000000.
+	 * We map the entire fake channel into the start of the PRAMIN BAR
+	 */
+	ret = nouveau_gpuobj_new_ref(dev, chan, NULL, 0, pt_size, 0x1000,
+				     0, &priv->pramin_pt);
+	if (ret)
+		return ret;
+
+	v = c_offset | 1;
+	if (dev_priv->vram_sys_base) {
+		v += dev_priv->vram_sys_base;
+		v |= 0x30;
+	}
+
+	i = 0;
+	while (v < dev_priv->vram_sys_base + c_offset + c_size) {
+		BAR0_WI32(priv->pramin_pt->gpuobj, i + 0, v);
+		BAR0_WI32(priv->pramin_pt->gpuobj, i + 4, 0x00000000);
+		v += 0x1000;
+		i += 8;
+	}
+
+	while (i < pt_size) {
+		BAR0_WI32(priv->pramin_pt->gpuobj, i + 0, 0x00000000);
+		BAR0_WI32(priv->pramin_pt->gpuobj, i + 4, 0x00000000);
+		i += 8;
+	}
+
+	BAR0_WI32(chan->vm_pd, 0x00, priv->pramin_pt->instance | 0x63);
+	BAR0_WI32(chan->vm_pd, 0x04, 0x00000000);
+
+	/* VRAM page table(s), mapped into VM at +1GiB  */
+	for (i = 0; i < dev_priv->vm_vram_pt_nr; i++) {
+		ret = nouveau_gpuobj_new_ref(dev, chan, NULL, 0,
+					     NV50_VM_BLOCK/65536*8, 0, 0,
+					     &chan->vm_vram_pt[i]);
+		if (ret) {
+			NV_ERROR(dev, "Error creating VRAM page tables: %d\n",
+									ret);
+			dev_priv->vm_vram_pt_nr = i;
+			return ret;
+		}
+		dev_priv->vm_vram_pt[i] = chan->vm_vram_pt[i]->gpuobj;
+
+		for (v = 0; v < dev_priv->vm_vram_pt[i]->im_pramin->size;
+								v += 4)
+			BAR0_WI32(dev_priv->vm_vram_pt[i], v, 0);
+
+		BAR0_WI32(chan->vm_pd, 0x10 + (i*8),
+			  chan->vm_vram_pt[i]->instance | 0x61);
+		BAR0_WI32(chan->vm_pd, 0x14 + (i*8), 0);
+	}
+
+	/* DMA object for PRAMIN BAR */
+	ret = nouveau_gpuobj_new_ref(dev, chan, chan, 0, 6*4, 16, 0,
+							&priv->pramin_bar);
+	if (ret)
+		return ret;
+	BAR0_WI32(priv->pramin_bar->gpuobj, 0x00, 0x7fc00000);
+	BAR0_WI32(priv->pramin_bar->gpuobj, 0x04, dev_priv->ramin_size - 1);
+	BAR0_WI32(priv->pramin_bar->gpuobj, 0x08, 0x00000000);
+	BAR0_WI32(priv->pramin_bar->gpuobj, 0x0c, 0x00000000);
+	BAR0_WI32(priv->pramin_bar->gpuobj, 0x10, 0x00000000);
+	BAR0_WI32(priv->pramin_bar->gpuobj, 0x14, 0x00000000);
+
+	/* DMA object for FB BAR */
+	ret = nouveau_gpuobj_new_ref(dev, chan, chan, 0, 6*4, 16, 0,
+							&priv->fb_bar);
+	if (ret)
+		return ret;
+	BAR0_WI32(priv->fb_bar->gpuobj, 0x00, 0x7fc00000);
+	BAR0_WI32(priv->fb_bar->gpuobj, 0x04, 0x40000000 +
+					      drm_get_resource_len(dev, 1) - 1);
+	BAR0_WI32(priv->fb_bar->gpuobj, 0x08, 0x40000000);
+	BAR0_WI32(priv->fb_bar->gpuobj, 0x0c, 0x00000000);
+	BAR0_WI32(priv->fb_bar->gpuobj, 0x10, 0x00000000);
+	BAR0_WI32(priv->fb_bar->gpuobj, 0x14, 0x00000000);
+
+	/* Poke the relevant regs, and pray it works :) */
+	nv_wr32(dev, NV50_PUNK_BAR_CFG_BASE, (chan->ramin->instance >> 12));
+	nv_wr32(dev, NV50_PUNK_UNK1710, 0);
+	nv_wr32(dev, NV50_PUNK_BAR_CFG_BASE, (chan->ramin->instance >> 12) |
+					 NV50_PUNK_BAR_CFG_BASE_VALID);
+	nv_wr32(dev, NV50_PUNK_BAR1_CTXDMA, (priv->fb_bar->instance >> 4) |
+					NV50_PUNK_BAR1_CTXDMA_VALID);
+	nv_wr32(dev, NV50_PUNK_BAR3_CTXDMA, (priv->pramin_bar->instance >> 4) |
+					NV50_PUNK_BAR3_CTXDMA_VALID);
+
+	for (i = 0; i < 8; i++)
+		nv_wr32(dev, 0x1900 + (i*4), 0);
+
+	/* Assume that praying isn't enough, check that we can re-read the
+	 * entire fake channel back from the PRAMIN BAR */
+	dev_priv->engine.instmem.prepare_access(dev, false);
+	for (i = 0; i < c_size; i += 4) {
+		if (nv_rd32(dev, NV_RAMIN + i) != nv_ri32(dev, i)) {
+			NV_ERROR(dev, "Error reading back PRAMIN at 0x%08x\n",
+									i);
+			dev_priv->engine.instmem.finish_access(dev);
+			return -EINVAL;
+		}
+	}
+	dev_priv->engine.instmem.finish_access(dev);
+
+	nv_wr32(dev, NV50_PUNK_BAR0_PRAMIN, save_nv001700);
+
+	/* Global PRAMIN heap */
+	if (nouveau_mem_init_heap(&dev_priv->ramin_heap,
+				  c_size, dev_priv->ramin_size - c_size)) {
+		dev_priv->ramin_heap = NULL;
+		NV_ERROR(dev, "Failed to init RAMIN heap\n");
+	}
+
+	/*XXX: incorrect, but needed to make hash func "work" */
+	dev_priv->ramht_offset = 0x10000;
+	dev_priv->ramht_bits   = 9;
+	dev_priv->ramht_size   = (1 << dev_priv->ramht_bits);
+	return 0;
+}
+
+void
+nv50_instmem_takedown(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nv50_instmem_priv *priv = dev_priv->engine.instmem.priv;
+	struct nouveau_channel *chan = dev_priv->fifos[0];
+	int i;
+
+	NV_DEBUG(dev, "\n");
+
+	if (!priv)
+		return;
+
+	/* Restore state from before init */
+	for (i = 0x1700; i <= 0x1710; i += 4)
+		nv_wr32(dev, i, priv->save1700[(i - 0x1700) / 4]);
+
+	nouveau_gpuobj_ref_del(dev, &priv->fb_bar);
+	nouveau_gpuobj_ref_del(dev, &priv->pramin_bar);
+	nouveau_gpuobj_ref_del(dev, &priv->pramin_pt);
+
+	/* Destroy dummy channel */
+	if (chan) {
+		for (i = 0; i < dev_priv->vm_vram_pt_nr; i++) {
+			nouveau_gpuobj_ref_del(dev, &chan->vm_vram_pt[i]);
+			dev_priv->vm_vram_pt[i] = NULL;
+		}
+		dev_priv->vm_vram_pt_nr = 0;
+
+		nouveau_gpuobj_del(dev, &chan->vm_pd);
+		nouveau_gpuobj_ref_del(dev, &chan->ramfc);
+		nouveau_gpuobj_ref_del(dev, &chan->ramin);
+		nouveau_mem_takedown(&chan->ramin_heap);
+
+		dev_priv->fifos[0] = dev_priv->fifos[127] = NULL;
+		kfree(chan);
+	}
+
+	dev_priv->engine.instmem.priv = NULL;
+	kfree(priv);
+}
+
+int
+nv50_instmem_suspend(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_channel *chan = dev_priv->fifos[0];
+	struct nouveau_gpuobj *ramin = chan->ramin->gpuobj;
+	int i;
+
+	ramin->im_backing_suspend = vmalloc(ramin->im_pramin->size);
+	if (!ramin->im_backing_suspend)
+		return -ENOMEM;
+
+	for (i = 0; i < ramin->im_pramin->size; i += 4)
+		ramin->im_backing_suspend[i/4] = nv_ri32(dev, i);
+	return 0;
+}
+
+void
+nv50_instmem_resume(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nv50_instmem_priv *priv = dev_priv->engine.instmem.priv;
+	struct nouveau_channel *chan = dev_priv->fifos[0];
+	struct nouveau_gpuobj *ramin = chan->ramin->gpuobj;
+	int i;
+
+	nv_wr32(dev, NV50_PUNK_BAR0_PRAMIN, (ramin->im_backing_start >> 16));
+	for (i = 0; i < ramin->im_pramin->size; i += 4)
+		BAR0_WI32(ramin, i, ramin->im_backing_suspend[i/4]);
+	vfree(ramin->im_backing_suspend);
+	ramin->im_backing_suspend = NULL;
+
+	/* Poke the relevant regs, and pray it works :) */
+	nv_wr32(dev, NV50_PUNK_BAR_CFG_BASE, (chan->ramin->instance >> 12));
+	nv_wr32(dev, NV50_PUNK_UNK1710, 0);
+	nv_wr32(dev, NV50_PUNK_BAR_CFG_BASE, (chan->ramin->instance >> 12) |
+					 NV50_PUNK_BAR_CFG_BASE_VALID);
+	nv_wr32(dev, NV50_PUNK_BAR1_CTXDMA, (priv->fb_bar->instance >> 4) |
+					NV50_PUNK_BAR1_CTXDMA_VALID);
+	nv_wr32(dev, NV50_PUNK_BAR3_CTXDMA, (priv->pramin_bar->instance >> 4) |
+					NV50_PUNK_BAR3_CTXDMA_VALID);
+
+	for (i = 0; i < 8; i++)
+		nv_wr32(dev, 0x1900 + (i*4), 0);
+}
+
+int
+nv50_instmem_populate(struct drm_device *dev, struct nouveau_gpuobj *gpuobj,
+		      uint32_t *sz)
+{
+	int ret;
+
+	if (gpuobj->im_backing)
+		return -EINVAL;
+
+	*sz = ALIGN(*sz, NV50_INSTMEM_PAGE_SIZE);
+	if (*sz == 0)
+		return -EINVAL;
+
+	ret = nouveau_bo_new(dev, NULL, *sz, 0, TTM_PL_FLAG_VRAM, 0, 0x0000,
+			     true, false, &gpuobj->im_backing);
+	if (ret) {
+		NV_ERROR(dev, "error getting PRAMIN backing pages: %d\n", ret);
+		return ret;
+	}
+
+	ret = nouveau_bo_pin(gpuobj->im_backing, TTM_PL_FLAG_VRAM);
+	if (ret) {
+		NV_ERROR(dev, "error pinning PRAMIN backing VRAM: %d\n", ret);
+		nouveau_bo_ref(NULL, &gpuobj->im_backing);
+		return ret;
+	}
+
+	gpuobj->im_backing_start = gpuobj->im_backing->bo.mem.mm_node->start;
+	gpuobj->im_backing_start <<= PAGE_SHIFT;
+
+	return 0;
+}
+
+void
+nv50_instmem_clear(struct drm_device *dev, struct nouveau_gpuobj *gpuobj)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	if (gpuobj && gpuobj->im_backing) {
+		if (gpuobj->im_bound)
+			dev_priv->engine.instmem.unbind(dev, gpuobj);
+		nouveau_bo_unpin(gpuobj->im_backing);
+		nouveau_bo_ref(NULL, &gpuobj->im_backing);
+		gpuobj->im_backing = NULL;
+	}
+}
+
+int
+nv50_instmem_bind(struct drm_device *dev, struct nouveau_gpuobj *gpuobj)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nv50_instmem_priv *priv = dev_priv->engine.instmem.priv;
+	struct nouveau_gpuobj *pramin_pt = priv->pramin_pt->gpuobj;
+	uint32_t pte, pte_end;
+	uint64_t vram;
+
+	if (!gpuobj->im_backing || !gpuobj->im_pramin || gpuobj->im_bound)
+		return -EINVAL;
+
+	NV_DEBUG(dev, "st=0x%0llx sz=0x%0llx\n",
+		 gpuobj->im_pramin->start, gpuobj->im_pramin->size);
+
+	pte     = (gpuobj->im_pramin->start >> 12) << 1;
+	pte_end = ((gpuobj->im_pramin->size >> 12) << 1) + pte;
+	vram    = gpuobj->im_backing_start;
+
+	NV_DEBUG(dev, "pramin=0x%llx, pte=%d, pte_end=%d\n",
+		 gpuobj->im_pramin->start, pte, pte_end);
+	NV_DEBUG(dev, "first vram page: 0x%08x\n", gpuobj->im_backing_start);
+
+	vram |= 1;
+	if (dev_priv->vram_sys_base) {
+		vram += dev_priv->vram_sys_base;
+		vram |= 0x30;
+	}
+
+	dev_priv->engine.instmem.prepare_access(dev, true);
+	while (pte < pte_end) {
+		nv_wo32(dev, pramin_pt, pte++, lower_32_bits(vram));
+		nv_wo32(dev, pramin_pt, pte++, upper_32_bits(vram));
+		vram += NV50_INSTMEM_PAGE_SIZE;
+	}
+	dev_priv->engine.instmem.finish_access(dev);
+
+	nv_wr32(dev, 0x100c80, 0x00040001);
+	if (!nv_wait(0x100c80, 0x00000001, 0x00000000)) {
+		NV_ERROR(dev, "timeout: (0x100c80 & 1) == 0 (1)\n");
+		NV_ERROR(dev, "0x100c80 = 0x%08x\n", nv_rd32(dev, 0x100c80));
+		return -EBUSY;
+	}
+
+	nv_wr32(dev, 0x100c80, 0x00060001);
+	if (!nv_wait(0x100c80, 0x00000001, 0x00000000)) {
+		NV_ERROR(dev, "timeout: (0x100c80 & 1) == 0 (2)\n");
+		NV_ERROR(dev, "0x100c80 = 0x%08x\n", nv_rd32(dev, 0x100c80));
+		return -EBUSY;
+	}
+
+	gpuobj->im_bound = 1;
+	return 0;
+}
+
+int
+nv50_instmem_unbind(struct drm_device *dev, struct nouveau_gpuobj *gpuobj)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nv50_instmem_priv *priv = dev_priv->engine.instmem.priv;
+	uint32_t pte, pte_end;
+
+	if (gpuobj->im_bound == 0)
+		return -EINVAL;
+
+	pte     = (gpuobj->im_pramin->start >> 12) << 1;
+	pte_end = ((gpuobj->im_pramin->size >> 12) << 1) + pte;
+
+	dev_priv->engine.instmem.prepare_access(dev, true);
+	while (pte < pte_end) {
+		nv_wo32(dev, priv->pramin_pt->gpuobj, pte++, 0x00000000);
+		nv_wo32(dev, priv->pramin_pt->gpuobj, pte++, 0x00000000);
+	}
+	dev_priv->engine.instmem.finish_access(dev);
+
+	gpuobj->im_bound = 0;
+	return 0;
+}
+
+void
+nv50_instmem_prepare_access(struct drm_device *dev, bool write)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nv50_instmem_priv *priv = dev_priv->engine.instmem.priv;
+
+	priv->last_access_wr = write;
+}
+
+void
+nv50_instmem_finish_access(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nv50_instmem_priv *priv = dev_priv->engine.instmem.priv;
+
+	if (priv->last_access_wr) {
+		nv_wr32(dev, 0x070000, 0x00000001);
+		if (!nv_wait(0x070000, 0x00000001, 0x00000000))
+			NV_ERROR(dev, "PRAMIN flush timeout\n");
+	}
+}
+
diff --git a/drivers/gpu/drm/nouveau/nv50_mc.c b/drivers/gpu/drm/nouveau/nv50_mc.c
new file mode 100644
index 000000000000..e0a9c3faa202
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv50_mc.c
@@ -0,0 +1,40 @@
+/*
+ * Copyright (C) 2007 Ben Skeggs.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "drmP.h"
+#include "drm.h"
+#include "nouveau_drv.h"
+
+int
+nv50_mc_init(struct drm_device *dev)
+{
+	nv_wr32(dev, NV03_PMC_ENABLE, 0xFFFFFFFF);
+	return 0;
+}
+
+void nv50_mc_takedown(struct drm_device *dev)
+{
+}
diff --git a/drivers/gpu/drm/nouveau/nv50_sor.c b/drivers/gpu/drm/nouveau/nv50_sor.c
new file mode 100644
index 000000000000..c2fff543b06f
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv50_sor.c
@@ -0,0 +1,323 @@
+/*
+ * Copyright (C) 2008 Maarten Maathuis.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "drmP.h"
+#include "drm_crtc_helper.h"
+
+#define NOUVEAU_DMA_DEBUG (nouveau_reg_debug & NOUVEAU_REG_DEBUG_EVO)
+#include "nouveau_reg.h"
+#include "nouveau_drv.h"
+#include "nouveau_dma.h"
+#include "nouveau_encoder.h"
+#include "nouveau_connector.h"
+#include "nouveau_crtc.h"
+#include "nv50_display.h"
+
+static void
+nv50_sor_disconnect(struct nouveau_encoder *nv_encoder)
+{
+	struct drm_device *dev = to_drm_encoder(nv_encoder)->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_channel *evo = dev_priv->evo;
+	int ret;
+
+	NV_DEBUG_KMS(dev, "Disconnecting SOR %d\n", nv_encoder->or);
+
+	ret = RING_SPACE(evo, 2);
+	if (ret) {
+		NV_ERROR(dev, "no space while disconnecting SOR\n");
+		return;
+	}
+	BEGIN_RING(evo, 0, NV50_EVO_SOR(nv_encoder->or, MODE_CTRL), 1);
+	OUT_RING(evo, 0);
+}
+
+static void
+nv50_sor_dp_link_train(struct drm_encoder *encoder)
+{
+	struct drm_device *dev = encoder->dev;
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	struct bit_displayport_encoder_table *dpe;
+	int dpe_headerlen;
+
+	dpe = nouveau_bios_dp_table(dev, nv_encoder->dcb, &dpe_headerlen);
+	if (!dpe) {
+		NV_ERROR(dev, "SOR-%d: no DP encoder table!\n", nv_encoder->or);
+		return;
+	}
+
+	if (dpe->script0) {
+		NV_DEBUG_KMS(dev, "SOR-%d: running DP script 0\n", nv_encoder->or);
+		nouveau_bios_run_init_table(dev, le16_to_cpu(dpe->script0),
+					    nv_encoder->dcb);
+	}
+
+	if (!nouveau_dp_link_train(encoder))
+		NV_ERROR(dev, "SOR-%d: link training failed\n", nv_encoder->or);
+
+	if (dpe->script1) {
+		NV_DEBUG_KMS(dev, "SOR-%d: running DP script 1\n", nv_encoder->or);
+		nouveau_bios_run_init_table(dev, le16_to_cpu(dpe->script1),
+					    nv_encoder->dcb);
+	}
+}
+
+static void
+nv50_sor_dpms(struct drm_encoder *encoder, int mode)
+{
+	struct drm_device *dev = encoder->dev;
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	struct drm_encoder *enc;
+	uint32_t val;
+	int or = nv_encoder->or;
+
+	NV_DEBUG_KMS(dev, "or %d mode %d\n", or, mode);
+
+	nv_encoder->last_dpms = mode;
+	list_for_each_entry(enc, &dev->mode_config.encoder_list, head) {
+		struct nouveau_encoder *nvenc = nouveau_encoder(enc);
+
+		if (nvenc == nv_encoder ||
+		    nvenc->disconnect != nv50_sor_disconnect ||
+		    nvenc->dcb->or != nv_encoder->dcb->or)
+			continue;
+
+		if (nvenc->last_dpms == DRM_MODE_DPMS_ON)
+			return;
+	}
+
+	/* wait for it to be done */
+	if (!nv_wait(NV50_PDISPLAY_SOR_DPMS_CTRL(or),
+		     NV50_PDISPLAY_SOR_DPMS_CTRL_PENDING, 0)) {
+		NV_ERROR(dev, "timeout: SOR_DPMS_CTRL_PENDING(%d) == 0\n", or);
+		NV_ERROR(dev, "SOR_DPMS_CTRL(%d) = 0x%08x\n", or,
+			 nv_rd32(dev, NV50_PDISPLAY_SOR_DPMS_CTRL(or)));
+	}
+
+	val = nv_rd32(dev, NV50_PDISPLAY_SOR_DPMS_CTRL(or));
+
+	if (mode == DRM_MODE_DPMS_ON)
+		val |= NV50_PDISPLAY_SOR_DPMS_CTRL_ON;
+	else
+		val &= ~NV50_PDISPLAY_SOR_DPMS_CTRL_ON;
+
+	nv_wr32(dev, NV50_PDISPLAY_SOR_DPMS_CTRL(or), val |
+		NV50_PDISPLAY_SOR_DPMS_CTRL_PENDING);
+	if (!nv_wait(NV50_PDISPLAY_SOR_DPMS_STATE(or),
+		     NV50_PDISPLAY_SOR_DPMS_STATE_WAIT, 0)) {
+		NV_ERROR(dev, "timeout: SOR_DPMS_STATE_WAIT(%d) == 0\n", or);
+		NV_ERROR(dev, "SOR_DPMS_STATE(%d) = 0x%08x\n", or,
+			 nv_rd32(dev, NV50_PDISPLAY_SOR_DPMS_STATE(or)));
+	}
+
+	if (nv_encoder->dcb->type == OUTPUT_DP && mode == DRM_MODE_DPMS_ON)
+		nv50_sor_dp_link_train(encoder);
+}
+
+static void
+nv50_sor_save(struct drm_encoder *encoder)
+{
+	NV_ERROR(encoder->dev, "!!\n");
+}
+
+static void
+nv50_sor_restore(struct drm_encoder *encoder)
+{
+	NV_ERROR(encoder->dev, "!!\n");
+}
+
+static bool
+nv50_sor_mode_fixup(struct drm_encoder *encoder, struct drm_display_mode *mode,
+		    struct drm_display_mode *adjusted_mode)
+{
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	struct nouveau_connector *connector;
+
+	NV_DEBUG_KMS(encoder->dev, "or %d\n", nv_encoder->or);
+
+	connector = nouveau_encoder_connector_get(nv_encoder);
+	if (!connector) {
+		NV_ERROR(encoder->dev, "Encoder has no connector\n");
+		return false;
+	}
+
+	if (connector->scaling_mode != DRM_MODE_SCALE_NONE &&
+	     connector->native_mode) {
+		int id = adjusted_mode->base.id;
+		*adjusted_mode = *connector->native_mode;
+		adjusted_mode->base.id = id;
+	}
+
+	return true;
+}
+
+static void
+nv50_sor_prepare(struct drm_encoder *encoder)
+{
+}
+
+static void
+nv50_sor_commit(struct drm_encoder *encoder)
+{
+}
+
+static void
+nv50_sor_mode_set(struct drm_encoder *encoder, struct drm_display_mode *mode,
+		  struct drm_display_mode *adjusted_mode)
+{
+	struct drm_nouveau_private *dev_priv = encoder->dev->dev_private;
+	struct nouveau_channel *evo = dev_priv->evo;
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+	struct drm_device *dev = encoder->dev;
+	struct nouveau_crtc *crtc = nouveau_crtc(encoder->crtc);
+	uint32_t mode_ctl = 0;
+	int ret;
+
+	NV_DEBUG_KMS(dev, "or %d\n", nv_encoder->or);
+
+	nv50_sor_dpms(encoder, DRM_MODE_DPMS_ON);
+
+	switch (nv_encoder->dcb->type) {
+	case OUTPUT_TMDS:
+		if (nv_encoder->dcb->sorconf.link & 1) {
+			if (adjusted_mode->clock < 165000)
+				mode_ctl = 0x0100;
+			else
+				mode_ctl = 0x0500;
+		} else
+			mode_ctl = 0x0200;
+		break;
+	case OUTPUT_DP:
+		mode_ctl |= 0x00050000;
+		if (nv_encoder->dcb->sorconf.link & 1)
+			mode_ctl |= 0x00000800;
+		else
+			mode_ctl |= 0x00000900;
+		break;
+	default:
+		break;
+	}
+
+	if (crtc->index == 1)
+		mode_ctl |= NV50_EVO_SOR_MODE_CTRL_CRTC1;
+	else
+		mode_ctl |= NV50_EVO_SOR_MODE_CTRL_CRTC0;
+
+	if (adjusted_mode->flags & DRM_MODE_FLAG_NHSYNC)
+		mode_ctl |= NV50_EVO_SOR_MODE_CTRL_NHSYNC;
+
+	if (adjusted_mode->flags & DRM_MODE_FLAG_NVSYNC)
+		mode_ctl |= NV50_EVO_SOR_MODE_CTRL_NVSYNC;
+
+	ret = RING_SPACE(evo, 2);
+	if (ret) {
+		NV_ERROR(dev, "no space while connecting SOR\n");
+		return;
+	}
+	BEGIN_RING(evo, 0, NV50_EVO_SOR(nv_encoder->or, MODE_CTRL), 1);
+	OUT_RING(evo, mode_ctl);
+}
+
+static const struct drm_encoder_helper_funcs nv50_sor_helper_funcs = {
+	.dpms = nv50_sor_dpms,
+	.save = nv50_sor_save,
+	.restore = nv50_sor_restore,
+	.mode_fixup = nv50_sor_mode_fixup,
+	.prepare = nv50_sor_prepare,
+	.commit = nv50_sor_commit,
+	.mode_set = nv50_sor_mode_set,
+	.detect = NULL
+};
+
+static void
+nv50_sor_destroy(struct drm_encoder *encoder)
+{
+	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
+
+	if (!encoder)
+		return;
+
+	NV_DEBUG_KMS(encoder->dev, "\n");
+
+	drm_encoder_cleanup(encoder);
+
+	kfree(nv_encoder);
+}
+
+static const struct drm_encoder_funcs nv50_sor_encoder_funcs = {
+	.destroy = nv50_sor_destroy,
+};
+
+int
+nv50_sor_create(struct drm_device *dev, struct dcb_entry *entry)
+{
+	struct nouveau_encoder *nv_encoder = NULL;
+	struct drm_encoder *encoder;
+	bool dum;
+	int type;
+
+	NV_DEBUG_KMS(dev, "\n");
+
+	switch (entry->type) {
+	case OUTPUT_TMDS:
+		NV_INFO(dev, "Detected a TMDS output\n");
+		type = DRM_MODE_ENCODER_TMDS;
+		break;
+	case OUTPUT_LVDS:
+		NV_INFO(dev, "Detected a LVDS output\n");
+		type = DRM_MODE_ENCODER_LVDS;
+
+		if (nouveau_bios_parse_lvds_table(dev, 0, &dum, &dum)) {
+			NV_ERROR(dev, "Failed parsing LVDS table\n");
+			return -EINVAL;
+		}
+		break;
+	case OUTPUT_DP:
+		NV_INFO(dev, "Detected a DP output\n");
+		type = DRM_MODE_ENCODER_TMDS;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	nv_encoder = kzalloc(sizeof(*nv_encoder), GFP_KERNEL);
+	if (!nv_encoder)
+		return -ENOMEM;
+	encoder = to_drm_encoder(nv_encoder);
+
+	nv_encoder->dcb = entry;
+	nv_encoder->or = ffs(entry->or) - 1;
+
+	nv_encoder->disconnect = nv50_sor_disconnect;
+
+	drm_encoder_init(dev, encoder, &nv50_sor_encoder_funcs, type);
+	drm_encoder_helper_add(encoder, &nv50_sor_helper_funcs);
+
+	encoder->possible_crtcs = entry->heads;
+	encoder->possible_clones = 0;
+
+	return 0;
+}
diff --git a/drivers/gpu/drm/nouveau/nvreg.h b/drivers/gpu/drm/nouveau/nvreg.h
new file mode 100644
index 000000000000..5998c35237b0
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nvreg.h
@@ -0,0 +1,535 @@
+/* $XConsortium: nvreg.h /main/2 1996/10/28 05:13:41 kaleb $ */
+/*
+ * Copyright 1996-1997  David J. McKay
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * DAVID J. MCKAY BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF
+ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/nv/nvreg.h,v 1.6 2002/01/25 21:56:06 tsi Exp $ */
+
+#ifndef __NVREG_H_
+#define __NVREG_H_
+
+#define NV_PMC_OFFSET               0x00000000
+#define NV_PMC_SIZE                 0x00001000
+
+#define NV_PBUS_OFFSET              0x00001000
+#define NV_PBUS_SIZE                0x00001000
+
+#define NV_PFIFO_OFFSET             0x00002000
+#define NV_PFIFO_SIZE               0x00002000
+
+#define NV_HDIAG_OFFSET             0x00005000
+#define NV_HDIAG_SIZE               0x00001000
+
+#define NV_PRAM_OFFSET              0x00006000
+#define NV_PRAM_SIZE                0x00001000
+
+#define NV_PVIDEO_OFFSET            0x00008000
+#define NV_PVIDEO_SIZE              0x00001000
+
+#define NV_PTIMER_OFFSET            0x00009000
+#define NV_PTIMER_SIZE              0x00001000
+
+#define NV_PPM_OFFSET               0x0000A000
+#define NV_PPM_SIZE                 0x00001000
+
+#define NV_PTV_OFFSET               0x0000D000
+#define NV_PTV_SIZE                 0x00001000
+
+#define NV_PRMVGA_OFFSET            0x000A0000
+#define NV_PRMVGA_SIZE              0x00020000
+
+#define NV_PRMVIO0_OFFSET           0x000C0000
+#define NV_PRMVIO_SIZE              0x00002000
+#define NV_PRMVIO1_OFFSET           0x000C2000
+
+#define NV_PFB_OFFSET               0x00100000
+#define NV_PFB_SIZE                 0x00001000
+
+#define NV_PEXTDEV_OFFSET           0x00101000
+#define NV_PEXTDEV_SIZE             0x00001000
+
+#define NV_PME_OFFSET               0x00200000
+#define NV_PME_SIZE                 0x00001000
+
+#define NV_PROM_OFFSET              0x00300000
+#define NV_PROM_SIZE                0x00010000
+
+#define NV_PGRAPH_OFFSET            0x00400000
+#define NV_PGRAPH_SIZE              0x00010000
+
+#define NV_PCRTC0_OFFSET            0x00600000
+#define NV_PCRTC0_SIZE              0x00002000 /* empirical */
+
+#define NV_PRMCIO0_OFFSET           0x00601000
+#define NV_PRMCIO_SIZE              0x00002000
+#define NV_PRMCIO1_OFFSET           0x00603000
+
+#define NV50_DISPLAY_OFFSET           0x00610000
+#define NV50_DISPLAY_SIZE             0x0000FFFF
+
+#define NV_PRAMDAC0_OFFSET          0x00680000
+#define NV_PRAMDAC0_SIZE            0x00002000
+
+#define NV_PRMDIO0_OFFSET           0x00681000
+#define NV_PRMDIO_SIZE              0x00002000
+#define NV_PRMDIO1_OFFSET           0x00683000
+
+#define NV_PRAMIN_OFFSET            0x00700000
+#define NV_PRAMIN_SIZE              0x00100000
+
+#define NV_FIFO_OFFSET              0x00800000
+#define NV_FIFO_SIZE                0x00800000
+
+#define NV_PMC_BOOT_0			0x00000000
+#define NV_PMC_ENABLE			0x00000200
+
+#define NV_VIO_VSE2			0x000003c3
+#define NV_VIO_SRX			0x000003c4
+
+#define NV_CIO_CRX__COLOR		0x000003d4
+#define NV_CIO_CR__COLOR		0x000003d5
+
+#define NV_PBUS_DEBUG_1			0x00001084
+#define NV_PBUS_DEBUG_4			0x00001098
+#define NV_PBUS_DEBUG_DUALHEAD_CTL	0x000010f0
+#define NV_PBUS_POWERCTRL_1		0x00001584
+#define NV_PBUS_POWERCTRL_2		0x00001588
+#define NV_PBUS_POWERCTRL_4		0x00001590
+#define NV_PBUS_PCI_NV_19		0x0000184C
+#define NV_PBUS_PCI_NV_20		0x00001850
+#	define NV_PBUS_PCI_NV_20_ROM_SHADOW_DISABLED	(0 << 0)
+#	define NV_PBUS_PCI_NV_20_ROM_SHADOW_ENABLED	(1 << 0)
+
+#define NV_PFIFO_RAMHT			0x00002210
+
+#define NV_PTV_TV_INDEX			0x0000d220
+#define NV_PTV_TV_DATA			0x0000d224
+#define NV_PTV_HFILTER			0x0000d310
+#define NV_PTV_HFILTER2			0x0000d390
+#define NV_PTV_VFILTER			0x0000d510
+
+#define NV_PRMVIO_MISC__WRITE		0x000c03c2
+#define NV_PRMVIO_SRX			0x000c03c4
+#define NV_PRMVIO_SR			0x000c03c5
+#	define NV_VIO_SR_RESET_INDEX		0x00
+#	define NV_VIO_SR_CLOCK_INDEX		0x01
+#	define NV_VIO_SR_PLANE_MASK_INDEX	0x02
+#	define NV_VIO_SR_CHAR_MAP_INDEX		0x03
+#	define NV_VIO_SR_MEM_MODE_INDEX		0x04
+#define NV_PRMVIO_MISC__READ		0x000c03cc
+#define NV_PRMVIO_GRX			0x000c03ce
+#define NV_PRMVIO_GX			0x000c03cf
+#	define NV_VIO_GX_SR_INDEX		0x00
+#	define NV_VIO_GX_SREN_INDEX		0x01
+#	define NV_VIO_GX_CCOMP_INDEX		0x02
+#	define NV_VIO_GX_ROP_INDEX		0x03
+#	define NV_VIO_GX_READ_MAP_INDEX		0x04
+#	define NV_VIO_GX_MODE_INDEX		0x05
+#	define NV_VIO_GX_MISC_INDEX		0x06
+#	define NV_VIO_GX_DONT_CARE_INDEX	0x07
+#	define NV_VIO_GX_BIT_MASK_INDEX		0x08
+
+#define NV_PFB_BOOT_0			0x00100000
+#define NV_PFB_CFG0			0x00100200
+#define NV_PFB_CFG1			0x00100204
+#define NV_PFB_CSTATUS			0x0010020C
+#define NV_PFB_REFCTRL			0x00100210
+#	define NV_PFB_REFCTRL_VALID_1			(1 << 31)
+#define NV_PFB_PAD			0x0010021C
+#	define NV_PFB_PAD_CKE_NORMAL			(1 << 0)
+#define NV_PFB_TILE_NV10		0x00100240
+#define NV_PFB_TILE_SIZE_NV10		0x00100244
+#define NV_PFB_REF			0x001002D0
+#	define NV_PFB_REF_CMD_REFRESH			(1 << 0)
+#define NV_PFB_PRE			0x001002D4
+#	define NV_PFB_PRE_CMD_PRECHARGE			(1 << 0)
+#define NV_PFB_CLOSE_PAGE2		0x0010033C
+#define NV_PFB_TILE_NV40		0x00100600
+#define NV_PFB_TILE_SIZE_NV40		0x00100604
+
+#define NV_PEXTDEV_BOOT_0		0x00101000
+#	define NV_PEXTDEV_BOOT_0_STRAP_FP_IFACE_12BIT	(8 << 12)
+#define NV_PEXTDEV_BOOT_3		0x0010100c
+
+#define NV_PCRTC_INTR_0					0x00600100
+#	define NV_PCRTC_INTR_0_VBLANK				(1 << 0)
+#define NV_PCRTC_INTR_EN_0				0x00600140
+#define NV_PCRTC_START					0x00600800
+#define NV_PCRTC_CONFIG					0x00600804
+#	define NV_PCRTC_CONFIG_START_ADDRESS_NON_VGA		(1 << 0)
+#	define NV_PCRTC_CONFIG_START_ADDRESS_HSYNC		(2 << 0)
+#define NV_PCRTC_CURSOR_CONFIG				0x00600810
+#	define NV_PCRTC_CURSOR_CONFIG_ENABLE_ENABLE		(1 << 0)
+#	define NV_PCRTC_CURSOR_CONFIG_DOUBLE_SCAN_ENABLE	(1 << 4)
+#	define NV_PCRTC_CURSOR_CONFIG_ADDRESS_SPACE_PNVM	(1 << 8)
+#	define NV_PCRTC_CURSOR_CONFIG_CUR_BPP_32		(1 << 12)
+#	define NV_PCRTC_CURSOR_CONFIG_CUR_PIXELS_64		(1 << 16)
+#	define NV_PCRTC_CURSOR_CONFIG_CUR_LINES_32		(2 << 24)
+#	define NV_PCRTC_CURSOR_CONFIG_CUR_LINES_64		(4 << 24)
+#	define NV_PCRTC_CURSOR_CONFIG_CUR_BLEND_ALPHA		(1 << 28)
+
+/* note: PCRTC_GPIO is not available on nv10, and in fact aliases 0x600810 */
+#define NV_PCRTC_GPIO					0x00600818
+#define NV_PCRTC_GPIO_EXT				0x0060081c
+#define NV_PCRTC_830					0x00600830
+#define NV_PCRTC_834					0x00600834
+#define NV_PCRTC_850					0x00600850
+#define NV_PCRTC_ENGINE_CTRL				0x00600860
+#	define NV_CRTC_FSEL_I2C					(1 << 4)
+#	define NV_CRTC_FSEL_OVERLAY				(1 << 12)
+
+#define NV_PRMCIO_ARX			0x006013c0
+#define NV_PRMCIO_AR__WRITE		0x006013c0
+#define NV_PRMCIO_AR__READ		0x006013c1
+#	define NV_CIO_AR_MODE_INDEX		0x10
+#	define NV_CIO_AR_OSCAN_INDEX		0x11
+#	define NV_CIO_AR_PLANE_INDEX		0x12
+#	define NV_CIO_AR_HPP_INDEX		0x13
+#	define NV_CIO_AR_CSEL_INDEX		0x14
+#define NV_PRMCIO_INP0			0x006013c2
+#define NV_PRMCIO_CRX__COLOR		0x006013d4
+#define NV_PRMCIO_CR__COLOR		0x006013d5
+	/* Standard VGA CRTC registers */
+#	define NV_CIO_CR_HDT_INDEX		0x00	/* horizontal display total */
+#	define NV_CIO_CR_HDE_INDEX		0x01	/* horizontal display end */
+#	define NV_CIO_CR_HBS_INDEX		0x02	/* horizontal blanking start */
+#	define NV_CIO_CR_HBE_INDEX		0x03	/* horizontal blanking end */
+#		define NV_CIO_CR_HBE_4_0		4:0
+#	define NV_CIO_CR_HRS_INDEX		0x04	/* horizontal retrace start */
+#	define NV_CIO_CR_HRE_INDEX		0x05	/* horizontal retrace end */
+#		define NV_CIO_CR_HRE_4_0		4:0
+#		define NV_CIO_CR_HRE_HBE_5		7:7
+#	define NV_CIO_CR_VDT_INDEX		0x06	/* vertical display total */
+#	define NV_CIO_CR_OVL_INDEX		0x07	/* overflow bits */
+#		define NV_CIO_CR_OVL_VDT_8		0:0
+#		define NV_CIO_CR_OVL_VDE_8		1:1
+#		define NV_CIO_CR_OVL_VRS_8		2:2
+#		define NV_CIO_CR_OVL_VBS_8		3:3
+#		define NV_CIO_CR_OVL_VDT_9		5:5
+#		define NV_CIO_CR_OVL_VDE_9		6:6
+#		define NV_CIO_CR_OVL_VRS_9		7:7
+#	define NV_CIO_CR_RSAL_INDEX		0x08	/* normally "preset row scan" */
+#	define NV_CIO_CR_CELL_HT_INDEX		0x09	/* cell height?! normally "max scan line" */
+#		define NV_CIO_CR_CELL_HT_VBS_9		5:5
+#		define NV_CIO_CR_CELL_HT_SCANDBL	7:7
+#	define NV_CIO_CR_CURS_ST_INDEX		0x0a	/* cursor start */
+#	define NV_CIO_CR_CURS_END_INDEX		0x0b	/* cursor end */
+#	define NV_CIO_CR_SA_HI_INDEX		0x0c	/* screen start address high */
+#	define NV_CIO_CR_SA_LO_INDEX		0x0d	/* screen start address low */
+#	define NV_CIO_CR_TCOFF_HI_INDEX		0x0e	/* cursor offset high */
+#	define NV_CIO_CR_TCOFF_LO_INDEX		0x0f	/* cursor offset low */
+#	define NV_CIO_CR_VRS_INDEX		0x10	/* vertical retrace start */
+#	define NV_CIO_CR_VRE_INDEX		0x11	/* vertical retrace end */
+#		define NV_CIO_CR_VRE_3_0		3:0
+#	define NV_CIO_CR_VDE_INDEX		0x12	/* vertical display end */
+#	define NV_CIO_CR_OFFSET_INDEX		0x13	/* sets screen pitch */
+#	define NV_CIO_CR_ULINE_INDEX		0x14	/* underline location */
+#	define NV_CIO_CR_VBS_INDEX		0x15	/* vertical blank start */
+#	define NV_CIO_CR_VBE_INDEX		0x16	/* vertical blank end */
+#	define NV_CIO_CR_MODE_INDEX		0x17	/* crtc mode control */
+#	define NV_CIO_CR_LCOMP_INDEX		0x18	/* line compare */
+	/* Extended VGA CRTC registers */
+#	define NV_CIO_CRE_RPC0_INDEX		0x19	/* repaint control 0 */
+#		define NV_CIO_CRE_RPC0_OFFSET_10_8	7:5
+#	define NV_CIO_CRE_RPC1_INDEX		0x1a	/* repaint control 1 */
+#		define NV_CIO_CRE_RPC1_LARGE		2:2
+#	define NV_CIO_CRE_FF_INDEX		0x1b	/* fifo control */
+#	define NV_CIO_CRE_ENH_INDEX		0x1c	/* enhanced? */
+#	define NV_CIO_SR_LOCK_INDEX		0x1f	/* crtc lock */
+#		define NV_CIO_SR_UNLOCK_RW_VALUE	0x57
+#		define NV_CIO_SR_LOCK_VALUE		0x99
+#	define NV_CIO_CRE_FFLWM__INDEX		0x20	/* fifo low water mark */
+#	define NV_CIO_CRE_21			0x21	/* vga shadow crtc lock */
+#	define NV_CIO_CRE_LSR_INDEX		0x25	/* ? */
+#		define NV_CIO_CRE_LSR_VDT_10		0:0
+#		define NV_CIO_CRE_LSR_VDE_10		1:1
+#		define NV_CIO_CRE_LSR_VRS_10		2:2
+#		define NV_CIO_CRE_LSR_VBS_10		3:3
+#		define NV_CIO_CRE_LSR_HBE_6		4:4
+#	define NV_CIO_CR_ARX_INDEX		0x26	/* attribute index -- ro copy of 0x60.3c0 */
+#	define NV_CIO_CRE_CHIP_ID_INDEX		0x27	/* chip revision */
+#	define NV_CIO_CRE_PIXEL_INDEX		0x28
+#		define NV_CIO_CRE_PIXEL_FORMAT		1:0
+#	define NV_CIO_CRE_HEB__INDEX		0x2d	/* horizontal extra bits? */
+#		define NV_CIO_CRE_HEB_HDT_8		0:0
+#		define NV_CIO_CRE_HEB_HDE_8		1:1
+#		define NV_CIO_CRE_HEB_HBS_8		2:2
+#		define NV_CIO_CRE_HEB_HRS_8		3:3
+#		define NV_CIO_CRE_HEB_ILC_8		4:4
+#	define NV_CIO_CRE_2E			0x2e	/* some scratch or dummy reg to force writes to sink in */
+#	define NV_CIO_CRE_HCUR_ADDR2_INDEX	0x2f	/* cursor */
+#	define NV_CIO_CRE_HCUR_ADDR0_INDEX	0x30		/* pixmap */
+#		define NV_CIO_CRE_HCUR_ADDR0_ADR	6:0
+#		define NV_CIO_CRE_HCUR_ASI		7:7
+#	define NV_CIO_CRE_HCUR_ADDR1_INDEX	0x31			/* address */
+#		define NV_CIO_CRE_HCUR_ADDR1_ENABLE	0:0
+#		define NV_CIO_CRE_HCUR_ADDR1_CUR_DBL	1:1
+#		define NV_CIO_CRE_HCUR_ADDR1_ADR	7:2
+#	define NV_CIO_CRE_LCD__INDEX		0x33
+#		define NV_CIO_CRE_LCD_LCD_SELECT	0:0
+#	define NV_CIO_CRE_DDC0_STATUS__INDEX	0x36
+#	define NV_CIO_CRE_DDC0_WR__INDEX	0x37
+#	define NV_CIO_CRE_ILACE__INDEX		0x39	/* interlace */
+#	define NV_CIO_CRE_SCRATCH3__INDEX	0x3b
+#	define NV_CIO_CRE_SCRATCH4__INDEX	0x3c
+#	define NV_CIO_CRE_DDC_STATUS__INDEX	0x3e
+#	define NV_CIO_CRE_DDC_WR__INDEX		0x3f
+#	define NV_CIO_CRE_EBR_INDEX		0x41	/* extra bits ? (vertical) */
+#		define NV_CIO_CRE_EBR_VDT_11		0:0
+#		define NV_CIO_CRE_EBR_VDE_11		2:2
+#		define NV_CIO_CRE_EBR_VRS_11		4:4
+#		define NV_CIO_CRE_EBR_VBS_11		6:6
+#	define NV_CIO_CRE_43			0x43
+#	define NV_CIO_CRE_44			0x44	/* head control */
+#	define NV_CIO_CRE_CSB			0x45	/* colour saturation boost */
+#	define NV_CIO_CRE_RCR			0x46
+#		define NV_CIO_CRE_RCR_ENDIAN_BIG	7:7
+#	define NV_CIO_CRE_47			0x47	/* extended fifo lwm, used on nv30+ */
+#	define NV_CIO_CRE_49			0x49
+#	define NV_CIO_CRE_4B			0x4b	/* given patterns in 0x[2-3][a-c] regs, probably scratch 6 */
+#	define NV_CIO_CRE_TVOUT_LATENCY		0x52
+#	define NV_CIO_CRE_53			0x53	/* `fp_htiming' according to Haiku */
+#	define NV_CIO_CRE_54			0x54	/* `fp_vtiming' according to Haiku */
+#	define NV_CIO_CRE_57			0x57	/* index reg for cr58 */
+#	define NV_CIO_CRE_58			0x58	/* data reg for cr57 */
+#	define NV_CIO_CRE_59			0x59	/* related to on/off-chip-ness of digital outputs */
+#	define NV_CIO_CRE_5B			0x5B	/* newer colour saturation reg */
+#	define NV_CIO_CRE_85			0x85
+#	define NV_CIO_CRE_86			0x86
+#define NV_PRMCIO_INP0__COLOR		0x006013da
+
+#define NV_PRAMDAC_CU_START_POS				0x00680300
+#	define NV_PRAMDAC_CU_START_POS_X			15:0
+#	define NV_PRAMDAC_CU_START_POS_Y			31:16
+#define NV_RAMDAC_NV10_CURSYNC				0x00680404
+
+#define NV_PRAMDAC_NVPLL_COEFF				0x00680500
+#define NV_PRAMDAC_MPLL_COEFF				0x00680504
+#define NV_PRAMDAC_VPLL_COEFF				0x00680508
+#	define NV30_RAMDAC_ENABLE_VCO2				(8 << 4)
+
+#define NV_PRAMDAC_PLL_COEFF_SELECT			0x0068050c
+#	define NV_PRAMDAC_PLL_COEFF_SELECT_USE_VPLL2_TRUE	(4 << 0)
+#	define NV_PRAMDAC_PLL_COEFF_SELECT_SOURCE_PROG_MPLL	(1 << 8)
+#	define NV_PRAMDAC_PLL_COEFF_SELECT_SOURCE_PROG_VPLL	(2 << 8)
+#	define NV_PRAMDAC_PLL_COEFF_SELECT_SOURCE_PROG_NVPLL	(4 << 8)
+#	define NV_PRAMDAC_PLL_COEFF_SELECT_PLL_SOURCE_VPLL2	(8 << 8)
+#	define NV_PRAMDAC_PLL_COEFF_SELECT_TV_VSCLK1		(1 << 16)
+#	define NV_PRAMDAC_PLL_COEFF_SELECT_TV_PCLK1		(2 << 16)
+#	define NV_PRAMDAC_PLL_COEFF_SELECT_TV_VSCLK2		(4 << 16)
+#	define NV_PRAMDAC_PLL_COEFF_SELECT_TV_PCLK2		(8 << 16)
+#	define NV_PRAMDAC_PLL_COEFF_SELECT_TV_CLK_SOURCE_VIP	(1 << 20)
+#	define NV_PRAMDAC_PLL_COEFF_SELECT_VCLK_RATIO_DB2	(1 << 28)
+#	define NV_PRAMDAC_PLL_COEFF_SELECT_VCLK2_RATIO_DB2	(2 << 28)
+
+#define NV_PRAMDAC_PLL_SETUP_CONTROL			0x00680510
+#define NV_RAMDAC_VPLL2					0x00680520
+#define NV_PRAMDAC_SEL_CLK				0x00680524
+#define NV_RAMDAC_DITHER_NV11				0x00680528
+#define NV_PRAMDAC_DACCLK				0x0068052c
+#	define NV_PRAMDAC_DACCLK_SEL_DACCLK			(1 << 0)
+
+#define NV_RAMDAC_NVPLL_B				0x00680570
+#define NV_RAMDAC_MPLL_B				0x00680574
+#define NV_RAMDAC_VPLL_B				0x00680578
+#define NV_RAMDAC_VPLL2_B				0x0068057c
+#	define NV31_RAMDAC_ENABLE_VCO2				(8 << 28)
+#define NV_PRAMDAC_580					0x00680580
+#	define NV_RAMDAC_580_VPLL1_ACTIVE			(1 << 8)
+#	define NV_RAMDAC_580_VPLL2_ACTIVE			(1 << 28)
+
+#define NV_PRAMDAC_GENERAL_CONTROL			0x00680600
+#	define NV_PRAMDAC_GENERAL_CONTROL_PIXMIX_ON		(3 << 4)
+#	define NV_PRAMDAC_GENERAL_CONTROL_VGA_STATE_SEL		(1 << 8)
+#	define NV_PRAMDAC_GENERAL_CONTROL_ALT_MODE_SEL		(1 << 12)
+#	define NV_PRAMDAC_GENERAL_CONTROL_TERMINATION_75OHM	(2 << 16)
+#	define NV_PRAMDAC_GENERAL_CONTROL_BPC_8BITS		(1 << 20)
+#	define NV_PRAMDAC_GENERAL_CONTROL_PIPE_LONG		(2 << 28)
+#define NV_PRAMDAC_TEST_CONTROL				0x00680608
+#	define NV_PRAMDAC_TEST_CONTROL_TP_INS_EN_ASSERTED	(1 << 12)
+#	define NV_PRAMDAC_TEST_CONTROL_PWRDWN_DAC_OFF		(1 << 16)
+#	define NV_PRAMDAC_TEST_CONTROL_SENSEB_ALLHI		(1 << 28)
+#define NV_PRAMDAC_TESTPOINT_DATA			0x00680610
+#	define NV_PRAMDAC_TESTPOINT_DATA_NOTBLANK		(8 << 28)
+#define NV_PRAMDAC_630					0x00680630
+#define NV_PRAMDAC_634					0x00680634
+
+#define NV_PRAMDAC_TV_SETUP				0x00680700
+#define NV_PRAMDAC_TV_VTOTAL				0x00680720
+#define NV_PRAMDAC_TV_VSKEW				0x00680724
+#define NV_PRAMDAC_TV_VSYNC_DELAY			0x00680728
+#define NV_PRAMDAC_TV_HTOTAL				0x0068072c
+#define NV_PRAMDAC_TV_HSKEW				0x00680730
+#define NV_PRAMDAC_TV_HSYNC_DELAY			0x00680734
+#define NV_PRAMDAC_TV_HSYNC_DELAY2			0x00680738
+
+#define NV_PRAMDAC_TV_SETUP                             0x00680700
+
+#define NV_PRAMDAC_FP_VDISPLAY_END			0x00680800
+#define NV_PRAMDAC_FP_VTOTAL				0x00680804
+#define NV_PRAMDAC_FP_VCRTC				0x00680808
+#define NV_PRAMDAC_FP_VSYNC_START			0x0068080c
+#define NV_PRAMDAC_FP_VSYNC_END				0x00680810
+#define NV_PRAMDAC_FP_VVALID_START			0x00680814
+#define NV_PRAMDAC_FP_VVALID_END			0x00680818
+#define NV_PRAMDAC_FP_HDISPLAY_END			0x00680820
+#define NV_PRAMDAC_FP_HTOTAL				0x00680824
+#define NV_PRAMDAC_FP_HCRTC				0x00680828
+#define NV_PRAMDAC_FP_HSYNC_START			0x0068082c
+#define NV_PRAMDAC_FP_HSYNC_END				0x00680830
+#define NV_PRAMDAC_FP_HVALID_START			0x00680834
+#define NV_PRAMDAC_FP_HVALID_END			0x00680838
+
+#define NV_RAMDAC_FP_DITHER				0x0068083c
+#define NV_PRAMDAC_FP_TG_CONTROL			0x00680848
+#	define NV_PRAMDAC_FP_TG_CONTROL_VSYNC_POS		(1 << 0)
+#	define NV_PRAMDAC_FP_TG_CONTROL_VSYNC_DISABLE		(2 << 0)
+#	define NV_PRAMDAC_FP_TG_CONTROL_HSYNC_POS		(1 << 4)
+#	define NV_PRAMDAC_FP_TG_CONTROL_HSYNC_DISABLE		(2 << 4)
+#	define NV_PRAMDAC_FP_TG_CONTROL_MODE_SCALE		(0 << 8)
+#	define NV_PRAMDAC_FP_TG_CONTROL_MODE_CENTER		(1 << 8)
+#	define NV_PRAMDAC_FP_TG_CONTROL_MODE_NATIVE		(2 << 8)
+#	define NV_PRAMDAC_FP_TG_CONTROL_READ_PROG		(1 << 20)
+#	define NV_PRAMDAC_FP_TG_CONTROL_WIDTH_12		(1 << 24)
+#	define NV_PRAMDAC_FP_TG_CONTROL_DISPEN_POS		(1 << 28)
+#	define NV_PRAMDAC_FP_TG_CONTROL_DISPEN_DISABLE		(2 << 28)
+#define NV_PRAMDAC_FP_MARGIN_COLOR			0x0068084c
+#define NV_PRAMDAC_850					0x00680850
+#define NV_PRAMDAC_85C					0x0068085c
+#define NV_PRAMDAC_FP_DEBUG_0				0x00680880
+#	define NV_PRAMDAC_FP_DEBUG_0_XSCALE_ENABLE		(1 << 0)
+#	define NV_PRAMDAC_FP_DEBUG_0_YSCALE_ENABLE		(1 << 4)
+/* This doesn't seem to be essential for tmds, but still often set */
+#	define NV_RAMDAC_FP_DEBUG_0_TMDS_ENABLED		(8 << 4)
+#	define NV_PRAMDAC_FP_DEBUG_0_XINTERP_BILINEAR		(1 << 8)
+#	define NV_PRAMDAC_FP_DEBUG_0_YINTERP_BILINEAR		(1 << 12)
+#	define NV_PRAMDAC_FP_DEBUG_0_XWEIGHT_ROUND		(1 << 20)
+#	define NV_PRAMDAC_FP_DEBUG_0_YWEIGHT_ROUND		(1 << 24)
+#       define NV_PRAMDAC_FP_DEBUG_0_PWRDOWN_FPCLK              (1 << 28)
+#define NV_PRAMDAC_FP_DEBUG_1				0x00680884
+#	define NV_PRAMDAC_FP_DEBUG_1_XSCALE_VALUE		11:0
+#	define NV_PRAMDAC_FP_DEBUG_1_XSCALE_TESTMODE_ENABLE	(1 << 12)
+#	define NV_PRAMDAC_FP_DEBUG_1_YSCALE_VALUE		27:16
+#	define NV_PRAMDAC_FP_DEBUG_1_YSCALE_TESTMODE_ENABLE	(1 << 28)
+#define NV_PRAMDAC_FP_DEBUG_2				0x00680888
+#define NV_PRAMDAC_FP_DEBUG_3				0x0068088C
+
+/* see NV_PRAMDAC_INDIR_TMDS in rules.xml */
+#define NV_PRAMDAC_FP_TMDS_CONTROL			0x006808b0
+#	define NV_PRAMDAC_FP_TMDS_CONTROL_WRITE_DISABLE		(1 << 16)
+#define NV_PRAMDAC_FP_TMDS_DATA				0x006808b4
+
+#define NV_PRAMDAC_8C0                                  0x006808c0
+
+/* Some kind of switch */
+#define NV_PRAMDAC_900					0x00680900
+#define NV_PRAMDAC_A20					0x00680A20
+#define NV_PRAMDAC_A24					0x00680A24
+#define NV_PRAMDAC_A34					0x00680A34
+
+#define NV_PRAMDAC_CTV					0x00680c00
+
+/* names fabricated from NV_USER_DAC info */
+#define NV_PRMDIO_PIXEL_MASK		0x006813c6
+#	define NV_PRMDIO_PIXEL_MASK_MASK	0xff
+#define NV_PRMDIO_READ_MODE_ADDRESS	0x006813c7
+#define NV_PRMDIO_WRITE_MODE_ADDRESS	0x006813c8
+#define NV_PRMDIO_PALETTE_DATA		0x006813c9
+
+#define NV_PGRAPH_DEBUG_0		0x00400080
+#define NV_PGRAPH_DEBUG_1		0x00400084
+#define NV_PGRAPH_DEBUG_2_NV04		0x00400088
+#define NV_PGRAPH_DEBUG_2		0x00400620
+#define NV_PGRAPH_DEBUG_3		0x0040008c
+#define NV_PGRAPH_DEBUG_4		0x00400090
+#define NV_PGRAPH_INTR			0x00400100
+#define NV_PGRAPH_INTR_EN		0x00400140
+#define NV_PGRAPH_CTX_CONTROL		0x00400144
+#define NV_PGRAPH_CTX_CONTROL_NV04	0x00400170
+#define NV_PGRAPH_ABS_UCLIP_XMIN	0x0040053C
+#define NV_PGRAPH_ABS_UCLIP_YMIN	0x00400540
+#define NV_PGRAPH_ABS_UCLIP_XMAX	0x00400544
+#define NV_PGRAPH_ABS_UCLIP_YMAX	0x00400548
+#define NV_PGRAPH_BETA_AND		0x00400608
+#define NV_PGRAPH_LIMIT_VIOL_PIX	0x00400610
+#define NV_PGRAPH_BOFFSET0		0x00400640
+#define NV_PGRAPH_BOFFSET1		0x00400644
+#define NV_PGRAPH_BOFFSET2		0x00400648
+#define NV_PGRAPH_BLIMIT0		0x00400684
+#define NV_PGRAPH_BLIMIT1		0x00400688
+#define NV_PGRAPH_BLIMIT2		0x0040068c
+#define NV_PGRAPH_STATUS		0x00400700
+#define NV_PGRAPH_SURFACE		0x00400710
+#define NV_PGRAPH_STATE			0x00400714
+#define NV_PGRAPH_FIFO			0x00400720
+#define NV_PGRAPH_PATTERN_SHAPE		0x00400810
+#define NV_PGRAPH_TILE			0x00400b00
+
+#define NV_PVIDEO_INTR_EN		0x00008140
+#define NV_PVIDEO_BUFFER		0x00008700
+#define NV_PVIDEO_STOP			0x00008704
+#define NV_PVIDEO_UVPLANE_BASE(buff)	(0x00008800+(buff)*4)
+#define NV_PVIDEO_UVPLANE_LIMIT(buff)	(0x00008808+(buff)*4)
+#define NV_PVIDEO_UVPLANE_OFFSET_BUFF(buff)	(0x00008820+(buff)*4)
+#define NV_PVIDEO_BASE(buff)		(0x00008900+(buff)*4)
+#define NV_PVIDEO_LIMIT(buff)		(0x00008908+(buff)*4)
+#define NV_PVIDEO_LUMINANCE(buff)	(0x00008910+(buff)*4)
+#define NV_PVIDEO_CHROMINANCE(buff)	(0x00008918+(buff)*4)
+#define NV_PVIDEO_OFFSET_BUFF(buff)	(0x00008920+(buff)*4)
+#define NV_PVIDEO_SIZE_IN(buff)		(0x00008928+(buff)*4)
+#define NV_PVIDEO_POINT_IN(buff)	(0x00008930+(buff)*4)
+#define NV_PVIDEO_DS_DX(buff)		(0x00008938+(buff)*4)
+#define NV_PVIDEO_DT_DY(buff)		(0x00008940+(buff)*4)
+#define NV_PVIDEO_POINT_OUT(buff)	(0x00008948+(buff)*4)
+#define NV_PVIDEO_SIZE_OUT(buff)	(0x00008950+(buff)*4)
+#define NV_PVIDEO_FORMAT(buff)		(0x00008958+(buff)*4)
+#	define NV_PVIDEO_FORMAT_PLANAR			(1 << 0)
+#	define NV_PVIDEO_FORMAT_COLOR_LE_CR8YB8CB8YA8	(1 << 16)
+#	define NV_PVIDEO_FORMAT_DISPLAY_COLOR_KEY	(1 << 20)
+#	define NV_PVIDEO_FORMAT_MATRIX_ITURBT709	(1 << 24)
+#define NV_PVIDEO_COLOR_KEY		0x00008B00
+
+/* NV04 overlay defines from VIDIX & Haiku */
+#define NV_PVIDEO_INTR_EN_0		0x00680140
+#define NV_PVIDEO_STEP_SIZE		0x00680200
+#define NV_PVIDEO_CONTROL_Y		0x00680204
+#define NV_PVIDEO_CONTROL_X		0x00680208
+#define NV_PVIDEO_BUFF0_START_ADDRESS	0x0068020c
+#define NV_PVIDEO_BUFF0_PITCH_LENGTH	0x00680214
+#define NV_PVIDEO_BUFF0_OFFSET		0x0068021c
+#define NV_PVIDEO_BUFF1_START_ADDRESS	0x00680210
+#define NV_PVIDEO_BUFF1_PITCH_LENGTH	0x00680218
+#define NV_PVIDEO_BUFF1_OFFSET		0x00680220
+#define NV_PVIDEO_OE_STATE		0x00680224
+#define NV_PVIDEO_SU_STATE		0x00680228
+#define NV_PVIDEO_RM_STATE		0x0068022c
+#define NV_PVIDEO_WINDOW_START		0x00680230
+#define NV_PVIDEO_WINDOW_SIZE		0x00680234
+#define NV_PVIDEO_FIFO_THRES_SIZE	0x00680238
+#define NV_PVIDEO_FIFO_BURST_LENGTH	0x0068023c
+#define NV_PVIDEO_KEY			0x00680240
+#define NV_PVIDEO_OVERLAY		0x00680244
+#define NV_PVIDEO_RED_CSC_OFFSET	0x00680280
+#define NV_PVIDEO_GREEN_CSC_OFFSET	0x00680284
+#define NV_PVIDEO_BLUE_CSC_OFFSET	0x00680288
+#define NV_PVIDEO_CSC_ADJUST		0x0068028c
+
+#endif
diff --git a/drivers/gpu/drm/r128/r128_drv.c b/drivers/gpu/drm/r128/r128_drv.c
index 601f4c0e5da5..b806fdcc7170 100644
--- a/drivers/gpu/drm/r128/r128_drv.c
+++ b/drivers/gpu/drm/r128/r128_drv.c
@@ -64,7 +64,7 @@ static struct drm_driver driver = {
 		.owner = THIS_MODULE,
 		.open = drm_open,
 		.release = drm_release,
-		.ioctl = drm_ioctl,
+		.unlocked_ioctl = drm_ioctl,
 		.mmap = drm_mmap,
 		.poll = drm_poll,
 		.fasync = drm_fasync,
diff --git a/drivers/gpu/drm/r128/r128_ioc32.c b/drivers/gpu/drm/r128/r128_ioc32.c
index d3cb676eee84..51c99fc4dd38 100644
--- a/drivers/gpu/drm/r128/r128_ioc32.c
+++ b/drivers/gpu/drm/r128/r128_ioc32.c
@@ -95,8 +95,7 @@ static int compat_r128_init(struct file *file, unsigned int cmd,
 			  &init->agp_textures_offset))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_R128_INIT, (unsigned long)init);
+	return drm_ioctl(file, DRM_IOCTL_R128_INIT, (unsigned long)init);
 }
 
 typedef struct drm_r128_depth32 {
@@ -129,8 +128,7 @@ static int compat_r128_depth(struct file *file, unsigned int cmd,
 			  &depth->mask))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_R128_DEPTH, (unsigned long)depth);
+	return drm_ioctl(file, DRM_IOCTL_R128_DEPTH, (unsigned long)depth);
 
 }
 
@@ -153,8 +151,7 @@ static int compat_r128_stipple(struct file *file, unsigned int cmd,
 			  &stipple->mask))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_R128_STIPPLE, (unsigned long)stipple);
+	return drm_ioctl(file, DRM_IOCTL_R128_STIPPLE, (unsigned long)stipple);
 }
 
 typedef struct drm_r128_getparam32 {
@@ -178,8 +175,7 @@ static int compat_r128_getparam(struct file *file, unsigned int cmd,
 			  &getparam->value))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_R128_GETPARAM, (unsigned long)getparam);
+	return drm_ioctl(file, DRM_IOCTL_R128_GETPARAM, (unsigned long)getparam);
 }
 
 drm_ioctl_compat_t *r128_compat_ioctls[] = {
@@ -210,12 +206,10 @@ long r128_compat_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
 	if (nr < DRM_COMMAND_BASE + DRM_ARRAY_SIZE(r128_compat_ioctls))
 		fn = r128_compat_ioctls[nr - DRM_COMMAND_BASE];
 
-	lock_kernel();		/* XXX for now */
 	if (fn != NULL)
 		ret = (*fn) (filp, cmd, arg);
 	else
-		ret = drm_ioctl(filp->f_path.dentry->d_inode, filp, cmd, arg);
-	unlock_kernel();
+		ret = drm_ioctl(filp, cmd, arg);
 
 	return ret;
 }
diff --git a/drivers/gpu/drm/radeon/.gitignore b/drivers/gpu/drm/radeon/.gitignore
new file mode 100644
index 000000000000..403eb3a5891f
--- /dev/null
+++ b/drivers/gpu/drm/radeon/.gitignore
@@ -0,0 +1,3 @@
+mkregtable
+*_reg_safe.h
+
diff --git a/drivers/gpu/drm/radeon/Kconfig b/drivers/gpu/drm/radeon/Kconfig
index 5982321be4d5..1c02d23f6fcc 100644
--- a/drivers/gpu/drm/radeon/Kconfig
+++ b/drivers/gpu/drm/radeon/Kconfig
@@ -1,10 +1,14 @@
 config DRM_RADEON_KMS
-	bool "Enable modesetting on radeon by default"
+	bool "Enable modesetting on radeon by default - NEW DRIVER"
 	depends on DRM_RADEON
 	help
-	  Choose this option if you want kernel modesetting enabled by default,
-	  and you have a new enough userspace to support this. Running old
-	  userspaces with this enabled will cause pain.
+	  Choose this option if you want kernel modesetting enabled by default.
+
+	  This is a completely new driver. It's only part of the existing drm
+	  for compatibility reasons. It requires an entirely different graphics
+	  stack above it and works very differently from the old drm stack.
+	  i.e. don't enable this unless you know what you are doing it may
+	  cause issues or bugs compared to the previous userspace driver stack.
 
 	  When kernel modesetting is enabled the IOCTL of radeon/drm
 	  driver are considered as invalid and an error message is printed
diff --git a/drivers/gpu/drm/radeon/Makefile b/drivers/gpu/drm/radeon/Makefile
index 09a28923f46e..ed38262d9985 100644
--- a/drivers/gpu/drm/radeon/Makefile
+++ b/drivers/gpu/drm/radeon/Makefile
@@ -24,9 +24,15 @@ $(obj)/rv515_reg_safe.h: $(src)/reg_srcs/rv515 $(obj)/mkregtable
 $(obj)/r300_reg_safe.h: $(src)/reg_srcs/r300 $(obj)/mkregtable
 	$(call if_changed,mkregtable)
 
+$(obj)/r420_reg_safe.h: $(src)/reg_srcs/r420 $(obj)/mkregtable
+	$(call if_changed,mkregtable)
+
 $(obj)/rs600_reg_safe.h: $(src)/reg_srcs/rs600 $(obj)/mkregtable
 	$(call if_changed,mkregtable)
 
+$(obj)/r600_reg_safe.h: $(src)/reg_srcs/r600 $(obj)/mkregtable
+	$(call if_changed,mkregtable)
+
 $(obj)/r100.o: $(obj)/r100_reg_safe.h $(obj)/rn50_reg_safe.h
 
 $(obj)/r200.o: $(obj)/r200_reg_safe.h
@@ -35,8 +41,12 @@ $(obj)/rv515.o: $(obj)/rv515_reg_safe.h
 
 $(obj)/r300.o: $(obj)/r300_reg_safe.h
 
+$(obj)/r420.o: $(obj)/r420_reg_safe.h
+
 $(obj)/rs600.o: $(obj)/rs600_reg_safe.h
 
+$(obj)/r600_cs.o: $(obj)/r600_reg_safe.h
+
 radeon-y := radeon_drv.o radeon_cp.o radeon_state.o radeon_mem.o \
 	radeon_irq.o r300_cmdbuf.o r600_cp.o
 # add KMS driver
@@ -49,8 +59,10 @@ radeon-y += radeon_device.o radeon_kms.o \
 	radeon_cs.o radeon_bios.o radeon_benchmark.o r100.o r300.o r420.o \
 	rs400.o rs600.o rs690.o rv515.o r520.o r600.o rv770.o radeon_test.o \
 	r200.o radeon_legacy_tv.o r600_cs.o r600_blit.o r600_blit_shaders.o \
-	r600_blit_kms.o
+	r600_blit_kms.o radeon_pm.o atombios_dp.o r600_audio.o r600_hdmi.o \
+	evergreen.o
 
 radeon-$(CONFIG_COMPAT) += radeon_ioc32.o
+radeon-$(CONFIG_VGA_SWITCHEROO) += radeon_atpx_handler.o
 
 obj-$(CONFIG_DRM_RADEON)+= radeon.o
diff --git a/drivers/gpu/drm/radeon/ObjectID.h b/drivers/gpu/drm/radeon/ObjectID.h
index 6d0183c61d3b..c714179d1bfa 100644
--- a/drivers/gpu/drm/radeon/ObjectID.h
+++ b/drivers/gpu/drm/radeon/ObjectID.h
@@ -1,5 +1,5 @@
 /*
-* Copyright 2006-2007 Advanced Micro Devices, Inc.
+* Copyright 2006-2007 Advanced Micro Devices, Inc.  
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
@@ -41,14 +41,14 @@
 /****************************************************/
 /* Encoder Object ID Definition                     */
 /****************************************************/
-#define ENCODER_OBJECT_ID_NONE                    0x00
+#define ENCODER_OBJECT_ID_NONE                    0x00 
 
 /* Radeon Class Display Hardware */
 #define ENCODER_OBJECT_ID_INTERNAL_LVDS           0x01
 #define ENCODER_OBJECT_ID_INTERNAL_TMDS1          0x02
 #define ENCODER_OBJECT_ID_INTERNAL_TMDS2          0x03
 #define ENCODER_OBJECT_ID_INTERNAL_DAC1           0x04
-#define ENCODER_OBJECT_ID_INTERNAL_DAC2           0x05	/* TV/CV DAC */
+#define ENCODER_OBJECT_ID_INTERNAL_DAC2           0x05     /* TV/CV DAC */
 #define ENCODER_OBJECT_ID_INTERNAL_SDVOA          0x06
 #define ENCODER_OBJECT_ID_INTERNAL_SDVOB          0x07
 
@@ -56,11 +56,11 @@
 #define ENCODER_OBJECT_ID_SI170B                  0x08
 #define ENCODER_OBJECT_ID_CH7303                  0x09
 #define ENCODER_OBJECT_ID_CH7301                  0x0A
-#define ENCODER_OBJECT_ID_INTERNAL_DVO1           0x0B	/* This belongs to Radeon Class Display Hardware */
+#define ENCODER_OBJECT_ID_INTERNAL_DVO1           0x0B    /* This belongs to Radeon Class Display Hardware */
 #define ENCODER_OBJECT_ID_EXTERNAL_SDVOA          0x0C
 #define ENCODER_OBJECT_ID_EXTERNAL_SDVOB          0x0D
 #define ENCODER_OBJECT_ID_TITFP513                0x0E
-#define ENCODER_OBJECT_ID_INTERNAL_LVTM1          0x0F	/* not used for Radeon */
+#define ENCODER_OBJECT_ID_INTERNAL_LVTM1          0x0F    /* not used for Radeon */
 #define ENCODER_OBJECT_ID_VT1623                  0x10
 #define ENCODER_OBJECT_ID_HDMI_SI1930             0x11
 #define ENCODER_OBJECT_ID_HDMI_INTERNAL           0x12
@@ -68,9 +68,9 @@
 #define ENCODER_OBJECT_ID_INTERNAL_KLDSCP_TMDS1   0x13
 #define ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DVO1    0x14
 #define ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DAC1    0x15
-#define ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DAC2    0x16	/* Shared with CV/TV and CRT */
-#define ENCODER_OBJECT_ID_SI178                   0X17	/* External TMDS (dual link, no HDCP.) */
-#define ENCODER_OBJECT_ID_MVPU_FPGA               0x18	/* MVPU FPGA chip */
+#define ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DAC2    0x16  /* Shared with CV/TV and CRT */
+#define ENCODER_OBJECT_ID_SI178                   0X17  /* External TMDS (dual link, no HDCP.) */
+#define ENCODER_OBJECT_ID_MVPU_FPGA               0x18  /* MVPU FPGA chip */
 #define ENCODER_OBJECT_ID_INTERNAL_DDI            0x19
 #define ENCODER_OBJECT_ID_VT1625                  0x1A
 #define ENCODER_OBJECT_ID_HDMI_SI1932             0x1B
@@ -86,7 +86,7 @@
 /****************************************************/
 /* Connector Object ID Definition                   */
 /****************************************************/
-#define CONNECTOR_OBJECT_ID_NONE                  0x00
+#define CONNECTOR_OBJECT_ID_NONE                  0x00 
 #define CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_I     0x01
 #define CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_I       0x02
 #define CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_D     0x03
@@ -96,7 +96,7 @@
 #define CONNECTOR_OBJECT_ID_SVIDEO                0x07
 #define CONNECTOR_OBJECT_ID_YPbPr                 0x08
 #define CONNECTOR_OBJECT_ID_D_CONNECTOR           0x09
-#define CONNECTOR_OBJECT_ID_9PIN_DIN              0x0A	/* Supports both CV & TV */
+#define CONNECTOR_OBJECT_ID_9PIN_DIN              0x0A  /* Supports both CV & TV */
 #define CONNECTOR_OBJECT_ID_SCART                 0x0B
 #define CONNECTOR_OBJECT_ID_HDMI_TYPE_A           0x0C
 #define CONNECTOR_OBJECT_ID_HDMI_TYPE_B           0x0D
@@ -106,6 +106,8 @@
 #define CONNECTOR_OBJECT_ID_CROSSFIRE             0x11
 #define CONNECTOR_OBJECT_ID_HARDCODE_DVI          0x12
 #define CONNECTOR_OBJECT_ID_DISPLAYPORT           0x13
+#define CONNECTOR_OBJECT_ID_eDP                   0x14
+#define CONNECTOR_OBJECT_ID_MXM                   0x15
 
 /* deleted */
 
@@ -116,6 +118,14 @@
 #define ROUTER_OBJECT_ID_I2C_EXTENDER_CNTL				0x01
 
 /****************************************************/
+/* Generic Object ID Definition                     */
+/****************************************************/
+#define GENERIC_OBJECT_ID_NONE                    0x00
+#define GENERIC_OBJECT_ID_GLSYNC                  0x01
+#define GENERIC_OBJECT_ID_PX2_NON_DRIVABLE        0x02
+#define GENERIC_OBJECT_ID_MXM_OPM                 0x03
+
+/****************************************************/
 /* Graphics Object ENUM ID Definition               */
 /****************************************************/
 #define GRAPH_OBJECT_ENUM_ID1                     0x01
@@ -124,6 +134,7 @@
 #define GRAPH_OBJECT_ENUM_ID4                     0x04
 #define GRAPH_OBJECT_ENUM_ID5                     0x05
 #define GRAPH_OBJECT_ENUM_ID6                     0x06
+#define GRAPH_OBJECT_ENUM_ID7                     0x07
 
 /****************************************************/
 /* Graphics Object ID Bit definition                */
@@ -133,35 +144,35 @@
 #define RESERVED1_ID_MASK                         0x0800
 #define OBJECT_TYPE_MASK                          0x7000
 #define RESERVED2_ID_MASK                         0x8000
-
+                                                  
 #define OBJECT_ID_SHIFT                           0x00
 #define ENUM_ID_SHIFT                             0x08
 #define OBJECT_TYPE_SHIFT                         0x0C
 
+
 /****************************************************/
 /* Graphics Object family definition                */
 /****************************************************/
-#define CONSTRUCTOBJECTFAMILYID(GRAPHICS_OBJECT_TYPE, GRAPHICS_OBJECT_ID) \
-	(GRAPHICS_OBJECT_TYPE << OBJECT_TYPE_SHIFT | \
-	 GRAPHICS_OBJECT_ID   << OBJECT_ID_SHIFT)
+#define CONSTRUCTOBJECTFAMILYID(GRAPHICS_OBJECT_TYPE, GRAPHICS_OBJECT_ID) (GRAPHICS_OBJECT_TYPE << OBJECT_TYPE_SHIFT | \
+                                                                           GRAPHICS_OBJECT_ID   << OBJECT_ID_SHIFT)
 /****************************************************/
 /* GPU Object ID definition - Shared with BIOS      */
 /****************************************************/
-#define GPU_ENUM_ID1	(GRAPH_OBJECT_TYPE_GPU << OBJECT_TYPE_SHIFT |\
-			 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT)
+#define GPU_ENUM_ID1                            ( GRAPH_OBJECT_TYPE_GPU << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT)
 
 /****************************************************/
 /* Encoder Object ID definition - Shared with BIOS  */
 /****************************************************/
 /*
-#define ENCODER_INTERNAL_LVDS_ENUM_ID1        0x2101
+#define ENCODER_INTERNAL_LVDS_ENUM_ID1        0x2101      
 #define ENCODER_INTERNAL_TMDS1_ENUM_ID1       0x2102
 #define ENCODER_INTERNAL_TMDS2_ENUM_ID1       0x2103
 #define ENCODER_INTERNAL_DAC1_ENUM_ID1        0x2104
 #define ENCODER_INTERNAL_DAC2_ENUM_ID1        0x2105
 #define ENCODER_INTERNAL_SDVOA_ENUM_ID1       0x2106
 #define ENCODER_INTERNAL_SDVOB_ENUM_ID1       0x2107
-#define ENCODER_SIL170B_ENUM_ID1              0x2108
+#define ENCODER_SIL170B_ENUM_ID1              0x2108  
 #define ENCODER_CH7303_ENUM_ID1               0x2109
 #define ENCODER_CH7301_ENUM_ID1               0x210A
 #define ENCODER_INTERNAL_DVO1_ENUM_ID1        0x210B
@@ -175,8 +186,8 @@
 #define ENCODER_INTERNAL_KLDSCP_TMDS1_ENUM_ID1   0x2113
 #define ENCODER_INTERNAL_KLDSCP_DVO1_ENUM_ID1    0x2114
 #define ENCODER_INTERNAL_KLDSCP_DAC1_ENUM_ID1    0x2115
-#define ENCODER_INTERNAL_KLDSCP_DAC2_ENUM_ID1    0x2116
-#define ENCODER_SI178_ENUM_ID1                   0x2117
+#define ENCODER_INTERNAL_KLDSCP_DAC2_ENUM_ID1    0x2116  
+#define ENCODER_SI178_ENUM_ID1                   0x2117 
 #define ENCODER_MVPU_FPGA_ENUM_ID1               0x2118
 #define ENCODER_INTERNAL_DDI_ENUM_ID1            0x2119
 #define ENCODER_VT1625_ENUM_ID1                  0x211A
@@ -185,205 +196,169 @@
 #define ENCODER_DP_DP501_ENUM_ID1                0x211D
 #define ENCODER_INTERNAL_UNIPHY_ENUM_ID1         0x211E
 */
-#define ENCODER_INTERNAL_LVDS_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_INTERNAL_LVDS << OBJECT_ID_SHIFT)
-
-#define ENCODER_INTERNAL_TMDS1_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_INTERNAL_TMDS1 << OBJECT_ID_SHIFT)
-
-#define ENCODER_INTERNAL_TMDS2_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_INTERNAL_TMDS2 << OBJECT_ID_SHIFT)
-
-#define ENCODER_INTERNAL_DAC1_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_INTERNAL_DAC1 << OBJECT_ID_SHIFT)
-
-#define ENCODER_INTERNAL_DAC2_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_INTERNAL_DAC2 << OBJECT_ID_SHIFT)
-
-#define ENCODER_INTERNAL_SDVOA_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_INTERNAL_SDVOA << OBJECT_ID_SHIFT)
-
-#define ENCODER_INTERNAL_SDVOA_ENUM_ID2 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_INTERNAL_SDVOA << OBJECT_ID_SHIFT)
-
-#define ENCODER_INTERNAL_SDVOB_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_INTERNAL_SDVOB << OBJECT_ID_SHIFT)
-
-#define ENCODER_SIL170B_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_SI170B << OBJECT_ID_SHIFT)
-
-#define ENCODER_CH7303_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_CH7303 << OBJECT_ID_SHIFT)
-
-#define ENCODER_CH7301_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_CH7301 << OBJECT_ID_SHIFT)
-
-#define ENCODER_INTERNAL_DVO1_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_INTERNAL_DVO1 << OBJECT_ID_SHIFT)
-
-#define ENCODER_EXTERNAL_SDVOA_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_EXTERNAL_SDVOA << OBJECT_ID_SHIFT)
-
-#define ENCODER_EXTERNAL_SDVOA_ENUM_ID2 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_EXTERNAL_SDVOA << OBJECT_ID_SHIFT)
-
-#define ENCODER_EXTERNAL_SDVOB_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_EXTERNAL_SDVOB << OBJECT_ID_SHIFT)
-
-#define ENCODER_TITFP513_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_TITFP513 << OBJECT_ID_SHIFT)
-
-#define ENCODER_INTERNAL_LVTM1_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_INTERNAL_LVTM1 << OBJECT_ID_SHIFT)
-
-#define ENCODER_VT1623_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_VT1623 << OBJECT_ID_SHIFT)
-
-#define ENCODER_HDMI_SI1930_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_HDMI_SI1930 << OBJECT_ID_SHIFT)
-
-#define ENCODER_HDMI_INTERNAL_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_HDMI_INTERNAL << OBJECT_ID_SHIFT)
-
-#define ENCODER_INTERNAL_KLDSCP_TMDS1_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_INTERNAL_KLDSCP_TMDS1 << OBJECT_ID_SHIFT)
-
-#define ENCODER_INTERNAL_KLDSCP_TMDS1_ENUM_ID2 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_INTERNAL_KLDSCP_TMDS1 << OBJECT_ID_SHIFT)
-
-#define ENCODER_INTERNAL_KLDSCP_DVO1_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DVO1 << OBJECT_ID_SHIFT)
-
-#define ENCODER_INTERNAL_KLDSCP_DAC1_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DAC1 << OBJECT_ID_SHIFT)
-
-#define ENCODER_INTERNAL_KLDSCP_DAC2_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DAC2 << OBJECT_ID_SHIFT) /* Shared with CV/TV and CRT */
-
-#define ENCODER_SI178_ENUM_ID1  \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_SI178 << OBJECT_ID_SHIFT)
-
-#define ENCODER_MVPU_FPGA_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_MVPU_FPGA << OBJECT_ID_SHIFT)
-
-#define ENCODER_INTERNAL_DDI_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_INTERNAL_DDI << OBJECT_ID_SHIFT)
-
-#define ENCODER_VT1625_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_VT1625 << OBJECT_ID_SHIFT)
-
-#define ENCODER_HDMI_SI1932_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_HDMI_SI1932 << OBJECT_ID_SHIFT)
-
-#define ENCODER_DP_DP501_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_DP_DP501 << OBJECT_ID_SHIFT)
-
-#define ENCODER_DP_AN9801_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_DP_AN9801 << OBJECT_ID_SHIFT)
-
-#define ENCODER_INTERNAL_UNIPHY_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_INTERNAL_UNIPHY << OBJECT_ID_SHIFT)
-
-#define ENCODER_INTERNAL_UNIPHY_ENUM_ID2 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_INTERNAL_UNIPHY << OBJECT_ID_SHIFT)
-
-#define ENCODER_INTERNAL_KLDSCP_LVTMA_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_INTERNAL_KLDSCP_LVTMA << OBJECT_ID_SHIFT)
-
-#define ENCODER_INTERNAL_UNIPHY1_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_INTERNAL_UNIPHY1 << OBJECT_ID_SHIFT)
-
-#define ENCODER_INTERNAL_UNIPHY1_ENUM_ID2 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_INTERNAL_UNIPHY1 << OBJECT_ID_SHIFT)
-
-#define ENCODER_INTERNAL_UNIPHY2_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_INTERNAL_UNIPHY2 << OBJECT_ID_SHIFT)
-
-#define ENCODER_INTERNAL_UNIPHY2_ENUM_ID2 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_INTERNAL_UNIPHY2 << OBJECT_ID_SHIFT)
-
-#define ENCODER_GENERAL_EXTERNAL_DVO_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ENCODER_OBJECT_ID_GENERAL_EXTERNAL_DVO << OBJECT_ID_SHIFT)
+#define ENCODER_INTERNAL_LVDS_ENUM_ID1     ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                             GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                             ENCODER_OBJECT_ID_INTERNAL_LVDS << OBJECT_ID_SHIFT)
+
+#define ENCODER_INTERNAL_TMDS1_ENUM_ID1    ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                             GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                             ENCODER_OBJECT_ID_INTERNAL_TMDS1 << OBJECT_ID_SHIFT)
+
+#define ENCODER_INTERNAL_TMDS2_ENUM_ID1    ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                             GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                             ENCODER_OBJECT_ID_INTERNAL_TMDS2 << OBJECT_ID_SHIFT)
+
+#define ENCODER_INTERNAL_DAC1_ENUM_ID1     ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                             GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                             ENCODER_OBJECT_ID_INTERNAL_DAC1 << OBJECT_ID_SHIFT)
+
+#define ENCODER_INTERNAL_DAC2_ENUM_ID1     ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                             GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                             ENCODER_OBJECT_ID_INTERNAL_DAC2 << OBJECT_ID_SHIFT)
+
+#define ENCODER_INTERNAL_SDVOA_ENUM_ID1    ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                             GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                             ENCODER_OBJECT_ID_INTERNAL_SDVOA << OBJECT_ID_SHIFT)
+
+#define ENCODER_INTERNAL_SDVOA_ENUM_ID2    ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                             GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
+                                             ENCODER_OBJECT_ID_INTERNAL_SDVOA << OBJECT_ID_SHIFT)
+
+#define ENCODER_INTERNAL_SDVOB_ENUM_ID1    ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                             GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                             ENCODER_OBJECT_ID_INTERNAL_SDVOB << OBJECT_ID_SHIFT)
+
+#define ENCODER_SIL170B_ENUM_ID1           ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                             GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                             ENCODER_OBJECT_ID_SI170B << OBJECT_ID_SHIFT)
+
+#define ENCODER_CH7303_ENUM_ID1            ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                             GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                             ENCODER_OBJECT_ID_CH7303 << OBJECT_ID_SHIFT)
+
+#define ENCODER_CH7301_ENUM_ID1            ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                             GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                             ENCODER_OBJECT_ID_CH7301 << OBJECT_ID_SHIFT)
+
+#define ENCODER_INTERNAL_DVO1_ENUM_ID1     ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                             GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                             ENCODER_OBJECT_ID_INTERNAL_DVO1 << OBJECT_ID_SHIFT)
+
+#define ENCODER_EXTERNAL_SDVOA_ENUM_ID1    ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                             GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                             ENCODER_OBJECT_ID_EXTERNAL_SDVOA << OBJECT_ID_SHIFT)
+
+#define ENCODER_EXTERNAL_SDVOA_ENUM_ID2    ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                             GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
+                                             ENCODER_OBJECT_ID_EXTERNAL_SDVOA << OBJECT_ID_SHIFT)
+
+
+#define ENCODER_EXTERNAL_SDVOB_ENUM_ID1    ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                             GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                             ENCODER_OBJECT_ID_EXTERNAL_SDVOB << OBJECT_ID_SHIFT)
+
+
+#define ENCODER_TITFP513_ENUM_ID1          ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                             GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                             ENCODER_OBJECT_ID_TITFP513 << OBJECT_ID_SHIFT)
+
+#define ENCODER_INTERNAL_LVTM1_ENUM_ID1    ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                             GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                             ENCODER_OBJECT_ID_INTERNAL_LVTM1 << OBJECT_ID_SHIFT)
+
+#define ENCODER_VT1623_ENUM_ID1            ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                             GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                             ENCODER_OBJECT_ID_VT1623 << OBJECT_ID_SHIFT)
+
+#define ENCODER_HDMI_SI1930_ENUM_ID1       ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                             GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                             ENCODER_OBJECT_ID_HDMI_SI1930 << OBJECT_ID_SHIFT)
+
+#define ENCODER_HDMI_INTERNAL_ENUM_ID1     ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                             GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                             ENCODER_OBJECT_ID_HDMI_INTERNAL << OBJECT_ID_SHIFT)
+
+#define ENCODER_INTERNAL_KLDSCP_TMDS1_ENUM_ID1   ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                                   GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                   ENCODER_OBJECT_ID_INTERNAL_KLDSCP_TMDS1 << OBJECT_ID_SHIFT)
+
+
+#define ENCODER_INTERNAL_KLDSCP_TMDS1_ENUM_ID2   ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                                   GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
+                                                   ENCODER_OBJECT_ID_INTERNAL_KLDSCP_TMDS1 << OBJECT_ID_SHIFT)
+
+
+#define ENCODER_INTERNAL_KLDSCP_DVO1_ENUM_ID1    ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                                   GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                   ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DVO1 << OBJECT_ID_SHIFT)
+
+#define ENCODER_INTERNAL_KLDSCP_DAC1_ENUM_ID1    ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                                   GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                   ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DAC1 << OBJECT_ID_SHIFT)
+
+#define ENCODER_INTERNAL_KLDSCP_DAC2_ENUM_ID1    ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                                   GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                   ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DAC2 << OBJECT_ID_SHIFT)  // Shared with CV/TV and CRT
+
+#define ENCODER_SI178_ENUM_ID1                    ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                                   GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                   ENCODER_OBJECT_ID_SI178 << OBJECT_ID_SHIFT)  
+
+#define ENCODER_MVPU_FPGA_ENUM_ID1                ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                                   GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                   ENCODER_OBJECT_ID_MVPU_FPGA << OBJECT_ID_SHIFT)
+
+#define ENCODER_INTERNAL_DDI_ENUM_ID1     (  GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                             GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                             ENCODER_OBJECT_ID_INTERNAL_DDI << OBJECT_ID_SHIFT) 
+
+#define ENCODER_VT1625_ENUM_ID1            ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                             GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                             ENCODER_OBJECT_ID_VT1625 << OBJECT_ID_SHIFT)
+
+#define ENCODER_HDMI_SI1932_ENUM_ID1       ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                             GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                             ENCODER_OBJECT_ID_HDMI_SI1932 << OBJECT_ID_SHIFT)
+
+#define ENCODER_DP_DP501_ENUM_ID1            ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                             GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                             ENCODER_OBJECT_ID_DP_DP501 << OBJECT_ID_SHIFT)
+
+#define ENCODER_DP_AN9801_ENUM_ID1            ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                             GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                             ENCODER_OBJECT_ID_DP_AN9801 << OBJECT_ID_SHIFT)
+
+#define ENCODER_INTERNAL_UNIPHY_ENUM_ID1         ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                 ENCODER_OBJECT_ID_INTERNAL_UNIPHY << OBJECT_ID_SHIFT)
+
+#define ENCODER_INTERNAL_UNIPHY_ENUM_ID2         ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
+                                                 ENCODER_OBJECT_ID_INTERNAL_UNIPHY << OBJECT_ID_SHIFT)
+
+#define ENCODER_INTERNAL_KLDSCP_LVTMA_ENUM_ID1   ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                 ENCODER_OBJECT_ID_INTERNAL_KLDSCP_LVTMA << OBJECT_ID_SHIFT)  
+
+#define ENCODER_INTERNAL_UNIPHY1_ENUM_ID1         ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                 ENCODER_OBJECT_ID_INTERNAL_UNIPHY1 << OBJECT_ID_SHIFT)
+
+#define ENCODER_INTERNAL_UNIPHY1_ENUM_ID2         ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
+                                                 ENCODER_OBJECT_ID_INTERNAL_UNIPHY1 << OBJECT_ID_SHIFT)
+
+#define ENCODER_INTERNAL_UNIPHY2_ENUM_ID1         ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                 ENCODER_OBJECT_ID_INTERNAL_UNIPHY2 << OBJECT_ID_SHIFT)
+
+#define ENCODER_INTERNAL_UNIPHY2_ENUM_ID2         ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
+                                                 ENCODER_OBJECT_ID_INTERNAL_UNIPHY2 << OBJECT_ID_SHIFT)
+
+#define ENCODER_GENERAL_EXTERNAL_DVO_ENUM_ID1    ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\
+                                                  GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                  ENCODER_OBJECT_ID_GENERAL_EXTERNAL_DVO << OBJECT_ID_SHIFT)
 
 /****************************************************/
 /* Connector Object ID definition - Shared with BIOS */
@@ -406,167 +381,253 @@
 #define CONNECTOR_7PIN_DIN_ENUM_ID1                 0x310F
 #define CONNECTOR_PCIE_CONNECTOR_ENUM_ID1           0x3110
 */
-#define CONNECTOR_LVDS_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 CONNECTOR_OBJECT_ID_LVDS << OBJECT_ID_SHIFT)
-
-#define CONNECTOR_SINGLE_LINK_DVI_I_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_I << OBJECT_ID_SHIFT)
-
-#define CONNECTOR_SINGLE_LINK_DVI_I_ENUM_ID2 \
-	(GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
-	 CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_I << OBJECT_ID_SHIFT)
-
-#define CONNECTOR_DUAL_LINK_DVI_I_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_I << OBJECT_ID_SHIFT)
-
-#define CONNECTOR_DUAL_LINK_DVI_I_ENUM_ID2 \
-	(GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
-	 CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_I << OBJECT_ID_SHIFT)
-
-#define CONNECTOR_SINGLE_LINK_DVI_D_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_D << OBJECT_ID_SHIFT)
-
-#define CONNECTOR_SINGLE_LINK_DVI_D_ENUM_ID2 \
-	(GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
-	 CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_D << OBJECT_ID_SHIFT)
-
-#define CONNECTOR_DUAL_LINK_DVI_D_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_D << OBJECT_ID_SHIFT)
-
-#define CONNECTOR_VGA_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 CONNECTOR_OBJECT_ID_VGA << OBJECT_ID_SHIFT)
-
-#define CONNECTOR_VGA_ENUM_ID2 \
-	(GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
-	 CONNECTOR_OBJECT_ID_VGA << OBJECT_ID_SHIFT)
-
-#define CONNECTOR_COMPOSITE_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 CONNECTOR_OBJECT_ID_COMPOSITE << OBJECT_ID_SHIFT)
-
-#define CONNECTOR_SVIDEO_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 CONNECTOR_OBJECT_ID_SVIDEO << OBJECT_ID_SHIFT)
-
-#define CONNECTOR_YPbPr_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 CONNECTOR_OBJECT_ID_YPbPr << OBJECT_ID_SHIFT)
-
-#define CONNECTOR_D_CONNECTOR_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 CONNECTOR_OBJECT_ID_D_CONNECTOR << OBJECT_ID_SHIFT)
-
-#define CONNECTOR_9PIN_DIN_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 CONNECTOR_OBJECT_ID_9PIN_DIN << OBJECT_ID_SHIFT)
-
-#define CONNECTOR_SCART_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 CONNECTOR_OBJECT_ID_SCART << OBJECT_ID_SHIFT)
-
-#define CONNECTOR_HDMI_TYPE_A_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 CONNECTOR_OBJECT_ID_HDMI_TYPE_A << OBJECT_ID_SHIFT)
-
-#define CONNECTOR_HDMI_TYPE_B_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 CONNECTOR_OBJECT_ID_HDMI_TYPE_B << OBJECT_ID_SHIFT)
-
-#define CONNECTOR_7PIN_DIN_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 CONNECTOR_OBJECT_ID_7PIN_DIN << OBJECT_ID_SHIFT)
-
-#define CONNECTOR_PCIE_CONNECTOR_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 CONNECTOR_OBJECT_ID_PCIE_CONNECTOR << OBJECT_ID_SHIFT)
-
-#define CONNECTOR_PCIE_CONNECTOR_ENUM_ID2 \
-	(GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
-	 CONNECTOR_OBJECT_ID_PCIE_CONNECTOR << OBJECT_ID_SHIFT)
-
-#define CONNECTOR_CROSSFIRE_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 CONNECTOR_OBJECT_ID_CROSSFIRE << OBJECT_ID_SHIFT)
-
-#define CONNECTOR_CROSSFIRE_ENUM_ID2 \
-	(GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
-	 CONNECTOR_OBJECT_ID_CROSSFIRE << OBJECT_ID_SHIFT)
-
-#define CONNECTOR_HARDCODE_DVI_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 CONNECTOR_OBJECT_ID_HARDCODE_DVI << OBJECT_ID_SHIFT)
-
-#define CONNECTOR_HARDCODE_DVI_ENUM_ID2 \
-	(GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
-	 CONNECTOR_OBJECT_ID_HARDCODE_DVI << OBJECT_ID_SHIFT)
-
-#define CONNECTOR_DISPLAYPORT_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 CONNECTOR_OBJECT_ID_DISPLAYPORT << OBJECT_ID_SHIFT)
-
-#define CONNECTOR_DISPLAYPORT_ENUM_ID2 \
-	(GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
-	 CONNECTOR_OBJECT_ID_DISPLAYPORT << OBJECT_ID_SHIFT)
-
-#define CONNECTOR_DISPLAYPORT_ENUM_ID3 \
-	(GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID3 << ENUM_ID_SHIFT |\
-	 CONNECTOR_OBJECT_ID_DISPLAYPORT << OBJECT_ID_SHIFT)
-
-#define CONNECTOR_DISPLAYPORT_ENUM_ID4 \
-	(GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID4 << ENUM_ID_SHIFT |\
-	 CONNECTOR_OBJECT_ID_DISPLAYPORT << OBJECT_ID_SHIFT)
+#define CONNECTOR_LVDS_ENUM_ID1                ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_LVDS << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_LVDS_ENUM_ID2                ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_LVDS << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_eDP_ENUM_ID1                 ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_eDP << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_eDP_ENUM_ID2                 ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_eDP << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_SINGLE_LINK_DVI_I_ENUM_ID1   ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_I << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_SINGLE_LINK_DVI_I_ENUM_ID2   ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_I << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_DUAL_LINK_DVI_I_ENUM_ID1     ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_I << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_DUAL_LINK_DVI_I_ENUM_ID2     ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_I << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_SINGLE_LINK_DVI_D_ENUM_ID1   ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_D << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_SINGLE_LINK_DVI_D_ENUM_ID2   ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_D << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_DUAL_LINK_DVI_D_ENUM_ID1     ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_D << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_DUAL_LINK_DVI_D_ENUM_ID2     ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_D << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_DUAL_LINK_DVI_D_ENUM_ID3     ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID3 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_D << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_VGA_ENUM_ID1                 ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_VGA << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_VGA_ENUM_ID2                 ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_VGA << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_COMPOSITE_ENUM_ID1           ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_COMPOSITE << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_COMPOSITE_ENUM_ID2           ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_COMPOSITE << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_SVIDEO_ENUM_ID1              ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_SVIDEO << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_SVIDEO_ENUM_ID2              ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_SVIDEO << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_YPbPr_ENUM_ID1               ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_YPbPr << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_YPbPr_ENUM_ID2               ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_YPbPr << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_D_CONNECTOR_ENUM_ID1         ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_D_CONNECTOR << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_D_CONNECTOR_ENUM_ID2         ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_D_CONNECTOR << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_9PIN_DIN_ENUM_ID1            ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_9PIN_DIN << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_9PIN_DIN_ENUM_ID2            ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_9PIN_DIN << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_SCART_ENUM_ID1               ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_SCART << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_SCART_ENUM_ID2               ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_SCART << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_HDMI_TYPE_A_ENUM_ID1         ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_HDMI_TYPE_A << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_HDMI_TYPE_A_ENUM_ID2         ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_HDMI_TYPE_A << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_HDMI_TYPE_A_ENUM_ID3         ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID3 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_HDMI_TYPE_A << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_HDMI_TYPE_B_ENUM_ID1         ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_HDMI_TYPE_B << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_HDMI_TYPE_B_ENUM_ID2         ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_HDMI_TYPE_B << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_7PIN_DIN_ENUM_ID1            ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_7PIN_DIN << OBJECT_ID_SHIFT)
+#define CONNECTOR_7PIN_DIN_ENUM_ID2            ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_7PIN_DIN << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_PCIE_CONNECTOR_ENUM_ID1      ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_PCIE_CONNECTOR << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_PCIE_CONNECTOR_ENUM_ID2      ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_PCIE_CONNECTOR << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_CROSSFIRE_ENUM_ID1           ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_CROSSFIRE << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_CROSSFIRE_ENUM_ID2           ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_CROSSFIRE << OBJECT_ID_SHIFT)
+
+
+#define CONNECTOR_HARDCODE_DVI_ENUM_ID1        ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_HARDCODE_DVI << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_HARDCODE_DVI_ENUM_ID2        ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_HARDCODE_DVI << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_DISPLAYPORT_ENUM_ID1         ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_DISPLAYPORT << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_DISPLAYPORT_ENUM_ID2         ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_DISPLAYPORT << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_DISPLAYPORT_ENUM_ID3         ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID3 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_DISPLAYPORT << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_DISPLAYPORT_ENUM_ID4         ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID4 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_DISPLAYPORT << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_DISPLAYPORT_ENUM_ID5         ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID5 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_DISPLAYPORT << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_DISPLAYPORT_ENUM_ID6         ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID6 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_DISPLAYPORT << OBJECT_ID_SHIFT)
+
+#define CONNECTOR_MXM_ENUM_ID1                 ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_MXM << OBJECT_ID_SHIFT)          //Mapping to MXM_DP_A
+
+#define CONNECTOR_MXM_ENUM_ID2                 ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_MXM << OBJECT_ID_SHIFT)          //Mapping to MXM_DP_B
+
+#define CONNECTOR_MXM_ENUM_ID3                 ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID3 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_MXM << OBJECT_ID_SHIFT)          //Mapping to MXM_DP_C
+
+#define CONNECTOR_MXM_ENUM_ID4                 ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID4 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_MXM << OBJECT_ID_SHIFT)          //Mapping to MXM_DP_D
+
+#define CONNECTOR_MXM_ENUM_ID5                 ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID5 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_MXM << OBJECT_ID_SHIFT)          //Mapping to MXM_LVDS_TXxx
+
+#define CONNECTOR_MXM_ENUM_ID6                 ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID6 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_MXM << OBJECT_ID_SHIFT)          //Mapping to MXM_LVDS_UXxx
+
+#define CONNECTOR_MXM_ENUM_ID7                 ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID7 << ENUM_ID_SHIFT |\
+                                                 CONNECTOR_OBJECT_ID_MXM << OBJECT_ID_SHIFT)          //Mapping to MXM_DAC
 
 /****************************************************/
 /* Router Object ID definition - Shared with BIOS   */
 /****************************************************/
-#define ROUTER_I2C_EXTENDER_CNTL_ENUM_ID1 \
-	(GRAPH_OBJECT_TYPE_ROUTER << OBJECT_TYPE_SHIFT |\
-	 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
-	 ROUTER_OBJECT_ID_I2C_EXTENDER_CNTL << OBJECT_ID_SHIFT)
+#define ROUTER_I2C_EXTENDER_CNTL_ENUM_ID1      ( GRAPH_OBJECT_TYPE_ROUTER << OBJECT_TYPE_SHIFT |\
+                                                GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                ROUTER_OBJECT_ID_I2C_EXTENDER_CNTL << OBJECT_ID_SHIFT)
 
 /* deleted */
 
 /****************************************************/
+/* Generic Object ID definition - Shared with BIOS  */
+/****************************************************/
+#define GENERICOBJECT_GLSYNC_ENUM_ID1           (GRAPH_OBJECT_TYPE_GENERIC << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                 GENERIC_OBJECT_ID_GLSYNC << OBJECT_ID_SHIFT)
+
+#define GENERICOBJECT_PX2_NON_DRIVABLE_ID1       (GRAPH_OBJECT_TYPE_GENERIC << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                 GENERIC_OBJECT_ID_PX2_NON_DRIVABLE<< OBJECT_ID_SHIFT)
+
+#define GENERICOBJECT_PX2_NON_DRIVABLE_ID2       (GRAPH_OBJECT_TYPE_GENERIC << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\
+                                                 GENERIC_OBJECT_ID_PX2_NON_DRIVABLE<< OBJECT_ID_SHIFT)
+
+#define GENERICOBJECT_MXM_OPM_ENUM_ID1           (GRAPH_OBJECT_TYPE_GENERIC << OBJECT_TYPE_SHIFT |\
+                                                 GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\
+                                                 GENERIC_OBJECT_ID_MXM_OPM << OBJECT_ID_SHIFT)
+
+/****************************************************/
 /* Object Cap definition - Shared with BIOS         */
 /****************************************************/
 #define GRAPHICS_OBJECT_CAP_I2C                 0x00000001L
 #define GRAPHICS_OBJECT_CAP_TABLE_ID            0x00000002L
 
+
 #define GRAPHICS_OBJECT_I2CCOMMAND_TABLE_ID                   0x01
 #define GRAPHICS_OBJECT_HOTPLUGDETECTIONINTERUPT_TABLE_ID     0x02
 #define GRAPHICS_OBJECT_ENCODER_OUTPUT_PROTECTION_TABLE_ID    0x03
@@ -575,4 +636,8 @@
 #pragma pack()
 #endif
 
-#endif /*GRAPHICTYPE */
+#endif  /*GRAPHICTYPE */
+
+
+
+
diff --git a/drivers/gpu/drm/radeon/atom.c b/drivers/gpu/drm/radeon/atom.c
index 901befe03da2..d75788feac6c 100644
--- a/drivers/gpu/drm/radeon/atom.c
+++ b/drivers/gpu/drm/radeon/atom.c
@@ -24,6 +24,7 @@
 
 #include <linux/module.h>
 #include <linux/sched.h>
+#include <asm/unaligned.h>
 
 #define ATOM_DEBUG
 
@@ -58,6 +59,7 @@ typedef struct {
 } atom_exec_context;
 
 int atom_debug = 0;
+static void atom_execute_table_locked(struct atom_context *ctx, int index, uint32_t * params);
 void atom_execute_table(struct atom_context *ctx, int index, uint32_t * params);
 
 static uint32_t atom_arg_mask[8] =
@@ -107,6 +109,7 @@ static uint32_t atom_iio_execute(struct atom_context *ctx, int base,
 			base += 3;
 			break;
 		case ATOM_IIO_WRITE:
+			(void)ctx->card->reg_read(ctx->card, CU16(base + 1));
 			ctx->card->reg_write(ctx->card, CU16(base + 1), temp);
 			base += 3;
 			break;
@@ -210,7 +213,9 @@ static uint32_t atom_get_src_int(atom_exec_context *ctx, uint8_t attr,
 	case ATOM_ARG_PS:
 		idx = U8(*ptr);
 		(*ptr)++;
-		val = le32_to_cpu(ctx->ps[idx]);
+		/* get_unaligned_le32 avoids unaligned accesses from atombios
+		 * tables, noticed on a DEC Alpha. */
+		val = get_unaligned_le32((u32 *)&ctx->ps[idx]);
 		if (print)
 			DEBUG("PS[0x%02X,0x%04X]", idx, val);
 		break;
@@ -244,6 +249,9 @@ static uint32_t atom_get_src_int(atom_exec_context *ctx, uint8_t attr,
 		case ATOM_WS_ATTRIBUTES:
 			val = gctx->io_attr;
 			break;
+		case ATOM_WS_REGPTR:
+			val = gctx->reg_block;
+			break;
 		default:
 			val = ctx->ws[idx];
 		}
@@ -262,10 +270,10 @@ static uint32_t atom_get_src_int(atom_exec_context *ctx, uint8_t attr,
 	case ATOM_ARG_FB:
 		idx = U8(*ptr);
 		(*ptr)++;
+		val = gctx->scratch[((gctx->fb_base + idx) / 4)];
 		if (print)
 			DEBUG("FB[0x%02X]", idx);
-		printk(KERN_INFO "FB access is not implemented.\n");
-		return 0;
+		break;
 	case ATOM_ARG_IMM:
 		switch (align) {
 		case ATOM_SRC_DWORD:
@@ -383,6 +391,32 @@ static uint32_t atom_get_src(atom_exec_context *ctx, uint8_t attr, int *ptr)
 	return atom_get_src_int(ctx, attr, ptr, NULL, 1);
 }
 
+static uint32_t atom_get_src_direct(atom_exec_context *ctx, uint8_t align, int *ptr)
+{
+	uint32_t val = 0xCDCDCDCD;
+
+	switch (align) {
+	case ATOM_SRC_DWORD:
+		val = U32(*ptr);
+		(*ptr) += 4;
+		break;
+	case ATOM_SRC_WORD0:
+	case ATOM_SRC_WORD8:
+	case ATOM_SRC_WORD16:
+		val = U16(*ptr);
+		(*ptr) += 2;
+		break;
+	case ATOM_SRC_BYTE0:
+	case ATOM_SRC_BYTE8:
+	case ATOM_SRC_BYTE16:
+	case ATOM_SRC_BYTE24:
+		val = U8(*ptr);
+		(*ptr)++;
+		break;
+	}
+	return val;
+}
+
 static uint32_t atom_get_dst(atom_exec_context *ctx, int arg, uint8_t attr,
 			     int *ptr, uint32_t *saved, int print)
 {
@@ -480,6 +514,9 @@ static void atom_put_dst(atom_exec_context *ctx, int arg, uint8_t attr,
 		case ATOM_WS_ATTRIBUTES:
 			gctx->io_attr = val;
 			break;
+		case ATOM_WS_REGPTR:
+			gctx->reg_block = val;
+			break;
 		default:
 			ctx->ws[idx] = val;
 		}
@@ -487,9 +524,9 @@ static void atom_put_dst(atom_exec_context *ctx, int arg, uint8_t attr,
 	case ATOM_ARG_FB:
 		idx = U8(*ptr);
 		(*ptr)++;
+		gctx->scratch[((gctx->fb_base + idx) / 4)] = val;
 		DEBUG("FB[0x%02X]", idx);
-		printk(KERN_INFO "FB access is not implemented.\n");
-		return;
+		break;
 	case ATOM_ARG_PLL:
 		idx = U8(*ptr);
 		(*ptr)++;
@@ -572,7 +609,7 @@ static void atom_op_calltable(atom_exec_context *ctx, int *ptr, int arg)
 	else
 		SDEBUG("   table: %d\n", idx);
 	if (U16(ctx->ctx->cmd_table + 4 + 2 * idx))
-		atom_execute_table(ctx->ctx, idx, ctx->ps + ctx->ps_shift);
+		atom_execute_table_locked(ctx->ctx, idx, ctx->ps + ctx->ps_shift);
 }
 
 static void atom_op_clear(atom_exec_context *ctx, int *ptr, int arg)
@@ -606,7 +643,7 @@ static void atom_op_delay(atom_exec_context *ctx, int *ptr, int arg)
 	uint8_t count = U8((*ptr)++);
 	SDEBUG("   count: %d\n", count);
 	if (arg == ATOM_UNIT_MICROSEC)
-		schedule_timeout_uninterruptible(usecs_to_jiffies(count));
+		udelay(count);
 	else
 		schedule_timeout_uninterruptible(msecs_to_jiffies(count));
 }
@@ -675,7 +712,7 @@ static void atom_op_mask(atom_exec_context *ctx, int *ptr, int arg)
 	SDEBUG("   dst: ");
 	dst = atom_get_dst(ctx, arg, attr, ptr, &saved, 1);
 	SDEBUG("   src1: ");
-	src1 = atom_get_src(ctx, attr, ptr);
+	src1 = atom_get_src_direct(ctx, ((attr >> 3) & 7), ptr);
 	SDEBUG("   src2: ");
 	src2 = atom_get_src(ctx, attr, ptr);
 	dst &= src1;
@@ -807,7 +844,7 @@ static void atom_op_setregblock(atom_exec_context *ctx, int *ptr, int arg)
 	SDEBUG("   base: 0x%04X\n", ctx->ctx->reg_block);
 }
 
-static void atom_op_shl(atom_exec_context *ctx, int *ptr, int arg)
+static void atom_op_shift_left(atom_exec_context *ctx, int *ptr, int arg)
 {
 	uint8_t attr = U8((*ptr)++), shift;
 	uint32_t saved, dst;
@@ -816,14 +853,14 @@ static void atom_op_shl(atom_exec_context *ctx, int *ptr, int arg)
 	attr |= atom_def_dst[attr >> 3] << 6;
 	SDEBUG("   dst: ");
 	dst = atom_get_dst(ctx, arg, attr, ptr, &saved, 1);
-	shift = U8((*ptr)++);
+	shift = atom_get_src_direct(ctx, ATOM_SRC_BYTE0, ptr);
 	SDEBUG("   shift: %d\n", shift);
 	dst <<= shift;
 	SDEBUG("   dst: ");
 	atom_put_dst(ctx, arg, attr, &dptr, dst, saved);
 }
 
-static void atom_op_shr(atom_exec_context *ctx, int *ptr, int arg)
+static void atom_op_shift_right(atom_exec_context *ctx, int *ptr, int arg)
 {
 	uint8_t attr = U8((*ptr)++), shift;
 	uint32_t saved, dst;
@@ -832,7 +869,35 @@ static void atom_op_shr(atom_exec_context *ctx, int *ptr, int arg)
 	attr |= atom_def_dst[attr >> 3] << 6;
 	SDEBUG("   dst: ");
 	dst = atom_get_dst(ctx, arg, attr, ptr, &saved, 1);
-	shift = U8((*ptr)++);
+	shift = atom_get_src_direct(ctx, ATOM_SRC_BYTE0, ptr);
+	SDEBUG("   shift: %d\n", shift);
+	dst >>= shift;
+	SDEBUG("   dst: ");
+	atom_put_dst(ctx, arg, attr, &dptr, dst, saved);
+}
+
+static void atom_op_shl(atom_exec_context *ctx, int *ptr, int arg)
+{
+	uint8_t attr = U8((*ptr)++), shift;
+	uint32_t saved, dst;
+	int dptr = *ptr;
+	SDEBUG("   dst: ");
+	dst = atom_get_dst(ctx, arg, attr, ptr, &saved, 1);
+	shift = atom_get_src(ctx, attr, ptr);
+	SDEBUG("   shift: %d\n", shift);
+	dst <<= shift;
+	SDEBUG("   dst: ");
+	atom_put_dst(ctx, arg, attr, &dptr, dst, saved);
+}
+
+static void atom_op_shr(atom_exec_context *ctx, int *ptr, int arg)
+{
+	uint8_t attr = U8((*ptr)++), shift;
+	uint32_t saved, dst;
+	int dptr = *ptr;
+	SDEBUG("   dst: ");
+	dst = atom_get_dst(ctx, arg, attr, ptr, &saved, 1);
+	shift = atom_get_src(ctx, attr, ptr);
 	SDEBUG("   shift: %d\n", shift);
 	dst >>= shift;
 	SDEBUG("   dst: ");
@@ -935,18 +1000,18 @@ static struct {
 	atom_op_or, ATOM_ARG_FB}, {
 	atom_op_or, ATOM_ARG_PLL}, {
 	atom_op_or, ATOM_ARG_MC}, {
-	atom_op_shl, ATOM_ARG_REG}, {
-	atom_op_shl, ATOM_ARG_PS}, {
-	atom_op_shl, ATOM_ARG_WS}, {
-	atom_op_shl, ATOM_ARG_FB}, {
-	atom_op_shl, ATOM_ARG_PLL}, {
-	atom_op_shl, ATOM_ARG_MC}, {
-	atom_op_shr, ATOM_ARG_REG}, {
-	atom_op_shr, ATOM_ARG_PS}, {
-	atom_op_shr, ATOM_ARG_WS}, {
-	atom_op_shr, ATOM_ARG_FB}, {
-	atom_op_shr, ATOM_ARG_PLL}, {
-	atom_op_shr, ATOM_ARG_MC}, {
+	atom_op_shift_left, ATOM_ARG_REG}, {
+	atom_op_shift_left, ATOM_ARG_PS}, {
+	atom_op_shift_left, ATOM_ARG_WS}, {
+	atom_op_shift_left, ATOM_ARG_FB}, {
+	atom_op_shift_left, ATOM_ARG_PLL}, {
+	atom_op_shift_left, ATOM_ARG_MC}, {
+	atom_op_shift_right, ATOM_ARG_REG}, {
+	atom_op_shift_right, ATOM_ARG_PS}, {
+	atom_op_shift_right, ATOM_ARG_WS}, {
+	atom_op_shift_right, ATOM_ARG_FB}, {
+	atom_op_shift_right, ATOM_ARG_PLL}, {
+	atom_op_shift_right, ATOM_ARG_MC}, {
 	atom_op_mul, ATOM_ARG_REG}, {
 	atom_op_mul, ATOM_ARG_PS}, {
 	atom_op_mul, ATOM_ARG_WS}, {
@@ -1039,7 +1104,7 @@ static struct {
 	atom_op_shr, ATOM_ARG_MC}, {
 atom_op_debug, 0},};
 
-void atom_execute_table(struct atom_context *ctx, int index, uint32_t * params)
+static void atom_execute_table_locked(struct atom_context *ctx, int index, uint32_t * params)
 {
 	int base = CU16(ctx->cmd_table + 4 + 2 * index);
 	int len, ws, ps, ptr;
@@ -1056,8 +1121,6 @@ void atom_execute_table(struct atom_context *ctx, int index, uint32_t * params)
 
 	SDEBUG(">> execute %04X (len %d, WS %d, PS %d)\n", base, len, ws, ps);
 
-	/* reset reg block */
-	ctx->reg_block = 0;
 	ectx.ctx = ctx;
 	ectx.ps_shift = ps / 4;
 	ectx.start = base;
@@ -1091,6 +1154,19 @@ void atom_execute_table(struct atom_context *ctx, int index, uint32_t * params)
 		kfree(ectx.ws);
 }
 
+void atom_execute_table(struct atom_context *ctx, int index, uint32_t * params)
+{
+	mutex_lock(&ctx->mutex);
+	/* reset reg block */
+	ctx->reg_block = 0;
+	/* reset fb window */
+	ctx->fb_base = 0;
+	/* reset io mode */
+	ctx->io_mode = ATOM_IO_MM;
+	atom_execute_table_locked(ctx, index, params);
+	mutex_unlock(&ctx->mutex);
+}
+
 static int atom_iio_len[] = { 1, 2, 3, 3, 3, 3, 4, 4, 4, 3 };
 
 static void atom_index_iio(struct atom_context *ctx, int base)
@@ -1213,3 +1289,28 @@ void atom_parse_cmd_header(struct atom_context *ctx, int index, uint8_t * frev,
 		*crev = CU8(idx + 3);
 	return;
 }
+
+int atom_allocate_fb_scratch(struct atom_context *ctx)
+{
+	int index = GetIndexIntoMasterTable(DATA, VRAM_UsageByFirmware);
+	uint16_t data_offset;
+	int usage_bytes;
+	struct _ATOM_VRAM_USAGE_BY_FIRMWARE *firmware_usage;
+
+	atom_parse_data_header(ctx, index, NULL, NULL, NULL, &data_offset);
+
+	firmware_usage = (struct _ATOM_VRAM_USAGE_BY_FIRMWARE *)(ctx->bios + data_offset);
+
+	DRM_DEBUG("atom firmware requested %08x %dkb\n",
+		  firmware_usage->asFirmwareVramReserveInfo[0].ulStartAddrUsedByFirmware,
+		  firmware_usage->asFirmwareVramReserveInfo[0].usFirmwareUseInKb);
+
+	usage_bytes = firmware_usage->asFirmwareVramReserveInfo[0].usFirmwareUseInKb * 1024;
+	if (usage_bytes == 0)
+		usage_bytes = 20 * 1024;
+	/* allocate some scratch memory */
+	ctx->scratch = kzalloc(usage_bytes, GFP_KERNEL);
+	if (!ctx->scratch)
+		return -ENOMEM;
+	return 0;
+}
diff --git a/drivers/gpu/drm/radeon/atom.h b/drivers/gpu/drm/radeon/atom.h
index e6eb38f2bcae..bc73781423a1 100644
--- a/drivers/gpu/drm/radeon/atom.h
+++ b/drivers/gpu/drm/radeon/atom.h
@@ -91,6 +91,7 @@
 #define ATOM_WS_AND_MASK	0x45
 #define ATOM_WS_FB_WINDOW	0x46
 #define ATOM_WS_ATTRIBUTES	0x47
+#define ATOM_WS_REGPTR  	0x48
 
 #define ATOM_IIO_NOP		0
 #define ATOM_IIO_START		1
@@ -120,6 +121,7 @@ struct card_info {
 
 struct atom_context {
 	struct card_info *card;
+	struct mutex mutex;
 	void *bios;
 	uint32_t cmd_table, data_table;
 	uint16_t *iio;
@@ -132,6 +134,7 @@ struct atom_context {
 	uint8_t shift;
 	int cs_equal, cs_above;
 	int io_mode;
+	uint32_t *scratch;
 };
 
 extern int atom_debug;
@@ -142,6 +145,7 @@ int atom_asic_init(struct atom_context *);
 void atom_destroy(struct atom_context *);
 void atom_parse_data_header(struct atom_context *ctx, int index, uint16_t *size, uint8_t *frev, uint8_t *crev, uint16_t *data_start);
 void atom_parse_cmd_header(struct atom_context *ctx, int index, uint8_t *frev, uint8_t *crev);
+int atom_allocate_fb_scratch(struct atom_context *ctx);
 #include "atom-types.h"
 #include "atombios.h"
 #include "ObjectID.h"
diff --git a/drivers/gpu/drm/radeon/atombios.h b/drivers/gpu/drm/radeon/atombios.h
index 5d402086bc47..6732b5dd8ff4 100644
--- a/drivers/gpu/drm/radeon/atombios.h
+++ b/drivers/gpu/drm/radeon/atombios.h
@@ -1,5 +1,5 @@
 /*
- * Copyright 2006-2007 Advanced Micro Devices, Inc.
+ * Copyright 2006-2007 Advanced Micro Devices, Inc.  
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -20,10 +20,12 @@
  * OTHER DEALINGS IN THE SOFTWARE.
  */
 
-/****************************************************************************/
+
+/****************************************************************************/	
 /*Portion I: Definitions  shared between VBIOS and Driver                   */
 /****************************************************************************/
 
+
 #ifndef _ATOMBIOS_H
 #define _ATOMBIOS_H
 
@@ -40,39 +42,46 @@
 #endif
 
 #ifdef _H2INC
-#ifndef ULONG
-typedef unsigned long ULONG;
-#endif
+  #ifndef ULONG 
+    typedef unsigned long ULONG;
+  #endif
 
-#ifndef UCHAR
-typedef unsigned char UCHAR;
-#endif
+  #ifndef UCHAR
+    typedef unsigned char UCHAR;
+  #endif
 
-#ifndef USHORT
-typedef unsigned short USHORT;
-#endif
+  #ifndef USHORT 
+    typedef unsigned short USHORT;
+  #endif
 #endif
-
-#define ATOM_DAC_A            0
+      
+#define ATOM_DAC_A            0 
 #define ATOM_DAC_B            1
 #define ATOM_EXT_DAC          2
 
 #define ATOM_CRTC1            0
 #define ATOM_CRTC2            1
+#define ATOM_CRTC3            2
+#define ATOM_CRTC4            3
+#define ATOM_CRTC5            4
+#define ATOM_CRTC6            5
+#define ATOM_CRTC_INVALID     0xFF
 
 #define ATOM_DIGA             0
 #define ATOM_DIGB             1
 
 #define ATOM_PPLL1            0
 #define ATOM_PPLL2            1
+#define ATOM_DCPLL            2
+#define ATOM_PPLL_INVALID     0xFF
 
 #define ATOM_SCALER1          0
 #define ATOM_SCALER2          1
 
-#define ATOM_SCALER_DISABLE   0
-#define ATOM_SCALER_CENTER    1
-#define ATOM_SCALER_EXPANSION 2
-#define ATOM_SCALER_MULTI_EX  3
+#define ATOM_SCALER_DISABLE   0   
+#define ATOM_SCALER_CENTER    1   
+#define ATOM_SCALER_EXPANSION 2   
+#define ATOM_SCALER_MULTI_EX  3   
 
 #define ATOM_DISABLE          0
 #define ATOM_ENABLE           1
@@ -82,6 +91,7 @@ typedef unsigned short USHORT;
 #define ATOM_LCD_SELFTEST_START									(ATOM_DISABLE+5)
 #define ATOM_LCD_SELFTEST_STOP									(ATOM_ENABLE+5)
 #define ATOM_ENCODER_INIT			                  (ATOM_DISABLE+7)
+#define ATOM_GET_STATUS                         (ATOM_DISABLE+8)
 
 #define ATOM_BLANKING         1
 #define ATOM_BLANKING_OFF     0
@@ -114,7 +124,7 @@ typedef unsigned short USHORT;
 #define ATOM_DAC2_CV          ATOM_DAC1_CV
 #define ATOM_DAC2_NTSC        ATOM_DAC1_NTSC
 #define ATOM_DAC2_PAL         ATOM_DAC1_PAL
-
+ 
 #define ATOM_PM_ON            0
 #define ATOM_PM_STANDBY       1
 #define ATOM_PM_SUSPEND       2
@@ -134,6 +144,7 @@ typedef unsigned short USHORT;
 #define ATOM_PANEL_MISC_TEMPORAL           0x00000040
 #define ATOM_PANEL_MISC_API_ENABLED        0x00000080
 
+
 #define MEMTYPE_DDR1              "DDR1"
 #define MEMTYPE_DDR2              "DDR2"
 #define MEMTYPE_DDR3              "DDR3"
@@ -145,19 +156,19 @@ typedef unsigned short USHORT;
 
 /* Maximum size of that FireGL flag string */
 
-#define ATOM_FIREGL_FLAG_STRING     "FGL"	/* Flag used to enable FireGL Support */
-#define ATOM_MAX_SIZE_OF_FIREGL_FLAG_STRING  3	/* sizeof( ATOM_FIREGL_FLAG_STRING ) */
+#define ATOM_FIREGL_FLAG_STRING     "FGL"             //Flag used to enable FireGL Support
+#define ATOM_MAX_SIZE_OF_FIREGL_FLAG_STRING  3        //sizeof( ATOM_FIREGL_FLAG_STRING )
 
-#define ATOM_FAKE_DESKTOP_STRING    "DSK"	/* Flag used to enable mobile ASIC on Desktop */
-#define ATOM_MAX_SIZE_OF_FAKE_DESKTOP_STRING  ATOM_MAX_SIZE_OF_FIREGL_FLAG_STRING
+#define ATOM_FAKE_DESKTOP_STRING    "DSK"             //Flag used to enable mobile ASIC on Desktop
+#define ATOM_MAX_SIZE_OF_FAKE_DESKTOP_STRING  ATOM_MAX_SIZE_OF_FIREGL_FLAG_STRING 
 
-#define ATOM_M54T_FLAG_STRING       "M54T"	/* Flag used to enable M54T Support */
-#define ATOM_MAX_SIZE_OF_M54T_FLAG_STRING    4	/* sizeof( ATOM_M54T_FLAG_STRING ) */
+#define ATOM_M54T_FLAG_STRING       "M54T"            //Flag used to enable M54T Support
+#define ATOM_MAX_SIZE_OF_M54T_FLAG_STRING    4        //sizeof( ATOM_M54T_FLAG_STRING )
 
 #define HW_ASSISTED_I2C_STATUS_FAILURE          2
 #define HW_ASSISTED_I2C_STATUS_SUCCESS          1
 
-#pragma pack(1)			/* BIOS data must use byte aligment */
+#pragma pack(1)                                       /* BIOS data must use byte aligment */
 
 /*  Define offset to location of ROM header. */
 
@@ -165,367 +176,410 @@ typedef unsigned short USHORT;
 #define OFFSET_TO_ATOM_ROM_IMAGE_SIZE				    0x00000002L
 
 #define OFFSET_TO_ATOMBIOS_ASIC_BUS_MEM_TYPE    0x94
-#define MAXSIZE_OF_ATOMBIOS_ASIC_BUS_MEM_TYPE   20	/* including the terminator 0x0! */
+#define MAXSIZE_OF_ATOMBIOS_ASIC_BUS_MEM_TYPE   20    /* including the terminator 0x0! */
 #define	OFFSET_TO_GET_ATOMBIOS_STRINGS_NUMBER		0x002f
 #define	OFFSET_TO_GET_ATOMBIOS_STRINGS_START		0x006e
 
 /* Common header for all ROM Data tables.
-  Every table pointed  _ATOM_MASTER_DATA_TABLE has this common header.
+  Every table pointed  _ATOM_MASTER_DATA_TABLE has this common header. 
   And the pointer actually points to this header. */
 
-typedef struct _ATOM_COMMON_TABLE_HEADER {
-	USHORT usStructureSize;
-	UCHAR ucTableFormatRevision;	/*Change it when the Parser is not backward compatible */
-	UCHAR ucTableContentRevision;	/*Change it only when the table needs to change but the firmware */
-	/*Image can't be updated, while Driver needs to carry the new table! */
-} ATOM_COMMON_TABLE_HEADER;
-
-typedef struct _ATOM_ROM_HEADER {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	UCHAR uaFirmWareSignature[4];	/*Signature to distinguish between Atombios and non-atombios,
-					   atombios should init it as "ATOM", don't change the position */
-	USHORT usBiosRuntimeSegmentAddress;
-	USHORT usProtectedModeInfoOffset;
-	USHORT usConfigFilenameOffset;
-	USHORT usCRC_BlockOffset;
-	USHORT usBIOS_BootupMessageOffset;
-	USHORT usInt10Offset;
-	USHORT usPciBusDevInitCode;
-	USHORT usIoBaseAddress;
-	USHORT usSubsystemVendorID;
-	USHORT usSubsystemID;
-	USHORT usPCI_InfoOffset;
-	USHORT usMasterCommandTableOffset;	/*Offset for SW to get all command table offsets, Don't change the position */
-	USHORT usMasterDataTableOffset;	/*Offset for SW to get all data table offsets, Don't change the position */
-	UCHAR ucExtendedFunctionCode;
-	UCHAR ucReserved;
-} ATOM_ROM_HEADER;
+typedef struct _ATOM_COMMON_TABLE_HEADER
+{
+  USHORT usStructureSize;
+  UCHAR  ucTableFormatRevision;   /*Change it when the Parser is not backward compatible */
+  UCHAR  ucTableContentRevision;  /*Change it only when the table needs to change but the firmware */
+                                  /*Image can't be updated, while Driver needs to carry the new table! */
+}ATOM_COMMON_TABLE_HEADER;
+
+typedef struct _ATOM_ROM_HEADER
+{
+  ATOM_COMMON_TABLE_HEADER		sHeader;
+  UCHAR	 uaFirmWareSignature[4];    /*Signature to distinguish between Atombios and non-atombios, 
+                                      atombios should init it as "ATOM", don't change the position */
+  USHORT usBiosRuntimeSegmentAddress;
+  USHORT usProtectedModeInfoOffset;
+  USHORT usConfigFilenameOffset;
+  USHORT usCRC_BlockOffset;
+  USHORT usBIOS_BootupMessageOffset;
+  USHORT usInt10Offset;
+  USHORT usPciBusDevInitCode;
+  USHORT usIoBaseAddress;
+  USHORT usSubsystemVendorID;
+  USHORT usSubsystemID;
+  USHORT usPCI_InfoOffset; 
+  USHORT usMasterCommandTableOffset; /*Offset for SW to get all command table offsets, Don't change the position */
+  USHORT usMasterDataTableOffset;   /*Offset for SW to get all data table offsets, Don't change the position */
+  UCHAR  ucExtendedFunctionCode;
+  UCHAR  ucReserved;
+}ATOM_ROM_HEADER;
 
 /*==============================Command Table Portion==================================== */
 
 #ifdef	UEFI_BUILD
-#define	UTEMP	USHORT
-#define	USHORT	void*
+	#define	UTEMP	USHORT
+	#define	USHORT	void*
 #endif
 
-typedef struct _ATOM_MASTER_LIST_OF_COMMAND_TABLES {
-	USHORT ASIC_Init;	/* Function Table, used by various SW components,latest version 1.1 */
-	USHORT GetDisplaySurfaceSize;	/* Atomic Table,  Used by Bios when enabling HW ICON */
-	USHORT ASIC_RegistersInit;	/* Atomic Table,  indirectly used by various SW components,called from ASIC_Init */
-	USHORT VRAM_BlockVenderDetection;	/* Atomic Table,  used only by Bios */
-	USHORT DIGxEncoderControl;	/* Only used by Bios */
-	USHORT MemoryControllerInit;	/* Atomic Table,  indirectly used by various SW components,called from ASIC_Init */
-	USHORT EnableCRTCMemReq;	/* Function Table,directly used by various SW components,latest version 2.1 */
-	USHORT MemoryParamAdjust;	/* Atomic Table,  indirectly used by various SW components,called from SetMemoryClock if needed */
-	USHORT DVOEncoderControl;	/* Function Table,directly used by various SW components,latest version 1.2 */
-	USHORT GPIOPinControl;	/* Atomic Table,  only used by Bios */
-	USHORT SetEngineClock;	/*Function Table,directly used by various SW components,latest version 1.1 */
-	USHORT SetMemoryClock;	/* Function Table,directly used by various SW components,latest version 1.1 */
-	USHORT SetPixelClock;	/*Function Table,directly used by various SW components,latest version 1.2 */
-	USHORT DynamicClockGating;	/* Atomic Table,  indirectly used by various SW components,called from ASIC_Init */
-	USHORT ResetMemoryDLL;	/* Atomic Table,  indirectly used by various SW components,called from SetMemoryClock */
-	USHORT ResetMemoryDevice;	/* Atomic Table,  indirectly used by various SW components,called from SetMemoryClock */
-	USHORT MemoryPLLInit;
-	USHORT AdjustDisplayPll;	/* only used by Bios */
-	USHORT AdjustMemoryController;	/* Atomic Table,  indirectly used by various SW components,called from SetMemoryClock */
-	USHORT EnableASIC_StaticPwrMgt;	/* Atomic Table,  only used by Bios */
-	USHORT ASIC_StaticPwrMgtStatusChange;	/* Obsolete, only used by Bios */
-	USHORT DAC_LoadDetection;	/* Atomic Table,  directly used by various SW components,latest version 1.2 */
-	USHORT LVTMAEncoderControl;	/* Atomic Table,directly used by various SW components,latest version 1.3 */
-	USHORT LCD1OutputControl;	/* Atomic Table,  directly used by various SW components,latest version 1.1 */
-	USHORT DAC1EncoderControl;	/* Atomic Table,  directly used by various SW components,latest version 1.1 */
-	USHORT DAC2EncoderControl;	/* Atomic Table,  directly used by various SW components,latest version 1.1 */
-	USHORT DVOOutputControl;	/* Atomic Table,  directly used by various SW components,latest version 1.1 */
-	USHORT CV1OutputControl;	/* Atomic Table,  directly used by various SW components,latest version 1.1 */
-	USHORT GetConditionalGoldenSetting;	/* only used by Bios */
-	USHORT TVEncoderControl;	/* Function Table,directly used by various SW components,latest version 1.1 */
-	USHORT TMDSAEncoderControl;	/* Atomic Table,  directly used by various SW components,latest version 1.3 */
-	USHORT LVDSEncoderControl;	/* Atomic Table,  directly used by various SW components,latest version 1.3 */
-	USHORT TV1OutputControl;	/* Atomic Table,  directly used by various SW components,latest version 1.1 */
-	USHORT EnableScaler;	/* Atomic Table,  used only by Bios */
-	USHORT BlankCRTC;	/* Atomic Table,  directly used by various SW components,latest version 1.1 */
-	USHORT EnableCRTC;	/* Atomic Table,  directly used by various SW components,latest version 1.1 */
-	USHORT GetPixelClock;	/* Atomic Table,  directly used by various SW components,latest version 1.1 */
-	USHORT EnableVGA_Render;	/* Function Table,directly used by various SW components,latest version 1.1 */
-	USHORT EnableVGA_Access;	/* Obsolete ,     only used by Bios */
-	USHORT SetCRTC_Timing;	/* Atomic Table,  directly used by various SW components,latest version 1.1 */
-	USHORT SetCRTC_OverScan;	/* Atomic Table,  used by various SW components,latest version 1.1 */
-	USHORT SetCRTC_Replication;	/* Atomic Table,  used only by Bios */
-	USHORT SelectCRTC_Source;	/* Atomic Table,  directly used by various SW components,latest version 1.1 */
-	USHORT EnableGraphSurfaces;	/* Atomic Table,  used only by Bios */
-	USHORT UpdateCRTC_DoubleBufferRegisters;
-	USHORT LUT_AutoFill;	/* Atomic Table,  only used by Bios */
-	USHORT EnableHW_IconCursor;	/* Atomic Table,  only used by Bios */
-	USHORT GetMemoryClock;	/* Atomic Table,  directly used by various SW components,latest version 1.1 */
-	USHORT GetEngineClock;	/* Atomic Table,  directly used by various SW components,latest version 1.1 */
-	USHORT SetCRTC_UsingDTDTiming;	/* Atomic Table,  directly used by various SW components,latest version 1.1 */
-	USHORT ExternalEncoderControl;	/* Atomic Table,  directly used by various SW components,latest version 2.1 */
-	USHORT LVTMAOutputControl;	/* Atomic Table,  directly used by various SW components,latest version 1.1 */
-	USHORT VRAM_BlockDetectionByStrap;	/* Atomic Table,  used only by Bios */
-	USHORT MemoryCleanUp;	/* Atomic Table,  only used by Bios */
-	USHORT ProcessI2cChannelTransaction;	/* Function Table,only used by Bios */
-	USHORT WriteOneByteToHWAssistedI2C;	/* Function Table,indirectly used by various SW components */
-	USHORT ReadHWAssistedI2CStatus;	/* Atomic Table,  indirectly used by various SW components */
-	USHORT SpeedFanControl;	/* Function Table,indirectly used by various SW components,called from ASIC_Init */
-	USHORT PowerConnectorDetection;	/* Atomic Table,  directly used by various SW components,latest version 1.1 */
-	USHORT MC_Synchronization;	/* Atomic Table,  indirectly used by various SW components,called from SetMemoryClock */
-	USHORT ComputeMemoryEnginePLL;	/* Atomic Table,  indirectly used by various SW components,called from SetMemory/EngineClock */
-	USHORT MemoryRefreshConversion;	/* Atomic Table,  indirectly used by various SW components,called from SetMemory or SetEngineClock */
-	USHORT VRAM_GetCurrentInfoBlock;	/* Atomic Table,  used only by Bios */
-	USHORT DynamicMemorySettings;	/* Atomic Table,  indirectly used by various SW components,called from SetMemoryClock */
-	USHORT MemoryTraining;	/* Atomic Table,  used only by Bios */
-	USHORT EnableSpreadSpectrumOnPPLL;	/* Atomic Table,  directly used by various SW components,latest version 1.2 */
-	USHORT TMDSAOutputControl;	/* Atomic Table,  directly used by various SW components,latest version 1.1 */
-	USHORT SetVoltage;	/* Function Table,directly and/or indirectly used by various SW components,latest version 1.1 */
-	USHORT DAC1OutputControl;	/* Atomic Table,  directly used by various SW components,latest version 1.1 */
-	USHORT DAC2OutputControl;	/* Atomic Table,  directly used by various SW components,latest version 1.1 */
-	USHORT SetupHWAssistedI2CStatus;	/* Function Table,only used by Bios, obsolete soon.Switch to use "ReadEDIDFromHWAssistedI2C" */
-	USHORT ClockSource;	/* Atomic Table,  indirectly used by various SW components,called from ASIC_Init */
-	USHORT MemoryDeviceInit;	/* Atomic Table,  indirectly used by various SW components,called from SetMemoryClock */
-	USHORT EnableYUV;	/* Atomic Table,  indirectly used by various SW components,called from EnableVGARender */
-	USHORT DIG1EncoderControl;	/* Atomic Table,directly used by various SW components,latest version 1.1 */
-	USHORT DIG2EncoderControl;	/* Atomic Table,directly used by various SW components,latest version 1.1 */
-	USHORT DIG1TransmitterControl;	/* Atomic Table,directly used by various SW components,latest version 1.1 */
-	USHORT DIG2TransmitterControl;	/* Atomic Table,directly used by various SW components,latest version 1.1 */
-	USHORT ProcessAuxChannelTransaction;	/* Function Table,only used by Bios */
-	USHORT DPEncoderService;	/* Function Table,only used by Bios */
-} ATOM_MASTER_LIST_OF_COMMAND_TABLES;
-
-/*  For backward compatible */
+typedef struct _ATOM_MASTER_LIST_OF_COMMAND_TABLES{
+  USHORT ASIC_Init;                              //Function Table, used by various SW components,latest version 1.1
+  USHORT GetDisplaySurfaceSize;                  //Atomic Table,  Used by Bios when enabling HW ICON
+  USHORT ASIC_RegistersInit;                     //Atomic Table,  indirectly used by various SW components,called from ASIC_Init
+  USHORT VRAM_BlockVenderDetection;              //Atomic Table,  used only by Bios
+  USHORT DIGxEncoderControl;										 //Only used by Bios
+  USHORT MemoryControllerInit;                   //Atomic Table,  indirectly used by various SW components,called from ASIC_Init
+  USHORT EnableCRTCMemReq;                       //Function Table,directly used by various SW components,latest version 2.1
+  USHORT MemoryParamAdjust; 										 //Atomic Table,  indirectly used by various SW components,called from SetMemoryClock if needed
+  USHORT DVOEncoderControl;                      //Function Table,directly used by various SW components,latest version 1.2
+  USHORT GPIOPinControl;												 //Atomic Table,  only used by Bios
+  USHORT SetEngineClock;                         //Function Table,directly used by various SW components,latest version 1.1
+  USHORT SetMemoryClock;                         //Function Table,directly used by various SW components,latest version 1.1
+  USHORT SetPixelClock;                          //Function Table,directly used by various SW components,latest version 1.2  
+  USHORT DynamicClockGating;                     //Atomic Table,  indirectly used by various SW components,called from ASIC_Init
+  USHORT ResetMemoryDLL;                         //Atomic Table,  indirectly used by various SW components,called from SetMemoryClock
+  USHORT ResetMemoryDevice;                      //Atomic Table,  indirectly used by various SW components,called from SetMemoryClock
+  USHORT MemoryPLLInit;
+  USHORT AdjustDisplayPll;												//only used by Bios
+  USHORT AdjustMemoryController;                 //Atomic Table,  indirectly used by various SW components,called from SetMemoryClock                
+  USHORT EnableASIC_StaticPwrMgt;                //Atomic Table,  only used by Bios
+  USHORT ASIC_StaticPwrMgtStatusChange;          //Obsolete ,     only used by Bios   
+  USHORT DAC_LoadDetection;                      //Atomic Table,  directly used by various SW components,latest version 1.2  
+  USHORT LVTMAEncoderControl;                    //Atomic Table,directly used by various SW components,latest version 1.3
+  USHORT LCD1OutputControl;                      //Atomic Table,  directly used by various SW components,latest version 1.1 
+  USHORT DAC1EncoderControl;                     //Atomic Table,  directly used by various SW components,latest version 1.1  
+  USHORT DAC2EncoderControl;                     //Atomic Table,  directly used by various SW components,latest version 1.1 
+  USHORT DVOOutputControl;                       //Atomic Table,  directly used by various SW components,latest version 1.1 
+  USHORT CV1OutputControl;                       //Atomic Table,  Atomic Table,  Obsolete from Ry6xx, use DAC2 Output instead 
+  USHORT GetConditionalGoldenSetting;            //only used by Bios
+  USHORT TVEncoderControl;                       //Function Table,directly used by various SW components,latest version 1.1
+  USHORT TMDSAEncoderControl;                    //Atomic Table,  directly used by various SW components,latest version 1.3
+  USHORT LVDSEncoderControl;                     //Atomic Table,  directly used by various SW components,latest version 1.3
+  USHORT TV1OutputControl;                       //Atomic Table,  Obsolete from Ry6xx, use DAC2 Output instead
+  USHORT EnableScaler;                           //Atomic Table,  used only by Bios
+  USHORT BlankCRTC;                              //Atomic Table,  directly used by various SW components,latest version 1.1 
+  USHORT EnableCRTC;                             //Atomic Table,  directly used by various SW components,latest version 1.1 
+  USHORT GetPixelClock;                          //Atomic Table,  directly used by various SW components,latest version 1.1 
+  USHORT EnableVGA_Render;                       //Function Table,directly used by various SW components,latest version 1.1
+  USHORT GetSCLKOverMCLKRatio;                   //Atomic Table,  only used by Bios
+  USHORT SetCRTC_Timing;                         //Atomic Table,  directly used by various SW components,latest version 1.1
+  USHORT SetCRTC_OverScan;                       //Atomic Table,  used by various SW components,latest version 1.1 
+  USHORT SetCRTC_Replication;                    //Atomic Table,  used only by Bios
+  USHORT SelectCRTC_Source;                      //Atomic Table,  directly used by various SW components,latest version 1.1 
+  USHORT EnableGraphSurfaces;                    //Atomic Table,  used only by Bios
+  USHORT UpdateCRTC_DoubleBufferRegisters;
+  USHORT LUT_AutoFill;                           //Atomic Table,  only used by Bios
+  USHORT EnableHW_IconCursor;                    //Atomic Table,  only used by Bios
+  USHORT GetMemoryClock;                         //Atomic Table,  directly used by various SW components,latest version 1.1 
+  USHORT GetEngineClock;                         //Atomic Table,  directly used by various SW components,latest version 1.1 
+  USHORT SetCRTC_UsingDTDTiming;                 //Atomic Table,  directly used by various SW components,latest version 1.1
+  USHORT ExternalEncoderControl;                 //Atomic Table,  directly used by various SW components,latest version 2.1
+  USHORT LVTMAOutputControl;                     //Atomic Table,  directly used by various SW components,latest version 1.1
+  USHORT VRAM_BlockDetectionByStrap;             //Atomic Table,  used only by Bios
+  USHORT MemoryCleanUp;                          //Atomic Table,  only used by Bios    
+  USHORT ProcessI2cChannelTransaction;           //Function Table,only used by Bios
+  USHORT WriteOneByteToHWAssistedI2C;            //Function Table,indirectly used by various SW components 
+  USHORT ReadHWAssistedI2CStatus;                //Atomic Table,  indirectly used by various SW components
+  USHORT SpeedFanControl;                        //Function Table,indirectly used by various SW components,called from ASIC_Init
+  USHORT PowerConnectorDetection;                //Atomic Table,  directly used by various SW components,latest version 1.1
+  USHORT MC_Synchronization;                     //Atomic Table,  indirectly used by various SW components,called from SetMemoryClock
+  USHORT ComputeMemoryEnginePLL;                 //Atomic Table,  indirectly used by various SW components,called from SetMemory/EngineClock
+  USHORT MemoryRefreshConversion;                //Atomic Table,  indirectly used by various SW components,called from SetMemory or SetEngineClock
+  USHORT VRAM_GetCurrentInfoBlock;               //Atomic Table,  used only by Bios
+  USHORT DynamicMemorySettings;                  //Atomic Table,  indirectly used by various SW components,called from SetMemoryClock
+  USHORT MemoryTraining;                         //Atomic Table,  used only by Bios
+  USHORT EnableSpreadSpectrumOnPPLL;             //Atomic Table,  directly used by various SW components,latest version 1.2
+  USHORT TMDSAOutputControl;                     //Atomic Table,  directly used by various SW components,latest version 1.1
+  USHORT SetVoltage;                             //Function Table,directly and/or indirectly used by various SW components,latest version 1.1
+  USHORT DAC1OutputControl;                      //Atomic Table,  directly used by various SW components,latest version 1.1
+  USHORT DAC2OutputControl;                      //Atomic Table,  directly used by various SW components,latest version 1.1
+  USHORT SetupHWAssistedI2CStatus;               //Function Table,only used by Bios, obsolete soon.Switch to use "ReadEDIDFromHWAssistedI2C"
+  USHORT ClockSource;                            //Atomic Table,  indirectly used by various SW components,called from ASIC_Init
+  USHORT MemoryDeviceInit;                       //Atomic Table,  indirectly used by various SW components,called from SetMemoryClock
+  USHORT EnableYUV;                              //Atomic Table,  indirectly used by various SW components,called from EnableVGARender
+  USHORT DIG1EncoderControl;                     //Atomic Table,directly used by various SW components,latest version 1.1
+  USHORT DIG2EncoderControl;                     //Atomic Table,directly used by various SW components,latest version 1.1
+  USHORT DIG1TransmitterControl;                 //Atomic Table,directly used by various SW components,latest version 1.1
+  USHORT DIG2TransmitterControl;	               //Atomic Table,directly used by various SW components,latest version 1.1 
+  USHORT ProcessAuxChannelTransaction;					 //Function Table,only used by Bios
+  USHORT DPEncoderService;											 //Function Table,only used by Bios
+}ATOM_MASTER_LIST_OF_COMMAND_TABLES;   
+
+// For backward compatible 
 #define ReadEDIDFromHWAssistedI2C                ProcessI2cChannelTransaction
 #define UNIPHYTransmitterControl						     DIG1TransmitterControl
 #define LVTMATransmitterControl							     DIG2TransmitterControl
 #define SetCRTC_DPM_State                        GetConditionalGoldenSetting
 #define SetUniphyInstance                        ASIC_StaticPwrMgtStatusChange
+#define HPDInterruptService                      ReadHWAssistedI2CStatus
+#define EnableVGA_Access                         GetSCLKOverMCLKRatio
 
-typedef struct _ATOM_MASTER_COMMAND_TABLE {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	ATOM_MASTER_LIST_OF_COMMAND_TABLES ListOfCommandTables;
-} ATOM_MASTER_COMMAND_TABLE;
-
-/****************************************************************************/
-/*  Structures used in every command table */
-/****************************************************************************/
-typedef struct _ATOM_TABLE_ATTRIBUTE {
+typedef struct _ATOM_MASTER_COMMAND_TABLE
+{
+  ATOM_COMMON_TABLE_HEADER           sHeader;
+  ATOM_MASTER_LIST_OF_COMMAND_TABLES ListOfCommandTables;
+}ATOM_MASTER_COMMAND_TABLE;
+
+/****************************************************************************/	
+// Structures used in every command table
+/****************************************************************************/	
+typedef struct _ATOM_TABLE_ATTRIBUTE
+{
 #if ATOM_BIG_ENDIAN
-	USHORT UpdatedByUtility:1;	/* [15]=Table updated by utility flag */
-	USHORT PS_SizeInBytes:7;	/* [14:8]=Size of parameter space in Bytes (multiple of a dword), */
-	USHORT WS_SizeInBytes:8;	/* [7:0]=Size of workspace in Bytes (in multiple of a dword), */
+  USHORT  UpdatedByUtility:1;         //[15]=Table updated by utility flag
+  USHORT  PS_SizeInBytes:7;           //[14:8]=Size of parameter space in Bytes (multiple of a dword), 
+  USHORT  WS_SizeInBytes:8;           //[7:0]=Size of workspace in Bytes (in multiple of a dword), 
 #else
-	USHORT WS_SizeInBytes:8;	/* [7:0]=Size of workspace in Bytes (in multiple of a dword), */
-	USHORT PS_SizeInBytes:7;	/* [14:8]=Size of parameter space in Bytes (multiple of a dword), */
-	USHORT UpdatedByUtility:1;	/* [15]=Table updated by utility flag */
+  USHORT  WS_SizeInBytes:8;           //[7:0]=Size of workspace in Bytes (in multiple of a dword), 
+  USHORT  PS_SizeInBytes:7;           //[14:8]=Size of parameter space in Bytes (multiple of a dword), 
+  USHORT  UpdatedByUtility:1;         //[15]=Table updated by utility flag
 #endif
-} ATOM_TABLE_ATTRIBUTE;
+}ATOM_TABLE_ATTRIBUTE;
 
-typedef union _ATOM_TABLE_ATTRIBUTE_ACCESS {
-	ATOM_TABLE_ATTRIBUTE sbfAccess;
-	USHORT susAccess;
-} ATOM_TABLE_ATTRIBUTE_ACCESS;
-
-/****************************************************************************/
-/*  Common header for all command tables. */
-/*  Every table pointed by _ATOM_MASTER_COMMAND_TABLE has this common header. */
-/*  And the pointer actually points to this header. */
-/****************************************************************************/
-typedef struct _ATOM_COMMON_ROM_COMMAND_TABLE_HEADER {
-	ATOM_COMMON_TABLE_HEADER CommonHeader;
-	ATOM_TABLE_ATTRIBUTE TableAttribute;
-} ATOM_COMMON_ROM_COMMAND_TABLE_HEADER;
+typedef union _ATOM_TABLE_ATTRIBUTE_ACCESS
+{
+  ATOM_TABLE_ATTRIBUTE sbfAccess;
+  USHORT               susAccess;
+}ATOM_TABLE_ATTRIBUTE_ACCESS;
+
+/****************************************************************************/	
+// Common header for all command tables.
+// Every table pointed by _ATOM_MASTER_COMMAND_TABLE has this common header. 
+// And the pointer actually points to this header.
+/****************************************************************************/	
+typedef struct _ATOM_COMMON_ROM_COMMAND_TABLE_HEADER
+{
+  ATOM_COMMON_TABLE_HEADER CommonHeader;
+  ATOM_TABLE_ATTRIBUTE     TableAttribute;	
+}ATOM_COMMON_ROM_COMMAND_TABLE_HEADER;
 
-/****************************************************************************/
-/*  Structures used by ComputeMemoryEnginePLLTable */
-/****************************************************************************/
+/****************************************************************************/	
+// Structures used by ComputeMemoryEnginePLLTable
+/****************************************************************************/	
 #define COMPUTE_MEMORY_PLL_PARAM        1
 #define COMPUTE_ENGINE_PLL_PARAM        2
 
-typedef struct _COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS {
-	ULONG ulClock;		/* When returen, it's the re-calculated clock based on given Fb_div Post_Div and ref_div */
-	UCHAR ucAction;		/* 0:reserved //1:Memory //2:Engine */
-	UCHAR ucReserved;	/* may expand to return larger Fbdiv later */
-	UCHAR ucFbDiv;		/* return value */
-	UCHAR ucPostDiv;	/* return value */
-} COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS;
-
-typedef struct _COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS_V2 {
-	ULONG ulClock;		/* When return, [23:0] return real clock */
-	UCHAR ucAction;		/* 0:reserved;COMPUTE_MEMORY_PLL_PARAM:Memory;COMPUTE_ENGINE_PLL_PARAM:Engine. it return ref_div to be written to register */
-	USHORT usFbDiv;		/* return Feedback value to be written to register */
-	UCHAR ucPostDiv;	/* return post div to be written to register */
-} COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS_V2;
+typedef struct _COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS
+{
+  ULONG   ulClock;        //When returen, it's the re-calculated clock based on given Fb_div Post_Div and ref_div
+  UCHAR   ucAction;       //0:reserved //1:Memory //2:Engine  
+  UCHAR   ucReserved;     //may expand to return larger Fbdiv later
+  UCHAR   ucFbDiv;        //return value
+  UCHAR   ucPostDiv;      //return value
+}COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS;
+
+typedef struct _COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS_V2
+{
+  ULONG   ulClock;        //When return, [23:0] return real clock 
+  UCHAR   ucAction;       //0:reserved;COMPUTE_MEMORY_PLL_PARAM:Memory;COMPUTE_ENGINE_PLL_PARAM:Engine. it return ref_div to be written to register
+  USHORT  usFbDiv;		    //return Feedback value to be written to register
+  UCHAR   ucPostDiv;      //return post div to be written to register
+}COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS_V2;
 #define COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS_PS_ALLOCATION   COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS
 
-#define SET_CLOCK_FREQ_MASK                     0x00FFFFFF	/* Clock change tables only take bit [23:0] as the requested clock value */
-#define USE_NON_BUS_CLOCK_MASK                  0x01000000	/* Applicable to both memory and engine clock change, when set, it uses another clock as the temporary clock (engine uses memory and vice versa) */
-#define USE_MEMORY_SELF_REFRESH_MASK            0x02000000	/* Only applicable to memory clock change, when set, using memory self refresh during clock transition */
-#define SKIP_INTERNAL_MEMORY_PARAMETER_CHANGE   0x04000000	/* Only applicable to memory clock change, when set, the table will skip predefined internal memory parameter change */
-#define FIRST_TIME_CHANGE_CLOCK									0x08000000	/* Applicable to both memory and engine clock change,when set, it means this is 1st time to change clock after ASIC bootup */
-#define SKIP_SW_PROGRAM_PLL											0x10000000	/* Applicable to both memory and engine clock change, when set, it means the table will not program SPLL/MPLL */
+
+#define SET_CLOCK_FREQ_MASK                     0x00FFFFFF  //Clock change tables only take bit [23:0] as the requested clock value
+#define USE_NON_BUS_CLOCK_MASK                  0x01000000  //Applicable to both memory and engine clock change, when set, it uses another clock as the temporary clock (engine uses memory and vice versa)
+#define USE_MEMORY_SELF_REFRESH_MASK            0x02000000	//Only applicable to memory clock change, when set, using memory self refresh during clock transition
+#define SKIP_INTERNAL_MEMORY_PARAMETER_CHANGE   0x04000000  //Only applicable to memory clock change, when set, the table will skip predefined internal memory parameter change
+#define FIRST_TIME_CHANGE_CLOCK									0x08000000	//Applicable to both memory and engine clock change,when set, it means this is 1st time to change clock after ASIC bootup
+#define SKIP_SW_PROGRAM_PLL											0x10000000	//Applicable to both memory and engine clock change, when set, it means the table will not program SPLL/MPLL
 #define USE_SS_ENABLED_PIXEL_CLOCK  USE_NON_BUS_CLOCK_MASK
 
-#define b3USE_NON_BUS_CLOCK_MASK                  0x01	/* Applicable to both memory and engine clock change, when set, it uses another clock as the temporary clock (engine uses memory and vice versa) */
-#define b3USE_MEMORY_SELF_REFRESH                 0x02	/* Only applicable to memory clock change, when set, using memory self refresh during clock transition */
-#define b3SKIP_INTERNAL_MEMORY_PARAMETER_CHANGE   0x04	/* Only applicable to memory clock change, when set, the table will skip predefined internal memory parameter change */
-#define b3FIRST_TIME_CHANGE_CLOCK									0x08	/* Applicable to both memory and engine clock change,when set, it means this is 1st time to change clock after ASIC bootup */
-#define b3SKIP_SW_PROGRAM_PLL											0x10	/* Applicable to both memory and engine clock change, when set, it means the table will not program SPLL/MPLL */
+#define b3USE_NON_BUS_CLOCK_MASK                  0x01       //Applicable to both memory and engine clock change, when set, it uses another clock as the temporary clock (engine uses memory and vice versa)
+#define b3USE_MEMORY_SELF_REFRESH                 0x02	     //Only applicable to memory clock change, when set, using memory self refresh during clock transition
+#define b3SKIP_INTERNAL_MEMORY_PARAMETER_CHANGE   0x04       //Only applicable to memory clock change, when set, the table will skip predefined internal memory parameter change
+#define b3FIRST_TIME_CHANGE_CLOCK									0x08       //Applicable to both memory and engine clock change,when set, it means this is 1st time to change clock after ASIC bootup
+#define b3SKIP_SW_PROGRAM_PLL											0x10			 //Applicable to both memory and engine clock change, when set, it means the table will not program SPLL/MPLL
 
-typedef struct _ATOM_COMPUTE_CLOCK_FREQ {
+typedef struct _ATOM_COMPUTE_CLOCK_FREQ
+{
 #if ATOM_BIG_ENDIAN
-	ULONG ulComputeClockFlag:8;	/*  =1: COMPUTE_MEMORY_PLL_PARAM, =2: COMPUTE_ENGINE_PLL_PARAM */
-	ULONG ulClockFreq:24;	/*  in unit of 10kHz */
+  ULONG ulComputeClockFlag:8;                 // =1: COMPUTE_MEMORY_PLL_PARAM, =2: COMPUTE_ENGINE_PLL_PARAM
+  ULONG ulClockFreq:24;                       // in unit of 10kHz
 #else
-	ULONG ulClockFreq:24;	/*  in unit of 10kHz */
-	ULONG ulComputeClockFlag:8;	/*  =1: COMPUTE_MEMORY_PLL_PARAM, =2: COMPUTE_ENGINE_PLL_PARAM */
+  ULONG ulClockFreq:24;                       // in unit of 10kHz
+  ULONG ulComputeClockFlag:8;                 // =1: COMPUTE_MEMORY_PLL_PARAM, =2: COMPUTE_ENGINE_PLL_PARAM
 #endif
-} ATOM_COMPUTE_CLOCK_FREQ;
+}ATOM_COMPUTE_CLOCK_FREQ;
 
-typedef struct _ATOM_S_MPLL_FB_DIVIDER {
-	USHORT usFbDivFrac;
-	USHORT usFbDiv;
-} ATOM_S_MPLL_FB_DIVIDER;
-
-typedef struct _COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS_V3 {
-	union {
-		ATOM_COMPUTE_CLOCK_FREQ ulClock;	/* Input Parameter */
-		ATOM_S_MPLL_FB_DIVIDER ulFbDiv;	/* Output Parameter */
-	};
-	UCHAR ucRefDiv;		/* Output Parameter */
-	UCHAR ucPostDiv;	/* Output Parameter */
-	UCHAR ucCntlFlag;	/* Output Parameter */
-	UCHAR ucReserved;
-} COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS_V3;
+typedef struct _ATOM_S_MPLL_FB_DIVIDER
+{
+  USHORT usFbDivFrac;  
+  USHORT usFbDiv;  
+}ATOM_S_MPLL_FB_DIVIDER;
 
-/*  ucCntlFlag */
+typedef struct _COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS_V3
+{
+  union
+  {
+    ATOM_COMPUTE_CLOCK_FREQ  ulClock;         //Input Parameter
+    ATOM_S_MPLL_FB_DIVIDER   ulFbDiv;         //Output Parameter
+  };
+  UCHAR   ucRefDiv;                           //Output Parameter      
+  UCHAR   ucPostDiv;                          //Output Parameter      
+  UCHAR   ucCntlFlag;                         //Output Parameter      
+  UCHAR   ucReserved;
+}COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS_V3;
+
+// ucCntlFlag
 #define ATOM_PLL_CNTL_FLAG_PLL_POST_DIV_EN          1
 #define ATOM_PLL_CNTL_FLAG_MPLL_VCO_MODE            2
 #define ATOM_PLL_CNTL_FLAG_FRACTION_DISABLE         4
+#define ATOM_PLL_CNTL_FLAG_SPLL_ISPARE_9						8
 
-typedef struct _DYNAMICE_MEMORY_SETTINGS_PARAMETER {
-	ATOM_COMPUTE_CLOCK_FREQ ulClock;
-	ULONG ulReserved[2];
-} DYNAMICE_MEMORY_SETTINGS_PARAMETER;
-
-typedef struct _DYNAMICE_ENGINE_SETTINGS_PARAMETER {
-	ATOM_COMPUTE_CLOCK_FREQ ulClock;
-	ULONG ulMemoryClock;
-	ULONG ulReserved;
-} DYNAMICE_ENGINE_SETTINGS_PARAMETER;
 
-/****************************************************************************/
-/*  Structures used by SetEngineClockTable */
-/****************************************************************************/
-typedef struct _SET_ENGINE_CLOCK_PARAMETERS {
-	ULONG ulTargetEngineClock;	/* In 10Khz unit */
-} SET_ENGINE_CLOCK_PARAMETERS;
-
-typedef struct _SET_ENGINE_CLOCK_PS_ALLOCATION {
-	ULONG ulTargetEngineClock;	/* In 10Khz unit */
-	COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS_PS_ALLOCATION sReserved;
-} SET_ENGINE_CLOCK_PS_ALLOCATION;
+// V4 are only used for APU which PLL outside GPU
+typedef struct _COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS_V4
+{
+#if ATOM_BIG_ENDIAN
+  ULONG  ucPostDiv;          //return parameter: post divider which is used to program to register directly
+  ULONG  ulClock:24;         //Input= target clock, output = actual clock 
+#else
+  ULONG  ulClock:24;         //Input= target clock, output = actual clock 
+  ULONG  ucPostDiv;          //return parameter: post divider which is used to program to register directly
+#endif
+}COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS_V4;
 
-/****************************************************************************/
-/*  Structures used by SetMemoryClockTable */
-/****************************************************************************/
-typedef struct _SET_MEMORY_CLOCK_PARAMETERS {
-	ULONG ulTargetMemoryClock;	/* In 10Khz unit */
-} SET_MEMORY_CLOCK_PARAMETERS;
+typedef struct _DYNAMICE_MEMORY_SETTINGS_PARAMETER
+{
+  ATOM_COMPUTE_CLOCK_FREQ ulClock;
+  ULONG ulReserved[2];
+}DYNAMICE_MEMORY_SETTINGS_PARAMETER;
 
-typedef struct _SET_MEMORY_CLOCK_PS_ALLOCATION {
-	ULONG ulTargetMemoryClock;	/* In 10Khz unit */
-	COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS_PS_ALLOCATION sReserved;
-} SET_MEMORY_CLOCK_PS_ALLOCATION;
+typedef struct _DYNAMICE_ENGINE_SETTINGS_PARAMETER
+{
+  ATOM_COMPUTE_CLOCK_FREQ ulClock;
+  ULONG ulMemoryClock;
+  ULONG ulReserved;
+}DYNAMICE_ENGINE_SETTINGS_PARAMETER;
+
+/****************************************************************************/	
+// Structures used by SetEngineClockTable
+/****************************************************************************/	
+typedef struct _SET_ENGINE_CLOCK_PARAMETERS
+{
+  ULONG ulTargetEngineClock;          //In 10Khz unit
+}SET_ENGINE_CLOCK_PARAMETERS;
 
-/****************************************************************************/
-/*  Structures used by ASIC_Init.ctb */
-/****************************************************************************/
-typedef struct _ASIC_INIT_PARAMETERS {
-	ULONG ulDefaultEngineClock;	/* In 10Khz unit */
-	ULONG ulDefaultMemoryClock;	/* In 10Khz unit */
-} ASIC_INIT_PARAMETERS;
+typedef struct _SET_ENGINE_CLOCK_PS_ALLOCATION
+{
+  ULONG ulTargetEngineClock;          //In 10Khz unit
+  COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS_PS_ALLOCATION sReserved;
+}SET_ENGINE_CLOCK_PS_ALLOCATION;
+
+/****************************************************************************/	
+// Structures used by SetMemoryClockTable
+/****************************************************************************/	
+typedef struct _SET_MEMORY_CLOCK_PARAMETERS
+{
+  ULONG ulTargetMemoryClock;          //In 10Khz unit
+}SET_MEMORY_CLOCK_PARAMETERS;
 
-typedef struct _ASIC_INIT_PS_ALLOCATION {
-	ASIC_INIT_PARAMETERS sASICInitClocks;
-	SET_ENGINE_CLOCK_PS_ALLOCATION sReserved;	/* Caller doesn't need to init this structure */
-} ASIC_INIT_PS_ALLOCATION;
+typedef struct _SET_MEMORY_CLOCK_PS_ALLOCATION
+{
+  ULONG ulTargetMemoryClock;          //In 10Khz unit
+  COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS_PS_ALLOCATION sReserved;
+}SET_MEMORY_CLOCK_PS_ALLOCATION;
+
+/****************************************************************************/	
+// Structures used by ASIC_Init.ctb
+/****************************************************************************/	
+typedef struct _ASIC_INIT_PARAMETERS
+{
+  ULONG ulDefaultEngineClock;         //In 10Khz unit
+  ULONG ulDefaultMemoryClock;         //In 10Khz unit
+}ASIC_INIT_PARAMETERS;
 
-/****************************************************************************/
-/*  Structure used by DynamicClockGatingTable.ctb */
-/****************************************************************************/
-typedef struct _DYNAMIC_CLOCK_GATING_PARAMETERS {
-	UCHAR ucEnable;		/*  ATOM_ENABLE or ATOM_DISABLE */
-	UCHAR ucPadding[3];
-} DYNAMIC_CLOCK_GATING_PARAMETERS;
+typedef struct _ASIC_INIT_PS_ALLOCATION
+{
+  ASIC_INIT_PARAMETERS sASICInitClocks;
+  SET_ENGINE_CLOCK_PS_ALLOCATION sReserved; //Caller doesn't need to init this structure
+}ASIC_INIT_PS_ALLOCATION;
+
+/****************************************************************************/	
+// Structure used by DynamicClockGatingTable.ctb
+/****************************************************************************/	
+typedef struct _DYNAMIC_CLOCK_GATING_PARAMETERS 
+{
+  UCHAR ucEnable;                     // ATOM_ENABLE or ATOM_DISABLE
+  UCHAR ucPadding[3];
+}DYNAMIC_CLOCK_GATING_PARAMETERS;
 #define  DYNAMIC_CLOCK_GATING_PS_ALLOCATION  DYNAMIC_CLOCK_GATING_PARAMETERS
 
-/****************************************************************************/
-/*  Structure used by EnableASIC_StaticPwrMgtTable.ctb */
-/****************************************************************************/
-typedef struct _ENABLE_ASIC_STATIC_PWR_MGT_PARAMETERS {
-	UCHAR ucEnable;		/*  ATOM_ENABLE or ATOM_DISABLE */
-	UCHAR ucPadding[3];
-} ENABLE_ASIC_STATIC_PWR_MGT_PARAMETERS;
+/****************************************************************************/	
+// Structure used by EnableASIC_StaticPwrMgtTable.ctb
+/****************************************************************************/	
+typedef struct _ENABLE_ASIC_STATIC_PWR_MGT_PARAMETERS
+{
+  UCHAR ucEnable;                     // ATOM_ENABLE or ATOM_DISABLE
+  UCHAR ucPadding[3];
+}ENABLE_ASIC_STATIC_PWR_MGT_PARAMETERS;
 #define ENABLE_ASIC_STATIC_PWR_MGT_PS_ALLOCATION  ENABLE_ASIC_STATIC_PWR_MGT_PARAMETERS
 
-/****************************************************************************/
-/*  Structures used by DAC_LoadDetectionTable.ctb */
-/****************************************************************************/
-typedef struct _DAC_LOAD_DETECTION_PARAMETERS {
-	USHORT usDeviceID;	/* {ATOM_DEVICE_CRTx_SUPPORT,ATOM_DEVICE_TVx_SUPPORT,ATOM_DEVICE_CVx_SUPPORT} */
-	UCHAR ucDacType;	/* {ATOM_DAC_A,ATOM_DAC_B, ATOM_EXT_DAC} */
-	UCHAR ucMisc;		/* Valid only when table revision =1.3 and above */
-} DAC_LOAD_DETECTION_PARAMETERS;
+/****************************************************************************/	
+// Structures used by DAC_LoadDetectionTable.ctb
+/****************************************************************************/	
+typedef struct _DAC_LOAD_DETECTION_PARAMETERS
+{
+  USHORT usDeviceID;                  //{ATOM_DEVICE_CRTx_SUPPORT,ATOM_DEVICE_TVx_SUPPORT,ATOM_DEVICE_CVx_SUPPORT}
+  UCHAR  ucDacType;                   //{ATOM_DAC_A,ATOM_DAC_B, ATOM_EXT_DAC}
+  UCHAR  ucMisc;											//Valid only when table revision =1.3 and above
+}DAC_LOAD_DETECTION_PARAMETERS;
 
-/*  DAC_LOAD_DETECTION_PARAMETERS.ucMisc */
+// DAC_LOAD_DETECTION_PARAMETERS.ucMisc
 #define DAC_LOAD_MISC_YPrPb						0x01
 
-typedef struct _DAC_LOAD_DETECTION_PS_ALLOCATION {
-	DAC_LOAD_DETECTION_PARAMETERS sDacload;
-	ULONG Reserved[2];	/*  Don't set this one, allocation for EXT DAC */
-} DAC_LOAD_DETECTION_PS_ALLOCATION;
-
-/****************************************************************************/
-/*  Structures used by DAC1EncoderControlTable.ctb and DAC2EncoderControlTable.ctb */
-/****************************************************************************/
-typedef struct _DAC_ENCODER_CONTROL_PARAMETERS {
-	USHORT usPixelClock;	/*  in 10KHz; for bios convenient */
-	UCHAR ucDacStandard;	/*  See definition of ATOM_DACx_xxx, For DEC3.0, bit 7 used as internal flag to indicate DAC2 (==1) or DAC1 (==0) */
-	UCHAR ucAction;		/*  0: turn off encoder */
-	/*  1: setup and turn on encoder */
-	/*  7: ATOM_ENCODER_INIT Initialize DAC */
-} DAC_ENCODER_CONTROL_PARAMETERS;
+typedef struct _DAC_LOAD_DETECTION_PS_ALLOCATION
+{
+  DAC_LOAD_DETECTION_PARAMETERS            sDacload;
+  ULONG                                    Reserved[2];// Don't set this one, allocation for EXT DAC
+}DAC_LOAD_DETECTION_PS_ALLOCATION;
+
+/****************************************************************************/	
+// Structures used by DAC1EncoderControlTable.ctb and DAC2EncoderControlTable.ctb
+/****************************************************************************/	
+typedef struct _DAC_ENCODER_CONTROL_PARAMETERS 
+{
+  USHORT usPixelClock;                // in 10KHz; for bios convenient
+  UCHAR  ucDacStandard;               // See definition of ATOM_DACx_xxx, For DEC3.0, bit 7 used as internal flag to indicate DAC2 (==1) or DAC1 (==0)
+  UCHAR  ucAction;                    // 0: turn off encoder
+                                      // 1: setup and turn on encoder
+                                      // 7: ATOM_ENCODER_INIT Initialize DAC
+}DAC_ENCODER_CONTROL_PARAMETERS;
 
 #define DAC_ENCODER_CONTROL_PS_ALLOCATION  DAC_ENCODER_CONTROL_PARAMETERS
 
-/****************************************************************************/
-/*  Structures used by DIG1EncoderControlTable */
-/*                     DIG2EncoderControlTable */
-/*                     ExternalEncoderControlTable */
-/****************************************************************************/
-typedef struct _DIG_ENCODER_CONTROL_PARAMETERS {
-	USHORT usPixelClock;	/*  in 10KHz; for bios convenient */
-	UCHAR ucConfig;
-	/*  [2] Link Select: */
-	/*  =0: PHY linkA if bfLane<3 */
-	/*  =1: PHY linkB if bfLanes<3 */
-	/*  =0: PHY linkA+B if bfLanes=3 */
-	/*  [3] Transmitter Sel */
-	/*  =0: UNIPHY or PCIEPHY */
-	/*  =1: LVTMA */
-	UCHAR ucAction;		/*  =0: turn off encoder */
-	/*  =1: turn on encoder */
-	UCHAR ucEncoderMode;
-	/*  =0: DP   encoder */
-	/*  =1: LVDS encoder */
-	/*  =2: DVI  encoder */
-	/*  =3: HDMI encoder */
-	/*  =4: SDVO encoder */
-	UCHAR ucLaneNum;	/*  how many lanes to enable */
-	UCHAR ucReserved[2];
-} DIG_ENCODER_CONTROL_PARAMETERS;
+/****************************************************************************/	
+// Structures used by DIG1EncoderControlTable
+//                    DIG2EncoderControlTable
+//                    ExternalEncoderControlTable
+/****************************************************************************/	
+typedef struct _DIG_ENCODER_CONTROL_PARAMETERS
+{
+  USHORT usPixelClock;		// in 10KHz; for bios convenient
+  UCHAR  ucConfig;		  
+                            // [2] Link Select:
+                            // =0: PHY linkA if bfLane<3
+                            // =1: PHY linkB if bfLanes<3
+                            // =0: PHY linkA+B if bfLanes=3
+                            // [3] Transmitter Sel
+                            // =0: UNIPHY or PCIEPHY
+                            // =1: LVTMA 					
+  UCHAR ucAction;           // =0: turn off encoder					
+                            // =1: turn on encoder			
+  UCHAR ucEncoderMode;
+                            // =0: DP   encoder      
+                            // =1: LVDS encoder          
+                            // =2: DVI  encoder  
+                            // =3: HDMI encoder
+                            // =4: SDVO encoder
+  UCHAR ucLaneNum;          // how many lanes to enable
+  UCHAR ucReserved[2];
+}DIG_ENCODER_CONTROL_PARAMETERS;
 #define DIG_ENCODER_CONTROL_PS_ALLOCATION			  DIG_ENCODER_CONTROL_PARAMETERS
 #define EXTERNAL_ENCODER_CONTROL_PARAMETER			DIG_ENCODER_CONTROL_PARAMETERS
 
-/* ucConfig */
+//ucConfig
 #define ATOM_ENCODER_CONFIG_DPLINKRATE_MASK				0x01
 #define ATOM_ENCODER_CONFIG_DPLINKRATE_1_62GHZ		0x00
 #define ATOM_ENCODER_CONFIG_DPLINKRATE_2_70GHZ		0x01
@@ -539,52 +593,57 @@ typedef struct _DIG_ENCODER_CONTROL_PARAMETERS {
 #define ATOM_ENCODER_CONFIG_LVTMA								  0x08
 #define ATOM_ENCODER_CONFIG_TRANSMITTER1				  0x00
 #define ATOM_ENCODER_CONFIG_TRANSMITTER2				  0x08
-#define ATOM_ENCODER_CONFIG_DIGB								  0x80	/*  VBIOS Internal use, outside SW should set this bit=0 */
-/*  ucAction */
-/*  ATOM_ENABLE:  Enable Encoder */
-/*  ATOM_DISABLE: Disable Encoder */
+#define ATOM_ENCODER_CONFIG_DIGB								  0x80			// VBIOS Internal use, outside SW should set this bit=0
+// ucAction
+// ATOM_ENABLE:  Enable Encoder
+// ATOM_DISABLE: Disable Encoder
 
-/* ucEncoderMode */
+//ucEncoderMode
 #define ATOM_ENCODER_MODE_DP											0
 #define ATOM_ENCODER_MODE_LVDS										1
 #define ATOM_ENCODER_MODE_DVI											2
 #define ATOM_ENCODER_MODE_HDMI										3
 #define ATOM_ENCODER_MODE_SDVO										4
+#define ATOM_ENCODER_MODE_DP_AUDIO                5
 #define ATOM_ENCODER_MODE_TV											13
 #define ATOM_ENCODER_MODE_CV											14
 #define ATOM_ENCODER_MODE_CRT											15
 
-typedef struct _ATOM_DIG_ENCODER_CONFIG_V2 {
+typedef struct _ATOM_DIG_ENCODER_CONFIG_V2
+{
 #if ATOM_BIG_ENDIAN
-	UCHAR ucReserved1:2;
-	UCHAR ucTransmitterSel:2;	/*  =0: UniphyAB, =1: UniphyCD  =2: UniphyEF */
-	UCHAR ucLinkSel:1;	/*  =0: linkA/C/E =1: linkB/D/F */
-	UCHAR ucReserved:1;
-	UCHAR ucDPLinkRate:1;	/*  =0: 1.62Ghz, =1: 2.7Ghz */
+    UCHAR ucReserved1:2;
+    UCHAR ucTransmitterSel:2;     // =0: UniphyAB, =1: UniphyCD  =2: UniphyEF
+    UCHAR ucLinkSel:1;            // =0: linkA/C/E =1: linkB/D/F
+    UCHAR ucReserved:1;
+    UCHAR ucDPLinkRate:1;         // =0: 1.62Ghz, =1: 2.7Ghz
 #else
-	UCHAR ucDPLinkRate:1;	/*  =0: 1.62Ghz, =1: 2.7Ghz */
-	UCHAR ucReserved:1;
-	UCHAR ucLinkSel:1;	/*  =0: linkA/C/E =1: linkB/D/F */
-	UCHAR ucTransmitterSel:2;	/*  =0: UniphyAB, =1: UniphyCD  =2: UniphyEF */
-	UCHAR ucReserved1:2;
+    UCHAR ucDPLinkRate:1;         // =0: 1.62Ghz, =1: 2.7Ghz
+    UCHAR ucReserved:1;
+    UCHAR ucLinkSel:1;            // =0: linkA/C/E =1: linkB/D/F
+    UCHAR ucTransmitterSel:2;     // =0: UniphyAB, =1: UniphyCD  =2: UniphyEF
+    UCHAR ucReserved1:2;
 #endif
-} ATOM_DIG_ENCODER_CONFIG_V2;
+}ATOM_DIG_ENCODER_CONFIG_V2;
 
-typedef struct _DIG_ENCODER_CONTROL_PARAMETERS_V2 {
-	USHORT usPixelClock;	/*  in 10KHz; for bios convenient */
-	ATOM_DIG_ENCODER_CONFIG_V2 acConfig;
-	UCHAR ucAction;
-	UCHAR ucEncoderMode;
-	/*  =0: DP   encoder */
-	/*  =1: LVDS encoder */
-	/*  =2: DVI  encoder */
-	/*  =3: HDMI encoder */
-	/*  =4: SDVO encoder */
-	UCHAR ucLaneNum;	/*  how many lanes to enable */
-	UCHAR ucReserved[2];
-} DIG_ENCODER_CONTROL_PARAMETERS_V2;
 
-/* ucConfig */
+typedef struct _DIG_ENCODER_CONTROL_PARAMETERS_V2
+{
+  USHORT usPixelClock;      // in 10KHz; for bios convenient
+  ATOM_DIG_ENCODER_CONFIG_V2 acConfig;
+  UCHAR ucAction;                                       
+  UCHAR ucEncoderMode;
+                            // =0: DP   encoder      
+                            // =1: LVDS encoder          
+                            // =2: DVI  encoder  
+                            // =3: HDMI encoder
+                            // =4: SDVO encoder
+  UCHAR ucLaneNum;          // how many lanes to enable
+  UCHAR ucStatus;           // = DP_LINK_TRAINING_COMPLETE or DP_LINK_TRAINING_INCOMPLETE, only used by VBIOS with command ATOM_ENCODER_CMD_QUERY_DP_LINK_TRAINING_STATUS
+  UCHAR ucReserved;
+}DIG_ENCODER_CONTROL_PARAMETERS_V2;
+
+//ucConfig
 #define ATOM_ENCODER_CONFIG_V2_DPLINKRATE_MASK				0x01
 #define ATOM_ENCODER_CONFIG_V2_DPLINKRATE_1_62GHZ		  0x00
 #define ATOM_ENCODER_CONFIG_V2_DPLINKRATE_2_70GHZ		  0x01
@@ -596,58 +655,122 @@ typedef struct _DIG_ENCODER_CONTROL_PARAMETERS_V2 {
 #define ATOM_ENCODER_CONFIG_V2_TRANSMITTER2				    0x08
 #define ATOM_ENCODER_CONFIG_V2_TRANSMITTER3				    0x10
 
-/****************************************************************************/
-/*  Structures used by UNIPHYTransmitterControlTable */
-/*                     LVTMATransmitterControlTable */
-/*                     DVOOutputControlTable */
-/****************************************************************************/
-typedef struct _ATOM_DP_VS_MODE {
-	UCHAR ucLaneSel;
-	UCHAR ucLaneSet;
-} ATOM_DP_VS_MODE;
-
-typedef struct _DIG_TRANSMITTER_CONTROL_PARAMETERS {
-	union {
-		USHORT usPixelClock;	/*  in 10KHz; for bios convenient */
-		USHORT usInitInfo;	/*  when init uniphy,lower 8bit is used for connector type defined in objectid.h */
-		ATOM_DP_VS_MODE asMode;	/*  DP Voltage swing mode */
+// ucAction:
+// ATOM_DISABLE
+// ATOM_ENABLE
+#define ATOM_ENCODER_CMD_DP_LINK_TRAINING_START       0x08
+#define ATOM_ENCODER_CMD_DP_LINK_TRAINING_PATTERN1    0x09
+#define ATOM_ENCODER_CMD_DP_LINK_TRAINING_PATTERN2    0x0a
+#define ATOM_ENCODER_CMD_DP_LINK_TRAINING_COMPLETE    0x0b
+#define ATOM_ENCODER_CMD_DP_VIDEO_OFF                 0x0c
+#define ATOM_ENCODER_CMD_DP_VIDEO_ON                  0x0d
+#define ATOM_ENCODER_CMD_QUERY_DP_LINK_TRAINING_STATUS    0x0e
+#define ATOM_ENCODER_CMD_SETUP                        0x0f
+
+// ucStatus
+#define ATOM_ENCODER_STATUS_LINK_TRAINING_COMPLETE    0x10
+#define ATOM_ENCODER_STATUS_LINK_TRAINING_INCOMPLETE  0x00
+
+// Following function ENABLE sub-function will be used by driver when TMDS/HDMI/LVDS is used, disable function will be used by driver
+typedef struct _ATOM_DIG_ENCODER_CONFIG_V3
+{
+#if ATOM_BIG_ENDIAN
+    UCHAR ucReserved1:1;
+    UCHAR ucDigSel:3;             // =0: DIGA/B/C/D/E/F
+    UCHAR ucReserved:3;
+    UCHAR ucDPLinkRate:1;         // =0: 1.62Ghz, =1: 2.7Ghz
+#else
+    UCHAR ucDPLinkRate:1;         // =0: 1.62Ghz, =1: 2.7Ghz
+    UCHAR ucReserved:3;
+    UCHAR ucDigSel:3;             // =0: DIGA/B/C/D/E/F
+    UCHAR ucReserved1:1;
+#endif
+}ATOM_DIG_ENCODER_CONFIG_V3;
+
+#define ATOM_ENCODER_CONFIG_V3_ENCODER_SEL					  0x70
+
+
+typedef struct _DIG_ENCODER_CONTROL_PARAMETERS_V3
+{
+  USHORT usPixelClock;      // in 10KHz; for bios convenient
+  ATOM_DIG_ENCODER_CONFIG_V3 acConfig;
+  UCHAR ucAction;                              
+  UCHAR ucEncoderMode;
+                            // =0: DP   encoder      
+                            // =1: LVDS encoder          
+                            // =2: DVI  encoder  
+                            // =3: HDMI encoder
+                            // =4: SDVO encoder
+                            // =5: DP audio
+  UCHAR ucLaneNum;          // how many lanes to enable
+  UCHAR ucBitPerColor;      // only valid for DP mode when ucAction = ATOM_ENCODER_CMD_SETUP
+  UCHAR ucReserved;
+}DIG_ENCODER_CONTROL_PARAMETERS_V3;
+
+
+// define ucBitPerColor: 
+#define PANEL_BPC_UNDEFINE                               0x00
+#define PANEL_6BIT_PER_COLOR                             0x01 
+#define PANEL_8BIT_PER_COLOR                             0x02
+#define PANEL_10BIT_PER_COLOR                            0x03
+#define PANEL_12BIT_PER_COLOR                            0x04
+#define PANEL_16BIT_PER_COLOR                            0x05
+
+/****************************************************************************/	
+// Structures used by UNIPHYTransmitterControlTable
+//                    LVTMATransmitterControlTable
+//                    DVOOutputControlTable
+/****************************************************************************/	
+typedef struct _ATOM_DP_VS_MODE
+{
+  UCHAR ucLaneSel;
+  UCHAR ucLaneSet;
+}ATOM_DP_VS_MODE;
+
+typedef struct _DIG_TRANSMITTER_CONTROL_PARAMETERS
+{
+	union
+	{
+  USHORT usPixelClock;		// in 10KHz; for bios convenient
+	USHORT usInitInfo;			// when init uniphy,lower 8bit is used for connector type defined in objectid.h
+  ATOM_DP_VS_MODE asMode; // DP Voltage swing mode
 	};
-	UCHAR ucConfig;
-	/*  [0]=0: 4 lane Link, */
-	/*     =1: 8 lane Link ( Dual Links TMDS ) */
-	/*  [1]=0: InCoherent mode */
-	/*     =1: Coherent Mode */
-	/*  [2] Link Select: */
-	/*  =0: PHY linkA   if bfLane<3 */
-	/*  =1: PHY linkB   if bfLanes<3 */
-	/*  =0: PHY linkA+B if bfLanes=3 */
-	/*  [5:4]PCIE lane Sel */
-	/*  =0: lane 0~3 or 0~7 */
-	/*  =1: lane 4~7 */
-	/*  =2: lane 8~11 or 8~15 */
-	/*  =3: lane 12~15 */
-	UCHAR ucAction;		/*  =0: turn off encoder */
-	/*  =1: turn on encoder */
-	UCHAR ucReserved[4];
-} DIG_TRANSMITTER_CONTROL_PARAMETERS;
-
-#define DIG_TRANSMITTER_CONTROL_PS_ALLOCATION		DIG_TRANSMITTER_CONTROL_PARAMETERS
-
-/* ucInitInfo */
-#define ATOM_TRAMITTER_INITINFO_CONNECTOR_MASK	0x00ff
-
-/* ucConfig */
+  UCHAR ucConfig;
+													// [0]=0: 4 lane Link,      
+													//    =1: 8 lane Link ( Dual Links TMDS ) 
+                          // [1]=0: InCoherent mode   
+													//    =1: Coherent Mode										
+													// [2] Link Select:
+  												// =0: PHY linkA   if bfLane<3
+													// =1: PHY linkB   if bfLanes<3
+		  										// =0: PHY linkA+B if bfLanes=3		
+                          // [5:4]PCIE lane Sel
+                          // =0: lane 0~3 or 0~7
+                          // =1: lane 4~7
+                          // =2: lane 8~11 or 8~15
+                          // =3: lane 12~15 
+	UCHAR ucAction;				  // =0: turn off encoder					
+	                        // =1: turn on encoder			
+  UCHAR ucReserved[4];
+}DIG_TRANSMITTER_CONTROL_PARAMETERS;
+
+#define DIG_TRANSMITTER_CONTROL_PS_ALLOCATION		DIG_TRANSMITTER_CONTROL_PARAMETERS					
+
+//ucInitInfo
+#define ATOM_TRAMITTER_INITINFO_CONNECTOR_MASK	0x00ff			
+
+//ucConfig 
 #define ATOM_TRANSMITTER_CONFIG_8LANE_LINK			0x01
 #define ATOM_TRANSMITTER_CONFIG_COHERENT				0x02
 #define ATOM_TRANSMITTER_CONFIG_LINK_SEL_MASK		0x04
 #define ATOM_TRANSMITTER_CONFIG_LINKA						0x00
 #define ATOM_TRANSMITTER_CONFIG_LINKB						0x04
-#define ATOM_TRANSMITTER_CONFIG_LINKA_B					0x00
+#define ATOM_TRANSMITTER_CONFIG_LINKA_B					0x00			
 #define ATOM_TRANSMITTER_CONFIG_LINKB_A					0x04
 
-#define ATOM_TRANSMITTER_CONFIG_ENCODER_SEL_MASK	0x08	/*  only used when ATOM_TRANSMITTER_ACTION_ENABLE */
-#define ATOM_TRANSMITTER_CONFIG_DIG1_ENCODER		0x00	/*  only used when ATOM_TRANSMITTER_ACTION_ENABLE */
-#define ATOM_TRANSMITTER_CONFIG_DIG2_ENCODER		0x08	/*  only used when ATOM_TRANSMITTER_ACTION_ENABLE */
+#define ATOM_TRANSMITTER_CONFIG_ENCODER_SEL_MASK	0x08			// only used when ATOM_TRANSMITTER_ACTION_ENABLE
+#define ATOM_TRANSMITTER_CONFIG_DIG1_ENCODER		0x00				// only used when ATOM_TRANSMITTER_ACTION_ENABLE
+#define ATOM_TRANSMITTER_CONFIG_DIG2_ENCODER		0x08				// only used when ATOM_TRANSMITTER_ACTION_ENABLE
 
 #define ATOM_TRANSMITTER_CONFIG_CLKSRC_MASK			0x30
 #define ATOM_TRANSMITTER_CONFIG_CLKSRC_PPLL			0x00
@@ -661,7 +784,7 @@ typedef struct _DIG_TRANSMITTER_CONTROL_PARAMETERS {
 #define ATOM_TRANSMITTER_CONFIG_LANE_8_15				0x80
 #define ATOM_TRANSMITTER_CONFIG_LANE_12_15			0xc0
 
-/* ucAction */
+//ucAction
 #define ATOM_TRANSMITTER_ACTION_DISABLE					       0
 #define ATOM_TRANSMITTER_ACTION_ENABLE					       1
 #define ATOM_TRANSMITTER_ACTION_LCD_BLOFF				       2
@@ -674,93 +797,168 @@ typedef struct _DIG_TRANSMITTER_CONTROL_PARAMETERS {
 #define ATOM_TRANSMITTER_ACTION_ENABLE_OUTPUT		       9
 #define ATOM_TRANSMITTER_ACTION_SETUP						       10
 #define ATOM_TRANSMITTER_ACTION_SETUP_VSEMPH           11
+#define ATOM_TRANSMITTER_ACTION_POWER_ON               12
+#define ATOM_TRANSMITTER_ACTION_POWER_OFF              13
 
-/*  Following are used for DigTransmitterControlTable ver1.2 */
-typedef struct _ATOM_DIG_TRANSMITTER_CONFIG_V2 {
+// Following are used for DigTransmitterControlTable ver1.2
+typedef struct _ATOM_DIG_TRANSMITTER_CONFIG_V2
+{
 #if ATOM_BIG_ENDIAN
-	UCHAR ucTransmitterSel:2;	/* bit7:6: =0 Dig Transmitter 1 ( Uniphy AB ) */
-	/*         =1 Dig Transmitter 2 ( Uniphy CD ) */
-	/*         =2 Dig Transmitter 3 ( Uniphy EF ) */
-	UCHAR ucReserved:1;
-	UCHAR fDPConnector:1;	/* bit4=0: DP connector  =1: None DP connector */
-	UCHAR ucEncoderSel:1;	/* bit3=0: Data/Clk path source from DIGA( DIG inst0 ). =1: Data/clk path source from DIGB ( DIG inst1 ) */
-	UCHAR ucLinkSel:1;	/* bit2=0: Uniphy LINKA or C or E when fDualLinkConnector=0. when fDualLinkConnector=1, it means master link of dual link is A or C or E */
-	/*     =1: Uniphy LINKB or D or F when fDualLinkConnector=0. when fDualLinkConnector=1, it means master link of dual link is B or D or F */
-
-	UCHAR fCoherentMode:1;	/* bit1=1: Coherent Mode ( for DVI/HDMI mode ) */
-	UCHAR fDualLinkConnector:1;	/* bit0=1: Dual Link DVI connector */
+  UCHAR ucTransmitterSel:2;         //bit7:6: =0 Dig Transmitter 1 ( Uniphy AB )
+                                    //        =1 Dig Transmitter 2 ( Uniphy CD )
+                                    //        =2 Dig Transmitter 3 ( Uniphy EF )
+  UCHAR ucReserved:1;               
+  UCHAR fDPConnector:1;             //bit4=0: DP connector  =1: None DP connector
+  UCHAR ucEncoderSel:1;             //bit3=0: Data/Clk path source from DIGA( DIG inst0 ). =1: Data/clk path source from DIGB ( DIG inst1 )
+  UCHAR ucLinkSel:1;                //bit2=0: Uniphy LINKA or C or E when fDualLinkConnector=0. when fDualLinkConnector=1, it means master link of dual link is A or C or E
+                                    //    =1: Uniphy LINKB or D or F when fDualLinkConnector=0. when fDualLinkConnector=1, it means master link of dual link is B or D or F
+
+  UCHAR fCoherentMode:1;            //bit1=1: Coherent Mode ( for DVI/HDMI mode )
+  UCHAR fDualLinkConnector:1;       //bit0=1: Dual Link DVI connector
 #else
-	UCHAR fDualLinkConnector:1;	/* bit0=1: Dual Link DVI connector */
-	UCHAR fCoherentMode:1;	/* bit1=1: Coherent Mode ( for DVI/HDMI mode ) */
-	UCHAR ucLinkSel:1;	/* bit2=0: Uniphy LINKA or C or E when fDualLinkConnector=0. when fDualLinkConnector=1, it means master link of dual link is A or C or E */
-	/*     =1: Uniphy LINKB or D or F when fDualLinkConnector=0. when fDualLinkConnector=1, it means master link of dual link is B or D or F */
-	UCHAR ucEncoderSel:1;	/* bit3=0: Data/Clk path source from DIGA( DIG inst0 ). =1: Data/clk path source from DIGB ( DIG inst1 ) */
-	UCHAR fDPConnector:1;	/* bit4=0: DP connector  =1: None DP connector */
-	UCHAR ucReserved:1;
-	UCHAR ucTransmitterSel:2;	/* bit7:6: =0 Dig Transmitter 1 ( Uniphy AB ) */
-	/*         =1 Dig Transmitter 2 ( Uniphy CD ) */
-	/*         =2 Dig Transmitter 3 ( Uniphy EF ) */
+  UCHAR fDualLinkConnector:1;       //bit0=1: Dual Link DVI connector
+  UCHAR fCoherentMode:1;            //bit1=1: Coherent Mode ( for DVI/HDMI mode )
+  UCHAR ucLinkSel:1;                //bit2=0: Uniphy LINKA or C or E when fDualLinkConnector=0. when fDualLinkConnector=1, it means master link of dual link is A or C or E
+                                    //    =1: Uniphy LINKB or D or F when fDualLinkConnector=0. when fDualLinkConnector=1, it means master link of dual link is B or D or F
+  UCHAR ucEncoderSel:1;             //bit3=0: Data/Clk path source from DIGA( DIG inst0 ). =1: Data/clk path source from DIGB ( DIG inst1 )
+  UCHAR fDPConnector:1;             //bit4=0: DP connector  =1: None DP connector
+  UCHAR ucReserved:1;               
+  UCHAR ucTransmitterSel:2;         //bit7:6: =0 Dig Transmitter 1 ( Uniphy AB )
+                                    //        =1 Dig Transmitter 2 ( Uniphy CD )
+                                    //        =2 Dig Transmitter 3 ( Uniphy EF )
 #endif
-} ATOM_DIG_TRANSMITTER_CONFIG_V2;
+}ATOM_DIG_TRANSMITTER_CONFIG_V2;
 
-/* ucConfig */
-/* Bit0 */
+//ucConfig 
+//Bit0
 #define ATOM_TRANSMITTER_CONFIG_V2_DUAL_LINK_CONNECTOR			0x01
 
-/* Bit1 */
+//Bit1
 #define ATOM_TRANSMITTER_CONFIG_V2_COHERENT				          0x02
 
-/* Bit2 */
+//Bit2
 #define ATOM_TRANSMITTER_CONFIG_V2_LINK_SEL_MASK		        0x04
-#define ATOM_TRANSMITTER_CONFIG_V2_LINKA			            0x00
+#define ATOM_TRANSMITTER_CONFIG_V2_LINKA  			            0x00
 #define ATOM_TRANSMITTER_CONFIG_V2_LINKB				            0x04
 
-/*  Bit3 */
+// Bit3
 #define ATOM_TRANSMITTER_CONFIG_V2_ENCODER_SEL_MASK	        0x08
-#define ATOM_TRANSMITTER_CONFIG_V2_DIG1_ENCODER		          0x00	/*  only used when ucAction == ATOM_TRANSMITTER_ACTION_ENABLE or ATOM_TRANSMITTER_ACTION_SETUP */
-#define ATOM_TRANSMITTER_CONFIG_V2_DIG2_ENCODER		          0x08	/*  only used when ucAction == ATOM_TRANSMITTER_ACTION_ENABLE or ATOM_TRANSMITTER_ACTION_SETUP */
+#define ATOM_TRANSMITTER_CONFIG_V2_DIG1_ENCODER		          0x00				// only used when ucAction == ATOM_TRANSMITTER_ACTION_ENABLE or ATOM_TRANSMITTER_ACTION_SETUP
+#define ATOM_TRANSMITTER_CONFIG_V2_DIG2_ENCODER		          0x08				// only used when ucAction == ATOM_TRANSMITTER_ACTION_ENABLE or ATOM_TRANSMITTER_ACTION_SETUP
 
-/*  Bit4 */
+// Bit4
 #define ATOM_TRASMITTER_CONFIG_V2_DP_CONNECTOR			        0x10
 
-/*  Bit7:6 */
+// Bit7:6
 #define ATOM_TRANSMITTER_CONFIG_V2_TRANSMITTER_SEL_MASK     0xC0
-#define ATOM_TRANSMITTER_CONFIG_V2_TRANSMITTER1			0x00	/* AB */
-#define ATOM_TRANSMITTER_CONFIG_V2_TRANSMITTER2			0x40	/* CD */
-#define ATOM_TRANSMITTER_CONFIG_V2_TRANSMITTER3			0x80	/* EF */
-
-typedef struct _DIG_TRANSMITTER_CONTROL_PARAMETERS_V2 {
-	union {
-		USHORT usPixelClock;	/*  in 10KHz; for bios convenient */
-		USHORT usInitInfo;	/*  when init uniphy,lower 8bit is used for connector type defined in objectid.h */
-		ATOM_DP_VS_MODE asMode;	/*  DP Voltage swing mode */
+#define ATOM_TRANSMITTER_CONFIG_V2_TRANSMITTER1           	0x00	//AB
+#define ATOM_TRANSMITTER_CONFIG_V2_TRANSMITTER2           	0x40	//CD
+#define ATOM_TRANSMITTER_CONFIG_V2_TRANSMITTER3           	0x80	//EF
+
+typedef struct _DIG_TRANSMITTER_CONTROL_PARAMETERS_V2
+{
+	union
+	{
+  USHORT usPixelClock;		// in 10KHz; for bios convenient
+	USHORT usInitInfo;			// when init uniphy,lower 8bit is used for connector type defined in objectid.h
+  ATOM_DP_VS_MODE asMode; // DP Voltage swing mode
 	};
-	ATOM_DIG_TRANSMITTER_CONFIG_V2 acConfig;
-	UCHAR ucAction;		/*  define as ATOM_TRANSMITER_ACTION_XXX */
-	UCHAR ucReserved[4];
-} DIG_TRANSMITTER_CONTROL_PARAMETERS_V2;
+  ATOM_DIG_TRANSMITTER_CONFIG_V2 acConfig;
+	UCHAR ucAction;				  // define as ATOM_TRANSMITER_ACTION_XXX
+  UCHAR ucReserved[4];
+}DIG_TRANSMITTER_CONTROL_PARAMETERS_V2;
 
-/****************************************************************************/
-/*  Structures used by DAC1OuputControlTable */
-/*                     DAC2OuputControlTable */
-/*                     LVTMAOutputControlTable  (Before DEC30) */
-/*                     TMDSAOutputControlTable  (Before DEC30) */
-/****************************************************************************/
-typedef struct _DISPLAY_DEVICE_OUTPUT_CONTROL_PARAMETERS {
-	UCHAR ucAction;		/*  Possible input:ATOM_ENABLE||ATOMDISABLE */
-	/*  When the display is LCD, in addition to above: */
-	/*  ATOM_LCD_BLOFF|| ATOM_LCD_BLON ||ATOM_LCD_BL_BRIGHTNESS_CONTROL||ATOM_LCD_SELFTEST_START|| */
-	/*  ATOM_LCD_SELFTEST_STOP */
+typedef struct _ATOM_DIG_TRANSMITTER_CONFIG_V3
+{
+#if ATOM_BIG_ENDIAN
+  UCHAR ucTransmitterSel:2;         //bit7:6: =0 Dig Transmitter 1 ( Uniphy AB )
+                                    //        =1 Dig Transmitter 2 ( Uniphy CD )
+                                    //        =2 Dig Transmitter 3 ( Uniphy EF )
+  UCHAR ucRefClkSource:2;           //bit5:4: PPLL1 =0, PPLL2=1, EXT_CLK=2
+  UCHAR ucEncoderSel:1;             //bit3=0: Data/Clk path source from DIGA/C/E. =1: Data/clk path source from DIGB/D/F
+  UCHAR ucLinkSel:1;                //bit2=0: Uniphy LINKA or C or E when fDualLinkConnector=0. when fDualLinkConnector=1, it means master link of dual link is A or C or E
+                                    //    =1: Uniphy LINKB or D or F when fDualLinkConnector=0. when fDualLinkConnector=1, it means master link of dual link is B or D or F
+  UCHAR fCoherentMode:1;            //bit1=1: Coherent Mode ( for DVI/HDMI mode )
+  UCHAR fDualLinkConnector:1;       //bit0=1: Dual Link DVI connector
+#else
+  UCHAR fDualLinkConnector:1;       //bit0=1: Dual Link DVI connector
+  UCHAR fCoherentMode:1;            //bit1=1: Coherent Mode ( for DVI/HDMI mode )
+  UCHAR ucLinkSel:1;                //bit2=0: Uniphy LINKA or C or E when fDualLinkConnector=0. when fDualLinkConnector=1, it means master link of dual link is A or C or E
+                                    //    =1: Uniphy LINKB or D or F when fDualLinkConnector=0. when fDualLinkConnector=1, it means master link of dual link is B or D or F
+  UCHAR ucEncoderSel:1;             //bit3=0: Data/Clk path source from DIGA/C/E. =1: Data/clk path source from DIGB/D/F
+  UCHAR ucRefClkSource:2;           //bit5:4: PPLL1 =0, PPLL2=1, EXT_CLK=2
+  UCHAR ucTransmitterSel:2;         //bit7:6: =0 Dig Transmitter 1 ( Uniphy AB )
+                                    //        =1 Dig Transmitter 2 ( Uniphy CD )
+                                    //        =2 Dig Transmitter 3 ( Uniphy EF )
+#endif
+}ATOM_DIG_TRANSMITTER_CONFIG_V3;
 
-	UCHAR aucPadding[3];	/*  padding to DWORD aligned */
-} DISPLAY_DEVICE_OUTPUT_CONTROL_PARAMETERS;
+typedef struct _DIG_TRANSMITTER_CONTROL_PARAMETERS_V3
+{
+	union
+	{
+    USHORT usPixelClock;		// in 10KHz; for bios convenient
+	  USHORT usInitInfo;			// when init uniphy,lower 8bit is used for connector type defined in objectid.h
+    ATOM_DP_VS_MODE asMode; // DP Voltage swing mode
+	};
+  ATOM_DIG_TRANSMITTER_CONFIG_V3 acConfig;
+	UCHAR ucAction;				    // define as ATOM_TRANSMITER_ACTION_XXX
+  UCHAR ucLaneNum;
+  UCHAR ucReserved[3];
+}DIG_TRANSMITTER_CONTROL_PARAMETERS_V3;
+
+//ucConfig 
+//Bit0
+#define ATOM_TRANSMITTER_CONFIG_V3_DUAL_LINK_CONNECTOR			0x01
+
+//Bit1
+#define ATOM_TRANSMITTER_CONFIG_V3_COHERENT				          0x02
+
+//Bit2
+#define ATOM_TRANSMITTER_CONFIG_V3_LINK_SEL_MASK		        0x04
+#define ATOM_TRANSMITTER_CONFIG_V3_LINKA  			            0x00
+#define ATOM_TRANSMITTER_CONFIG_V3_LINKB				            0x04
+
+// Bit3
+#define ATOM_TRANSMITTER_CONFIG_V3_ENCODER_SEL_MASK	        0x08
+#define ATOM_TRANSMITTER_CONFIG_V3_DIG1_ENCODER		          0x00
+#define ATOM_TRANSMITTER_CONFIG_V3_DIG2_ENCODER		          0x08
+
+// Bit5:4
+#define ATOM_TRASMITTER_CONFIG_V3_REFCLK_SEL_MASK 	        0x30
+#define ATOM_TRASMITTER_CONFIG_V3_P1PLL          		        0x00
+#define ATOM_TRASMITTER_CONFIG_V3_P2PLL		                  0x10
+#define ATOM_TRASMITTER_CONFIG_V3_REFCLK_SRC_EXT            0x20
+
+// Bit7:6
+#define ATOM_TRANSMITTER_CONFIG_V3_TRANSMITTER_SEL_MASK     0xC0
+#define ATOM_TRANSMITTER_CONFIG_V3_TRANSMITTER1           	0x00	//AB
+#define ATOM_TRANSMITTER_CONFIG_V3_TRANSMITTER2           	0x40	//CD
+#define ATOM_TRANSMITTER_CONFIG_V3_TRANSMITTER3           	0x80	//EF
+
+/****************************************************************************/	
+// Structures used by DAC1OuputControlTable
+//                    DAC2OuputControlTable
+//                    LVTMAOutputControlTable  (Before DEC30)
+//                    TMDSAOutputControlTable  (Before DEC30)
+/****************************************************************************/	
+typedef struct _DISPLAY_DEVICE_OUTPUT_CONTROL_PARAMETERS
+{
+  UCHAR  ucAction;                    // Possible input:ATOM_ENABLE||ATOMDISABLE
+                                      // When the display is LCD, in addition to above:
+                                      // ATOM_LCD_BLOFF|| ATOM_LCD_BLON ||ATOM_LCD_BL_BRIGHTNESS_CONTROL||ATOM_LCD_SELFTEST_START||
+                                      // ATOM_LCD_SELFTEST_STOP
+                                      
+  UCHAR  aucPadding[3];               // padding to DWORD aligned
+}DISPLAY_DEVICE_OUTPUT_CONTROL_PARAMETERS;
 
 #define DISPLAY_DEVICE_OUTPUT_CONTROL_PS_ALLOCATION DISPLAY_DEVICE_OUTPUT_CONTROL_PARAMETERS
 
-#define CRT1_OUTPUT_CONTROL_PARAMETERS     DISPLAY_DEVICE_OUTPUT_CONTROL_PARAMETERS
+
+#define CRT1_OUTPUT_CONTROL_PARAMETERS     DISPLAY_DEVICE_OUTPUT_CONTROL_PARAMETERS 
 #define CRT1_OUTPUT_CONTROL_PS_ALLOCATION  DISPLAY_DEVICE_OUTPUT_CONTROL_PS_ALLOCATION
 
-#define CRT2_OUTPUT_CONTROL_PARAMETERS     DISPLAY_DEVICE_OUTPUT_CONTROL_PARAMETERS
+#define CRT2_OUTPUT_CONTROL_PARAMETERS     DISPLAY_DEVICE_OUTPUT_CONTROL_PARAMETERS 
 #define CRT2_OUTPUT_CONTROL_PS_ALLOCATION  DISPLAY_DEVICE_OUTPUT_CONTROL_PS_ALLOCATION
 
 #define CV1_OUTPUT_CONTROL_PARAMETERS      DISPLAY_DEVICE_OUTPUT_CONTROL_PARAMETERS
@@ -782,397 +980,550 @@ typedef struct _DISPLAY_DEVICE_OUTPUT_CONTROL_PARAMETERS {
 #define DVO_OUTPUT_CONTROL_PS_ALLOCATION   DIG_TRANSMITTER_CONTROL_PS_ALLOCATION
 #define DVO_OUTPUT_CONTROL_PARAMETERS_V3	 DIG_TRANSMITTER_CONTROL_PARAMETERS
 
-/****************************************************************************/
-/*  Structures used by BlankCRTCTable */
-/****************************************************************************/
-typedef struct _BLANK_CRTC_PARAMETERS {
-	UCHAR ucCRTC;		/*  ATOM_CRTC1 or ATOM_CRTC2 */
-	UCHAR ucBlanking;	/*  ATOM_BLANKING or ATOM_BLANKINGOFF */
-	USHORT usBlackColorRCr;
-	USHORT usBlackColorGY;
-	USHORT usBlackColorBCb;
-} BLANK_CRTC_PARAMETERS;
+/****************************************************************************/	
+// Structures used by BlankCRTCTable
+/****************************************************************************/	
+typedef struct _BLANK_CRTC_PARAMETERS
+{
+  UCHAR  ucCRTC;                    	// ATOM_CRTC1 or ATOM_CRTC2
+  UCHAR  ucBlanking;                  // ATOM_BLANKING or ATOM_BLANKINGOFF
+  USHORT usBlackColorRCr;
+  USHORT usBlackColorGY;
+  USHORT usBlackColorBCb;
+}BLANK_CRTC_PARAMETERS;
 #define BLANK_CRTC_PS_ALLOCATION    BLANK_CRTC_PARAMETERS
 
-/****************************************************************************/
-/*  Structures used by EnableCRTCTable */
-/*                     EnableCRTCMemReqTable */
-/*                     UpdateCRTC_DoubleBufferRegistersTable */
-/****************************************************************************/
-typedef struct _ENABLE_CRTC_PARAMETERS {
-	UCHAR ucCRTC;		/*  ATOM_CRTC1 or ATOM_CRTC2 */
-	UCHAR ucEnable;		/*  ATOM_ENABLE or ATOM_DISABLE */
-	UCHAR ucPadding[2];
-} ENABLE_CRTC_PARAMETERS;
+/****************************************************************************/	
+// Structures used by EnableCRTCTable
+//                    EnableCRTCMemReqTable
+//                    UpdateCRTC_DoubleBufferRegistersTable
+/****************************************************************************/	
+typedef struct _ENABLE_CRTC_PARAMETERS
+{
+  UCHAR ucCRTC;                    	  // ATOM_CRTC1 or ATOM_CRTC2
+  UCHAR ucEnable;                     // ATOM_ENABLE or ATOM_DISABLE 
+  UCHAR ucPadding[2];
+}ENABLE_CRTC_PARAMETERS;
 #define ENABLE_CRTC_PS_ALLOCATION   ENABLE_CRTC_PARAMETERS
 
-/****************************************************************************/
-/*  Structures used by SetCRTC_OverScanTable */
-/****************************************************************************/
-typedef struct _SET_CRTC_OVERSCAN_PARAMETERS {
-	USHORT usOverscanRight;	/*  right */
-	USHORT usOverscanLeft;	/*  left */
-	USHORT usOverscanBottom;	/*  bottom */
-	USHORT usOverscanTop;	/*  top */
-	UCHAR ucCRTC;		/*  ATOM_CRTC1 or ATOM_CRTC2 */
-	UCHAR ucPadding[3];
-} SET_CRTC_OVERSCAN_PARAMETERS;
+/****************************************************************************/	
+// Structures used by SetCRTC_OverScanTable
+/****************************************************************************/	
+typedef struct _SET_CRTC_OVERSCAN_PARAMETERS
+{
+  USHORT usOverscanRight;             // right
+  USHORT usOverscanLeft;              // left
+  USHORT usOverscanBottom;            // bottom
+  USHORT usOverscanTop;               // top
+  UCHAR  ucCRTC;                      // ATOM_CRTC1 or ATOM_CRTC2
+  UCHAR  ucPadding[3];
+}SET_CRTC_OVERSCAN_PARAMETERS;
 #define SET_CRTC_OVERSCAN_PS_ALLOCATION  SET_CRTC_OVERSCAN_PARAMETERS
 
-/****************************************************************************/
-/*  Structures used by SetCRTC_ReplicationTable */
-/****************************************************************************/
-typedef struct _SET_CRTC_REPLICATION_PARAMETERS {
-	UCHAR ucH_Replication;	/*  horizontal replication */
-	UCHAR ucV_Replication;	/*  vertical replication */
-	UCHAR usCRTC;		/*  ATOM_CRTC1 or ATOM_CRTC2 */
-	UCHAR ucPadding;
-} SET_CRTC_REPLICATION_PARAMETERS;
+/****************************************************************************/	
+// Structures used by SetCRTC_ReplicationTable
+/****************************************************************************/	
+typedef struct _SET_CRTC_REPLICATION_PARAMETERS
+{
+  UCHAR ucH_Replication;              // horizontal replication
+  UCHAR ucV_Replication;              // vertical replication
+  UCHAR usCRTC;                       // ATOM_CRTC1 or ATOM_CRTC2
+  UCHAR ucPadding;
+}SET_CRTC_REPLICATION_PARAMETERS;
 #define SET_CRTC_REPLICATION_PS_ALLOCATION  SET_CRTC_REPLICATION_PARAMETERS
 
-/****************************************************************************/
-/*  Structures used by SelectCRTC_SourceTable */
-/****************************************************************************/
-typedef struct _SELECT_CRTC_SOURCE_PARAMETERS {
-	UCHAR ucCRTC;		/*  ATOM_CRTC1 or ATOM_CRTC2 */
-	UCHAR ucDevice;		/*  ATOM_DEVICE_CRT1|ATOM_DEVICE_CRT2|.... */
-	UCHAR ucPadding[2];
-} SELECT_CRTC_SOURCE_PARAMETERS;
+/****************************************************************************/	
+// Structures used by SelectCRTC_SourceTable
+/****************************************************************************/	
+typedef struct _SELECT_CRTC_SOURCE_PARAMETERS
+{
+  UCHAR ucCRTC;                    	  // ATOM_CRTC1 or ATOM_CRTC2
+  UCHAR ucDevice;                     // ATOM_DEVICE_CRT1|ATOM_DEVICE_CRT2|....
+  UCHAR ucPadding[2];
+}SELECT_CRTC_SOURCE_PARAMETERS;
 #define SELECT_CRTC_SOURCE_PS_ALLOCATION  SELECT_CRTC_SOURCE_PARAMETERS
 
-typedef struct _SELECT_CRTC_SOURCE_PARAMETERS_V2 {
-	UCHAR ucCRTC;		/*  ATOM_CRTC1 or ATOM_CRTC2 */
-	UCHAR ucEncoderID;	/*  DAC1/DAC2/TVOUT/DIG1/DIG2/DVO */
-	UCHAR ucEncodeMode;	/*  Encoding mode, only valid when using DIG1/DIG2/DVO */
-	UCHAR ucPadding;
-} SELECT_CRTC_SOURCE_PARAMETERS_V2;
-
-/* ucEncoderID */
-/* #define ASIC_INT_DAC1_ENCODER_ID                                              0x00 */
-/* #define ASIC_INT_TV_ENCODER_ID                                                                        0x02 */
-/* #define ASIC_INT_DIG1_ENCODER_ID                                                              0x03 */
-/* #define ASIC_INT_DAC2_ENCODER_ID                                                              0x04 */
-/* #define ASIC_EXT_TV_ENCODER_ID                                                                        0x06 */
-/* #define ASIC_INT_DVO_ENCODER_ID                                                                       0x07 */
-/* #define ASIC_INT_DIG2_ENCODER_ID                                                              0x09 */
-/* #define ASIC_EXT_DIG_ENCODER_ID                                                                       0x05 */
-
-/* ucEncodeMode */
-/* #define ATOM_ENCODER_MODE_DP                                                                          0 */
-/* #define ATOM_ENCODER_MODE_LVDS                                                                        1 */
-/* #define ATOM_ENCODER_MODE_DVI                                                                         2 */
-/* #define ATOM_ENCODER_MODE_HDMI                                                                        3 */
-/* #define ATOM_ENCODER_MODE_SDVO                                                                        4 */
-/* #define ATOM_ENCODER_MODE_TV                                                                          13 */
-/* #define ATOM_ENCODER_MODE_CV                                                                          14 */
-/* #define ATOM_ENCODER_MODE_CRT                                                                         15 */
-
-/****************************************************************************/
-/*  Structures used by SetPixelClockTable */
-/*                     GetPixelClockTable */
-/****************************************************************************/
-/* Major revision=1., Minor revision=1 */
-typedef struct _PIXEL_CLOCK_PARAMETERS {
-	USHORT usPixelClock;	/*  in 10kHz unit; for bios convenient = (RefClk*FB_Div)/(Ref_Div*Post_Div) */
-	/*  0 means disable PPLL */
-	USHORT usRefDiv;	/*  Reference divider */
-	USHORT usFbDiv;		/*  feedback divider */
-	UCHAR ucPostDiv;	/*  post divider */
-	UCHAR ucFracFbDiv;	/*  fractional feedback divider */
-	UCHAR ucPpll;		/*  ATOM_PPLL1 or ATOM_PPL2 */
-	UCHAR ucRefDivSrc;	/*  ATOM_PJITTER or ATO_NONPJITTER */
-	UCHAR ucCRTC;		/*  Which CRTC uses this Ppll */
-	UCHAR ucPadding;
-} PIXEL_CLOCK_PARAMETERS;
-
-/* Major revision=1., Minor revision=2, add ucMiscIfno */
-/* ucMiscInfo: */
+typedef struct _SELECT_CRTC_SOURCE_PARAMETERS_V2
+{
+  UCHAR ucCRTC;                    	  // ATOM_CRTC1 or ATOM_CRTC2
+  UCHAR ucEncoderID;                  // DAC1/DAC2/TVOUT/DIG1/DIG2/DVO
+  UCHAR ucEncodeMode;									// Encoding mode, only valid when using DIG1/DIG2/DVO
+  UCHAR ucPadding;
+}SELECT_CRTC_SOURCE_PARAMETERS_V2;
+
+//ucEncoderID
+//#define ASIC_INT_DAC1_ENCODER_ID    						0x00 
+//#define ASIC_INT_TV_ENCODER_ID									0x02
+//#define ASIC_INT_DIG1_ENCODER_ID								0x03
+//#define ASIC_INT_DAC2_ENCODER_ID								0x04
+//#define ASIC_EXT_TV_ENCODER_ID									0x06
+//#define ASIC_INT_DVO_ENCODER_ID									0x07
+//#define ASIC_INT_DIG2_ENCODER_ID								0x09
+//#define ASIC_EXT_DIG_ENCODER_ID									0x05
+
+//ucEncodeMode
+//#define ATOM_ENCODER_MODE_DP										0
+//#define ATOM_ENCODER_MODE_LVDS									1
+//#define ATOM_ENCODER_MODE_DVI										2
+//#define ATOM_ENCODER_MODE_HDMI									3
+//#define ATOM_ENCODER_MODE_SDVO									4
+//#define ATOM_ENCODER_MODE_TV										13
+//#define ATOM_ENCODER_MODE_CV										14
+//#define ATOM_ENCODER_MODE_CRT										15
+
+/****************************************************************************/	
+// Structures used by SetPixelClockTable
+//                    GetPixelClockTable 
+/****************************************************************************/	
+//Major revision=1., Minor revision=1
+typedef struct _PIXEL_CLOCK_PARAMETERS
+{
+  USHORT usPixelClock;                // in 10kHz unit; for bios convenient = (RefClk*FB_Div)/(Ref_Div*Post_Div)
+                                      // 0 means disable PPLL
+  USHORT usRefDiv;                    // Reference divider
+  USHORT usFbDiv;                     // feedback divider
+  UCHAR  ucPostDiv;                   // post divider	
+  UCHAR  ucFracFbDiv;                 // fractional feedback divider
+  UCHAR  ucPpll;                      // ATOM_PPLL1 or ATOM_PPL2
+  UCHAR  ucRefDivSrc;                 // ATOM_PJITTER or ATO_NONPJITTER
+  UCHAR  ucCRTC;                      // Which CRTC uses this Ppll
+  UCHAR  ucPadding;
+}PIXEL_CLOCK_PARAMETERS;
+
+//Major revision=1., Minor revision=2, add ucMiscIfno
+//ucMiscInfo:
 #define MISC_FORCE_REPROG_PIXEL_CLOCK 0x1
 #define MISC_DEVICE_INDEX_MASK        0xF0
 #define MISC_DEVICE_INDEX_SHIFT       4
 
-typedef struct _PIXEL_CLOCK_PARAMETERS_V2 {
-	USHORT usPixelClock;	/*  in 10kHz unit; for bios convenient = (RefClk*FB_Div)/(Ref_Div*Post_Div) */
-	/*  0 means disable PPLL */
-	USHORT usRefDiv;	/*  Reference divider */
-	USHORT usFbDiv;		/*  feedback divider */
-	UCHAR ucPostDiv;	/*  post divider */
-	UCHAR ucFracFbDiv;	/*  fractional feedback divider */
-	UCHAR ucPpll;		/*  ATOM_PPLL1 or ATOM_PPL2 */
-	UCHAR ucRefDivSrc;	/*  ATOM_PJITTER or ATO_NONPJITTER */
-	UCHAR ucCRTC;		/*  Which CRTC uses this Ppll */
-	UCHAR ucMiscInfo;	/*  Different bits for different purpose, bit [7:4] as device index, bit[0]=Force prog */
-} PIXEL_CLOCK_PARAMETERS_V2;
-
-/* Major revision=1., Minor revision=3, structure/definition change */
-/* ucEncoderMode: */
-/* ATOM_ENCODER_MODE_DP */
-/* ATOM_ENOCDER_MODE_LVDS */
-/* ATOM_ENOCDER_MODE_DVI */
-/* ATOM_ENOCDER_MODE_HDMI */
-/* ATOM_ENOCDER_MODE_SDVO */
-/* ATOM_ENCODER_MODE_TV                                                                          13 */
-/* ATOM_ENCODER_MODE_CV                                                                          14 */
-/* ATOM_ENCODER_MODE_CRT                                                                         15 */
-
-/* ucDVOConfig */
-/* #define DVO_ENCODER_CONFIG_RATE_SEL                                                   0x01 */
-/* #define DVO_ENCODER_CONFIG_DDR_SPEED                                          0x00 */
-/* #define DVO_ENCODER_CONFIG_SDR_SPEED                                          0x01 */
-/* #define DVO_ENCODER_CONFIG_OUTPUT_SEL                                         0x0c */
-/* #define DVO_ENCODER_CONFIG_LOW12BIT                                                   0x00 */
-/* #define DVO_ENCODER_CONFIG_UPPER12BIT                                         0x04 */
-/* #define DVO_ENCODER_CONFIG_24BIT                                                              0x08 */
-
-/* ucMiscInfo: also changed, see below */
+typedef struct _PIXEL_CLOCK_PARAMETERS_V2
+{
+  USHORT usPixelClock;                // in 10kHz unit; for bios convenient = (RefClk*FB_Div)/(Ref_Div*Post_Div)
+                                      // 0 means disable PPLL
+  USHORT usRefDiv;                    // Reference divider
+  USHORT usFbDiv;                     // feedback divider
+  UCHAR  ucPostDiv;                   // post divider	
+  UCHAR  ucFracFbDiv;                 // fractional feedback divider
+  UCHAR  ucPpll;                      // ATOM_PPLL1 or ATOM_PPL2
+  UCHAR  ucRefDivSrc;                 // ATOM_PJITTER or ATO_NONPJITTER
+  UCHAR  ucCRTC;                      // Which CRTC uses this Ppll
+  UCHAR  ucMiscInfo;                  // Different bits for different purpose, bit [7:4] as device index, bit[0]=Force prog
+}PIXEL_CLOCK_PARAMETERS_V2;
+
+//Major revision=1., Minor revision=3, structure/definition change
+//ucEncoderMode:
+//ATOM_ENCODER_MODE_DP
+//ATOM_ENOCDER_MODE_LVDS
+//ATOM_ENOCDER_MODE_DVI
+//ATOM_ENOCDER_MODE_HDMI
+//ATOM_ENOCDER_MODE_SDVO
+//ATOM_ENCODER_MODE_TV										13
+//ATOM_ENCODER_MODE_CV										14
+//ATOM_ENCODER_MODE_CRT										15
+
+//ucDVOConfig
+//#define DVO_ENCODER_CONFIG_RATE_SEL							0x01
+//#define DVO_ENCODER_CONFIG_DDR_SPEED						0x00
+//#define DVO_ENCODER_CONFIG_SDR_SPEED						0x01
+//#define DVO_ENCODER_CONFIG_OUTPUT_SEL						0x0c
+//#define DVO_ENCODER_CONFIG_LOW12BIT							0x00
+//#define DVO_ENCODER_CONFIG_UPPER12BIT						0x04
+//#define DVO_ENCODER_CONFIG_24BIT								0x08
+
+//ucMiscInfo: also changed, see below
 #define PIXEL_CLOCK_MISC_FORCE_PROG_PPLL						0x01
 #define PIXEL_CLOCK_MISC_VGA_MODE										0x02
 #define PIXEL_CLOCK_MISC_CRTC_SEL_MASK							0x04
 #define PIXEL_CLOCK_MISC_CRTC_SEL_CRTC1							0x00
 #define PIXEL_CLOCK_MISC_CRTC_SEL_CRTC2							0x04
 #define PIXEL_CLOCK_MISC_USE_ENGINE_FOR_DISPCLK			0x08
+#define PIXEL_CLOCK_MISC_REF_DIV_SRC                    0x10
+// V1.4 for RoadRunner
+#define PIXEL_CLOCK_V4_MISC_SS_ENABLE               0x10
+#define PIXEL_CLOCK_V4_MISC_COHERENT_MODE           0x20
 
-typedef struct _PIXEL_CLOCK_PARAMETERS_V3 {
-	USHORT usPixelClock;	/*  in 10kHz unit; for bios convenient = (RefClk*FB_Div)/(Ref_Div*Post_Div) */
-	/*  0 means disable PPLL. For VGA PPLL,make sure this value is not 0. */
-	USHORT usRefDiv;	/*  Reference divider */
-	USHORT usFbDiv;		/*  feedback divider */
-	UCHAR ucPostDiv;	/*  post divider */
-	UCHAR ucFracFbDiv;	/*  fractional feedback divider */
-	UCHAR ucPpll;		/*  ATOM_PPLL1 or ATOM_PPL2 */
-	UCHAR ucTransmitterId;	/*  graphic encoder id defined in objectId.h */
-	union {
-		UCHAR ucEncoderMode;	/*  encoder type defined as ATOM_ENCODER_MODE_DP/DVI/HDMI/ */
-		UCHAR ucDVOConfig;	/*  when use DVO, need to know SDR/DDR, 12bit or 24bit */
+typedef struct _PIXEL_CLOCK_PARAMETERS_V3
+{
+  USHORT usPixelClock;                // in 10kHz unit; for bios convenient = (RefClk*FB_Div)/(Ref_Div*Post_Div)
+                                      // 0 means disable PPLL. For VGA PPLL,make sure this value is not 0.
+  USHORT usRefDiv;                    // Reference divider
+  USHORT usFbDiv;                     // feedback divider
+  UCHAR  ucPostDiv;                   // post divider	
+  UCHAR  ucFracFbDiv;                 // fractional feedback divider
+  UCHAR  ucPpll;                      // ATOM_PPLL1 or ATOM_PPL2
+  UCHAR  ucTransmitterId;             // graphic encoder id defined in objectId.h
+	union
+	{
+  UCHAR  ucEncoderMode;               // encoder type defined as ATOM_ENCODER_MODE_DP/DVI/HDMI/
+	UCHAR  ucDVOConfig;									// when use DVO, need to know SDR/DDR, 12bit or 24bit
 	};
-	UCHAR ucMiscInfo;	/*  bit[0]=Force program, bit[1]= set pclk for VGA, b[2]= CRTC sel */
-	/*  bit[3]=0:use PPLL for dispclk source, =1: use engine clock for dispclock source */
-} PIXEL_CLOCK_PARAMETERS_V3;
+  UCHAR  ucMiscInfo;                  // bit[0]=Force program, bit[1]= set pclk for VGA, b[2]= CRTC sel
+                                      // bit[3]=0:use PPLL for dispclk source, =1: use engine clock for dispclock source
+                                      // bit[4]=0:use XTALIN as the source of reference divider,=1 use the pre-defined clock as the source of reference divider
+}PIXEL_CLOCK_PARAMETERS_V3;
 
 #define PIXEL_CLOCK_PARAMETERS_LAST			PIXEL_CLOCK_PARAMETERS_V2
 #define GET_PIXEL_CLOCK_PS_ALLOCATION		PIXEL_CLOCK_PARAMETERS_LAST
 
-/****************************************************************************/
-/*  Structures used by AdjustDisplayPllTable */
-/****************************************************************************/
-typedef struct _ADJUST_DISPLAY_PLL_PARAMETERS {
+typedef struct _PIXEL_CLOCK_PARAMETERS_V5
+{
+  UCHAR  ucCRTC;             // ATOM_CRTC1~6, indicate the CRTC controller to 
+                             // drive the pixel clock. not used for DCPLL case.
+  union{
+  UCHAR  ucReserved;
+  UCHAR  ucFracFbDiv;        // [gphan] temporary to prevent build problem.  remove it after driver code is changed.
+  };
+  USHORT usPixelClock;       // target the pixel clock to drive the CRTC timing
+                             // 0 means disable PPLL/DCPLL. 
+  USHORT usFbDiv;            // feedback divider integer part. 
+  UCHAR  ucPostDiv;          // post divider. 
+  UCHAR  ucRefDiv;           // Reference divider
+  UCHAR  ucPpll;             // ATOM_PPLL1/ATOM_PPLL2/ATOM_DCPLL
+  UCHAR  ucTransmitterID;    // ASIC encoder id defined in objectId.h, 
+                             // indicate which graphic encoder will be used. 
+  UCHAR  ucEncoderMode;      // Encoder mode: 
+  UCHAR  ucMiscInfo;         // bit[0]= Force program PPLL 
+                             // bit[1]= when VGA timing is used. 
+                             // bit[3:2]= HDMI panel bit depth: =0: 24bpp =1:30bpp, =2:32bpp
+                             // bit[4]= RefClock source for PPLL. 
+                             // =0: XTLAIN( default mode )
+	                           // =1: other external clock source, which is pre-defined
+                             //     by VBIOS depend on the feature required.
+                             // bit[7:5]: reserved.
+  ULONG  ulFbDivDecFrac;     // 20 bit feedback divider decimal fraction part, range from 1~999999 ( 0.000001 to 0.999999 )
+
+}PIXEL_CLOCK_PARAMETERS_V5;
+
+#define PIXEL_CLOCK_V5_MISC_FORCE_PROG_PPLL					0x01
+#define PIXEL_CLOCK_V5_MISC_VGA_MODE								0x02
+#define PIXEL_CLOCK_V5_MISC_HDMI_BPP_MASK           0x0c
+#define PIXEL_CLOCK_V5_MISC_HDMI_24BPP              0x00
+#define PIXEL_CLOCK_V5_MISC_HDMI_30BPP              0x04
+#define PIXEL_CLOCK_V5_MISC_HDMI_32BPP              0x08
+#define PIXEL_CLOCK_V5_MISC_REF_DIV_SRC             0x10
+
+typedef struct _GET_DISP_PLL_STATUS_INPUT_PARAMETERS_V2
+{
+  PIXEL_CLOCK_PARAMETERS_V3 sDispClkInput;
+}GET_DISP_PLL_STATUS_INPUT_PARAMETERS_V2;
+
+typedef struct _GET_DISP_PLL_STATUS_OUTPUT_PARAMETERS_V2
+{
+  UCHAR  ucStatus;
+  UCHAR  ucRefDivSrc;                 // =1: reference clock source from XTALIN, =0: source from PCIE ref clock
+  UCHAR  ucReserved[2];
+}GET_DISP_PLL_STATUS_OUTPUT_PARAMETERS_V2;
+
+typedef struct _GET_DISP_PLL_STATUS_INPUT_PARAMETERS_V3
+{
+  PIXEL_CLOCK_PARAMETERS_V5 sDispClkInput;
+}GET_DISP_PLL_STATUS_INPUT_PARAMETERS_V3;
+
+/****************************************************************************/	
+// Structures used by AdjustDisplayPllTable
+/****************************************************************************/	
+typedef struct _ADJUST_DISPLAY_PLL_PARAMETERS
+{
 	USHORT usPixelClock;
 	UCHAR ucTransmitterID;
 	UCHAR ucEncodeMode;
-	union {
-		UCHAR ucDVOConfig;	/* if DVO, need passing link rate and output 12bitlow or 24bit */
-		UCHAR ucConfig;	/* if none DVO, not defined yet */
+	union
+	{
+		UCHAR ucDVOConfig;									//if DVO, need passing link rate and output 12bitlow or 24bit
+		UCHAR ucConfig;											//if none DVO, not defined yet
 	};
 	UCHAR ucReserved[3];
-} ADJUST_DISPLAY_PLL_PARAMETERS;
+}ADJUST_DISPLAY_PLL_PARAMETERS;
 
 #define ADJUST_DISPLAY_CONFIG_SS_ENABLE       0x10
-
 #define ADJUST_DISPLAY_PLL_PS_ALLOCATION			ADJUST_DISPLAY_PLL_PARAMETERS
 
-/****************************************************************************/
-/*  Structures used by EnableYUVTable */
-/****************************************************************************/
-typedef struct _ENABLE_YUV_PARAMETERS {
-	UCHAR ucEnable;		/*  ATOM_ENABLE:Enable YUV or ATOM_DISABLE:Disable YUV (RGB) */
-	UCHAR ucCRTC;		/*  Which CRTC needs this YUV or RGB format */
-	UCHAR ucPadding[2];
-} ENABLE_YUV_PARAMETERS;
+typedef struct _ADJUST_DISPLAY_PLL_INPUT_PARAMETERS_V3
+{
+	USHORT usPixelClock;                    // target pixel clock
+	UCHAR ucTransmitterID;                  // transmitter id defined in objectid.h
+	UCHAR ucEncodeMode;                     // encoder mode: CRT, LVDS, DP, TMDS or HDMI
+  UCHAR ucDispPllConfig;                 // display pll configure parameter defined as following DISPPLL_CONFIG_XXXX
+	UCHAR ucReserved[3];
+}ADJUST_DISPLAY_PLL_INPUT_PARAMETERS_V3;
+
+// usDispPllConfig v1.2 for RoadRunner
+#define DISPPLL_CONFIG_DVO_RATE_SEL                0x0001     // need only when ucTransmitterID = DVO
+#define DISPPLL_CONFIG_DVO_DDR_SPEED               0x0000     // need only when ucTransmitterID = DVO
+#define DISPPLL_CONFIG_DVO_SDR_SPEED               0x0001     // need only when ucTransmitterID = DVO
+#define DISPPLL_CONFIG_DVO_OUTPUT_SEL              0x000c     // need only when ucTransmitterID = DVO
+#define DISPPLL_CONFIG_DVO_LOW12BIT                0x0000     // need only when ucTransmitterID = DVO
+#define DISPPLL_CONFIG_DVO_UPPER12BIT              0x0004     // need only when ucTransmitterID = DVO
+#define DISPPLL_CONFIG_DVO_24BIT                   0x0008     // need only when ucTransmitterID = DVO
+#define DISPPLL_CONFIG_SS_ENABLE                   0x0010     // Only used when ucEncoderMode = DP or LVDS
+#define DISPPLL_CONFIG_COHERENT_MODE               0x0020     // Only used when ucEncoderMode = TMDS or HDMI
+#define DISPPLL_CONFIG_DUAL_LINK                   0x0040     // Only used when ucEncoderMode = TMDS or LVDS
+
+
+typedef struct _ADJUST_DISPLAY_PLL_OUTPUT_PARAMETERS_V3
+{
+  ULONG ulDispPllFreq;                 // return display PPLL freq which is used to generate the pixclock, and related idclk, symclk etc
+  UCHAR ucRefDiv;                      // if it is none-zero, it is used to be calculated the other ppll parameter fb_divider and post_div ( if it is not given )
+  UCHAR ucPostDiv;                     // if it is none-zero, it is used to be calculated the other ppll parameter fb_divider
+  UCHAR ucReserved[2];  
+}ADJUST_DISPLAY_PLL_OUTPUT_PARAMETERS_V3;
+
+typedef struct _ADJUST_DISPLAY_PLL_PS_ALLOCATION_V3
+{
+  union 
+  {
+    ADJUST_DISPLAY_PLL_INPUT_PARAMETERS_V3  sInput;
+    ADJUST_DISPLAY_PLL_OUTPUT_PARAMETERS_V3 sOutput;
+  };
+} ADJUST_DISPLAY_PLL_PS_ALLOCATION_V3;
+
+/****************************************************************************/	
+// Structures used by EnableYUVTable
+/****************************************************************************/	
+typedef struct _ENABLE_YUV_PARAMETERS
+{
+  UCHAR ucEnable;                     // ATOM_ENABLE:Enable YUV or ATOM_DISABLE:Disable YUV (RGB)
+  UCHAR ucCRTC;                       // Which CRTC needs this YUV or RGB format
+  UCHAR ucPadding[2];
+}ENABLE_YUV_PARAMETERS;
 #define ENABLE_YUV_PS_ALLOCATION ENABLE_YUV_PARAMETERS
 
-/****************************************************************************/
-/*  Structures used by GetMemoryClockTable */
-/****************************************************************************/
-typedef struct _GET_MEMORY_CLOCK_PARAMETERS {
-	ULONG ulReturnMemoryClock;	/*  current memory speed in 10KHz unit */
+/****************************************************************************/	
+// Structures used by GetMemoryClockTable
+/****************************************************************************/	
+typedef struct _GET_MEMORY_CLOCK_PARAMETERS
+{
+  ULONG ulReturnMemoryClock;          // current memory speed in 10KHz unit
 } GET_MEMORY_CLOCK_PARAMETERS;
 #define GET_MEMORY_CLOCK_PS_ALLOCATION  GET_MEMORY_CLOCK_PARAMETERS
 
-/****************************************************************************/
-/*  Structures used by GetEngineClockTable */
-/****************************************************************************/
-typedef struct _GET_ENGINE_CLOCK_PARAMETERS {
-	ULONG ulReturnEngineClock;	/*  current engine speed in 10KHz unit */
+/****************************************************************************/	
+// Structures used by GetEngineClockTable
+/****************************************************************************/	
+typedef struct _GET_ENGINE_CLOCK_PARAMETERS
+{
+  ULONG ulReturnEngineClock;          // current engine speed in 10KHz unit
 } GET_ENGINE_CLOCK_PARAMETERS;
 #define GET_ENGINE_CLOCK_PS_ALLOCATION  GET_ENGINE_CLOCK_PARAMETERS
 
-/****************************************************************************/
-/*  Following Structures and constant may be obsolete */
-/****************************************************************************/
-/* Maxium 8 bytes,the data read in will be placed in the parameter space. */
-/* Read operaion successeful when the paramter space is non-zero, otherwise read operation failed */
-typedef struct _READ_EDID_FROM_HW_I2C_DATA_PARAMETERS {
-	USHORT usPrescale;	/* Ratio between Engine clock and I2C clock */
-	USHORT usVRAMAddress;	/* Adress in Frame Buffer where to pace raw EDID */
-	USHORT usStatus;	/* When use output: lower byte EDID checksum, high byte hardware status */
-	/* WHen use input:  lower byte as 'byte to read':currently limited to 128byte or 1byte */
-	UCHAR ucSlaveAddr;	/* Read from which slave */
-	UCHAR ucLineNumber;	/* Read from which HW assisted line */
-} READ_EDID_FROM_HW_I2C_DATA_PARAMETERS;
+/****************************************************************************/	
+// Following Structures and constant may be obsolete
+/****************************************************************************/	
+//Maxium 8 bytes,the data read in will be placed in the parameter space.
+//Read operaion successeful when the paramter space is non-zero, otherwise read operation failed
+typedef struct _READ_EDID_FROM_HW_I2C_DATA_PARAMETERS
+{
+  USHORT    usPrescale;         //Ratio between Engine clock and I2C clock
+  USHORT    usVRAMAddress;      //Adress in Frame Buffer where to pace raw EDID
+  USHORT    usStatus;           //When use output: lower byte EDID checksum, high byte hardware status
+                                //WHen use input:  lower byte as 'byte to read':currently limited to 128byte or 1byte
+  UCHAR     ucSlaveAddr;        //Read from which slave
+  UCHAR     ucLineNumber;       //Read from which HW assisted line
+}READ_EDID_FROM_HW_I2C_DATA_PARAMETERS;
 #define READ_EDID_FROM_HW_I2C_DATA_PS_ALLOCATION  READ_EDID_FROM_HW_I2C_DATA_PARAMETERS
 
+
 #define  ATOM_WRITE_I2C_FORMAT_PSOFFSET_PSDATABYTE                  0
 #define  ATOM_WRITE_I2C_FORMAT_PSOFFSET_PSTWODATABYTES              1
 #define  ATOM_WRITE_I2C_FORMAT_PSCOUNTER_PSOFFSET_IDDATABLOCK       2
 #define  ATOM_WRITE_I2C_FORMAT_PSCOUNTER_IDOFFSET_PLUS_IDDATABLOCK  3
 #define  ATOM_WRITE_I2C_FORMAT_IDCOUNTER_IDOFFSET_IDDATABLOCK       4
 
-typedef struct _WRITE_ONE_BYTE_HW_I2C_DATA_PARAMETERS {
-	USHORT usPrescale;	/* Ratio between Engine clock and I2C clock */
-	USHORT usByteOffset;	/* Write to which byte */
-	/* Upper portion of usByteOffset is Format of data */
-	/* 1bytePS+offsetPS */
-	/* 2bytesPS+offsetPS */
-	/* blockID+offsetPS */
-	/* blockID+offsetID */
-	/* blockID+counterID+offsetID */
-	UCHAR ucData;		/* PS data1 */
-	UCHAR ucStatus;		/* Status byte 1=success, 2=failure, Also is used as PS data2 */
-	UCHAR ucSlaveAddr;	/* Write to which slave */
-	UCHAR ucLineNumber;	/* Write from which HW assisted line */
-} WRITE_ONE_BYTE_HW_I2C_DATA_PARAMETERS;
+typedef struct _WRITE_ONE_BYTE_HW_I2C_DATA_PARAMETERS
+{
+  USHORT    usPrescale;         //Ratio between Engine clock and I2C clock
+  USHORT    usByteOffset;       //Write to which byte
+                                //Upper portion of usByteOffset is Format of data 
+                                //1bytePS+offsetPS
+                                //2bytesPS+offsetPS
+                                //blockID+offsetPS
+                                //blockID+offsetID
+                                //blockID+counterID+offsetID
+  UCHAR     ucData;             //PS data1
+  UCHAR     ucStatus;           //Status byte 1=success, 2=failure, Also is used as PS data2
+  UCHAR     ucSlaveAddr;        //Write to which slave
+  UCHAR     ucLineNumber;       //Write from which HW assisted line
+}WRITE_ONE_BYTE_HW_I2C_DATA_PARAMETERS;
 
 #define WRITE_ONE_BYTE_HW_I2C_DATA_PS_ALLOCATION  WRITE_ONE_BYTE_HW_I2C_DATA_PARAMETERS
 
-typedef struct _SET_UP_HW_I2C_DATA_PARAMETERS {
-	USHORT usPrescale;	/* Ratio between Engine clock and I2C clock */
-	UCHAR ucSlaveAddr;	/* Write to which slave */
-	UCHAR ucLineNumber;	/* Write from which HW assisted line */
-} SET_UP_HW_I2C_DATA_PARAMETERS;
+typedef struct _SET_UP_HW_I2C_DATA_PARAMETERS
+{
+  USHORT    usPrescale;         //Ratio between Engine clock and I2C clock
+  UCHAR     ucSlaveAddr;        //Write to which slave
+  UCHAR     ucLineNumber;       //Write from which HW assisted line
+}SET_UP_HW_I2C_DATA_PARAMETERS;
+
 
 /**************************************************************************/
 #define SPEED_FAN_CONTROL_PS_ALLOCATION   WRITE_ONE_BYTE_HW_I2C_DATA_PARAMETERS
 
-/****************************************************************************/
-/*  Structures used by PowerConnectorDetectionTable */
-/****************************************************************************/
-typedef struct _POWER_CONNECTOR_DETECTION_PARAMETERS {
-	UCHAR ucPowerConnectorStatus;	/* Used for return value 0: detected, 1:not detected */
-	UCHAR ucPwrBehaviorId;
-	USHORT usPwrBudget;	/* how much power currently boot to in unit of watt */
-} POWER_CONNECTOR_DETECTION_PARAMETERS;
-
-typedef struct POWER_CONNECTOR_DETECTION_PS_ALLOCATION {
-	UCHAR ucPowerConnectorStatus;	/* Used for return value 0: detected, 1:not detected */
-	UCHAR ucReserved;
-	USHORT usPwrBudget;	/* how much power currently boot to in unit of watt */
-	WRITE_ONE_BYTE_HW_I2C_DATA_PS_ALLOCATION sReserved;
-} POWER_CONNECTOR_DETECTION_PS_ALLOCATION;
+/****************************************************************************/	
+// Structures used by PowerConnectorDetectionTable
+/****************************************************************************/	
+typedef struct	_POWER_CONNECTOR_DETECTION_PARAMETERS
+{
+  UCHAR   ucPowerConnectorStatus;      //Used for return value 0: detected, 1:not detected
+	UCHAR   ucPwrBehaviorId;							
+	USHORT	usPwrBudget;								 //how much power currently boot to in unit of watt
+}POWER_CONNECTOR_DETECTION_PARAMETERS;
+
+typedef struct POWER_CONNECTOR_DETECTION_PS_ALLOCATION
+{                               
+  UCHAR   ucPowerConnectorStatus;      //Used for return value 0: detected, 1:not detected
+	UCHAR   ucReserved;
+	USHORT	usPwrBudget;								 //how much power currently boot to in unit of watt
+  WRITE_ONE_BYTE_HW_I2C_DATA_PS_ALLOCATION    sReserved;
+}POWER_CONNECTOR_DETECTION_PS_ALLOCATION;
 
 /****************************LVDS SS Command Table Definitions**********************/
 
-/****************************************************************************/
-/*  Structures used by EnableSpreadSpectrumOnPPLLTable */
-/****************************************************************************/
-typedef struct _ENABLE_LVDS_SS_PARAMETERS {
-	USHORT usSpreadSpectrumPercentage;
-	UCHAR ucSpreadSpectrumType;	/* Bit1=0 Down Spread,=1 Center Spread. Bit1=1 Ext. =0 Int. Others:TBD */
-	UCHAR ucSpreadSpectrumStepSize_Delay;	/* bits3:2 SS_STEP_SIZE; bit 6:4 SS_DELAY */
-	UCHAR ucEnable;		/* ATOM_ENABLE or ATOM_DISABLE */
-	UCHAR ucPadding[3];
-} ENABLE_LVDS_SS_PARAMETERS;
-
-/* ucTableFormatRevision=1,ucTableContentRevision=2 */
-typedef struct _ENABLE_LVDS_SS_PARAMETERS_V2 {
-	USHORT usSpreadSpectrumPercentage;
-	UCHAR ucSpreadSpectrumType;	/* Bit1=0 Down Spread,=1 Center Spread. Bit1=1 Ext. =0 Int. Others:TBD */
-	UCHAR ucSpreadSpectrumStep;	/*  */
-	UCHAR ucEnable;		/* ATOM_ENABLE or ATOM_DISABLE */
-	UCHAR ucSpreadSpectrumDelay;
-	UCHAR ucSpreadSpectrumRange;
-	UCHAR ucPadding;
-} ENABLE_LVDS_SS_PARAMETERS_V2;
-
-/* This new structure is based on ENABLE_LVDS_SS_PARAMETERS but expands to SS on PPLL, so other devices can use SS. */
-typedef struct _ENABLE_SPREAD_SPECTRUM_ON_PPLL {
-	USHORT usSpreadSpectrumPercentage;
-	UCHAR ucSpreadSpectrumType;	/*  Bit1=0 Down Spread,=1 Center Spread. Bit1=1 Ext. =0 Int. Others:TBD */
-	UCHAR ucSpreadSpectrumStep;	/*  */
-	UCHAR ucEnable;		/*  ATOM_ENABLE or ATOM_DISABLE */
-	UCHAR ucSpreadSpectrumDelay;
-	UCHAR ucSpreadSpectrumRange;
-	UCHAR ucPpll;		/*  ATOM_PPLL1/ATOM_PPLL2 */
-} ENABLE_SPREAD_SPECTRUM_ON_PPLL;
+/****************************************************************************/	
+// Structures used by EnableSpreadSpectrumOnPPLLTable
+/****************************************************************************/	
+typedef struct	_ENABLE_LVDS_SS_PARAMETERS
+{
+  USHORT  usSpreadSpectrumPercentage;       
+  UCHAR   ucSpreadSpectrumType;           //Bit1=0 Down Spread,=1 Center Spread. Bit1=1 Ext. =0 Int. Others:TBD
+  UCHAR   ucSpreadSpectrumStepSize_Delay; //bits3:2 SS_STEP_SIZE; bit 6:4 SS_DELAY
+  UCHAR   ucEnable;                       //ATOM_ENABLE or ATOM_DISABLE
+  UCHAR   ucPadding[3];
+}ENABLE_LVDS_SS_PARAMETERS;
+
+//ucTableFormatRevision=1,ucTableContentRevision=2
+typedef struct	_ENABLE_LVDS_SS_PARAMETERS_V2
+{
+  USHORT  usSpreadSpectrumPercentage;       
+  UCHAR   ucSpreadSpectrumType;           //Bit1=0 Down Spread,=1 Center Spread. Bit1=1 Ext. =0 Int. Others:TBD
+  UCHAR   ucSpreadSpectrumStep;           //
+  UCHAR   ucEnable;                       //ATOM_ENABLE or ATOM_DISABLE
+  UCHAR   ucSpreadSpectrumDelay;
+  UCHAR   ucSpreadSpectrumRange;
+  UCHAR   ucPadding;
+}ENABLE_LVDS_SS_PARAMETERS_V2;
+
+//This new structure is based on ENABLE_LVDS_SS_PARAMETERS but expands to SS on PPLL, so other devices can use SS.
+typedef struct	_ENABLE_SPREAD_SPECTRUM_ON_PPLL
+{
+  USHORT  usSpreadSpectrumPercentage;
+  UCHAR   ucSpreadSpectrumType;           // Bit1=0 Down Spread,=1 Center Spread. Bit1=1 Ext. =0 Int. Others:TBD
+  UCHAR   ucSpreadSpectrumStep;           //
+  UCHAR   ucEnable;                       // ATOM_ENABLE or ATOM_DISABLE
+  UCHAR   ucSpreadSpectrumDelay;
+  UCHAR   ucSpreadSpectrumRange;
+  UCHAR   ucPpll;												  // ATOM_PPLL1/ATOM_PPLL2
+}ENABLE_SPREAD_SPECTRUM_ON_PPLL;
+
+typedef struct _ENABLE_SPREAD_SPECTRUM_ON_PPLL_V2
+{
+  USHORT  usSpreadSpectrumPercentage;
+  UCHAR   ucSpreadSpectrumType;	        // Bit[0]: 0-Down Spread,1-Center Spread. 
+                                        // Bit[1]: 1-Ext. 0-Int. 
+                                        // Bit[3:2]: =0 P1PLL =1 P2PLL =2 DCPLL
+                                        // Bits[7:4] reserved
+  UCHAR   ucEnable;	                    // ATOM_ENABLE or ATOM_DISABLE
+  USHORT  usSpreadSpectrumAmount;      	// Includes SS_AMOUNT_FBDIV[7:0] and SS_AMOUNT_NFRAC_SLIP[11:8]    
+  USHORT  usSpreadSpectrumStep;	        // SS_STEP_SIZE_DSFRAC
+}ENABLE_SPREAD_SPECTRUM_ON_PPLL_V2;
+
+#define ATOM_PPLL_SS_TYPE_V2_DOWN_SPREAD      0x00
+#define ATOM_PPLL_SS_TYPE_V2_CENTRE_SPREAD    0x01
+#define ATOM_PPLL_SS_TYPE_V2_EXT_SPREAD       0x02
+#define ATOM_PPLL_SS_TYPE_V2_PPLL_SEL_MASK    0x0c
+#define ATOM_PPLL_SS_TYPE_V2_P1PLL            0x00
+#define ATOM_PPLL_SS_TYPE_V2_P2PLL            0x04
+#define ATOM_PPLL_SS_TYPE_V2_DCPLL            0x08
+#define ATOM_PPLL_SS_AMOUNT_V2_FBDIV_MASK     0x00FF
+#define ATOM_PPLL_SS_AMOUNT_V2_FBDIV_SHIFT    0
+#define ATOM_PPLL_SS_AMOUNT_V2_NFRAC_MASK     0x0F00
+#define ATOM_PPLL_SS_AMOUNT_V2_NFRAC_SHIFT    8
 
 #define ENABLE_SPREAD_SPECTRUM_ON_PPLL_PS_ALLOCATION  ENABLE_SPREAD_SPECTRUM_ON_PPLL
 
 /**************************************************************************/
 
-typedef struct _SET_PIXEL_CLOCK_PS_ALLOCATION {
-	PIXEL_CLOCK_PARAMETERS sPCLKInput;
-	ENABLE_SPREAD_SPECTRUM_ON_PPLL sReserved;	/* Caller doesn't need to init this portion */
-} SET_PIXEL_CLOCK_PS_ALLOCATION;
+typedef struct _SET_PIXEL_CLOCK_PS_ALLOCATION
+{
+  PIXEL_CLOCK_PARAMETERS sPCLKInput;
+  ENABLE_SPREAD_SPECTRUM_ON_PPLL sReserved;//Caller doesn't need to init this portion 
+}SET_PIXEL_CLOCK_PS_ALLOCATION;
 
 #define ENABLE_VGA_RENDER_PS_ALLOCATION   SET_PIXEL_CLOCK_PS_ALLOCATION
 
-/****************************************************************************/
-/*  Structures used by ### */
-/****************************************************************************/
-typedef struct _MEMORY_TRAINING_PARAMETERS {
-	ULONG ulTargetMemoryClock;	/* In 10Khz unit */
-} MEMORY_TRAINING_PARAMETERS;
+/****************************************************************************/	
+// Structures used by ###
+/****************************************************************************/	
+typedef struct	_MEMORY_TRAINING_PARAMETERS
+{
+  ULONG ulTargetMemoryClock;          //In 10Khz unit
+}MEMORY_TRAINING_PARAMETERS;
 #define MEMORY_TRAINING_PS_ALLOCATION MEMORY_TRAINING_PARAMETERS
 
+
 /****************************LVDS and other encoder command table definitions **********************/
 
-/****************************************************************************/
-/*  Structures used by LVDSEncoderControlTable   (Before DCE30) */
-/*                     LVTMAEncoderControlTable  (Before DCE30) */
-/*                     TMDSAEncoderControlTable  (Before DCE30) */
-/****************************************************************************/
-typedef struct _LVDS_ENCODER_CONTROL_PARAMETERS {
-	USHORT usPixelClock;	/*  in 10KHz; for bios convenient */
-	UCHAR ucMisc;		/*  bit0=0: Enable single link */
-	/*      =1: Enable dual link */
-	/*  Bit1=0: 666RGB */
-	/*      =1: 888RGB */
-	UCHAR ucAction;		/*  0: turn off encoder */
-	/*  1: setup and turn on encoder */
-} LVDS_ENCODER_CONTROL_PARAMETERS;
 
-#define LVDS_ENCODER_CONTROL_PS_ALLOCATION  LVDS_ENCODER_CONTROL_PARAMETERS
+/****************************************************************************/	
+// Structures used by LVDSEncoderControlTable   (Before DCE30)
+//                    LVTMAEncoderControlTable  (Before DCE30)
+//                    TMDSAEncoderControlTable  (Before DCE30)
+/****************************************************************************/	
+typedef struct _LVDS_ENCODER_CONTROL_PARAMETERS
+{
+  USHORT usPixelClock;  // in 10KHz; for bios convenient
+  UCHAR  ucMisc;        // bit0=0: Enable single link
+                        //     =1: Enable dual link
+                        // Bit1=0: 666RGB
+                        //     =1: 888RGB
+  UCHAR  ucAction;      // 0: turn off encoder
+                        // 1: setup and turn on encoder
+}LVDS_ENCODER_CONTROL_PARAMETERS;
 
+#define LVDS_ENCODER_CONTROL_PS_ALLOCATION  LVDS_ENCODER_CONTROL_PARAMETERS
+   
 #define TMDS1_ENCODER_CONTROL_PARAMETERS    LVDS_ENCODER_CONTROL_PARAMETERS
 #define TMDS1_ENCODER_CONTROL_PS_ALLOCATION TMDS1_ENCODER_CONTROL_PARAMETERS
 
 #define TMDS2_ENCODER_CONTROL_PARAMETERS    TMDS1_ENCODER_CONTROL_PARAMETERS
 #define TMDS2_ENCODER_CONTROL_PS_ALLOCATION TMDS2_ENCODER_CONTROL_PARAMETERS
 
-/* ucTableFormatRevision=1,ucTableContentRevision=2 */
-typedef struct _LVDS_ENCODER_CONTROL_PARAMETERS_V2 {
-	USHORT usPixelClock;	/*  in 10KHz; for bios convenient */
-	UCHAR ucMisc;		/*  see PANEL_ENCODER_MISC_xx defintions below */
-	UCHAR ucAction;		/*  0: turn off encoder */
-	/*  1: setup and turn on encoder */
-	UCHAR ucTruncate;	/*  bit0=0: Disable truncate */
-	/*      =1: Enable truncate */
-	/*  bit4=0: 666RGB */
-	/*      =1: 888RGB */
-	UCHAR ucSpatial;	/*  bit0=0: Disable spatial dithering */
-	/*      =1: Enable spatial dithering */
-	/*  bit4=0: 666RGB */
-	/*      =1: 888RGB */
-	UCHAR ucTemporal;	/*  bit0=0: Disable temporal dithering */
-	/*      =1: Enable temporal dithering */
-	/*  bit4=0: 666RGB */
-	/*      =1: 888RGB */
-	/*  bit5=0: Gray level 2 */
-	/*      =1: Gray level 4 */
-	UCHAR ucFRC;		/*  bit4=0: 25FRC_SEL pattern E */
-	/*      =1: 25FRC_SEL pattern F */
-	/*  bit6:5=0: 50FRC_SEL pattern A */
-	/*        =1: 50FRC_SEL pattern B */
-	/*        =2: 50FRC_SEL pattern C */
-	/*        =3: 50FRC_SEL pattern D */
-	/*  bit7=0: 75FRC_SEL pattern E */
-	/*      =1: 75FRC_SEL pattern F */
-} LVDS_ENCODER_CONTROL_PARAMETERS_V2;
 
-#define LVDS_ENCODER_CONTROL_PS_ALLOCATION_V2  LVDS_ENCODER_CONTROL_PARAMETERS_V2
+//ucTableFormatRevision=1,ucTableContentRevision=2
+typedef struct _LVDS_ENCODER_CONTROL_PARAMETERS_V2
+{
+  USHORT usPixelClock;  // in 10KHz; for bios convenient
+  UCHAR  ucMisc;        // see PANEL_ENCODER_MISC_xx defintions below
+  UCHAR  ucAction;      // 0: turn off encoder
+                        // 1: setup and turn on encoder
+  UCHAR  ucTruncate;    // bit0=0: Disable truncate
+                        //     =1: Enable truncate
+                        // bit4=0: 666RGB
+                        //     =1: 888RGB
+  UCHAR  ucSpatial;     // bit0=0: Disable spatial dithering
+                        //     =1: Enable spatial dithering
+                        // bit4=0: 666RGB
+                        //     =1: 888RGB
+  UCHAR  ucTemporal;    // bit0=0: Disable temporal dithering
+                        //     =1: Enable temporal dithering
+                        // bit4=0: 666RGB
+                        //     =1: 888RGB
+                        // bit5=0: Gray level 2
+                        //     =1: Gray level 4
+  UCHAR  ucFRC;         // bit4=0: 25FRC_SEL pattern E
+                        //     =1: 25FRC_SEL pattern F
+                        // bit6:5=0: 50FRC_SEL pattern A
+                        //       =1: 50FRC_SEL pattern B
+                        //       =2: 50FRC_SEL pattern C
+                        //       =3: 50FRC_SEL pattern D
+                        // bit7=0: 75FRC_SEL pattern E
+                        //     =1: 75FRC_SEL pattern F
+}LVDS_ENCODER_CONTROL_PARAMETERS_V2;
 
+#define LVDS_ENCODER_CONTROL_PS_ALLOCATION_V2  LVDS_ENCODER_CONTROL_PARAMETERS_V2
+   
 #define TMDS1_ENCODER_CONTROL_PARAMETERS_V2    LVDS_ENCODER_CONTROL_PARAMETERS_V2
 #define TMDS1_ENCODER_CONTROL_PS_ALLOCATION_V2 TMDS1_ENCODER_CONTROL_PARAMETERS_V2
-
+  
 #define TMDS2_ENCODER_CONTROL_PARAMETERS_V2    TMDS1_ENCODER_CONTROL_PARAMETERS_V2
 #define TMDS2_ENCODER_CONTROL_PS_ALLOCATION_V2 TMDS2_ENCODER_CONTROL_PARAMETERS_V2
 
@@ -1185,38 +1536,42 @@ typedef struct _LVDS_ENCODER_CONTROL_PARAMETERS_V2 {
 #define TMDS2_ENCODER_CONTROL_PARAMETERS_V3    LVDS_ENCODER_CONTROL_PARAMETERS_V3
 #define TMDS2_ENCODER_CONTROL_PS_ALLOCATION_V3 TMDS2_ENCODER_CONTROL_PARAMETERS_V3
 
-/****************************************************************************/
-/*  Structures used by ### */
-/****************************************************************************/
-typedef struct _ENABLE_EXTERNAL_TMDS_ENCODER_PARAMETERS {
-	UCHAR ucEnable;		/*  Enable or Disable External TMDS encoder */
-	UCHAR ucMisc;		/*  Bit0=0:Enable Single link;=1:Enable Dual link;Bit1 {=0:666RGB, =1:888RGB} */
-	UCHAR ucPadding[2];
-} ENABLE_EXTERNAL_TMDS_ENCODER_PARAMETERS;
-
-typedef struct _ENABLE_EXTERNAL_TMDS_ENCODER_PS_ALLOCATION {
-	ENABLE_EXTERNAL_TMDS_ENCODER_PARAMETERS sXTmdsEncoder;
-	WRITE_ONE_BYTE_HW_I2C_DATA_PS_ALLOCATION sReserved;	/* Caller doesn't need to init this portion */
-} ENABLE_EXTERNAL_TMDS_ENCODER_PS_ALLOCATION;
+/****************************************************************************/	
+// Structures used by ###
+/****************************************************************************/	
+typedef struct _ENABLE_EXTERNAL_TMDS_ENCODER_PARAMETERS
+{                               
+  UCHAR    ucEnable;            // Enable or Disable External TMDS encoder
+  UCHAR    ucMisc;              // Bit0=0:Enable Single link;=1:Enable Dual link;Bit1 {=0:666RGB, =1:888RGB}
+  UCHAR    ucPadding[2];
+}ENABLE_EXTERNAL_TMDS_ENCODER_PARAMETERS;
+
+typedef struct _ENABLE_EXTERNAL_TMDS_ENCODER_PS_ALLOCATION
+{                               
+  ENABLE_EXTERNAL_TMDS_ENCODER_PARAMETERS    sXTmdsEncoder;
+  WRITE_ONE_BYTE_HW_I2C_DATA_PS_ALLOCATION   sReserved;     //Caller doesn't need to init this portion
+}ENABLE_EXTERNAL_TMDS_ENCODER_PS_ALLOCATION;
 
 #define ENABLE_EXTERNAL_TMDS_ENCODER_PARAMETERS_V2  LVDS_ENCODER_CONTROL_PARAMETERS_V2
 
-typedef struct _ENABLE_EXTERNAL_TMDS_ENCODER_PS_ALLOCATION_V2 {
-	ENABLE_EXTERNAL_TMDS_ENCODER_PARAMETERS_V2 sXTmdsEncoder;
-	WRITE_ONE_BYTE_HW_I2C_DATA_PS_ALLOCATION sReserved;	/* Caller doesn't need to init this portion */
-} ENABLE_EXTERNAL_TMDS_ENCODER_PS_ALLOCATION_V2;
+typedef struct _ENABLE_EXTERNAL_TMDS_ENCODER_PS_ALLOCATION_V2
+{                               
+  ENABLE_EXTERNAL_TMDS_ENCODER_PARAMETERS_V2    sXTmdsEncoder;
+  WRITE_ONE_BYTE_HW_I2C_DATA_PS_ALLOCATION      sReserved;     //Caller doesn't need to init this portion
+}ENABLE_EXTERNAL_TMDS_ENCODER_PS_ALLOCATION_V2;
 
-typedef struct _EXTERNAL_ENCODER_CONTROL_PS_ALLOCATION {
-	DIG_ENCODER_CONTROL_PARAMETERS sDigEncoder;
-	WRITE_ONE_BYTE_HW_I2C_DATA_PS_ALLOCATION sReserved;
-} EXTERNAL_ENCODER_CONTROL_PS_ALLOCATION;
+typedef struct _EXTERNAL_ENCODER_CONTROL_PS_ALLOCATION
+{
+  DIG_ENCODER_CONTROL_PARAMETERS            sDigEncoder;
+  WRITE_ONE_BYTE_HW_I2C_DATA_PS_ALLOCATION sReserved;
+}EXTERNAL_ENCODER_CONTROL_PS_ALLOCATION;
 
-/****************************************************************************/
-/*  Structures used by DVOEncoderControlTable */
-/****************************************************************************/
-/* ucTableFormatRevision=1,ucTableContentRevision=3 */
+/****************************************************************************/	
+// Structures used by DVOEncoderControlTable
+/****************************************************************************/	
+//ucTableFormatRevision=1,ucTableContentRevision=3
 
-/* ucDVOConfig: */
+//ucDVOConfig:
 #define DVO_ENCODER_CONFIG_RATE_SEL							0x01
 #define DVO_ENCODER_CONFIG_DDR_SPEED						0x00
 #define DVO_ENCODER_CONFIG_SDR_SPEED						0x01
@@ -1225,21 +1580,22 @@ typedef struct _EXTERNAL_ENCODER_CONTROL_PS_ALLOCATION {
 #define DVO_ENCODER_CONFIG_UPPER12BIT						0x04
 #define DVO_ENCODER_CONFIG_24BIT								0x08
 
-typedef struct _DVO_ENCODER_CONTROL_PARAMETERS_V3 {
-	USHORT usPixelClock;
-	UCHAR ucDVOConfig;
-	UCHAR ucAction;		/* ATOM_ENABLE/ATOM_DISABLE/ATOM_HPD_INIT */
-	UCHAR ucReseved[4];
-} DVO_ENCODER_CONTROL_PARAMETERS_V3;
+typedef struct _DVO_ENCODER_CONTROL_PARAMETERS_V3
+{
+  USHORT usPixelClock; 
+  UCHAR  ucDVOConfig;
+  UCHAR  ucAction;														//ATOM_ENABLE/ATOM_DISABLE/ATOM_HPD_INIT
+  UCHAR  ucReseved[4];
+}DVO_ENCODER_CONTROL_PARAMETERS_V3;
 #define DVO_ENCODER_CONTROL_PS_ALLOCATION_V3	DVO_ENCODER_CONTROL_PARAMETERS_V3
 
-/* ucTableFormatRevision=1 */
-/* ucTableContentRevision=3 structure is not changed but usMisc add bit 1 as another input for */
-/*  bit1=0: non-coherent mode */
-/*      =1: coherent mode */
+//ucTableFormatRevision=1
+//ucTableContentRevision=3 structure is not changed but usMisc add bit 1 as another input for 
+// bit1=0: non-coherent mode
+//     =1: coherent mode
 
-/* ========================================================================================== */
-/* Only change is here next time when changing encoder parameter definitions again! */
+//==========================================================================================
+//Only change is here next time when changing encoder parameter definitions again!
 #define LVDS_ENCODER_CONTROL_PARAMETERS_LAST     LVDS_ENCODER_CONTROL_PARAMETERS_V3
 #define LVDS_ENCODER_CONTROL_PS_ALLOCATION_LAST  LVDS_ENCODER_CONTROL_PARAMETERS_LAST
 
@@ -1252,7 +1608,7 @@ typedef struct _DVO_ENCODER_CONTROL_PARAMETERS_V3 {
 #define DVO_ENCODER_CONTROL_PARAMETERS_LAST      DVO_ENCODER_CONTROL_PARAMETERS
 #define DVO_ENCODER_CONTROL_PS_ALLOCATION_LAST   DVO_ENCODER_CONTROL_PS_ALLOCATION
 
-/* ========================================================================================== */
+//==========================================================================================
 #define PANEL_ENCODER_MISC_DUAL                0x01
 #define PANEL_ENCODER_MISC_COHERENT            0x02
 #define	PANEL_ENCODER_MISC_TMDS_LINKB					 0x04
@@ -1281,159 +1637,159 @@ typedef struct _DVO_ENCODER_CONTROL_PARAMETERS_V3 {
 #define PANEL_ENCODER_75FRC_E                  0x00
 #define PANEL_ENCODER_75FRC_F                  0x80
 
-/****************************************************************************/
-/*  Structures used by SetVoltageTable */
-/****************************************************************************/
+/****************************************************************************/	
+// Structures used by SetVoltageTable
+/****************************************************************************/	
 #define SET_VOLTAGE_TYPE_ASIC_VDDC             1
 #define SET_VOLTAGE_TYPE_ASIC_MVDDC            2
 #define SET_VOLTAGE_TYPE_ASIC_MVDDQ            3
 #define SET_VOLTAGE_TYPE_ASIC_VDDCI            4
 #define SET_VOLTAGE_INIT_MODE                  5
-#define SET_VOLTAGE_GET_MAX_VOLTAGE            6	/* Gets the Max. voltage for the soldered Asic */
+#define SET_VOLTAGE_GET_MAX_VOLTAGE            6					//Gets the Max. voltage for the soldered Asic
 
 #define SET_ASIC_VOLTAGE_MODE_ALL_SOURCE       0x1
 #define SET_ASIC_VOLTAGE_MODE_SOURCE_A         0x2
 #define SET_ASIC_VOLTAGE_MODE_SOURCE_B         0x4
 
 #define	SET_ASIC_VOLTAGE_MODE_SET_VOLTAGE      0x0
-#define	SET_ASIC_VOLTAGE_MODE_GET_GPIOVAL      0x1
+#define	SET_ASIC_VOLTAGE_MODE_GET_GPIOVAL      0x1	
 #define	SET_ASIC_VOLTAGE_MODE_GET_GPIOMASK     0x2
 
-typedef struct _SET_VOLTAGE_PARAMETERS {
-	UCHAR ucVoltageType;	/*  To tell which voltage to set up, VDDC/MVDDC/MVDDQ */
-	UCHAR ucVoltageMode;	/*  To set all, to set source A or source B or ... */
-	UCHAR ucVoltageIndex;	/*  An index to tell which voltage level */
-	UCHAR ucReserved;
-} SET_VOLTAGE_PARAMETERS;
-
-typedef struct _SET_VOLTAGE_PARAMETERS_V2 {
-	UCHAR ucVoltageType;	/*  To tell which voltage to set up, VDDC/MVDDC/MVDDQ */
-	UCHAR ucVoltageMode;	/*  Not used, maybe use for state machine for differen power mode */
-	USHORT usVoltageLevel;	/*  real voltage level */
-} SET_VOLTAGE_PARAMETERS_V2;
-
-typedef struct _SET_VOLTAGE_PS_ALLOCATION {
-	SET_VOLTAGE_PARAMETERS sASICSetVoltage;
-	WRITE_ONE_BYTE_HW_I2C_DATA_PS_ALLOCATION sReserved;
-} SET_VOLTAGE_PS_ALLOCATION;
-
-/****************************************************************************/
-/*  Structures used by TVEncoderControlTable */
-/****************************************************************************/
-typedef struct _TV_ENCODER_CONTROL_PARAMETERS {
-	USHORT usPixelClock;	/*  in 10KHz; for bios convenient */
-	UCHAR ucTvStandard;	/*  See definition "ATOM_TV_NTSC ..." */
-	UCHAR ucAction;		/*  0: turn off encoder */
-	/*  1: setup and turn on encoder */
-} TV_ENCODER_CONTROL_PARAMETERS;
-
-typedef struct _TV_ENCODER_CONTROL_PS_ALLOCATION {
-	TV_ENCODER_CONTROL_PARAMETERS sTVEncoder;
-	WRITE_ONE_BYTE_HW_I2C_DATA_PS_ALLOCATION sReserved;	/*  Don't set this one */
-} TV_ENCODER_CONTROL_PS_ALLOCATION;
-
-/* ==============================Data Table Portion==================================== */
-
-#ifdef	UEFI_BUILD
-#define	UTEMP	USHORT
-#define	USHORT	void*
-#endif
-
-/****************************************************************************/
-/*  Structure used in Data.mtb */
-/****************************************************************************/
-typedef struct _ATOM_MASTER_LIST_OF_DATA_TABLES {
-	USHORT UtilityPipeLine;	/*  Offest for the utility to get parser info,Don't change this position! */
-	USHORT MultimediaCapabilityInfo;	/*  Only used by MM Lib,latest version 1.1, not configuable from Bios, need to include the table to build Bios */
-	USHORT MultimediaConfigInfo;	/*  Only used by MM Lib,latest version 2.1, not configuable from Bios, need to include the table to build Bios */
-	USHORT StandardVESA_Timing;	/*  Only used by Bios */
-	USHORT FirmwareInfo;	/*  Shared by various SW components,latest version 1.4 */
-	USHORT DAC_Info;	/*  Will be obsolete from R600 */
-	USHORT LVDS_Info;	/*  Shared by various SW components,latest version 1.1 */
-	USHORT TMDS_Info;	/*  Will be obsolete from R600 */
-	USHORT AnalogTV_Info;	/*  Shared by various SW components,latest version 1.1 */
-	USHORT SupportedDevicesInfo;	/*  Will be obsolete from R600 */
-	USHORT GPIO_I2C_Info;	/*  Shared by various SW components,latest version 1.2 will be used from R600 */
-	USHORT VRAM_UsageByFirmware;	/*  Shared by various SW components,latest version 1.3 will be used from R600 */
-	USHORT GPIO_Pin_LUT;	/*  Shared by various SW components,latest version 1.1 */
-	USHORT VESA_ToInternalModeLUT;	/*  Only used by Bios */
-	USHORT ComponentVideoInfo;	/*  Shared by various SW components,latest version 2.1 will be used from R600 */
-	USHORT PowerPlayInfo;	/*  Shared by various SW components,latest version 2.1,new design from R600 */
-	USHORT CompassionateData;	/*  Will be obsolete from R600 */
-	USHORT SaveRestoreInfo;	/*  Only used by Bios */
-	USHORT PPLL_SS_Info;	/*  Shared by various SW components,latest version 1.2, used to call SS_Info, change to new name because of int ASIC SS info */
-	USHORT OemInfo;		/*  Defined and used by external SW, should be obsolete soon */
-	USHORT XTMDS_Info;	/*  Will be obsolete from R600 */
-	USHORT MclkSS_Info;	/*  Shared by various SW components,latest version 1.1, only enabled when ext SS chip is used */
-	USHORT Object_Header;	/*  Shared by various SW components,latest version 1.1 */
-	USHORT IndirectIOAccess;	/*  Only used by Bios,this table position can't change at all!! */
-	USHORT MC_InitParameter;	/*  Only used by command table */
-	USHORT ASIC_VDDC_Info;	/*  Will be obsolete from R600 */
-	USHORT ASIC_InternalSS_Info;	/*  New tabel name from R600, used to be called "ASIC_MVDDC_Info" */
-	USHORT TV_VideoMode;	/*  Only used by command table */
-	USHORT VRAM_Info;	/*  Only used by command table, latest version 1.3 */
-	USHORT MemoryTrainingInfo;	/*  Used for VBIOS and Diag utility for memory training purpose since R600. the new table rev start from 2.1 */
-	USHORT IntegratedSystemInfo;	/*  Shared by various SW components */
-	USHORT ASIC_ProfilingInfo;	/*  New table name from R600, used to be called "ASIC_VDDCI_Info" for pre-R600 */
-	USHORT VoltageObjectInfo;	/*  Shared by various SW components, latest version 1.1 */
-	USHORT PowerSourceInfo;	/*  Shared by various SW components, latest versoin 1.1 */
-} ATOM_MASTER_LIST_OF_DATA_TABLES;
-
-#ifdef	UEFI_BUILD
-#define	USHORT	UTEMP
-#endif
+typedef struct	_SET_VOLTAGE_PARAMETERS
+{
+  UCHAR    ucVoltageType;               // To tell which voltage to set up, VDDC/MVDDC/MVDDQ
+  UCHAR    ucVoltageMode;               // To set all, to set source A or source B or ...
+  UCHAR    ucVoltageIndex;              // An index to tell which voltage level
+  UCHAR    ucReserved;          
+}SET_VOLTAGE_PARAMETERS;
 
-typedef struct _ATOM_MASTER_DATA_TABLE {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	ATOM_MASTER_LIST_OF_DATA_TABLES ListOfDataTables;
-} ATOM_MASTER_DATA_TABLE;
+typedef struct	_SET_VOLTAGE_PARAMETERS_V2
+{
+  UCHAR    ucVoltageType;               // To tell which voltage to set up, VDDC/MVDDC/MVDDQ
+  UCHAR    ucVoltageMode;               // Not used, maybe use for state machine for differen power mode
+  USHORT   usVoltageLevel;              // real voltage level
+}SET_VOLTAGE_PARAMETERS_V2;
 
-/****************************************************************************/
-/*  Structure used in MultimediaCapabilityInfoTable */
-/****************************************************************************/
-typedef struct _ATOM_MULTIMEDIA_CAPABILITY_INFO {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	ULONG ulSignature;	/*  HW info table signature string "$ATI" */
-	UCHAR ucI2C_Type;	/*  I2C type (normal GP_IO, ImpactTV GP_IO, Dedicated I2C pin, etc) */
-	UCHAR ucTV_OutInfo;	/*  Type of TV out supported (3:0) and video out crystal frequency (6:4) and TV data port (7) */
-	UCHAR ucVideoPortInfo;	/*  Provides the video port capabilities */
-	UCHAR ucHostPortInfo;	/*  Provides host port configuration information */
-} ATOM_MULTIMEDIA_CAPABILITY_INFO;
+typedef struct _SET_VOLTAGE_PS_ALLOCATION
+{
+  SET_VOLTAGE_PARAMETERS sASICSetVoltage;
+  WRITE_ONE_BYTE_HW_I2C_DATA_PS_ALLOCATION sReserved;
+}SET_VOLTAGE_PS_ALLOCATION;
+
+/****************************************************************************/	
+// Structures used by TVEncoderControlTable
+/****************************************************************************/	
+typedef struct _TV_ENCODER_CONTROL_PARAMETERS
+{
+  USHORT usPixelClock;                // in 10KHz; for bios convenient
+  UCHAR  ucTvStandard;                // See definition "ATOM_TV_NTSC ..."
+  UCHAR  ucAction;                    // 0: turn off encoder
+                                      // 1: setup and turn on encoder
+}TV_ENCODER_CONTROL_PARAMETERS;
 
-/****************************************************************************/
-/*  Structure used in MultimediaConfigInfoTable */
-/****************************************************************************/
-typedef struct _ATOM_MULTIMEDIA_CONFIG_INFO {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	ULONG ulSignature;	/*  MM info table signature sting "$MMT" */
-	UCHAR ucTunerInfo;	/*  Type of tuner installed on the adapter (4:0) and video input for tuner (7:5) */
-	UCHAR ucAudioChipInfo;	/*  List the audio chip type (3:0) product type (4) and OEM revision (7:5) */
-	UCHAR ucProductID;	/*  Defines as OEM ID or ATI board ID dependent on product type setting */
-	UCHAR ucMiscInfo1;	/*  Tuner voltage (1:0) HW teletext support (3:2) FM audio decoder (5:4) reserved (6) audio scrambling (7) */
-	UCHAR ucMiscInfo2;	/*  I2S input config (0) I2S output config (1) I2S Audio Chip (4:2) SPDIF Output Config (5) reserved (7:6) */
-	UCHAR ucMiscInfo3;	/*  Video Decoder Type (3:0) Video In Standard/Crystal (7:4) */
-	UCHAR ucMiscInfo4;	/*  Video Decoder Host Config (2:0) reserved (7:3) */
-	UCHAR ucVideoInput0Info;	/*  Video Input 0 Type (1:0) F/B setting (2) physical connector ID (5:3) reserved (7:6) */
-	UCHAR ucVideoInput1Info;	/*  Video Input 1 Type (1:0) F/B setting (2) physical connector ID (5:3) reserved (7:6) */
-	UCHAR ucVideoInput2Info;	/*  Video Input 2 Type (1:0) F/B setting (2) physical connector ID (5:3) reserved (7:6) */
-	UCHAR ucVideoInput3Info;	/*  Video Input 3 Type (1:0) F/B setting (2) physical connector ID (5:3) reserved (7:6) */
-	UCHAR ucVideoInput4Info;	/*  Video Input 4 Type (1:0) F/B setting (2) physical connector ID (5:3) reserved (7:6) */
-} ATOM_MULTIMEDIA_CONFIG_INFO;
+typedef struct _TV_ENCODER_CONTROL_PS_ALLOCATION
+{
+  TV_ENCODER_CONTROL_PARAMETERS sTVEncoder;          
+  WRITE_ONE_BYTE_HW_I2C_DATA_PS_ALLOCATION    sReserved; // Don't set this one
+}TV_ENCODER_CONTROL_PS_ALLOCATION;
 
-/****************************************************************************/
-/*  Structures used in FirmwareInfoTable */
-/****************************************************************************/
+//==============================Data Table Portion====================================
 
-/*  usBIOSCapability Defintion: */
-/*  Bit 0 = 0: Bios image is not Posted, =1:Bios image is Posted; */
-/*  Bit 1 = 0: Dual CRTC is not supported, =1: Dual CRTC is supported; */
-/*  Bit 2 = 0: Extended Desktop is not supported, =1: Extended Desktop is supported; */
-/*  Others: Reserved */
+/****************************************************************************/	
+// Structure used in Data.mtb
+/****************************************************************************/	
+typedef struct _ATOM_MASTER_LIST_OF_DATA_TABLES
+{
+  USHORT        UtilityPipeLine;	        // Offest for the utility to get parser info,Don't change this position!
+  USHORT        MultimediaCapabilityInfo; // Only used by MM Lib,latest version 1.1, not configuable from Bios, need to include the table to build Bios 
+  USHORT        MultimediaConfigInfo;     // Only used by MM Lib,latest version 2.1, not configuable from Bios, need to include the table to build Bios
+  USHORT        StandardVESA_Timing;      // Only used by Bios
+  USHORT        FirmwareInfo;             // Shared by various SW components,latest version 1.4
+  USHORT        DAC_Info;                 // Will be obsolete from R600
+  USHORT        LVDS_Info;                // Shared by various SW components,latest version 1.1 
+  USHORT        TMDS_Info;                // Will be obsolete from R600
+  USHORT        AnalogTV_Info;            // Shared by various SW components,latest version 1.1 
+  USHORT        SupportedDevicesInfo;     // Will be obsolete from R600
+  USHORT        GPIO_I2C_Info;            // Shared by various SW components,latest version 1.2 will be used from R600           
+  USHORT        VRAM_UsageByFirmware;     // Shared by various SW components,latest version 1.3 will be used from R600
+  USHORT        GPIO_Pin_LUT;             // Shared by various SW components,latest version 1.1
+  USHORT        VESA_ToInternalModeLUT;   // Only used by Bios
+  USHORT        ComponentVideoInfo;       // Shared by various SW components,latest version 2.1 will be used from R600
+  USHORT        PowerPlayInfo;            // Shared by various SW components,latest version 2.1,new design from R600
+  USHORT        CompassionateData;        // Will be obsolete from R600
+  USHORT        SaveRestoreInfo;          // Only used by Bios
+  USHORT        PPLL_SS_Info;             // Shared by various SW components,latest version 1.2, used to call SS_Info, change to new name because of int ASIC SS info
+  USHORT        OemInfo;                  // Defined and used by external SW, should be obsolete soon
+  USHORT        XTMDS_Info;               // Will be obsolete from R600
+  USHORT        MclkSS_Info;              // Shared by various SW components,latest version 1.1, only enabled when ext SS chip is used
+  USHORT        Object_Header;            // Shared by various SW components,latest version 1.1
+  USHORT        IndirectIOAccess;         // Only used by Bios,this table position can't change at all!!
+  USHORT        MC_InitParameter;         // Only used by command table
+  USHORT        ASIC_VDDC_Info;						// Will be obsolete from R600
+  USHORT        ASIC_InternalSS_Info;			// New tabel name from R600, used to be called "ASIC_MVDDC_Info"
+  USHORT        TV_VideoMode;							// Only used by command table
+  USHORT        VRAM_Info;								// Only used by command table, latest version 1.3
+  USHORT        MemoryTrainingInfo;				// Used for VBIOS and Diag utility for memory training purpose since R600. the new table rev start from 2.1
+  USHORT        IntegratedSystemInfo;			// Shared by various SW components
+  USHORT        ASIC_ProfilingInfo;				// New table name from R600, used to be called "ASIC_VDDCI_Info" for pre-R600
+  USHORT        VoltageObjectInfo;				// Shared by various SW components, latest version 1.1
+	USHORT				PowerSourceInfo;					// Shared by various SW components, latest versoin 1.1
+}ATOM_MASTER_LIST_OF_DATA_TABLES;
+
+typedef struct _ATOM_MASTER_DATA_TABLE
+{ 
+  ATOM_COMMON_TABLE_HEADER sHeader;  
+  ATOM_MASTER_LIST_OF_DATA_TABLES   ListOfDataTables;
+}ATOM_MASTER_DATA_TABLE;
+
+/****************************************************************************/	
+// Structure used in MultimediaCapabilityInfoTable
+/****************************************************************************/	
+typedef struct _ATOM_MULTIMEDIA_CAPABILITY_INFO
+{
+  ATOM_COMMON_TABLE_HEADER sHeader;  
+  ULONG                    ulSignature;      // HW info table signature string "$ATI"
+  UCHAR                    ucI2C_Type;       // I2C type (normal GP_IO, ImpactTV GP_IO, Dedicated I2C pin, etc)
+  UCHAR                    ucTV_OutInfo;     // Type of TV out supported (3:0) and video out crystal frequency (6:4) and TV data port (7)
+  UCHAR                    ucVideoPortInfo;  // Provides the video port capabilities
+  UCHAR                    ucHostPortInfo;   // Provides host port configuration information
+}ATOM_MULTIMEDIA_CAPABILITY_INFO;
+
+/****************************************************************************/	
+// Structure used in MultimediaConfigInfoTable
+/****************************************************************************/	
+typedef struct _ATOM_MULTIMEDIA_CONFIG_INFO
+{
+  ATOM_COMMON_TABLE_HEADER sHeader;
+  ULONG                    ulSignature;      // MM info table signature sting "$MMT"
+  UCHAR                    ucTunerInfo;      // Type of tuner installed on the adapter (4:0) and video input for tuner (7:5)
+  UCHAR                    ucAudioChipInfo;  // List the audio chip type (3:0) product type (4) and OEM revision (7:5)
+  UCHAR                    ucProductID;      // Defines as OEM ID or ATI board ID dependent on product type setting
+  UCHAR                    ucMiscInfo1;      // Tuner voltage (1:0) HW teletext support (3:2) FM audio decoder (5:4) reserved (6) audio scrambling (7)
+  UCHAR                    ucMiscInfo2;      // I2S input config (0) I2S output config (1) I2S Audio Chip (4:2) SPDIF Output Config (5) reserved (7:6)
+  UCHAR                    ucMiscInfo3;      // Video Decoder Type (3:0) Video In Standard/Crystal (7:4)
+  UCHAR                    ucMiscInfo4;      // Video Decoder Host Config (2:0) reserved (7:3)
+  UCHAR                    ucVideoInput0Info;// Video Input 0 Type (1:0) F/B setting (2) physical connector ID (5:3) reserved (7:6)
+  UCHAR                    ucVideoInput1Info;// Video Input 1 Type (1:0) F/B setting (2) physical connector ID (5:3) reserved (7:6)
+  UCHAR                    ucVideoInput2Info;// Video Input 2 Type (1:0) F/B setting (2) physical connector ID (5:3) reserved (7:6)
+  UCHAR                    ucVideoInput3Info;// Video Input 3 Type (1:0) F/B setting (2) physical connector ID (5:3) reserved (7:6)
+  UCHAR                    ucVideoInput4Info;// Video Input 4 Type (1:0) F/B setting (2) physical connector ID (5:3) reserved (7:6)
+}ATOM_MULTIMEDIA_CONFIG_INFO;
+
+/****************************************************************************/	
+// Structures used in FirmwareInfoTable
+/****************************************************************************/	
+
+// usBIOSCapability Defintion:
+// Bit 0 = 0: Bios image is not Posted, =1:Bios image is Posted; 
+// Bit 1 = 0: Dual CRTC is not supported, =1: Dual CRTC is supported; 
+// Bit 2 = 0: Extended Desktop is not supported, =1: Extended Desktop is supported; 
+// Others: Reserved
 #define ATOM_BIOS_INFO_ATOM_FIRMWARE_POSTED         0x0001
 #define ATOM_BIOS_INFO_DUAL_CRTC_SUPPORT            0x0002
 #define ATOM_BIOS_INFO_EXTENDED_DESKTOP_SUPPORT     0x0004
-#define ATOM_BIOS_INFO_MEMORY_CLOCK_SS_SUPPORT      0x0008
-#define ATOM_BIOS_INFO_ENGINE_CLOCK_SS_SUPPORT      0x0010
+#define ATOM_BIOS_INFO_MEMORY_CLOCK_SS_SUPPORT      0x0008		// (valid from v1.1 ~v1.4):=1: memclk SS enable, =0 memclk SS disable. 
+#define ATOM_BIOS_INFO_ENGINE_CLOCK_SS_SUPPORT      0x0010		// (valid from v1.1 ~v1.4):=1: engclk SS enable, =0 engclk SS disable. 
 #define ATOM_BIOS_INFO_BL_CONTROLLED_BY_GPU         0x0020
 #define ATOM_BIOS_INFO_WMI_SUPPORT                  0x0040
 #define ATOM_BIOS_INFO_PPMODE_ASSIGNGED_BY_SYSTEM   0x0080
@@ -1441,242 +1797,292 @@ typedef struct _ATOM_MULTIMEDIA_CONFIG_INFO {
 #define ATOM_BIOS_INFO_HYPERMEMORY_SIZE_MASK        0x1E00
 #define ATOM_BIOS_INFO_VPOST_WITHOUT_FIRST_MODE_SET 0x2000
 #define ATOM_BIOS_INFO_BIOS_SCRATCH6_SCL2_REDEFINE  0x4000
+#define ATOM_BIOS_INFO_MEMORY_CLOCK_EXT_SS_SUPPORT  0x0008		// (valid from v2.1 ): =1: memclk ss enable with external ss chip
+#define ATOM_BIOS_INFO_ENGINE_CLOCK_EXT_SS_SUPPORT  0x0010		// (valid from v2.1 ): =1: engclk ss enable with external ss chip
 
 #ifndef _H2INC
 
-/* Please don't add or expand this bitfield structure below, this one will retire soon.! */
-typedef struct _ATOM_FIRMWARE_CAPABILITY {
+//Please don't add or expand this bitfield structure below, this one will retire soon.!
+typedef struct _ATOM_FIRMWARE_CAPABILITY
+{
 #if ATOM_BIG_ENDIAN
-	USHORT Reserved:3;
-	USHORT HyperMemory_Size:4;
-	USHORT HyperMemory_Support:1;
-	USHORT PPMode_Assigned:1;
-	USHORT WMI_SUPPORT:1;
-	USHORT GPUControlsBL:1;
-	USHORT EngineClockSS_Support:1;
-	USHORT MemoryClockSS_Support:1;
-	USHORT ExtendedDesktopSupport:1;
-	USHORT DualCRTC_Support:1;
-	USHORT FirmwarePosted:1;
+  USHORT Reserved:3;
+  USHORT HyperMemory_Size:4;
+  USHORT HyperMemory_Support:1;
+  USHORT PPMode_Assigned:1;
+  USHORT WMI_SUPPORT:1;
+  USHORT GPUControlsBL:1;
+  USHORT EngineClockSS_Support:1;
+  USHORT MemoryClockSS_Support:1;
+  USHORT ExtendedDesktopSupport:1;
+  USHORT DualCRTC_Support:1;
+  USHORT FirmwarePosted:1;
 #else
-	USHORT FirmwarePosted:1;
-	USHORT DualCRTC_Support:1;
-	USHORT ExtendedDesktopSupport:1;
-	USHORT MemoryClockSS_Support:1;
-	USHORT EngineClockSS_Support:1;
-	USHORT GPUControlsBL:1;
-	USHORT WMI_SUPPORT:1;
-	USHORT PPMode_Assigned:1;
-	USHORT HyperMemory_Support:1;
-	USHORT HyperMemory_Size:4;
-	USHORT Reserved:3;
+  USHORT FirmwarePosted:1;
+  USHORT DualCRTC_Support:1;
+  USHORT ExtendedDesktopSupport:1;
+  USHORT MemoryClockSS_Support:1;
+  USHORT EngineClockSS_Support:1;
+  USHORT GPUControlsBL:1;
+  USHORT WMI_SUPPORT:1;
+  USHORT PPMode_Assigned:1;
+  USHORT HyperMemory_Support:1;
+  USHORT HyperMemory_Size:4;
+  USHORT Reserved:3;
 #endif
-} ATOM_FIRMWARE_CAPABILITY;
+}ATOM_FIRMWARE_CAPABILITY;
 
-typedef union _ATOM_FIRMWARE_CAPABILITY_ACCESS {
-	ATOM_FIRMWARE_CAPABILITY sbfAccess;
-	USHORT susAccess;
-} ATOM_FIRMWARE_CAPABILITY_ACCESS;
+typedef union _ATOM_FIRMWARE_CAPABILITY_ACCESS
+{
+  ATOM_FIRMWARE_CAPABILITY sbfAccess;
+  USHORT                   susAccess;
+}ATOM_FIRMWARE_CAPABILITY_ACCESS;
 
 #else
 
-typedef union _ATOM_FIRMWARE_CAPABILITY_ACCESS {
-	USHORT susAccess;
-} ATOM_FIRMWARE_CAPABILITY_ACCESS;
+typedef union _ATOM_FIRMWARE_CAPABILITY_ACCESS
+{
+  USHORT                   susAccess;
+}ATOM_FIRMWARE_CAPABILITY_ACCESS;
 
 #endif
 
-typedef struct _ATOM_FIRMWARE_INFO {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	ULONG ulFirmwareRevision;
-	ULONG ulDefaultEngineClock;	/* In 10Khz unit */
-	ULONG ulDefaultMemoryClock;	/* In 10Khz unit */
-	ULONG ulDriverTargetEngineClock;	/* In 10Khz unit */
-	ULONG ulDriverTargetMemoryClock;	/* In 10Khz unit */
-	ULONG ulMaxEngineClockPLL_Output;	/* In 10Khz unit */
-	ULONG ulMaxMemoryClockPLL_Output;	/* In 10Khz unit */
-	ULONG ulMaxPixelClockPLL_Output;	/* In 10Khz unit */
-	ULONG ulASICMaxEngineClock;	/* In 10Khz unit */
-	ULONG ulASICMaxMemoryClock;	/* In 10Khz unit */
-	UCHAR ucASICMaxTemperature;
-	UCHAR ucPadding[3];	/* Don't use them */
-	ULONG aulReservedForBIOS[3];	/* Don't use them */
-	USHORT usMinEngineClockPLL_Input;	/* In 10Khz unit */
-	USHORT usMaxEngineClockPLL_Input;	/* In 10Khz unit */
-	USHORT usMinEngineClockPLL_Output;	/* In 10Khz unit */
-	USHORT usMinMemoryClockPLL_Input;	/* In 10Khz unit */
-	USHORT usMaxMemoryClockPLL_Input;	/* In 10Khz unit */
-	USHORT usMinMemoryClockPLL_Output;	/* In 10Khz unit */
-	USHORT usMaxPixelClock;	/* In 10Khz unit, Max.  Pclk */
-	USHORT usMinPixelClockPLL_Input;	/* In 10Khz unit */
-	USHORT usMaxPixelClockPLL_Input;	/* In 10Khz unit */
-	USHORT usMinPixelClockPLL_Output;	/* In 10Khz unit, the definitions above can't change!!! */
-	ATOM_FIRMWARE_CAPABILITY_ACCESS usFirmwareCapability;
-	USHORT usReferenceClock;	/* In 10Khz unit */
-	USHORT usPM_RTS_Location;	/* RTS PM4 starting location in ROM in 1Kb unit */
-	UCHAR ucPM_RTS_StreamSize;	/* RTS PM4 packets in Kb unit */
-	UCHAR ucDesign_ID;	/* Indicate what is the board design */
-	UCHAR ucMemoryModule_ID;	/* Indicate what is the board design */
-} ATOM_FIRMWARE_INFO;
-
-typedef struct _ATOM_FIRMWARE_INFO_V1_2 {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	ULONG ulFirmwareRevision;
-	ULONG ulDefaultEngineClock;	/* In 10Khz unit */
-	ULONG ulDefaultMemoryClock;	/* In 10Khz unit */
-	ULONG ulDriverTargetEngineClock;	/* In 10Khz unit */
-	ULONG ulDriverTargetMemoryClock;	/* In 10Khz unit */
-	ULONG ulMaxEngineClockPLL_Output;	/* In 10Khz unit */
-	ULONG ulMaxMemoryClockPLL_Output;	/* In 10Khz unit */
-	ULONG ulMaxPixelClockPLL_Output;	/* In 10Khz unit */
-	ULONG ulASICMaxEngineClock;	/* In 10Khz unit */
-	ULONG ulASICMaxMemoryClock;	/* In 10Khz unit */
-	UCHAR ucASICMaxTemperature;
-	UCHAR ucMinAllowedBL_Level;
-	UCHAR ucPadding[2];	/* Don't use them */
-	ULONG aulReservedForBIOS[2];	/* Don't use them */
-	ULONG ulMinPixelClockPLL_Output;	/* In 10Khz unit */
-	USHORT usMinEngineClockPLL_Input;	/* In 10Khz unit */
-	USHORT usMaxEngineClockPLL_Input;	/* In 10Khz unit */
-	USHORT usMinEngineClockPLL_Output;	/* In 10Khz unit */
-	USHORT usMinMemoryClockPLL_Input;	/* In 10Khz unit */
-	USHORT usMaxMemoryClockPLL_Input;	/* In 10Khz unit */
-	USHORT usMinMemoryClockPLL_Output;	/* In 10Khz unit */
-	USHORT usMaxPixelClock;	/* In 10Khz unit, Max.  Pclk */
-	USHORT usMinPixelClockPLL_Input;	/* In 10Khz unit */
-	USHORT usMaxPixelClockPLL_Input;	/* In 10Khz unit */
-	USHORT usMinPixelClockPLL_Output;	/* In 10Khz unit - lower 16bit of ulMinPixelClockPLL_Output */
-	ATOM_FIRMWARE_CAPABILITY_ACCESS usFirmwareCapability;
-	USHORT usReferenceClock;	/* In 10Khz unit */
-	USHORT usPM_RTS_Location;	/* RTS PM4 starting location in ROM in 1Kb unit */
-	UCHAR ucPM_RTS_StreamSize;	/* RTS PM4 packets in Kb unit */
-	UCHAR ucDesign_ID;	/* Indicate what is the board design */
-	UCHAR ucMemoryModule_ID;	/* Indicate what is the board design */
-} ATOM_FIRMWARE_INFO_V1_2;
-
-typedef struct _ATOM_FIRMWARE_INFO_V1_3 {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	ULONG ulFirmwareRevision;
-	ULONG ulDefaultEngineClock;	/* In 10Khz unit */
-	ULONG ulDefaultMemoryClock;	/* In 10Khz unit */
-	ULONG ulDriverTargetEngineClock;	/* In 10Khz unit */
-	ULONG ulDriverTargetMemoryClock;	/* In 10Khz unit */
-	ULONG ulMaxEngineClockPLL_Output;	/* In 10Khz unit */
-	ULONG ulMaxMemoryClockPLL_Output;	/* In 10Khz unit */
-	ULONG ulMaxPixelClockPLL_Output;	/* In 10Khz unit */
-	ULONG ulASICMaxEngineClock;	/* In 10Khz unit */
-	ULONG ulASICMaxMemoryClock;	/* In 10Khz unit */
-	UCHAR ucASICMaxTemperature;
-	UCHAR ucMinAllowedBL_Level;
-	UCHAR ucPadding[2];	/* Don't use them */
-	ULONG aulReservedForBIOS;	/* Don't use them */
-	ULONG ul3DAccelerationEngineClock;	/* In 10Khz unit */
-	ULONG ulMinPixelClockPLL_Output;	/* In 10Khz unit */
-	USHORT usMinEngineClockPLL_Input;	/* In 10Khz unit */
-	USHORT usMaxEngineClockPLL_Input;	/* In 10Khz unit */
-	USHORT usMinEngineClockPLL_Output;	/* In 10Khz unit */
-	USHORT usMinMemoryClockPLL_Input;	/* In 10Khz unit */
-	USHORT usMaxMemoryClockPLL_Input;	/* In 10Khz unit */
-	USHORT usMinMemoryClockPLL_Output;	/* In 10Khz unit */
-	USHORT usMaxPixelClock;	/* In 10Khz unit, Max.  Pclk */
-	USHORT usMinPixelClockPLL_Input;	/* In 10Khz unit */
-	USHORT usMaxPixelClockPLL_Input;	/* In 10Khz unit */
-	USHORT usMinPixelClockPLL_Output;	/* In 10Khz unit - lower 16bit of ulMinPixelClockPLL_Output */
-	ATOM_FIRMWARE_CAPABILITY_ACCESS usFirmwareCapability;
-	USHORT usReferenceClock;	/* In 10Khz unit */
-	USHORT usPM_RTS_Location;	/* RTS PM4 starting location in ROM in 1Kb unit */
-	UCHAR ucPM_RTS_StreamSize;	/* RTS PM4 packets in Kb unit */
-	UCHAR ucDesign_ID;	/* Indicate what is the board design */
-	UCHAR ucMemoryModule_ID;	/* Indicate what is the board design */
-} ATOM_FIRMWARE_INFO_V1_3;
-
-typedef struct _ATOM_FIRMWARE_INFO_V1_4 {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	ULONG ulFirmwareRevision;
-	ULONG ulDefaultEngineClock;	/* In 10Khz unit */
-	ULONG ulDefaultMemoryClock;	/* In 10Khz unit */
-	ULONG ulDriverTargetEngineClock;	/* In 10Khz unit */
-	ULONG ulDriverTargetMemoryClock;	/* In 10Khz unit */
-	ULONG ulMaxEngineClockPLL_Output;	/* In 10Khz unit */
-	ULONG ulMaxMemoryClockPLL_Output;	/* In 10Khz unit */
-	ULONG ulMaxPixelClockPLL_Output;	/* In 10Khz unit */
-	ULONG ulASICMaxEngineClock;	/* In 10Khz unit */
-	ULONG ulASICMaxMemoryClock;	/* In 10Khz unit */
-	UCHAR ucASICMaxTemperature;
-	UCHAR ucMinAllowedBL_Level;
-	USHORT usBootUpVDDCVoltage;	/* In MV unit */
-	USHORT usLcdMinPixelClockPLL_Output;	/*  In MHz unit */
-	USHORT usLcdMaxPixelClockPLL_Output;	/*  In MHz unit */
-	ULONG ul3DAccelerationEngineClock;	/* In 10Khz unit */
-	ULONG ulMinPixelClockPLL_Output;	/* In 10Khz unit */
-	USHORT usMinEngineClockPLL_Input;	/* In 10Khz unit */
-	USHORT usMaxEngineClockPLL_Input;	/* In 10Khz unit */
-	USHORT usMinEngineClockPLL_Output;	/* In 10Khz unit */
-	USHORT usMinMemoryClockPLL_Input;	/* In 10Khz unit */
-	USHORT usMaxMemoryClockPLL_Input;	/* In 10Khz unit */
-	USHORT usMinMemoryClockPLL_Output;	/* In 10Khz unit */
-	USHORT usMaxPixelClock;	/* In 10Khz unit, Max.  Pclk */
-	USHORT usMinPixelClockPLL_Input;	/* In 10Khz unit */
-	USHORT usMaxPixelClockPLL_Input;	/* In 10Khz unit */
-	USHORT usMinPixelClockPLL_Output;	/* In 10Khz unit - lower 16bit of ulMinPixelClockPLL_Output */
-	ATOM_FIRMWARE_CAPABILITY_ACCESS usFirmwareCapability;
-	USHORT usReferenceClock;	/* In 10Khz unit */
-	USHORT usPM_RTS_Location;	/* RTS PM4 starting location in ROM in 1Kb unit */
-	UCHAR ucPM_RTS_StreamSize;	/* RTS PM4 packets in Kb unit */
-	UCHAR ucDesign_ID;	/* Indicate what is the board design */
-	UCHAR ucMemoryModule_ID;	/* Indicate what is the board design */
-} ATOM_FIRMWARE_INFO_V1_4;
-
-#define ATOM_FIRMWARE_INFO_LAST  ATOM_FIRMWARE_INFO_V1_4
-
-/****************************************************************************/
-/*  Structures used in IntegratedSystemInfoTable */
-/****************************************************************************/
+typedef struct _ATOM_FIRMWARE_INFO
+{
+  ATOM_COMMON_TABLE_HEADER        sHeader; 
+  ULONG                           ulFirmwareRevision;
+  ULONG                           ulDefaultEngineClock;       //In 10Khz unit
+  ULONG                           ulDefaultMemoryClock;       //In 10Khz unit
+  ULONG                           ulDriverTargetEngineClock;  //In 10Khz unit
+  ULONG                           ulDriverTargetMemoryClock;  //In 10Khz unit
+  ULONG                           ulMaxEngineClockPLL_Output; //In 10Khz unit
+  ULONG                           ulMaxMemoryClockPLL_Output; //In 10Khz unit
+  ULONG                           ulMaxPixelClockPLL_Output;  //In 10Khz unit
+  ULONG                           ulASICMaxEngineClock;       //In 10Khz unit
+  ULONG                           ulASICMaxMemoryClock;       //In 10Khz unit
+  UCHAR                           ucASICMaxTemperature;
+  UCHAR                           ucPadding[3];               //Don't use them
+  ULONG                           aulReservedForBIOS[3];      //Don't use them
+  USHORT                          usMinEngineClockPLL_Input;  //In 10Khz unit
+  USHORT                          usMaxEngineClockPLL_Input;  //In 10Khz unit
+  USHORT                          usMinEngineClockPLL_Output; //In 10Khz unit
+  USHORT                          usMinMemoryClockPLL_Input;  //In 10Khz unit
+  USHORT                          usMaxMemoryClockPLL_Input;  //In 10Khz unit
+  USHORT                          usMinMemoryClockPLL_Output; //In 10Khz unit
+  USHORT                          usMaxPixelClock;            //In 10Khz unit, Max.  Pclk
+  USHORT                          usMinPixelClockPLL_Input;   //In 10Khz unit
+  USHORT                          usMaxPixelClockPLL_Input;   //In 10Khz unit
+  USHORT                          usMinPixelClockPLL_Output;  //In 10Khz unit, the definitions above can't change!!!
+  ATOM_FIRMWARE_CAPABILITY_ACCESS usFirmwareCapability;
+  USHORT                          usReferenceClock;           //In 10Khz unit	
+  USHORT                          usPM_RTS_Location;          //RTS PM4 starting location in ROM in 1Kb unit 
+  UCHAR                           ucPM_RTS_StreamSize;        //RTS PM4 packets in Kb unit
+  UCHAR                           ucDesign_ID;                //Indicate what is the board design
+  UCHAR                           ucMemoryModule_ID;          //Indicate what is the board design
+}ATOM_FIRMWARE_INFO;
+
+typedef struct _ATOM_FIRMWARE_INFO_V1_2
+{
+  ATOM_COMMON_TABLE_HEADER        sHeader; 
+  ULONG                           ulFirmwareRevision;
+  ULONG                           ulDefaultEngineClock;       //In 10Khz unit
+  ULONG                           ulDefaultMemoryClock;       //In 10Khz unit
+  ULONG                           ulDriverTargetEngineClock;  //In 10Khz unit
+  ULONG                           ulDriverTargetMemoryClock;  //In 10Khz unit
+  ULONG                           ulMaxEngineClockPLL_Output; //In 10Khz unit
+  ULONG                           ulMaxMemoryClockPLL_Output; //In 10Khz unit
+  ULONG                           ulMaxPixelClockPLL_Output;  //In 10Khz unit
+  ULONG                           ulASICMaxEngineClock;       //In 10Khz unit
+  ULONG                           ulASICMaxMemoryClock;       //In 10Khz unit
+  UCHAR                           ucASICMaxTemperature;
+  UCHAR                           ucMinAllowedBL_Level;
+  UCHAR                           ucPadding[2];               //Don't use them
+  ULONG                           aulReservedForBIOS[2];      //Don't use them
+  ULONG                           ulMinPixelClockPLL_Output;  //In 10Khz unit
+  USHORT                          usMinEngineClockPLL_Input;  //In 10Khz unit
+  USHORT                          usMaxEngineClockPLL_Input;  //In 10Khz unit
+  USHORT                          usMinEngineClockPLL_Output; //In 10Khz unit
+  USHORT                          usMinMemoryClockPLL_Input;  //In 10Khz unit
+  USHORT                          usMaxMemoryClockPLL_Input;  //In 10Khz unit
+  USHORT                          usMinMemoryClockPLL_Output; //In 10Khz unit
+  USHORT                          usMaxPixelClock;            //In 10Khz unit, Max.  Pclk
+  USHORT                          usMinPixelClockPLL_Input;   //In 10Khz unit
+  USHORT                          usMaxPixelClockPLL_Input;   //In 10Khz unit
+  USHORT                          usMinPixelClockPLL_Output;  //In 10Khz unit - lower 16bit of ulMinPixelClockPLL_Output
+  ATOM_FIRMWARE_CAPABILITY_ACCESS usFirmwareCapability;
+  USHORT                          usReferenceClock;           //In 10Khz unit	
+  USHORT                          usPM_RTS_Location;          //RTS PM4 starting location in ROM in 1Kb unit 
+  UCHAR                           ucPM_RTS_StreamSize;        //RTS PM4 packets in Kb unit
+  UCHAR                           ucDesign_ID;                //Indicate what is the board design
+  UCHAR                           ucMemoryModule_ID;          //Indicate what is the board design
+}ATOM_FIRMWARE_INFO_V1_2;
+
+typedef struct _ATOM_FIRMWARE_INFO_V1_3
+{
+  ATOM_COMMON_TABLE_HEADER        sHeader; 
+  ULONG                           ulFirmwareRevision;
+  ULONG                           ulDefaultEngineClock;       //In 10Khz unit
+  ULONG                           ulDefaultMemoryClock;       //In 10Khz unit
+  ULONG                           ulDriverTargetEngineClock;  //In 10Khz unit
+  ULONG                           ulDriverTargetMemoryClock;  //In 10Khz unit
+  ULONG                           ulMaxEngineClockPLL_Output; //In 10Khz unit
+  ULONG                           ulMaxMemoryClockPLL_Output; //In 10Khz unit
+  ULONG                           ulMaxPixelClockPLL_Output;  //In 10Khz unit
+  ULONG                           ulASICMaxEngineClock;       //In 10Khz unit
+  ULONG                           ulASICMaxMemoryClock;       //In 10Khz unit
+  UCHAR                           ucASICMaxTemperature;
+  UCHAR                           ucMinAllowedBL_Level;
+  UCHAR                           ucPadding[2];               //Don't use them
+  ULONG                           aulReservedForBIOS;         //Don't use them
+  ULONG                           ul3DAccelerationEngineClock;//In 10Khz unit
+  ULONG                           ulMinPixelClockPLL_Output;  //In 10Khz unit
+  USHORT                          usMinEngineClockPLL_Input;  //In 10Khz unit
+  USHORT                          usMaxEngineClockPLL_Input;  //In 10Khz unit
+  USHORT                          usMinEngineClockPLL_Output; //In 10Khz unit
+  USHORT                          usMinMemoryClockPLL_Input;  //In 10Khz unit
+  USHORT                          usMaxMemoryClockPLL_Input;  //In 10Khz unit
+  USHORT                          usMinMemoryClockPLL_Output; //In 10Khz unit
+  USHORT                          usMaxPixelClock;            //In 10Khz unit, Max.  Pclk
+  USHORT                          usMinPixelClockPLL_Input;   //In 10Khz unit
+  USHORT                          usMaxPixelClockPLL_Input;   //In 10Khz unit
+  USHORT                          usMinPixelClockPLL_Output;  //In 10Khz unit - lower 16bit of ulMinPixelClockPLL_Output
+  ATOM_FIRMWARE_CAPABILITY_ACCESS usFirmwareCapability;
+  USHORT                          usReferenceClock;           //In 10Khz unit	
+  USHORT                          usPM_RTS_Location;          //RTS PM4 starting location in ROM in 1Kb unit 
+  UCHAR                           ucPM_RTS_StreamSize;        //RTS PM4 packets in Kb unit
+  UCHAR                           ucDesign_ID;                //Indicate what is the board design
+  UCHAR                           ucMemoryModule_ID;          //Indicate what is the board design
+}ATOM_FIRMWARE_INFO_V1_3;
+
+typedef struct _ATOM_FIRMWARE_INFO_V1_4
+{
+  ATOM_COMMON_TABLE_HEADER        sHeader; 
+  ULONG                           ulFirmwareRevision;
+  ULONG                           ulDefaultEngineClock;       //In 10Khz unit
+  ULONG                           ulDefaultMemoryClock;       //In 10Khz unit
+  ULONG                           ulDriverTargetEngineClock;  //In 10Khz unit
+  ULONG                           ulDriverTargetMemoryClock;  //In 10Khz unit
+  ULONG                           ulMaxEngineClockPLL_Output; //In 10Khz unit
+  ULONG                           ulMaxMemoryClockPLL_Output; //In 10Khz unit
+  ULONG                           ulMaxPixelClockPLL_Output;  //In 10Khz unit
+  ULONG                           ulASICMaxEngineClock;       //In 10Khz unit
+  ULONG                           ulASICMaxMemoryClock;       //In 10Khz unit
+  UCHAR                           ucASICMaxTemperature;
+  UCHAR                           ucMinAllowedBL_Level;
+  USHORT                          usBootUpVDDCVoltage;        //In MV unit
+  USHORT                          usLcdMinPixelClockPLL_Output; // In MHz unit
+  USHORT                          usLcdMaxPixelClockPLL_Output; // In MHz unit
+  ULONG                           ul3DAccelerationEngineClock;//In 10Khz unit
+  ULONG                           ulMinPixelClockPLL_Output;  //In 10Khz unit
+  USHORT                          usMinEngineClockPLL_Input;  //In 10Khz unit
+  USHORT                          usMaxEngineClockPLL_Input;  //In 10Khz unit
+  USHORT                          usMinEngineClockPLL_Output; //In 10Khz unit
+  USHORT                          usMinMemoryClockPLL_Input;  //In 10Khz unit
+  USHORT                          usMaxMemoryClockPLL_Input;  //In 10Khz unit
+  USHORT                          usMinMemoryClockPLL_Output; //In 10Khz unit
+  USHORT                          usMaxPixelClock;            //In 10Khz unit, Max.  Pclk
+  USHORT                          usMinPixelClockPLL_Input;   //In 10Khz unit
+  USHORT                          usMaxPixelClockPLL_Input;   //In 10Khz unit
+  USHORT                          usMinPixelClockPLL_Output;  //In 10Khz unit - lower 16bit of ulMinPixelClockPLL_Output
+  ATOM_FIRMWARE_CAPABILITY_ACCESS usFirmwareCapability;
+  USHORT                          usReferenceClock;           //In 10Khz unit	
+  USHORT                          usPM_RTS_Location;          //RTS PM4 starting location in ROM in 1Kb unit 
+  UCHAR                           ucPM_RTS_StreamSize;        //RTS PM4 packets in Kb unit
+  UCHAR                           ucDesign_ID;                //Indicate what is the board design
+  UCHAR                           ucMemoryModule_ID;          //Indicate what is the board design
+}ATOM_FIRMWARE_INFO_V1_4;
+
+//the structure below to be used from Cypress
+typedef struct _ATOM_FIRMWARE_INFO_V2_1
+{
+  ATOM_COMMON_TABLE_HEADER        sHeader; 
+  ULONG                           ulFirmwareRevision;
+  ULONG                           ulDefaultEngineClock;       //In 10Khz unit
+  ULONG                           ulDefaultMemoryClock;       //In 10Khz unit
+  ULONG                           ulReserved1;
+  ULONG                           ulReserved2;
+  ULONG                           ulMaxEngineClockPLL_Output; //In 10Khz unit
+  ULONG                           ulMaxMemoryClockPLL_Output; //In 10Khz unit
+  ULONG                           ulMaxPixelClockPLL_Output;  //In 10Khz unit
+  ULONG                           ulBinaryAlteredInfo;        //Was ulASICMaxEngineClock
+  ULONG                           ulDefaultDispEngineClkFreq; //In 10Khz unit
+  UCHAR                           ucReserved1;                //Was ucASICMaxTemperature;
+  UCHAR                           ucMinAllowedBL_Level;
+  USHORT                          usBootUpVDDCVoltage;        //In MV unit
+  USHORT                          usLcdMinPixelClockPLL_Output; // In MHz unit
+  USHORT                          usLcdMaxPixelClockPLL_Output; // In MHz unit
+  ULONG                           ulReserved4;                //Was ulAsicMaximumVoltage
+  ULONG                           ulMinPixelClockPLL_Output;  //In 10Khz unit
+  USHORT                          usMinEngineClockPLL_Input;  //In 10Khz unit
+  USHORT                          usMaxEngineClockPLL_Input;  //In 10Khz unit
+  USHORT                          usMinEngineClockPLL_Output; //In 10Khz unit
+  USHORT                          usMinMemoryClockPLL_Input;  //In 10Khz unit
+  USHORT                          usMaxMemoryClockPLL_Input;  //In 10Khz unit
+  USHORT                          usMinMemoryClockPLL_Output; //In 10Khz unit
+  USHORT                          usMaxPixelClock;            //In 10Khz unit, Max.  Pclk
+  USHORT                          usMinPixelClockPLL_Input;   //In 10Khz unit
+  USHORT                          usMaxPixelClockPLL_Input;   //In 10Khz unit
+  USHORT                          usMinPixelClockPLL_Output;  //In 10Khz unit - lower 16bit of ulMinPixelClockPLL_Output
+  ATOM_FIRMWARE_CAPABILITY_ACCESS usFirmwareCapability;
+  USHORT                          usCoreReferenceClock;       //In 10Khz unit	
+  USHORT                          usMemoryReferenceClock;     //In 10Khz unit	
+  USHORT                          usUniphyDPModeExtClkFreq;   //In 10Khz unit, if it is 0, In DP Mode Uniphy Input clock from internal PPLL, otherwise Input clock from external Spread clock
+  UCHAR                           ucMemoryModule_ID;          //Indicate what is the board design
+  UCHAR                           ucReserved4[3];
+}ATOM_FIRMWARE_INFO_V2_1;
+
+
+#define ATOM_FIRMWARE_INFO_LAST  ATOM_FIRMWARE_INFO_V2_1
+
+/****************************************************************************/	
+// Structures used in IntegratedSystemInfoTable
+/****************************************************************************/	
 #define IGP_CAP_FLAG_DYNAMIC_CLOCK_EN      0x2
 #define IGP_CAP_FLAG_AC_CARD               0x4
 #define IGP_CAP_FLAG_SDVO_CARD             0x8
 #define IGP_CAP_FLAG_POSTDIV_BY_2_MODE     0x10
 
-typedef struct _ATOM_INTEGRATED_SYSTEM_INFO {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	ULONG ulBootUpEngineClock;	/* in 10kHz unit */
-	ULONG ulBootUpMemoryClock;	/* in 10kHz unit */
-	ULONG ulMaxSystemMemoryClock;	/* in 10kHz unit */
-	ULONG ulMinSystemMemoryClock;	/* in 10kHz unit */
-	UCHAR ucNumberOfCyclesInPeriodHi;
-	UCHAR ucLCDTimingSel;	/* =0:not valid.!=0 sel this timing descriptor from LCD EDID. */
-	USHORT usReserved1;
-	USHORT usInterNBVoltageLow;	/* An intermidiate PMW value to set the voltage */
-	USHORT usInterNBVoltageHigh;	/* Another intermidiate PMW value to set the voltage */
-	ULONG ulReserved[2];
-
-	USHORT usFSBClock;	/* In MHz unit */
-	USHORT usCapabilityFlag;	/* Bit0=1 indicates the fake HDMI support,Bit1=0/1 for Dynamic clocking dis/enable */
-	/* Bit[3:2]== 0:No PCIE card, 1:AC card, 2:SDVO card */
-	/* Bit[4]==1: P/2 mode, ==0: P/1 mode */
-	USHORT usPCIENBCfgReg7;	/* bit[7:0]=MUX_Sel, bit[9:8]=MUX_SEL_LEVEL2, bit[10]=Lane_Reversal */
-	USHORT usK8MemoryClock;	/* in MHz unit */
-	USHORT usK8SyncStartDelay;	/* in 0.01 us unit */
-	USHORT usK8DataReturnTime;	/* in 0.01 us unit */
-	UCHAR ucMaxNBVoltage;
-	UCHAR ucMinNBVoltage;
-	UCHAR ucMemoryType;	/* [7:4]=1:DDR1;=2:DDR2;=3:DDR3.[3:0] is reserved */
-	UCHAR ucNumberOfCyclesInPeriod;	/* CG.FVTHROT_PWM_CTRL_REG0.NumberOfCyclesInPeriod */
-	UCHAR ucStartingPWM_HighTime;	/* CG.FVTHROT_PWM_CTRL_REG0.StartingPWM_HighTime */
-	UCHAR ucHTLinkWidth;	/* 16 bit vs. 8 bit */
-	UCHAR ucMaxNBVoltageHigh;
-	UCHAR ucMinNBVoltageHigh;
-} ATOM_INTEGRATED_SYSTEM_INFO;
+typedef struct _ATOM_INTEGRATED_SYSTEM_INFO
+{
+  ATOM_COMMON_TABLE_HEADER        sHeader; 
+  ULONG	                          ulBootUpEngineClock;		    //in 10kHz unit
+  ULONG	                          ulBootUpMemoryClock;		    //in 10kHz unit
+  ULONG	                          ulMaxSystemMemoryClock;	    //in 10kHz unit
+  ULONG	                          ulMinSystemMemoryClock;	    //in 10kHz unit
+  UCHAR                           ucNumberOfCyclesInPeriodHi;
+  UCHAR                           ucLCDTimingSel;             //=0:not valid.!=0 sel this timing descriptor from LCD EDID.
+  USHORT                          usReserved1;
+  USHORT                          usInterNBVoltageLow;        //An intermidiate PMW value to set the voltage 
+  USHORT                          usInterNBVoltageHigh;       //Another intermidiate PMW value to set the voltage 
+  ULONG	                          ulReserved[2];
+
+  USHORT	                        usFSBClock;			            //In MHz unit
+  USHORT                          usCapabilityFlag;		        //Bit0=1 indicates the fake HDMI support,Bit1=0/1 for Dynamic clocking dis/enable
+																                              //Bit[3:2]== 0:No PCIE card, 1:AC card, 2:SDVO card
+                                                              //Bit[4]==1: P/2 mode, ==0: P/1 mode
+  USHORT	                        usPCIENBCfgReg7;				    //bit[7:0]=MUX_Sel, bit[9:8]=MUX_SEL_LEVEL2, bit[10]=Lane_Reversal
+  USHORT	                        usK8MemoryClock;            //in MHz unit
+  USHORT	                        usK8SyncStartDelay;         //in 0.01 us unit
+  USHORT	                        usK8DataReturnTime;         //in 0.01 us unit
+  UCHAR                           ucMaxNBVoltage;
+  UCHAR                           ucMinNBVoltage;
+  UCHAR                           ucMemoryType;					      //[7:4]=1:DDR1;=2:DDR2;=3:DDR3.[3:0] is reserved
+  UCHAR                           ucNumberOfCyclesInPeriod;		//CG.FVTHROT_PWM_CTRL_REG0.NumberOfCyclesInPeriod 
+  UCHAR                           ucStartingPWM_HighTime;     //CG.FVTHROT_PWM_CTRL_REG0.StartingPWM_HighTime
+  UCHAR                           ucHTLinkWidth;              //16 bit vs. 8 bit
+  UCHAR                           ucMaxNBVoltageHigh;    
+  UCHAR                           ucMinNBVoltageHigh;
+}ATOM_INTEGRATED_SYSTEM_INFO;
 
 /* Explanation on entries in ATOM_INTEGRATED_SYSTEM_INFO
-ulBootUpMemoryClock:    For Intel IGP,it's the UMA system memory clock
+ulBootUpMemoryClock:    For Intel IGP,it's the UMA system memory clock 
                         For AMD IGP,it's 0 if no SidePort memory installed or it's the boot-up SidePort memory clock
 ulMaxSystemMemoryClock: For Intel IGP,it's the Max freq from memory SPD if memory runs in ASYNC mode or otherwise (SYNC mode) it's 0
                         For AMD IGP,for now this can be 0
-ulMinSystemMemoryClock: For Intel IGP,it's 133MHz if memory runs in ASYNC mode or otherwise (SYNC mode) it's 0
+ulMinSystemMemoryClock: For Intel IGP,it's 133MHz if memory runs in ASYNC mode or otherwise (SYNC mode) it's 0 
                         For AMD IGP,for now this can be 0
 
-usFSBClock:             For Intel IGP,it's FSB Freq
+usFSBClock:             For Intel IGP,it's FSB Freq 
                         For AMD IGP,it's HT Link Speed
 
 usK8MemoryClock:        For AMD IGP only. For RevF CPU, set it to 200
@@ -1687,98 +2093,113 @@ VC:Voltage Control
 ucMaxNBVoltage:         Voltage regulator dependent PWM value. Low 8 bits of the value for the max voltage.Set this one to 0xFF if VC without PWM. Set this to 0x0 if no VC at all.
 ucMinNBVoltage:         Voltage regulator dependent PWM value. Low 8 bits of the value for the min voltage.Set this one to 0x00 if VC without PWM or no VC at all.
 
-ucNumberOfCyclesInPeriod:   Indicate how many cycles when PWM duty is 100%. low 8 bits of the value.
-ucNumberOfCyclesInPeriodHi: Indicate how many cycles when PWM duty is 100%. high 8 bits of the value.If the PWM has an inverter,set bit [7]==1,otherwise set it 0
+ucNumberOfCyclesInPeriod:   Indicate how many cycles when PWM duty is 100%. low 8 bits of the value. 
+ucNumberOfCyclesInPeriodHi: Indicate how many cycles when PWM duty is 100%. high 8 bits of the value.If the PWM has an inverter,set bit [7]==1,otherwise set it 0 
 
 ucMaxNBVoltageHigh:     Voltage regulator dependent PWM value. High 8 bits of  the value for the max voltage.Set this one to 0xFF if VC without PWM. Set this to 0x0 if no VC at all.
 ucMinNBVoltageHigh:     Voltage regulator dependent PWM value. High 8 bits of the value for the min voltage.Set this one to 0x00 if VC without PWM or no VC at all.
 
+
 usInterNBVoltageLow:    Voltage regulator dependent PWM value. The value makes the the voltage >=Min NB voltage but <=InterNBVoltageHigh. Set this to 0x0000 if VC without PWM or no VC at all.
 usInterNBVoltageHigh:   Voltage regulator dependent PWM value. The value makes the the voltage >=InterNBVoltageLow but <=Max NB voltage.Set this to 0x0000 if VC without PWM or no VC at all.
 */
 
+
 /*
 The following IGP table is introduced from RS780, which is supposed to be put by SBIOS in FB before IGP VBIOS starts VPOST;
-Then VBIOS will copy the whole structure to its image so all GPU SW components can access this data structure to get whatever they need.
+Then VBIOS will copy the whole structure to its image so all GPU SW components can access this data structure to get whatever they need. 
 The enough reservation should allow us to never change table revisions. Whenever needed, a GPU SW component can use reserved portion for new data entries.
 
 SW components can access the IGP system infor structure in the same way as before
 */
 
-typedef struct _ATOM_INTEGRATED_SYSTEM_INFO_V2 {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	ULONG ulBootUpEngineClock;	/* in 10kHz unit */
-	ULONG ulReserved1[2];	/* must be 0x0 for the reserved */
-	ULONG ulBootUpUMAClock;	/* in 10kHz unit */
-	ULONG ulBootUpSidePortClock;	/* in 10kHz unit */
-	ULONG ulMinSidePortClock;	/* in 10kHz unit */
-	ULONG ulReserved2[6];	/* must be 0x0 for the reserved */
-	ULONG ulSystemConfig;	/* see explanation below */
-	ULONG ulBootUpReqDisplayVector;
-	ULONG ulOtherDisplayMisc;
-	ULONG ulDDISlot1Config;
-	ULONG ulDDISlot2Config;
-	UCHAR ucMemoryType;	/* [3:0]=1:DDR1;=2:DDR2;=3:DDR3.[7:4] is reserved */
-	UCHAR ucUMAChannelNumber;
-	UCHAR ucDockingPinBit;
-	UCHAR ucDockingPinPolarity;
-	ULONG ulDockingPinCFGInfo;
-	ULONG ulCPUCapInfo;
-	USHORT usNumberOfCyclesInPeriod;
-	USHORT usMaxNBVoltage;
-	USHORT usMinNBVoltage;
-	USHORT usBootUpNBVoltage;
-	ULONG ulHTLinkFreq;	/* in 10Khz */
-	USHORT usMinHTLinkWidth;
-	USHORT usMaxHTLinkWidth;
-	USHORT usUMASyncStartDelay;
-	USHORT usUMADataReturnTime;
-	USHORT usLinkStatusZeroTime;
-	USHORT usReserved;
-	ULONG ulHighVoltageHTLinkFreq;	/*  in 10Khz */
-	ULONG ulLowVoltageHTLinkFreq;	/*  in 10Khz */
-	USHORT usMaxUpStreamHTLinkWidth;
-	USHORT usMaxDownStreamHTLinkWidth;
-	USHORT usMinUpStreamHTLinkWidth;
-	USHORT usMinDownStreamHTLinkWidth;
-	ULONG ulReserved3[97];	/* must be 0x0 */
-} ATOM_INTEGRATED_SYSTEM_INFO_V2;
+
+typedef struct _ATOM_INTEGRATED_SYSTEM_INFO_V2
+{
+  ATOM_COMMON_TABLE_HEADER   sHeader;
+  ULONG	                     ulBootUpEngineClock;       //in 10kHz unit
+  ULONG			     ulReserved1[2];            //must be 0x0 for the reserved
+  ULONG	                     ulBootUpUMAClock;          //in 10kHz unit
+  ULONG	                     ulBootUpSidePortClock;     //in 10kHz unit
+  ULONG	                     ulMinSidePortClock;        //in 10kHz unit
+  ULONG			     ulReserved2[6];            //must be 0x0 for the reserved
+  ULONG                      ulSystemConfig;            //see explanation below
+  ULONG                      ulBootUpReqDisplayVector;
+  ULONG                      ulOtherDisplayMisc;
+  ULONG                      ulDDISlot1Config;
+  ULONG                      ulDDISlot2Config;
+  UCHAR                      ucMemoryType;              //[3:0]=1:DDR1;=2:DDR2;=3:DDR3.[7:4] is reserved
+  UCHAR                      ucUMAChannelNumber;
+  UCHAR                      ucDockingPinBit;
+  UCHAR                      ucDockingPinPolarity;
+  ULONG                      ulDockingPinCFGInfo;
+  ULONG                      ulCPUCapInfo;
+  USHORT                     usNumberOfCyclesInPeriod;
+  USHORT                     usMaxNBVoltage;
+  USHORT                     usMinNBVoltage;
+  USHORT                     usBootUpNBVoltage;
+  ULONG                      ulHTLinkFreq;              //in 10Khz
+  USHORT                     usMinHTLinkWidth;
+  USHORT                     usMaxHTLinkWidth;
+  USHORT                     usUMASyncStartDelay;
+  USHORT                     usUMADataReturnTime;
+  USHORT                     usLinkStatusZeroTime;
+  USHORT                     usDACEfuse;				//for storing badgap value (for RS880 only)
+  ULONG                      ulHighVoltageHTLinkFreq;     // in 10Khz
+  ULONG                      ulLowVoltageHTLinkFreq;      // in 10Khz
+  USHORT                     usMaxUpStreamHTLinkWidth;
+  USHORT                     usMaxDownStreamHTLinkWidth;
+  USHORT                     usMinUpStreamHTLinkWidth;
+  USHORT                     usMinDownStreamHTLinkWidth;
+  USHORT                     usFirmwareVersion;         //0 means FW is not supported. Otherwise it's the FW version loaded by SBIOS and driver should enable FW.
+  USHORT                     usFullT0Time;             // Input to calculate minimum HT link change time required by NB P-State. Unit is 0.01us.
+  ULONG                      ulReserved3[96];          //must be 0x0
+}ATOM_INTEGRATED_SYSTEM_INFO_V2;   
 
 /*
 ulBootUpEngineClock:   Boot-up Engine Clock in 10Khz;
 ulBootUpUMAClock:      Boot-up UMA Clock in 10Khz; it must be 0x0 when UMA is not present
 ulBootUpSidePortClock: Boot-up SidePort Clock in 10Khz; it must be 0x0 when SidePort Memory is not present,this could be equal to or less than maximum supported Sideport memory clock
 
-ulSystemConfig:
-Bit[0]=1: PowerExpress mode =0 Non-PowerExpress mode;
+ulSystemConfig:  
+Bit[0]=1: PowerExpress mode =0 Non-PowerExpress mode; 
 Bit[1]=1: system boots up at AMD overdrived state or user customized  mode. In this case, driver will just stick to this boot-up mode. No other PowerPlay state
       =0: system boots up at driver control state. Power state depends on PowerPlay table.
 Bit[2]=1: PWM method is used on NB voltage control. =0: GPIO method is used.
 Bit[3]=1: Only one power state(Performance) will be supported.
       =0: Multiple power states supported from PowerPlay table.
-Bit[4]=1: CLMC is supported and enabled on current system.
-      =0: CLMC is not supported or enabled on current system. SBIOS need to support HT link/freq change through ATIF interface.
-Bit[5]=1: Enable CDLW for all driver control power states. Max HT width is from SBIOS, while Min HT width is determined by display requirement.
+Bit[4]=1: CLMC is supported and enabled on current system. 
+      =0: CLMC is not supported or enabled on current system. SBIOS need to support HT link/freq change through ATIF interface.  
+Bit[5]=1: Enable CDLW for all driver control power states. Max HT width is from SBIOS, while Min HT width is determined by display requirement.  
       =0: CDLW is disabled. If CLMC is enabled case, Min HT width will be set equal to Max HT width. If CLMC disabled case, Max HT width will be applied.
 Bit[6]=1: High Voltage requested for all power states. In this case, voltage will be forced at 1.1v and powerplay table voltage drop/throttling request will be ignored.
       =0: Voltage settings is determined by powerplay table.
 Bit[7]=1: Enable CLMC as hybrid Mode. CDLD and CILR will be disabled in this case and we're using legacy C1E. This is workaround for CPU(Griffin) performance issue.
       =0: Enable CLMC as regular mode, CDLD and CILR will be enabled.
+Bit[8]=1: CDLF is supported and enabled on current system.
+      =0: CDLF is not supported or enabled on current system.
+Bit[9]=1: DLL Shut Down feature is enabled on current system.
+      =0: DLL Shut Down feature is not enabled or supported on current system.
 
 ulBootUpReqDisplayVector: This dword is a bit vector indicates what display devices are requested during boot-up. Refer to ATOM_DEVICE_xxx_SUPPORT for the bit vector definitions.
 
 ulOtherDisplayMisc: [15:8]- Bootup LCD Expansion selection; 0-center, 1-full panel size expansion;
-			              [7:0] - BootupTV standard selection; This is a bit vector to indicate what TV standards are supported by the system. Refer to ucTVSuppportedStd definition;
+			              [7:0] - BootupTV standard selection; This is a bit vector to indicate what TV standards are supported by the system. Refer to ucTVSupportedStd definition;
 
 ulDDISlot1Config: Describes the PCIE lane configuration on this DDI PCIE slot (ADD2 card) or connector (Mobile design).
       [3:0]  - Bit vector to indicate PCIE lane config of the DDI slot/connector on chassis (bit 0=1 lane 3:0; bit 1=1 lane 7:4; bit 2=1 lane 11:8; bit 3=1 lane 15:12)
-			[7:4]  - Bit vector to indicate PCIE lane config of the same DDI slot/connector on docking station (bit 0=1 lane 3:0; bit 1=1 lane 7:4; bit 2=1 lane 11:8; bit 3=1 lane 15:12)
-			[15:8] - Lane configuration attribute;
+			[7:4]  - Bit vector to indicate PCIE lane config of the same DDI slot/connector on docking station (bit 4=1 lane 3:0; bit 5=1 lane 7:4; bit 6=1 lane 11:8; bit 7=1 lane 15:12)
+      When a DDI connector is not "paired" (meaming two connections mutualexclusive on chassis or docking, only one of them can be connected at one time.
+      in both chassis and docking, SBIOS has to duplicate the same PCIE lane info from chassis to docking or vice versa. For example:
+      one DDI connector is only populated in docking with PCIE lane 8-11, but there is no paired connection on chassis, SBIOS has to copy bit 6 to bit 2.
+
+			[15:8] - Lane configuration attribute; 
       [23:16]- Connector type, possible value:
                CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_D
                CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_D
                CONNECTOR_OBJECT_ID_HDMI_TYPE_A
                CONNECTOR_OBJECT_ID_DISPLAYPORT
+               CONNECTOR_OBJECT_ID_eDP
 			[31:24]- Reserved
 
 ulDDISlot2Config: Same as Slot1.
@@ -1787,29 +2208,31 @@ For IGP, Hypermemory is the only memory type showed in CCC.
 
 ucUMAChannelNumber:  how many channels for the UMA;
 
-ulDockingPinCFGInfo: [15:0]-Bus/Device/Function # to CFG to read this Docking Pin; [31:16]-reg offset in CFG to read this pin
+ulDockingPinCFGInfo: [15:0]-Bus/Device/Function # to CFG to read this Docking Pin; [31:16]-reg offset in CFG to read this pin 
 ucDockingPinBit:     which bit in this register to read the pin status;
 ucDockingPinPolarity:Polarity of the pin when docked;
 
 ulCPUCapInfo:        [7:0]=1:Griffin;[7:0]=2:Greyhound;[7:0]=3:K8, other bits reserved for now and must be 0x0
 
 usNumberOfCyclesInPeriod:Indicate how many cycles when PWM duty is 100%.
-usMaxNBVoltage:Max. voltage control value in either PWM or GPIO mode.
+
+usMaxNBVoltage:Max. voltage control value in either PWM or GPIO mode. 
 usMinNBVoltage:Min. voltage control value in either PWM or GPIO mode.
                     GPIO mode: both usMaxNBVoltage & usMinNBVoltage have a valid value ulSystemConfig.SYSTEM_CONFIG_USE_PWM_ON_VOLTAGE=0
                     PWM mode: both usMaxNBVoltage & usMinNBVoltage have a valid value ulSystemConfig.SYSTEM_CONFIG_USE_PWM_ON_VOLTAGE=1
                     GPU SW don't control mode: usMaxNBVoltage & usMinNBVoltage=0 and no care about ulSystemConfig.SYSTEM_CONFIG_USE_PWM_ON_VOLTAGE
+
 usBootUpNBVoltage:Boot-up voltage regulator dependent PWM value.
 
 ulHTLinkFreq:       Bootup HT link Frequency in 10Khz.
-usMinHTLinkWidth:   Bootup minimum HT link width. If CDLW disabled, this is equal to usMaxHTLinkWidth.
-                    If CDLW enabled, both upstream and downstream width should be the same during bootup.
-usMaxHTLinkWidth:   Bootup maximum HT link width. If CDLW disabled, this is equal to usMinHTLinkWidth.
+usMinHTLinkWidth:   Bootup minimum HT link width. If CDLW disabled, this is equal to usMaxHTLinkWidth. 
                     If CDLW enabled, both upstream and downstream width should be the same during bootup.
+usMaxHTLinkWidth:   Bootup maximum HT link width. If CDLW disabled, this is equal to usMinHTLinkWidth. 
+                    If CDLW enabled, both upstream and downstream width should be the same during bootup.  
 
-usUMASyncStartDelay: Memory access latency, required for watermark calculation
+usUMASyncStartDelay: Memory access latency, required for watermark calculation 
 usUMADataReturnTime: Memory access latency, required for watermark calculation
-usLinkStatusZeroTime:Memory access latency required for watermark calculation, set this to 0x0 for K8 CPU, set a proper value in 0.01 the unit of us
+usLinkStatusZeroTime:Memory access latency required for watermark calculation, set this to 0x0 for K8 CPU, set a proper value in 0.01 the unit of us 
 for Griffin or Greyhound. SBIOS needs to convert to actual time by:
                      if T0Ttime [5:4]=00b, then usLinkStatusZeroTime=T0Ttime [3:0]*0.1us (0.0 to 1.5us)
                      if T0Ttime [5:4]=01b, then usLinkStatusZeroTime=T0Ttime [3:0]*0.5us (0.0 to 7.5us)
@@ -1817,7 +2240,7 @@ for Griffin or Greyhound. SBIOS needs to convert to actual time by:
                      if T0Ttime [5:4]=11b, and T0Ttime [3:0]=0x0 to 0xa, then usLinkStatusZeroTime=T0Ttime [3:0]*20us (0.0 to 200us)
 
 ulHighVoltageHTLinkFreq:     HT link frequency for power state with low voltage. If boot up runs in HT1, this must be 0.
-                             This must be less than or equal to ulHTLinkFreq(bootup frequency).
+                             This must be less than or equal to ulHTLinkFreq(bootup frequency). 
 ulLowVoltageHTLinkFreq:      HT link frequency for power state with low voltage or voltage scaling 1.0v~1.1v. If boot up runs in HT1, this must be 0.
                              This must be less than or equal to ulHighVoltageHTLinkFreq.
 
@@ -1827,14 +2250,17 @@ usMinUpStreamHTLinkWidth:    Asymmetric link width support in the future, to rep
 usMinDownStreamHTLinkWidth:  same as above.
 */
 
+
 #define SYSTEM_CONFIG_POWEREXPRESS_ENABLE                 0x00000001
 #define SYSTEM_CONFIG_RUN_AT_OVERDRIVE_ENGINE             0x00000002
-#define SYSTEM_CONFIG_USE_PWM_ON_VOLTAGE                  0x00000004
+#define SYSTEM_CONFIG_USE_PWM_ON_VOLTAGE                  0x00000004 
 #define SYSTEM_CONFIG_PERFORMANCE_POWERSTATE_ONLY         0x00000008
 #define SYSTEM_CONFIG_CLMC_ENABLED                        0x00000010
 #define SYSTEM_CONFIG_CDLW_ENABLED                        0x00000020
 #define SYSTEM_CONFIG_HIGH_VOLTAGE_REQUESTED              0x00000040
 #define SYSTEM_CONFIG_CLMC_HYBRID_MODE_ENABLED            0x00000080
+#define SYSTEM_CONFIG_CDLF_ENABLED                        0x00000100
+#define SYSTEM_CONFIG_DLL_SHUTDOWN_ENABLED                0x00000200
 
 #define IGP_DDI_SLOT_LANE_CONFIG_MASK                     0x000000FF
 
@@ -1851,6 +2277,41 @@ usMinDownStreamHTLinkWidth:  same as above.
 
 #define IGP_DDI_SLOT_CONNECTOR_TYPE_MASK                  0x00FF0000
 
+// IntegratedSystemInfoTable new Rev is V5 after V2, because of the real rev of V2 is v1.4. This rev is used for RR
+typedef struct _ATOM_INTEGRATED_SYSTEM_INFO_V5
+{
+  ATOM_COMMON_TABLE_HEADER   sHeader;
+  ULONG	                     ulBootUpEngineClock;       //in 10kHz unit
+  ULONG                      ulDentistVCOFreq;          //Dentist VCO clock in 10kHz unit, the source of GPU SCLK, LCLK, UCLK and VCLK. 
+  ULONG                      ulLClockFreq;              //GPU Lclk freq in 10kHz unit, have relationship with NCLK in NorthBridge
+  ULONG	                     ulBootUpUMAClock;          //in 10kHz unit
+  ULONG                      ulReserved1[8];            //must be 0x0 for the reserved
+  ULONG                      ulBootUpReqDisplayVector;
+  ULONG                      ulOtherDisplayMisc;
+  ULONG                      ulReserved2[4];            //must be 0x0 for the reserved
+  ULONG                      ulSystemConfig;            //TBD
+  ULONG                      ulCPUCapInfo;              //TBD
+  USHORT                     usMaxNBVoltage;            //high NB voltage, calculated using current VDDNB (D24F2xDC) and VDDNB offset fuse;
+  USHORT                     usMinNBVoltage;            //low NB voltage, calculated using current VDDNB (D24F2xDC) and VDDNB offset fuse;
+  USHORT                     usBootUpNBVoltage;         //boot up NB voltage
+  UCHAR                      ucHtcTmpLmt;               //bit [22:16] of D24F3x64 Hardware Thermal Control (HTC) Register, may not be needed, TBD
+  UCHAR                      ucTjOffset;                //bit [28:22] of D24F3xE4 Thermtrip Status Register,may not be needed, TBD
+  ULONG                      ulReserved3[4];            //must be 0x0 for the reserved
+  ULONG                      ulDDISlot1Config;          //see above ulDDISlot1Config definition
+  ULONG                      ulDDISlot2Config;
+  ULONG                      ulDDISlot3Config;
+  ULONG                      ulDDISlot4Config;
+  ULONG                      ulReserved4[4];            //must be 0x0 for the reserved
+  UCHAR                      ucMemoryType;              //[3:0]=1:DDR1;=2:DDR2;=3:DDR3.[7:4] is reserved
+  UCHAR                      ucUMAChannelNumber;
+  USHORT                     usReserved;
+  ULONG                      ulReserved5[4];            //must be 0x0 for the reserved
+  ULONG                      ulCSR_M3_ARB_CNTL_DEFAULT[10];//arrays with values for CSR M3 arbiter for default
+  ULONG                      ulCSR_M3_ARB_CNTL_UVD[10]; //arrays with values for CSR M3 arbiter for UVD playback
+  ULONG                      ulCSR_M3_ARB_CNTL_FS3D[10];//arrays with values for CSR M3 arbiter for Full Screen 3D applications
+  ULONG                      ulReserved6[61];           //must be 0x0
+}ATOM_INTEGRATED_SYSTEM_INFO_V5;   
+
 #define ATOM_CRT_INT_ENCODER1_INDEX                       0x00000000
 #define ATOM_LCD_INT_ENCODER1_INDEX                       0x00000001
 #define ATOM_TV_INT_ENCODER1_INDEX                        0x00000002
@@ -1866,8 +2327,8 @@ usMinDownStreamHTLinkWidth:  same as above.
 #define ATOM_DFP_INT_ENCODER3_INDEX                       0x0000000C
 #define ATOM_DFP_INT_ENCODER4_INDEX                       0x0000000D
 
-/*  define ASIC internal encoder id ( bit vector ) */
-#define ASIC_INT_DAC1_ENCODER_ID											0x00
+// define ASIC internal encoder id ( bit vector ), used for CRTC_SourceSelTable
+#define ASIC_INT_DAC1_ENCODER_ID    											0x00 
 #define ASIC_INT_TV_ENCODER_ID														0x02
 #define ASIC_INT_DIG1_ENCODER_ID													0x03
 #define ASIC_INT_DAC2_ENCODER_ID													0x04
@@ -1875,10 +2336,24 @@ usMinDownStreamHTLinkWidth:  same as above.
 #define ASIC_INT_DVO_ENCODER_ID														0x07
 #define ASIC_INT_DIG2_ENCODER_ID													0x09
 #define ASIC_EXT_DIG_ENCODER_ID														0x05
+#define ASIC_EXT_DIG2_ENCODER_ID													0x08
+#define ASIC_INT_DIG3_ENCODER_ID													0x0a
+#define ASIC_INT_DIG4_ENCODER_ID													0x0b
+#define ASIC_INT_DIG5_ENCODER_ID													0x0c
+#define ASIC_INT_DIG6_ENCODER_ID													0x0d
 
-/* define Encoder attribute */
+//define Encoder attribute
 #define ATOM_ANALOG_ENCODER																0
-#define ATOM_DIGITAL_ENCODER															1
+#define ATOM_DIGITAL_ENCODER															1		
+#define ATOM_DP_ENCODER															      2		
+
+#define ATOM_ENCODER_ENUM_MASK                            0x70
+#define ATOM_ENCODER_ENUM_ID1                             0x00
+#define ATOM_ENCODER_ENUM_ID2                             0x10
+#define ATOM_ENCODER_ENUM_ID3                             0x20
+#define ATOM_ENCODER_ENUM_ID4                             0x30
+#define ATOM_ENCODER_ENUM_ID5                             0x40 
+#define ATOM_ENCODER_ENUM_ID6                             0x50
 
 #define ATOM_DEVICE_CRT1_INDEX                            0x00000000
 #define ATOM_DEVICE_LCD1_INDEX                            0x00000001
@@ -1886,45 +2361,40 @@ usMinDownStreamHTLinkWidth:  same as above.
 #define ATOM_DEVICE_DFP1_INDEX                            0x00000003
 #define ATOM_DEVICE_CRT2_INDEX                            0x00000004
 #define ATOM_DEVICE_LCD2_INDEX                            0x00000005
-#define ATOM_DEVICE_TV2_INDEX                             0x00000006
+#define ATOM_DEVICE_DFP6_INDEX                            0x00000006
 #define ATOM_DEVICE_DFP2_INDEX                            0x00000007
 #define ATOM_DEVICE_CV_INDEX                              0x00000008
-#define ATOM_DEVICE_DFP3_INDEX														0x00000009
-#define ATOM_DEVICE_DFP4_INDEX														0x0000000A
-#define ATOM_DEVICE_DFP5_INDEX														0x0000000B
+#define ATOM_DEVICE_DFP3_INDEX                            0x00000009
+#define ATOM_DEVICE_DFP4_INDEX                            0x0000000A
+#define ATOM_DEVICE_DFP5_INDEX                            0x0000000B
+
 #define ATOM_DEVICE_RESERVEDC_INDEX                       0x0000000C
 #define ATOM_DEVICE_RESERVEDD_INDEX                       0x0000000D
 #define ATOM_DEVICE_RESERVEDE_INDEX                       0x0000000E
 #define ATOM_DEVICE_RESERVEDF_INDEX                       0x0000000F
 #define ATOM_MAX_SUPPORTED_DEVICE_INFO                    (ATOM_DEVICE_DFP3_INDEX+1)
 #define ATOM_MAX_SUPPORTED_DEVICE_INFO_2                  ATOM_MAX_SUPPORTED_DEVICE_INFO
-#define ATOM_MAX_SUPPORTED_DEVICE_INFO_3                  (ATOM_DEVICE_DFP5_INDEX + 1)
+#define ATOM_MAX_SUPPORTED_DEVICE_INFO_3                  (ATOM_DEVICE_DFP5_INDEX + 1 )
 
 #define ATOM_MAX_SUPPORTED_DEVICE                         (ATOM_DEVICE_RESERVEDF_INDEX+1)
 
-#define ATOM_DEVICE_CRT1_SUPPORT                          (0x1L << ATOM_DEVICE_CRT1_INDEX)
-#define ATOM_DEVICE_LCD1_SUPPORT                          (0x1L << ATOM_DEVICE_LCD1_INDEX)
-#define ATOM_DEVICE_TV1_SUPPORT                           (0x1L << ATOM_DEVICE_TV1_INDEX)
-#define ATOM_DEVICE_DFP1_SUPPORT                          (0x1L << ATOM_DEVICE_DFP1_INDEX)
-#define ATOM_DEVICE_CRT2_SUPPORT                          (0x1L << ATOM_DEVICE_CRT2_INDEX)
-#define ATOM_DEVICE_LCD2_SUPPORT                          (0x1L << ATOM_DEVICE_LCD2_INDEX)
-#define ATOM_DEVICE_TV2_SUPPORT                           (0x1L << ATOM_DEVICE_TV2_INDEX)
-#define ATOM_DEVICE_DFP2_SUPPORT                          (0x1L << ATOM_DEVICE_DFP2_INDEX)
-#define ATOM_DEVICE_CV_SUPPORT                            (0x1L << ATOM_DEVICE_CV_INDEX)
-#define ATOM_DEVICE_DFP3_SUPPORT													(0x1L << ATOM_DEVICE_DFP3_INDEX)
-#define ATOM_DEVICE_DFP4_SUPPORT													(0x1L << ATOM_DEVICE_DFP4_INDEX )
-#define ATOM_DEVICE_DFP5_SUPPORT													(0x1L << ATOM_DEVICE_DFP5_INDEX)
-
-#define ATOM_DEVICE_CRT_SUPPORT \
-	(ATOM_DEVICE_CRT1_SUPPORT | ATOM_DEVICE_CRT2_SUPPORT)
-#define ATOM_DEVICE_DFP_SUPPORT \
-	(ATOM_DEVICE_DFP1_SUPPORT | ATOM_DEVICE_DFP2_SUPPORT | \
-	 ATOM_DEVICE_DFP3_SUPPORT | ATOM_DEVICE_DFP4_SUPPORT | \
-	 ATOM_DEVICE_DFP5_SUPPORT)
-#define ATOM_DEVICE_TV_SUPPORT \
-	(ATOM_DEVICE_TV1_SUPPORT  | ATOM_DEVICE_TV2_SUPPORT)
-#define ATOM_DEVICE_LCD_SUPPORT \
-	(ATOM_DEVICE_LCD1_SUPPORT | ATOM_DEVICE_LCD2_SUPPORT)
+#define ATOM_DEVICE_CRT1_SUPPORT                          (0x1L << ATOM_DEVICE_CRT1_INDEX )
+#define ATOM_DEVICE_LCD1_SUPPORT                          (0x1L << ATOM_DEVICE_LCD1_INDEX )
+#define ATOM_DEVICE_TV1_SUPPORT                           (0x1L << ATOM_DEVICE_TV1_INDEX  )
+#define ATOM_DEVICE_DFP1_SUPPORT                          (0x1L << ATOM_DEVICE_DFP1_INDEX )
+#define ATOM_DEVICE_CRT2_SUPPORT                          (0x1L << ATOM_DEVICE_CRT2_INDEX )
+#define ATOM_DEVICE_LCD2_SUPPORT                          (0x1L << ATOM_DEVICE_LCD2_INDEX )
+#define ATOM_DEVICE_DFP6_SUPPORT                          (0x1L << ATOM_DEVICE_DFP6_INDEX )
+#define ATOM_DEVICE_DFP2_SUPPORT                          (0x1L << ATOM_DEVICE_DFP2_INDEX )
+#define ATOM_DEVICE_CV_SUPPORT                            (0x1L << ATOM_DEVICE_CV_INDEX   )
+#define ATOM_DEVICE_DFP3_SUPPORT                          (0x1L << ATOM_DEVICE_DFP3_INDEX )
+#define ATOM_DEVICE_DFP4_SUPPORT                          (0x1L << ATOM_DEVICE_DFP4_INDEX )
+#define ATOM_DEVICE_DFP5_SUPPORT                          (0x1L << ATOM_DEVICE_DFP5_INDEX )
+
+#define ATOM_DEVICE_CRT_SUPPORT                           (ATOM_DEVICE_CRT1_SUPPORT | ATOM_DEVICE_CRT2_SUPPORT)
+#define ATOM_DEVICE_DFP_SUPPORT                           (ATOM_DEVICE_DFP1_SUPPORT | ATOM_DEVICE_DFP2_SUPPORT |  ATOM_DEVICE_DFP3_SUPPORT | ATOM_DEVICE_DFP4_SUPPORT | ATOM_DEVICE_DFP5_SUPPORT | ATOM_DEVICE_DFP6_SUPPORT)
+#define ATOM_DEVICE_TV_SUPPORT                            (ATOM_DEVICE_TV1_SUPPORT)
+#define ATOM_DEVICE_LCD_SUPPORT                           (ATOM_DEVICE_LCD1_SUPPORT | ATOM_DEVICE_LCD2_SUPPORT)
 
 #define ATOM_DEVICE_CONNECTOR_TYPE_MASK                   0x000000F0
 #define ATOM_DEVICE_CONNECTOR_TYPE_SHIFT                  0x00000004
@@ -1942,6 +2412,7 @@ usMinDownStreamHTLinkWidth:  same as above.
 #define ATOM_DEVICE_CONNECTOR_CASE_1                      0x0000000E
 #define ATOM_DEVICE_CONNECTOR_DISPLAYPORT                 0x0000000F
 
+
 #define ATOM_DEVICE_DAC_INFO_MASK                         0x0000000F
 #define ATOM_DEVICE_DAC_INFO_SHIFT                        0x00000000
 #define ATOM_DEVICE_DAC_INFO_NODAC                        0x00000000
@@ -1958,139 +2429,150 @@ usMinDownStreamHTLinkWidth:  same as above.
 #define ATOM_DEVICE_I2C_ID_SHIFT                          0x00000004
 #define ATOM_DEVICE_I2C_ID_IS_FOR_NON_MM_USE              0x00000001
 #define ATOM_DEVICE_I2C_ID_IS_FOR_MM_USE                  0x00000002
-#define ATOM_DEVICE_I2C_ID_IS_FOR_SDVO_USE                0x00000003	/* For IGP RS600 */
-#define ATOM_DEVICE_I2C_ID_IS_FOR_DAC_SCL                 0x00000004	/* For IGP RS690 */
+#define ATOM_DEVICE_I2C_ID_IS_FOR_SDVO_USE                0x00000003    //For IGP RS600
+#define ATOM_DEVICE_I2C_ID_IS_FOR_DAC_SCL                 0x00000004    //For IGP RS690
 
 #define ATOM_DEVICE_I2C_HARDWARE_CAP_MASK                 0x00000080
 #define ATOM_DEVICE_I2C_HARDWARE_CAP_SHIFT                0x00000007
 #define	ATOM_DEVICE_USES_SOFTWARE_ASSISTED_I2C            0x00000000
 #define	ATOM_DEVICE_USES_HARDWARE_ASSISTED_I2C            0x00000001
 
-/*   usDeviceSupport: */
-/*   Bits0       = 0 - no CRT1 support= 1- CRT1 is supported */
-/*   Bit 1       = 0 - no LCD1 support= 1- LCD1 is supported */
-/*   Bit 2       = 0 - no TV1  support= 1- TV1  is supported */
-/*   Bit 3       = 0 - no DFP1 support= 1- DFP1 is supported */
-/*   Bit 4       = 0 - no CRT2 support= 1- CRT2 is supported */
-/*   Bit 5       = 0 - no LCD2 support= 1- LCD2 is supported */
-/*   Bit 6       = 0 - no TV2  support= 1- TV2  is supported */
-/*   Bit 7       = 0 - no DFP2 support= 1- DFP2 is supported */
-/*   Bit 8       = 0 - no CV   support= 1- CV   is supported */
-/*   Bit 9       = 0 - no DFP3 support= 1- DFP3 is supported */
-/*   Byte1 (Supported Device Info) */
-/*   Bit 0       = = 0 - no CV support= 1- CV is supported */
-/*  */
-/*  */
-
-/*               ucI2C_ConfigID */
-/*     [7:0] - I2C LINE Associate ID */
-/*           = 0   - no I2C */
-/*     [7]               -       HW_Cap        = 1,  [6:0]=HW assisted I2C ID(HW line selection) */
-/*                           =   0,  [6:0]=SW assisted I2C ID */
-/*     [6-4]     - HW_ENGINE_ID  =       1,  HW engine for NON multimedia use */
-/*                           =   2,      HW engine for Multimedia use */
-/*                           =   3-7     Reserved for future I2C engines */
-/*               [3-0] - I2C_LINE_MUX  = A Mux number when it's HW assisted I2C or GPIO ID when it's SW I2C */
-
-typedef struct _ATOM_I2C_ID_CONFIG {
-#if ATOM_BIG_ENDIAN
-	UCHAR bfHW_Capable:1;
-	UCHAR bfHW_EngineID:3;
-	UCHAR bfI2C_LineMux:4;
-#else
-	UCHAR bfI2C_LineMux:4;
-	UCHAR bfHW_EngineID:3;
-	UCHAR bfHW_Capable:1;
-#endif
-} ATOM_I2C_ID_CONFIG;
-
-typedef union _ATOM_I2C_ID_CONFIG_ACCESS {
-	ATOM_I2C_ID_CONFIG sbfAccess;
-	UCHAR ucAccess;
-} ATOM_I2C_ID_CONFIG_ACCESS;
+//  usDeviceSupport:
+//  Bits0	= 0 - no CRT1 support= 1- CRT1 is supported
+//  Bit 1	= 0 - no LCD1 support= 1- LCD1 is supported
+//  Bit 2	= 0 - no TV1  support= 1- TV1  is supported
+//  Bit 3	= 0 - no DFP1 support= 1- DFP1 is supported
+//  Bit 4	= 0 - no CRT2 support= 1- CRT2 is supported
+//  Bit 5	= 0 - no LCD2 support= 1- LCD2 is supported
+//  Bit 6	= 0 - no DFP6 support= 1- DFP6 is supported
+//  Bit 7	= 0 - no DFP2 support= 1- DFP2 is supported
+//  Bit 8	= 0 - no CV   support= 1- CV   is supported
+//  Bit 9	= 0 - no DFP3 support= 1- DFP3 is supported
+//  Bit 10      = 0 - no DFP4 support= 1- DFP4 is supported
+//  Bit 11      = 0 - no DFP5 support= 1- DFP5 is supported
+//   
+//  
 
 /****************************************************************************/
-/*  Structure used in GPIO_I2C_InfoTable */
+/* Structure used in MclkSS_InfoTable                                       */
 /****************************************************************************/
-typedef struct _ATOM_GPIO_I2C_ASSIGMENT {
-	USHORT usClkMaskRegisterIndex;
-	USHORT usClkEnRegisterIndex;
-	USHORT usClkY_RegisterIndex;
-	USHORT usClkA_RegisterIndex;
-	USHORT usDataMaskRegisterIndex;
-	USHORT usDataEnRegisterIndex;
-	USHORT usDataY_RegisterIndex;
-	USHORT usDataA_RegisterIndex;
-	ATOM_I2C_ID_CONFIG_ACCESS sucI2cId;
-	UCHAR ucClkMaskShift;
-	UCHAR ucClkEnShift;
-	UCHAR ucClkY_Shift;
-	UCHAR ucClkA_Shift;
-	UCHAR ucDataMaskShift;
-	UCHAR ucDataEnShift;
-	UCHAR ucDataY_Shift;
-	UCHAR ucDataA_Shift;
-	UCHAR ucReserved1;
-	UCHAR ucReserved2;
-} ATOM_GPIO_I2C_ASSIGMENT;
-
-typedef struct _ATOM_GPIO_I2C_INFO {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	ATOM_GPIO_I2C_ASSIGMENT asGPIO_Info[ATOM_MAX_SUPPORTED_DEVICE];
-} ATOM_GPIO_I2C_INFO;
+//		ucI2C_ConfigID
+//    [7:0] - I2C LINE Associate ID
+//          = 0   - no I2C
+//    [7]		-	HW_Cap        =	1,  [6:0]=HW assisted I2C ID(HW line selection)
+//                          =	0,  [6:0]=SW assisted I2C ID
+//    [6-4]	- HW_ENGINE_ID  =	1,  HW engine for NON multimedia use
+//                          =	2,	HW engine for Multimedia use
+//                          =	3-7	Reserved for future I2C engines
+//		[3-0] - I2C_LINE_MUX  = A Mux number when it's HW assisted I2C or GPIO ID when it's SW I2C
+
+typedef struct _ATOM_I2C_ID_CONFIG
+{
+#if ATOM_BIG_ENDIAN
+  UCHAR   bfHW_Capable:1;
+  UCHAR   bfHW_EngineID:3;
+  UCHAR   bfI2C_LineMux:4;
+#else
+  UCHAR   bfI2C_LineMux:4;
+  UCHAR   bfHW_EngineID:3;
+  UCHAR   bfHW_Capable:1;
+#endif
+}ATOM_I2C_ID_CONFIG;
 
-/****************************************************************************/
-/*  Common Structure used in other structures */
-/****************************************************************************/
+typedef union _ATOM_I2C_ID_CONFIG_ACCESS
+{
+  ATOM_I2C_ID_CONFIG sbfAccess;
+  UCHAR              ucAccess;
+}ATOM_I2C_ID_CONFIG_ACCESS;
+   
+
+/****************************************************************************/	
+// Structure used in GPIO_I2C_InfoTable
+/****************************************************************************/	
+typedef struct _ATOM_GPIO_I2C_ASSIGMENT
+{
+  USHORT                    usClkMaskRegisterIndex;
+  USHORT                    usClkEnRegisterIndex;
+  USHORT                    usClkY_RegisterIndex;
+  USHORT                    usClkA_RegisterIndex;
+  USHORT                    usDataMaskRegisterIndex;
+  USHORT                    usDataEnRegisterIndex;
+  USHORT                    usDataY_RegisterIndex;
+  USHORT                    usDataA_RegisterIndex;
+  ATOM_I2C_ID_CONFIG_ACCESS sucI2cId;
+  UCHAR                     ucClkMaskShift;
+  UCHAR                     ucClkEnShift;
+  UCHAR                     ucClkY_Shift;
+  UCHAR                     ucClkA_Shift;
+  UCHAR                     ucDataMaskShift;
+  UCHAR                     ucDataEnShift;
+  UCHAR                     ucDataY_Shift;
+  UCHAR                     ucDataA_Shift;
+  UCHAR                     ucReserved1;
+  UCHAR                     ucReserved2;
+}ATOM_GPIO_I2C_ASSIGMENT;
+
+typedef struct _ATOM_GPIO_I2C_INFO
+{ 
+  ATOM_COMMON_TABLE_HEADER	sHeader;
+  ATOM_GPIO_I2C_ASSIGMENT   asGPIO_Info[ATOM_MAX_SUPPORTED_DEVICE];
+}ATOM_GPIO_I2C_INFO;
+
+/****************************************************************************/	
+// Common Structure used in other structures
+/****************************************************************************/	
 
 #ifndef _H2INC
-
-/* Please don't add or expand this bitfield structure below, this one will retire soon.! */
-typedef struct _ATOM_MODE_MISC_INFO {
+  
+//Please don't add or expand this bitfield structure below, this one will retire soon.!
+typedef struct _ATOM_MODE_MISC_INFO
+{ 
 #if ATOM_BIG_ENDIAN
-	USHORT Reserved:6;
-	USHORT RGB888:1;
-	USHORT DoubleClock:1;
-	USHORT Interlace:1;
-	USHORT CompositeSync:1;
-	USHORT V_ReplicationBy2:1;
-	USHORT H_ReplicationBy2:1;
-	USHORT VerticalCutOff:1;
-	USHORT VSyncPolarity:1;	/* 0=Active High, 1=Active Low */
-	USHORT HSyncPolarity:1;	/* 0=Active High, 1=Active Low */
-	USHORT HorizontalCutOff:1;
+  USHORT Reserved:6;
+  USHORT RGB888:1;
+  USHORT DoubleClock:1;
+  USHORT Interlace:1;
+  USHORT CompositeSync:1;
+  USHORT V_ReplicationBy2:1;
+  USHORT H_ReplicationBy2:1;
+  USHORT VerticalCutOff:1;
+  USHORT VSyncPolarity:1;      //0=Active High, 1=Active Low
+  USHORT HSyncPolarity:1;      //0=Active High, 1=Active Low
+  USHORT HorizontalCutOff:1;
 #else
-	USHORT HorizontalCutOff:1;
-	USHORT HSyncPolarity:1;	/* 0=Active High, 1=Active Low */
-	USHORT VSyncPolarity:1;	/* 0=Active High, 1=Active Low */
-	USHORT VerticalCutOff:1;
-	USHORT H_ReplicationBy2:1;
-	USHORT V_ReplicationBy2:1;
-	USHORT CompositeSync:1;
-	USHORT Interlace:1;
-	USHORT DoubleClock:1;
-	USHORT RGB888:1;
-	USHORT Reserved:6;
+  USHORT HorizontalCutOff:1;
+  USHORT HSyncPolarity:1;      //0=Active High, 1=Active Low
+  USHORT VSyncPolarity:1;      //0=Active High, 1=Active Low
+  USHORT VerticalCutOff:1;
+  USHORT H_ReplicationBy2:1;
+  USHORT V_ReplicationBy2:1;
+  USHORT CompositeSync:1;
+  USHORT Interlace:1;
+  USHORT DoubleClock:1;
+  USHORT RGB888:1;
+  USHORT Reserved:6;           
 #endif
-} ATOM_MODE_MISC_INFO;
-
-typedef union _ATOM_MODE_MISC_INFO_ACCESS {
-	ATOM_MODE_MISC_INFO sbfAccess;
-	USHORT usAccess;
-} ATOM_MODE_MISC_INFO_ACCESS;
-
+}ATOM_MODE_MISC_INFO;
+  
+typedef union _ATOM_MODE_MISC_INFO_ACCESS
+{ 
+  ATOM_MODE_MISC_INFO sbfAccess;
+  USHORT              usAccess;
+}ATOM_MODE_MISC_INFO_ACCESS;
+  
 #else
-
-typedef union _ATOM_MODE_MISC_INFO_ACCESS {
-	USHORT usAccess;
-} ATOM_MODE_MISC_INFO_ACCESS;
-
+  
+typedef union _ATOM_MODE_MISC_INFO_ACCESS
+{ 
+  USHORT              usAccess;
+}ATOM_MODE_MISC_INFO_ACCESS;
+   
 #endif
 
-/*  usModeMiscInfo- */
+// usModeMiscInfo-
 #define ATOM_H_CUTOFF           0x01
-#define ATOM_HSYNC_POLARITY     0x02	/* 0=Active High, 1=Active Low */
-#define ATOM_VSYNC_POLARITY     0x04	/* 0=Active High, 1=Active Low */
+#define ATOM_HSYNC_POLARITY     0x02             //0=Active High, 1=Active Low
+#define ATOM_VSYNC_POLARITY     0x04             //0=Active High, 1=Active Low
 #define ATOM_V_CUTOFF           0x08
 #define ATOM_H_REPLICATIONBY2   0x10
 #define ATOM_V_REPLICATIONBY2   0x20
@@ -2099,10 +2581,10 @@ typedef union _ATOM_MODE_MISC_INFO_ACCESS {
 #define ATOM_DOUBLE_CLOCK_MODE  0x100
 #define ATOM_RGB888_MODE        0x200
 
-/* usRefreshRate- */
+//usRefreshRate-
 #define ATOM_REFRESH_43         43
 #define ATOM_REFRESH_47         47
-#define ATOM_REFRESH_56         56
+#define ATOM_REFRESH_56         56	
 #define ATOM_REFRESH_60         60
 #define ATOM_REFRESH_65         65
 #define ATOM_REFRESH_70         70
@@ -2110,192 +2592,233 @@ typedef union _ATOM_MODE_MISC_INFO_ACCESS {
 #define ATOM_REFRESH_75         75
 #define ATOM_REFRESH_85         85
 
-/*  ATOM_MODE_TIMING data are exactly the same as VESA timing data. */
-/*  Translation from EDID to ATOM_MODE_TIMING, use the following formula. */
-/*  */
-/*       VESA_HTOTAL                     =       VESA_ACTIVE + 2* VESA_BORDER + VESA_BLANK */
-/*                                               =       EDID_HA + EDID_HBL */
-/*       VESA_HDISP                      =       VESA_ACTIVE     =       EDID_HA */
-/*       VESA_HSYNC_START        =       VESA_ACTIVE + VESA_BORDER + VESA_FRONT_PORCH */
-/*                                               =       EDID_HA + EDID_HSO */
-/*       VESA_HSYNC_WIDTH        =       VESA_HSYNC_TIME =       EDID_HSPW */
-/*       VESA_BORDER                     =       EDID_BORDER */
-
-/****************************************************************************/
-/*  Structure used in SetCRTC_UsingDTDTimingTable */
-/****************************************************************************/
-typedef struct _SET_CRTC_USING_DTD_TIMING_PARAMETERS {
-	USHORT usH_Size;
-	USHORT usH_Blanking_Time;
-	USHORT usV_Size;
-	USHORT usV_Blanking_Time;
-	USHORT usH_SyncOffset;
-	USHORT usH_SyncWidth;
-	USHORT usV_SyncOffset;
-	USHORT usV_SyncWidth;
-	ATOM_MODE_MISC_INFO_ACCESS susModeMiscInfo;
-	UCHAR ucH_Border;	/*  From DFP EDID */
-	UCHAR ucV_Border;
-	UCHAR ucCRTC;		/*  ATOM_CRTC1 or ATOM_CRTC2 */
-	UCHAR ucPadding[3];
-} SET_CRTC_USING_DTD_TIMING_PARAMETERS;
-
-/****************************************************************************/
-/*  Structure used in SetCRTC_TimingTable */
-/****************************************************************************/
-typedef struct _SET_CRTC_TIMING_PARAMETERS {
-	USHORT usH_Total;	/*  horizontal total */
-	USHORT usH_Disp;	/*  horizontal display */
-	USHORT usH_SyncStart;	/*  horozontal Sync start */
-	USHORT usH_SyncWidth;	/*  horizontal Sync width */
-	USHORT usV_Total;	/*  vertical total */
-	USHORT usV_Disp;	/*  vertical display */
-	USHORT usV_SyncStart;	/*  vertical Sync start */
-	USHORT usV_SyncWidth;	/*  vertical Sync width */
-	ATOM_MODE_MISC_INFO_ACCESS susModeMiscInfo;
-	UCHAR ucCRTC;		/*  ATOM_CRTC1 or ATOM_CRTC2 */
-	UCHAR ucOverscanRight;	/*  right */
-	UCHAR ucOverscanLeft;	/*  left */
-	UCHAR ucOverscanBottom;	/*  bottom */
-	UCHAR ucOverscanTop;	/*  top */
-	UCHAR ucReserved;
-} SET_CRTC_TIMING_PARAMETERS;
+// ATOM_MODE_TIMING data are exactly the same as VESA timing data.
+// Translation from EDID to ATOM_MODE_TIMING, use the following formula.
+//
+//	VESA_HTOTAL			=	VESA_ACTIVE + 2* VESA_BORDER + VESA_BLANK
+//						=	EDID_HA + EDID_HBL
+//	VESA_HDISP			=	VESA_ACTIVE	=	EDID_HA
+//	VESA_HSYNC_START	=	VESA_ACTIVE + VESA_BORDER + VESA_FRONT_PORCH
+//						=	EDID_HA + EDID_HSO
+//	VESA_HSYNC_WIDTH	=	VESA_HSYNC_TIME	=	EDID_HSPW
+//	VESA_BORDER			=	EDID_BORDER
+
+/****************************************************************************/	
+// Structure used in SetCRTC_UsingDTDTimingTable
+/****************************************************************************/	
+typedef struct _SET_CRTC_USING_DTD_TIMING_PARAMETERS
+{
+  USHORT  usH_Size;
+  USHORT  usH_Blanking_Time;
+  USHORT  usV_Size;
+  USHORT  usV_Blanking_Time;			
+  USHORT  usH_SyncOffset;
+  USHORT  usH_SyncWidth;
+  USHORT  usV_SyncOffset;
+  USHORT  usV_SyncWidth;
+  ATOM_MODE_MISC_INFO_ACCESS  susModeMiscInfo;  
+  UCHAR   ucH_Border;         // From DFP EDID
+  UCHAR   ucV_Border;
+  UCHAR   ucCRTC;             // ATOM_CRTC1 or ATOM_CRTC2  
+  UCHAR   ucPadding[3];
+}SET_CRTC_USING_DTD_TIMING_PARAMETERS;
+
+/****************************************************************************/	
+// Structure used in SetCRTC_TimingTable
+/****************************************************************************/	
+typedef struct _SET_CRTC_TIMING_PARAMETERS
+{
+  USHORT                      usH_Total;        // horizontal total
+  USHORT                      usH_Disp;         // horizontal display
+  USHORT                      usH_SyncStart;    // horozontal Sync start
+  USHORT                      usH_SyncWidth;    // horizontal Sync width
+  USHORT                      usV_Total;        // vertical total
+  USHORT                      usV_Disp;         // vertical display
+  USHORT                      usV_SyncStart;    // vertical Sync start
+  USHORT                      usV_SyncWidth;    // vertical Sync width
+  ATOM_MODE_MISC_INFO_ACCESS  susModeMiscInfo;
+  UCHAR                       ucCRTC;           // ATOM_CRTC1 or ATOM_CRTC2
+  UCHAR                       ucOverscanRight;  // right
+  UCHAR                       ucOverscanLeft;   // left
+  UCHAR                       ucOverscanBottom; // bottom
+  UCHAR                       ucOverscanTop;    // top
+  UCHAR                       ucReserved;
+}SET_CRTC_TIMING_PARAMETERS;
 #define SET_CRTC_TIMING_PARAMETERS_PS_ALLOCATION SET_CRTC_TIMING_PARAMETERS
 
-/****************************************************************************/
-/*  Structure used in StandardVESA_TimingTable */
-/*                    AnalogTV_InfoTable */
-/*                    ComponentVideoInfoTable */
-/****************************************************************************/
-typedef struct _ATOM_MODE_TIMING {
-	USHORT usCRTC_H_Total;
-	USHORT usCRTC_H_Disp;
-	USHORT usCRTC_H_SyncStart;
-	USHORT usCRTC_H_SyncWidth;
-	USHORT usCRTC_V_Total;
-	USHORT usCRTC_V_Disp;
-	USHORT usCRTC_V_SyncStart;
-	USHORT usCRTC_V_SyncWidth;
-	USHORT usPixelClock;	/* in 10Khz unit */
-	ATOM_MODE_MISC_INFO_ACCESS susModeMiscInfo;
-	USHORT usCRTC_OverscanRight;
-	USHORT usCRTC_OverscanLeft;
-	USHORT usCRTC_OverscanBottom;
-	USHORT usCRTC_OverscanTop;
-	USHORT usReserve;
-	UCHAR ucInternalModeNumber;
-	UCHAR ucRefreshRate;
-} ATOM_MODE_TIMING;
-
-typedef struct _ATOM_DTD_FORMAT {
-	USHORT usPixClk;
-	USHORT usHActive;
-	USHORT usHBlanking_Time;
-	USHORT usVActive;
-	USHORT usVBlanking_Time;
-	USHORT usHSyncOffset;
-	USHORT usHSyncWidth;
-	USHORT usVSyncOffset;
-	USHORT usVSyncWidth;
-	USHORT usImageHSize;
-	USHORT usImageVSize;
-	UCHAR ucHBorder;
-	UCHAR ucVBorder;
-	ATOM_MODE_MISC_INFO_ACCESS susModeMiscInfo;
-	UCHAR ucInternalModeNumber;
-	UCHAR ucRefreshRate;
-} ATOM_DTD_FORMAT;
-
-/****************************************************************************/
-/*  Structure used in LVDS_InfoTable */
-/*   * Need a document to describe this table */
-/****************************************************************************/
+/****************************************************************************/	
+// Structure used in StandardVESA_TimingTable
+//                   AnalogTV_InfoTable 
+//                   ComponentVideoInfoTable
+/****************************************************************************/	
+typedef struct _ATOM_MODE_TIMING
+{
+  USHORT  usCRTC_H_Total;
+  USHORT  usCRTC_H_Disp;
+  USHORT  usCRTC_H_SyncStart;
+  USHORT  usCRTC_H_SyncWidth;
+  USHORT  usCRTC_V_Total;
+  USHORT  usCRTC_V_Disp;
+  USHORT  usCRTC_V_SyncStart;
+  USHORT  usCRTC_V_SyncWidth;
+  USHORT  usPixelClock;					                 //in 10Khz unit
+  ATOM_MODE_MISC_INFO_ACCESS  susModeMiscInfo;
+  USHORT  usCRTC_OverscanRight;
+  USHORT  usCRTC_OverscanLeft;
+  USHORT  usCRTC_OverscanBottom;
+  USHORT  usCRTC_OverscanTop;
+  USHORT  usReserve;
+  UCHAR   ucInternalModeNumber;
+  UCHAR   ucRefreshRate;
+}ATOM_MODE_TIMING;
+
+typedef struct _ATOM_DTD_FORMAT
+{
+  USHORT  usPixClk;
+  USHORT  usHActive;
+  USHORT  usHBlanking_Time;
+  USHORT  usVActive;
+  USHORT  usVBlanking_Time;			
+  USHORT  usHSyncOffset;
+  USHORT  usHSyncWidth;
+  USHORT  usVSyncOffset;
+  USHORT  usVSyncWidth;
+  USHORT  usImageHSize;
+  USHORT  usImageVSize;
+  UCHAR   ucHBorder;
+  UCHAR   ucVBorder;
+  ATOM_MODE_MISC_INFO_ACCESS susModeMiscInfo;
+  UCHAR   ucInternalModeNumber;
+  UCHAR   ucRefreshRate;
+}ATOM_DTD_FORMAT;
+
+/****************************************************************************/	
+// Structure used in LVDS_InfoTable 
+//  * Need a document to describe this table
+/****************************************************************************/	
 #define SUPPORTED_LCD_REFRESHRATE_30Hz          0x0004
 #define SUPPORTED_LCD_REFRESHRATE_40Hz          0x0008
 #define SUPPORTED_LCD_REFRESHRATE_50Hz          0x0010
 #define SUPPORTED_LCD_REFRESHRATE_60Hz          0x0020
 
-/* Once DAL sees this CAP is set, it will read EDID from LCD on its own instead of using sLCDTiming in ATOM_LVDS_INFO_V12. */
-/* Other entries in ATOM_LVDS_INFO_V12 are still valid/useful to DAL */
-#define	LCDPANEL_CAP_READ_EDID									0x1
-
-/* ucTableFormatRevision=1 */
-/* ucTableContentRevision=1 */
-typedef struct _ATOM_LVDS_INFO {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	ATOM_DTD_FORMAT sLCDTiming;
-	USHORT usModePatchTableOffset;
-	USHORT usSupportedRefreshRate;	/* Refer to panel info table in ATOMBIOS extension Spec. */
-	USHORT usOffDelayInMs;
-	UCHAR ucPowerSequenceDigOntoDEin10Ms;
-	UCHAR ucPowerSequenceDEtoBLOnin10Ms;
-	UCHAR ucLVDS_Misc;	/*  Bit0:{=0:single, =1:dual},Bit1 {=0:666RGB, =1:888RGB},Bit2:3:{Grey level} */
-	/*  Bit4:{=0:LDI format for RGB888, =1 FPDI format for RGB888} */
-	/*  Bit5:{=0:Spatial Dithering disabled;1 Spatial Dithering enabled} */
-	/*  Bit6:{=0:Temporal Dithering disabled;1 Temporal Dithering enabled} */
-	UCHAR ucPanelDefaultRefreshRate;
-	UCHAR ucPanelIdentification;
-	UCHAR ucSS_Id;
-} ATOM_LVDS_INFO;
-
-/* ucTableFormatRevision=1 */
-/* ucTableContentRevision=2 */
-typedef struct _ATOM_LVDS_INFO_V12 {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	ATOM_DTD_FORMAT sLCDTiming;
-	USHORT usExtInfoTableOffset;
-	USHORT usSupportedRefreshRate;	/* Refer to panel info table in ATOMBIOS extension Spec. */
-	USHORT usOffDelayInMs;
-	UCHAR ucPowerSequenceDigOntoDEin10Ms;
-	UCHAR ucPowerSequenceDEtoBLOnin10Ms;
-	UCHAR ucLVDS_Misc;	/*  Bit0:{=0:single, =1:dual},Bit1 {=0:666RGB, =1:888RGB},Bit2:3:{Grey level} */
-	/*  Bit4:{=0:LDI format for RGB888, =1 FPDI format for RGB888} */
-	/*  Bit5:{=0:Spatial Dithering disabled;1 Spatial Dithering enabled} */
-	/*  Bit6:{=0:Temporal Dithering disabled;1 Temporal Dithering enabled} */
-	UCHAR ucPanelDefaultRefreshRate;
-	UCHAR ucPanelIdentification;
-	UCHAR ucSS_Id;
-	USHORT usLCDVenderID;
-	USHORT usLCDProductID;
-	UCHAR ucLCDPanel_SpecialHandlingCap;
-	UCHAR ucPanelInfoSize;	/*   start from ATOM_DTD_FORMAT to end of panel info, include ExtInfoTable */
-	UCHAR ucReserved[2];
-} ATOM_LVDS_INFO_V12;
+//ucTableFormatRevision=1
+//ucTableContentRevision=1
+typedef struct _ATOM_LVDS_INFO
+{
+  ATOM_COMMON_TABLE_HEADER sHeader;  
+  ATOM_DTD_FORMAT     sLCDTiming;
+  USHORT              usModePatchTableOffset;
+  USHORT              usSupportedRefreshRate;     //Refer to panel info table in ATOMBIOS extension Spec.
+  USHORT              usOffDelayInMs;
+  UCHAR               ucPowerSequenceDigOntoDEin10Ms;
+  UCHAR               ucPowerSequenceDEtoBLOnin10Ms;
+  UCHAR               ucLVDS_Misc;               // Bit0:{=0:single, =1:dual},Bit1 {=0:666RGB, =1:888RGB},Bit2:3:{Grey level}
+                                                 // Bit4:{=0:LDI format for RGB888, =1 FPDI format for RGB888}
+                                                 // Bit5:{=0:Spatial Dithering disabled;1 Spatial Dithering enabled}
+                                                 // Bit6:{=0:Temporal Dithering disabled;1 Temporal Dithering enabled}
+  UCHAR               ucPanelDefaultRefreshRate;
+  UCHAR               ucPanelIdentification;
+  UCHAR               ucSS_Id;
+}ATOM_LVDS_INFO;
+
+//ucTableFormatRevision=1
+//ucTableContentRevision=2
+typedef struct _ATOM_LVDS_INFO_V12
+{
+  ATOM_COMMON_TABLE_HEADER sHeader;  
+  ATOM_DTD_FORMAT     sLCDTiming;
+  USHORT              usExtInfoTableOffset;
+  USHORT              usSupportedRefreshRate;     //Refer to panel info table in ATOMBIOS extension Spec.
+  USHORT              usOffDelayInMs;
+  UCHAR               ucPowerSequenceDigOntoDEin10Ms;
+  UCHAR               ucPowerSequenceDEtoBLOnin10Ms;
+  UCHAR               ucLVDS_Misc;               // Bit0:{=0:single, =1:dual},Bit1 {=0:666RGB, =1:888RGB},Bit2:3:{Grey level}
+                                                 // Bit4:{=0:LDI format for RGB888, =1 FPDI format for RGB888}
+                                                 // Bit5:{=0:Spatial Dithering disabled;1 Spatial Dithering enabled}
+                                                 // Bit6:{=0:Temporal Dithering disabled;1 Temporal Dithering enabled}
+  UCHAR               ucPanelDefaultRefreshRate;
+  UCHAR               ucPanelIdentification;
+  UCHAR               ucSS_Id;
+  USHORT              usLCDVenderID;
+  USHORT              usLCDProductID;
+  UCHAR               ucLCDPanel_SpecialHandlingCap; 
+	UCHAR								ucPanelInfoSize;					//  start from ATOM_DTD_FORMAT to end of panel info, include ExtInfoTable
+  UCHAR               ucReserved[2];
+}ATOM_LVDS_INFO_V12;
+
+//Definitions for ucLCDPanel_SpecialHandlingCap:
+
+//Once DAL sees this CAP is set, it will read EDID from LCD on its own instead of using sLCDTiming in ATOM_LVDS_INFO_V12. 
+//Other entries in ATOM_LVDS_INFO_V12 are still valid/useful to DAL 
+#define	LCDPANEL_CAP_READ_EDID                  0x1
+
+//If a design supports DRR (dynamic refresh rate) on internal panels (LVDS or EDP), this cap is set in ucLCDPanel_SpecialHandlingCap together
+//with multiple supported refresh rates@usSupportedRefreshRate. This cap should not be set when only slow refresh rate is supported (static
+//refresh rate switch by SW. This is only valid from ATOM_LVDS_INFO_V12
+#define	LCDPANEL_CAP_DRR_SUPPORTED              0x2
+
+//Use this cap bit for a quick reference whether an embadded panel (LCD1 ) is LVDS or eDP.
+#define	LCDPANEL_CAP_eDP                        0x4
+
+
+//Color Bit Depth definition in EDID V1.4 @BYTE 14h
+//Bit 6  5  4
+                              //      0  0  0  -  Color bit depth is undefined
+                              //      0  0  1  -  6 Bits per Primary Color
+                              //      0  1  0  -  8 Bits per Primary Color
+                              //      0  1  1  - 10 Bits per Primary Color
+                              //      1  0  0  - 12 Bits per Primary Color
+                              //      1  0  1  - 14 Bits per Primary Color
+                              //      1  1  0  - 16 Bits per Primary Color
+                              //      1  1  1  - Reserved
+
+#define PANEL_COLOR_BIT_DEPTH_MASK    0x70
+
+// Bit7:{=0:Random Dithering disabled;1 Random Dithering enabled}   
+#define PANEL_RANDOM_DITHER   0x80
+#define PANEL_RANDOM_DITHER_MASK   0x80
+
 
 #define ATOM_LVDS_INFO_LAST  ATOM_LVDS_INFO_V12
 
-typedef struct _ATOM_PATCH_RECORD_MODE {
-	UCHAR ucRecordType;
-	USHORT usHDisp;
-	USHORT usVDisp;
-} ATOM_PATCH_RECORD_MODE;
+typedef struct  _ATOM_PATCH_RECORD_MODE
+{
+  UCHAR     ucRecordType;
+  USHORT    usHDisp;
+  USHORT    usVDisp;
+}ATOM_PATCH_RECORD_MODE;
 
-typedef struct _ATOM_LCD_RTS_RECORD {
-	UCHAR ucRecordType;
-	UCHAR ucRTSValue;
-} ATOM_LCD_RTS_RECORD;
+typedef struct  _ATOM_LCD_RTS_RECORD
+{
+  UCHAR     ucRecordType;
+  UCHAR     ucRTSValue;
+}ATOM_LCD_RTS_RECORD;
 
-/* !! If the record below exits, it shoud always be the first record for easy use in command table!!! */
-typedef struct _ATOM_LCD_MODE_CONTROL_CAP {
-	UCHAR ucRecordType;
-	USHORT usLCDCap;
-} ATOM_LCD_MODE_CONTROL_CAP;
+//!! If the record below exits, it shoud always be the first record for easy use in command table!!! 
+// The record below is only used when LVDS_Info is present. From ATOM_LVDS_INFO_V12, use ucLCDPanel_SpecialHandlingCap instead.
+typedef struct  _ATOM_LCD_MODE_CONTROL_CAP
+{
+  UCHAR     ucRecordType;
+  USHORT    usLCDCap;
+}ATOM_LCD_MODE_CONTROL_CAP;
 
 #define LCD_MODE_CAP_BL_OFF                   1
 #define LCD_MODE_CAP_CRTC_OFF                 2
 #define LCD_MODE_CAP_PANEL_OFF                4
 
-typedef struct _ATOM_FAKE_EDID_PATCH_RECORD {
-	UCHAR ucRecordType;
-	UCHAR ucFakeEDIDLength;
-	UCHAR ucFakeEDIDString[1];	/*  This actually has ucFakeEdidLength elements. */
+typedef struct _ATOM_FAKE_EDID_PATCH_RECORD
+{
+  UCHAR ucRecordType;
+  UCHAR ucFakeEDIDLength;
+  UCHAR ucFakeEDIDString[1];    // This actually has ucFakeEdidLength elements.
 } ATOM_FAKE_EDID_PATCH_RECORD;
 
-typedef struct _ATOM_PANEL_RESOLUTION_PATCH_RECORD {
-	UCHAR ucRecordType;
-	USHORT usHSize;
-	USHORT usVSize;
-} ATOM_PANEL_RESOLUTION_PATCH_RECORD;
+typedef struct  _ATOM_PANEL_RESOLUTION_PATCH_RECORD
+{
+   UCHAR    ucRecordType;
+   USHORT		usHSize;
+   USHORT		usVSize;
+}ATOM_PANEL_RESOLUTION_PATCH_RECORD;
 
 #define LCD_MODE_PATCH_RECORD_MODE_TYPE       1
 #define LCD_RTS_RECORD_TYPE                   2
@@ -2306,21 +2829,25 @@ typedef struct _ATOM_PANEL_RESOLUTION_PATCH_RECORD {
 
 /****************************Spread Spectrum Info Table Definitions **********************/
 
-/* ucTableFormatRevision=1 */
-/* ucTableContentRevision=2 */
-typedef struct _ATOM_SPREAD_SPECTRUM_ASSIGNMENT {
-	USHORT usSpreadSpectrumPercentage;
-	UCHAR ucSpreadSpectrumType;	/* Bit1=0 Down Spread,=1 Center Spread. Bit1=1 Ext. =0 Int. Others:TBD */
-	UCHAR ucSS_Step;
-	UCHAR ucSS_Delay;
-	UCHAR ucSS_Id;
-	UCHAR ucRecommandedRef_Div;
-	UCHAR ucSS_Range;	/* it was reserved for V11 */
-} ATOM_SPREAD_SPECTRUM_ASSIGNMENT;
+//ucTableFormatRevision=1
+//ucTableContentRevision=2
+typedef struct _ATOM_SPREAD_SPECTRUM_ASSIGNMENT
+{
+  USHORT              usSpreadSpectrumPercentage; 
+  UCHAR               ucSpreadSpectrumType;	    //Bit1=0 Down Spread,=1 Center Spread. Bit1=1 Ext. =0 Int. Bit2=1: PCIE REFCLK SS =0 iternal PPLL SS  Others:TBD
+  UCHAR               ucSS_Step;
+  UCHAR               ucSS_Delay;
+  UCHAR               ucSS_Id;
+  UCHAR               ucRecommendedRef_Div;
+  UCHAR               ucSS_Range;               //it was reserved for V11
+}ATOM_SPREAD_SPECTRUM_ASSIGNMENT;
 
 #define ATOM_MAX_SS_ENTRY                      16
-#define ATOM_DP_SS_ID1												 0x0f1	/*  SS modulation freq=30k */
-#define ATOM_DP_SS_ID2												 0x0f2	/*  SS modulation freq=33k */
+#define ATOM_DP_SS_ID1												 0x0f1			// SS ID for internal DP stream at 2.7Ghz. if ATOM_DP_SS_ID2 does not exist in SS_InfoTable, it is used for internal DP stream at 1.62Ghz as well. 
+#define ATOM_DP_SS_ID2												 0x0f2			// SS ID for internal DP stream at 1.62Ghz, if it exists in SS_InfoTable. 
+#define ATOM_LVLINK_2700MHz_SS_ID              0x0f3      // SS ID for LV link translator chip at 2.7Ghz
+#define ATOM_LVLINK_1620MHz_SS_ID              0x0f4      // SS ID for LV link translator chip at 1.62Ghz
+
 
 #define ATOM_SS_DOWN_SPREAD_MODE_MASK          0x00000000
 #define ATOM_SS_DOWN_SPREAD_MODE               0x00000000
@@ -2329,29 +2856,30 @@ typedef struct _ATOM_SPREAD_SPECTRUM_ASSIGNMENT {
 #define ATOM_INTERNAL_SS_MASK                  0x00000000
 #define ATOM_EXTERNAL_SS_MASK                  0x00000002
 #define EXEC_SS_STEP_SIZE_SHIFT                2
-#define EXEC_SS_DELAY_SHIFT                    4
+#define EXEC_SS_DELAY_SHIFT                    4    
 #define ACTIVEDATA_TO_BLON_DELAY_SHIFT         4
 
-typedef struct _ATOM_SPREAD_SPECTRUM_INFO {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	ATOM_SPREAD_SPECTRUM_ASSIGNMENT asSS_Info[ATOM_MAX_SS_ENTRY];
-} ATOM_SPREAD_SPECTRUM_INFO;
-
-/****************************************************************************/
-/*  Structure used in AnalogTV_InfoTable (Top level) */
-/****************************************************************************/
-/* ucTVBootUpDefaultStd definiton: */
-
-/* ATOM_TV_NTSC                1 */
-/* ATOM_TV_NTSCJ               2 */
-/* ATOM_TV_PAL                 3 */
-/* ATOM_TV_PALM                4 */
-/* ATOM_TV_PALCN               5 */
-/* ATOM_TV_PALN                6 */
-/* ATOM_TV_PAL60               7 */
-/* ATOM_TV_SECAM               8 */
-
-/* ucTVSuppportedStd definition: */
+typedef struct _ATOM_SPREAD_SPECTRUM_INFO
+{ 
+  ATOM_COMMON_TABLE_HEADER	sHeader;
+  ATOM_SPREAD_SPECTRUM_ASSIGNMENT   asSS_Info[ATOM_MAX_SS_ENTRY];
+}ATOM_SPREAD_SPECTRUM_INFO;
+
+/****************************************************************************/	
+// Structure used in AnalogTV_InfoTable (Top level)
+/****************************************************************************/	
+//ucTVBootUpDefaultStd definiton:
+
+//ATOM_TV_NTSC                1
+//ATOM_TV_NTSCJ               2
+//ATOM_TV_PAL                 3
+//ATOM_TV_PALM                4
+//ATOM_TV_PALCN               5
+//ATOM_TV_PALN                6
+//ATOM_TV_PAL60               7
+//ATOM_TV_SECAM               8
+
+//ucTVSupportedStd definition:
 #define NTSC_SUPPORT          0x1
 #define NTSCJ_SUPPORT         0x2
 
@@ -2364,46 +2892,58 @@ typedef struct _ATOM_SPREAD_SPECTRUM_INFO {
 
 #define MAX_SUPPORTED_TV_TIMING    2
 
-typedef struct _ATOM_ANALOG_TV_INFO {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	UCHAR ucTV_SupportedStandard;
-	UCHAR ucTV_BootUpDefaultStandard;
-	UCHAR ucExt_TV_ASIC_ID;
-	UCHAR ucExt_TV_ASIC_SlaveAddr;
-	/*ATOM_DTD_FORMAT          aModeTimings[MAX_SUPPORTED_TV_TIMING]; */
-	ATOM_MODE_TIMING aModeTimings[MAX_SUPPORTED_TV_TIMING];
-} ATOM_ANALOG_TV_INFO;
+typedef struct _ATOM_ANALOG_TV_INFO
+{
+  ATOM_COMMON_TABLE_HEADER sHeader;  
+  UCHAR                    ucTV_SupportedStandard;
+  UCHAR                    ucTV_BootUpDefaultStandard; 
+  UCHAR                    ucExt_TV_ASIC_ID;
+  UCHAR                    ucExt_TV_ASIC_SlaveAddr;
+  /*ATOM_DTD_FORMAT          aModeTimings[MAX_SUPPORTED_TV_TIMING];*/
+  ATOM_MODE_TIMING         aModeTimings[MAX_SUPPORTED_TV_TIMING];
+}ATOM_ANALOG_TV_INFO;
 
 #define MAX_SUPPORTED_TV_TIMING_V1_2    3
 
-typedef struct _ATOM_ANALOG_TV_INFO_V1_2 {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	UCHAR                    ucTV_SupportedStandard;
-	UCHAR                    ucTV_BootUpDefaultStandard;
-	UCHAR                    ucExt_TV_ASIC_ID;
-	UCHAR                    ucExt_TV_ASIC_SlaveAddr;
-	ATOM_DTD_FORMAT          aModeTimings[MAX_SUPPORTED_TV_TIMING];
-} ATOM_ANALOG_TV_INFO_V1_2;
+typedef struct _ATOM_ANALOG_TV_INFO_V1_2
+{
+  ATOM_COMMON_TABLE_HEADER sHeader;  
+  UCHAR                    ucTV_SupportedStandard;
+  UCHAR                    ucTV_BootUpDefaultStandard; 
+  UCHAR                    ucExt_TV_ASIC_ID;
+  UCHAR                    ucExt_TV_ASIC_SlaveAddr;
+  ATOM_DTD_FORMAT          aModeTimings[MAX_SUPPORTED_TV_TIMING];
+}ATOM_ANALOG_TV_INFO_V1_2;
+
+typedef struct _ATOM_DPCD_INFO
+{
+  UCHAR   ucRevisionNumber;        //10h : Revision 1.0; 11h : Revision 1.1   
+  UCHAR   ucMaxLinkRate;           //06h : 1.62Gbps per lane; 0Ah = 2.7Gbps per lane
+  UCHAR   ucMaxLane;               //Bits 4:0 = MAX_LANE_COUNT (1/2/4). Bit 7 = ENHANCED_FRAME_CAP 
+  UCHAR   ucMaxDownSpread;         //Bit0 = 0: No Down spread; Bit0 = 1: 0.5% (Subject to change according to DP spec)
+}ATOM_DPCD_INFO;
+
+#define ATOM_DPCD_MAX_LANE_MASK    0x1F
 
 /**************************************************************************/
-/*  VRAM usage and their defintions */
+// VRAM usage and their defintions
 
-/*  One chunk of VRAM used by Bios are for HWICON surfaces,EDID data. */
-/*  Current Mode timing and Dail Timing and/or STD timing data EACH device. They can be broken down as below. */
-/*  All the addresses below are the offsets from the frame buffer start.They all MUST be Dword aligned! */
-/*  To driver: The physical address of this memory portion=mmFB_START(4K aligned)+ATOMBIOS_VRAM_USAGE_START_ADDR+ATOM_x_ADDR */
-/*  To Bios:  ATOMBIOS_VRAM_USAGE_START_ADDR+ATOM_x_ADDR->MM_INDEX */
+// One chunk of VRAM used by Bios are for HWICON surfaces,EDID data.
+// Current Mode timing and Dail Timing and/or STD timing data EACH device. They can be broken down as below.
+// All the addresses below are the offsets from the frame buffer start.They all MUST be Dword aligned!
+// To driver: The physical address of this memory portion=mmFB_START(4K aligned)+ATOMBIOS_VRAM_USAGE_START_ADDR+ATOM_x_ADDR
+// To Bios:  ATOMBIOS_VRAM_USAGE_START_ADDR+ATOM_x_ADDR->MM_INDEX 
 
 #ifndef VESA_MEMORY_IN_64K_BLOCK
-#define VESA_MEMORY_IN_64K_BLOCK        0x100	/* 256*64K=16Mb (Max. VESA memory is 16Mb!) */
+#define VESA_MEMORY_IN_64K_BLOCK        0x100       //256*64K=16Mb (Max. VESA memory is 16Mb!)
 #endif
 
-#define ATOM_EDID_RAW_DATASIZE          256	/* In Bytes */
-#define ATOM_HWICON_SURFACE_SIZE        4096	/* In Bytes */
+#define ATOM_EDID_RAW_DATASIZE          256         //In Bytes
+#define ATOM_HWICON_SURFACE_SIZE        4096        //In Bytes
 #define ATOM_HWICON_INFOTABLE_SIZE      32
 #define MAX_DTD_MODE_IN_VRAM            6
-#define ATOM_DTD_MODE_SUPPORT_TBL_SIZE  (MAX_DTD_MODE_IN_VRAM*28)	/* 28= (SIZEOF ATOM_DTD_FORMAT) */
-#define ATOM_STD_MODE_SUPPORT_TBL_SIZE  (32*8)	/* 32 is a predefined number,8= (SIZEOF ATOM_STD_FORMAT) */
+#define ATOM_DTD_MODE_SUPPORT_TBL_SIZE  (MAX_DTD_MODE_IN_VRAM*28)    //28= (SIZEOF ATOM_DTD_FORMAT) 
+#define ATOM_STD_MODE_SUPPORT_TBL_SIZE  32*8                         //32 is a predefined number,8= (SIZEOF ATOM_STD_FORMAT)
 #define DFP_ENCODER_TYPE_OFFSET					0x80
 #define DP_ENCODER_LANE_NUM_OFFSET			0x84
 #define DP_ENCODER_LINK_RATE_OFFSET			0x88
@@ -2417,7 +2957,7 @@ typedef struct _ATOM_ANALOG_TV_INFO_V1_2 {
 
 #define ATOM_LCD1_EDID_ADDR             (ATOM_CRT1_STD_MODE_TBL_ADDR + ATOM_STD_MODE_SUPPORT_TBL_SIZE)
 #define ATOM_LCD1_DTD_MODE_TBL_ADDR     (ATOM_LCD1_EDID_ADDR + ATOM_EDID_RAW_DATASIZE)
-#define ATOM_LCD1_STD_MODE_TBL_ADDR	(ATOM_LCD1_DTD_MODE_TBL_ADDR + ATOM_DTD_MODE_SUPPORT_TBL_SIZE)
+#define ATOM_LCD1_STD_MODE_TBL_ADDR   	(ATOM_LCD1_DTD_MODE_TBL_ADDR + ATOM_DTD_MODE_SUPPORT_TBL_SIZE)
 
 #define ATOM_TV1_DTD_MODE_TBL_ADDR      (ATOM_LCD1_STD_MODE_TBL_ADDR + ATOM_STD_MODE_SUPPORT_TBL_SIZE)
 
@@ -2431,13 +2971,13 @@ typedef struct _ATOM_ANALOG_TV_INFO_V1_2 {
 
 #define ATOM_LCD2_EDID_ADDR             (ATOM_CRT2_STD_MODE_TBL_ADDR + ATOM_STD_MODE_SUPPORT_TBL_SIZE)
 #define ATOM_LCD2_DTD_MODE_TBL_ADDR     (ATOM_LCD2_EDID_ADDR + ATOM_EDID_RAW_DATASIZE)
-#define ATOM_LCD2_STD_MODE_TBL_ADDR	(ATOM_LCD2_DTD_MODE_TBL_ADDR + ATOM_DTD_MODE_SUPPORT_TBL_SIZE)
+#define ATOM_LCD2_STD_MODE_TBL_ADDR   	(ATOM_LCD2_DTD_MODE_TBL_ADDR + ATOM_DTD_MODE_SUPPORT_TBL_SIZE)
 
-#define ATOM_TV2_EDID_ADDR              (ATOM_LCD2_STD_MODE_TBL_ADDR + ATOM_STD_MODE_SUPPORT_TBL_SIZE)
-#define ATOM_TV2_DTD_MODE_TBL_ADDR      (ATOM_TV2_EDID_ADDR + ATOM_EDID_RAW_DATASIZE)
-#define ATOM_TV2_STD_MODE_TBL_ADDR	  (ATOM_TV2_DTD_MODE_TBL_ADDR + ATOM_DTD_MODE_SUPPORT_TBL_SIZE)
+#define ATOM_DFP6_EDID_ADDR             (ATOM_LCD2_STD_MODE_TBL_ADDR + ATOM_STD_MODE_SUPPORT_TBL_SIZE)
+#define ATOM_DFP6_DTD_MODE_TBL_ADDR     (ATOM_DFP6_EDID_ADDR + ATOM_EDID_RAW_DATASIZE)
+#define ATOM_DFP6_STD_MODE_TBL_ADDR     (ATOM_DFP6_DTD_MODE_TBL_ADDR + ATOM_DTD_MODE_SUPPORT_TBL_SIZE)
 
-#define ATOM_DFP2_EDID_ADDR             (ATOM_TV2_STD_MODE_TBL_ADDR + ATOM_STD_MODE_SUPPORT_TBL_SIZE)
+#define ATOM_DFP2_EDID_ADDR             (ATOM_DFP6_STD_MODE_TBL_ADDR + ATOM_STD_MODE_SUPPORT_TBL_SIZE)
 #define ATOM_DFP2_DTD_MODE_TBL_ADDR     (ATOM_DFP2_EDID_ADDR + ATOM_EDID_RAW_DATASIZE)
 #define ATOM_DFP2_STD_MODE_TBL_ADDR     (ATOM_DFP2_DTD_MODE_TBL_ADDR + ATOM_DTD_MODE_SUPPORT_TBL_SIZE)
 
@@ -2457,533 +2997,850 @@ typedef struct _ATOM_ANALOG_TV_INFO_V1_2 {
 #define ATOM_DFP5_DTD_MODE_TBL_ADDR     (ATOM_DFP5_EDID_ADDR + ATOM_EDID_RAW_DATASIZE)
 #define ATOM_DFP5_STD_MODE_TBL_ADDR     (ATOM_DFP5_DTD_MODE_TBL_ADDR + ATOM_DTD_MODE_SUPPORT_TBL_SIZE)
 
-#define ATOM_DP_TRAINING_TBL_ADDR	(ATOM_DFP5_STD_MODE_TBL_ADDR+ATOM_STD_MODE_SUPPORT_TBL_SIZE)
+#define ATOM_DP_TRAINING_TBL_ADDR				(ATOM_DFP5_STD_MODE_TBL_ADDR+ATOM_STD_MODE_SUPPORT_TBL_SIZE)       
 
-#define ATOM_STACK_STORAGE_START        (ATOM_DP_TRAINING_TBL_ADDR + 256)
-#define ATOM_STACK_STORAGE_END          (ATOM_STACK_STORAGE_START + 512)
+#define ATOM_STACK_STORAGE_START        (ATOM_DP_TRAINING_TBL_ADDR+256)       
+#define ATOM_STACK_STORAGE_END          ATOM_STACK_STORAGE_START+512        
 
-/* The size below is in Kb! */
+//The size below is in Kb!
 #define ATOM_VRAM_RESERVE_SIZE         ((((ATOM_STACK_STORAGE_END - ATOM_HWICON1_SURFACE_ADDR)>>10)+4)&0xFFFC)
-
+   
 #define	ATOM_VRAM_OPERATION_FLAGS_MASK         0xC0000000L
 #define ATOM_VRAM_OPERATION_FLAGS_SHIFT        30
 #define	ATOM_VRAM_BLOCK_NEEDS_NO_RESERVATION   0x1
 #define	ATOM_VRAM_BLOCK_NEEDS_RESERVATION      0x0
 
-/***********************************************************************************/
-/*  Structure used in VRAM_UsageByFirmwareTable */
-/*  Note1: This table is filled by SetBiosReservationStartInFB in CoreCommSubs.asm */
-/*         at running time. */
-/*  note2: From RV770, the memory is more than 32bit addressable, so we will change */
-/*         ucTableFormatRevision=1,ucTableContentRevision=4, the strcuture remains */
-/*         exactly same as 1.1 and 1.2 (1.3 is never in use), but ulStartAddrUsedByFirmware */
-/*         (in offset to start of memory address) is KB aligned instead of byte aligend. */
-/***********************************************************************************/
+/***********************************************************************************/	
+// Structure used in VRAM_UsageByFirmwareTable
+// Note1: This table is filled by SetBiosReservationStartInFB in CoreCommSubs.asm
+//        at running time.   
+// note2: From RV770, the memory is more than 32bit addressable, so we will change 
+//        ucTableFormatRevision=1,ucTableContentRevision=4, the strcuture remains 
+//        exactly same as 1.1 and 1.2 (1.3 is never in use), but ulStartAddrUsedByFirmware 
+//        (in offset to start of memory address) is KB aligned instead of byte aligend.
+/***********************************************************************************/	
+// Note3:
+/* If we change usReserved to "usFBUsedbyDrvInKB", then to VBIOS this usFBUsedbyDrvInKB is a predefined, unchanged constant across VGA or non VGA adapter,
+for CAIL, The size of FB access area is known, only thing missing is the Offset of FB Access area, so we can  have:
+
+If (ulStartAddrUsedByFirmware!=0)
+FBAccessAreaOffset= ulStartAddrUsedByFirmware - usFBUsedbyDrvInKB;
+Reserved area has been claimed by VBIOS including this FB access area; CAIL doesn't need to reserve any extra area for this purpose
+else	//Non VGA case
+ if (FB_Size<=2Gb)
+    FBAccessAreaOffset= FB_Size - usFBUsedbyDrvInKB;
+ else
+	  FBAccessAreaOffset= Aper_Size - usFBUsedbyDrvInKB
+
+CAIL needs to claim an reserved area defined by FBAccessAreaOffset and usFBUsedbyDrvInKB in non VGA case.*/
+
 #define ATOM_MAX_FIRMWARE_VRAM_USAGE_INFO			1
 
-typedef struct _ATOM_FIRMWARE_VRAM_RESERVE_INFO {
-	ULONG ulStartAddrUsedByFirmware;
-	USHORT usFirmwareUseInKb;
-	USHORT usReserved;
-} ATOM_FIRMWARE_VRAM_RESERVE_INFO;
+typedef struct _ATOM_FIRMWARE_VRAM_RESERVE_INFO
+{
+  ULONG   ulStartAddrUsedByFirmware;
+  USHORT  usFirmwareUseInKb;
+  USHORT  usReserved;
+}ATOM_FIRMWARE_VRAM_RESERVE_INFO;
+
+typedef struct _ATOM_VRAM_USAGE_BY_FIRMWARE
+{
+  ATOM_COMMON_TABLE_HEADER sHeader;  
+  ATOM_FIRMWARE_VRAM_RESERVE_INFO	asFirmwareVramReserveInfo[ATOM_MAX_FIRMWARE_VRAM_USAGE_INFO];
+}ATOM_VRAM_USAGE_BY_FIRMWARE;
 
-typedef struct _ATOM_VRAM_USAGE_BY_FIRMWARE {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	ATOM_FIRMWARE_VRAM_RESERVE_INFO
-	    asFirmwareVramReserveInfo[ATOM_MAX_FIRMWARE_VRAM_USAGE_INFO];
-} ATOM_VRAM_USAGE_BY_FIRMWARE;
+// change verion to 1.5, when allow driver to allocate the vram area for command table access. 
+typedef struct _ATOM_FIRMWARE_VRAM_RESERVE_INFO_V1_5
+{
+  ULONG   ulStartAddrUsedByFirmware;
+  USHORT  usFirmwareUseInKb;
+  USHORT  usFBUsedByDrvInKb;
+}ATOM_FIRMWARE_VRAM_RESERVE_INFO_V1_5;
 
-/****************************************************************************/
-/*  Structure used in GPIO_Pin_LUTTable */
-/****************************************************************************/
-typedef struct _ATOM_GPIO_PIN_ASSIGNMENT {
-	USHORT usGpioPin_AIndex;
-	UCHAR ucGpioPinBitShift;
-	UCHAR ucGPIO_ID;
-} ATOM_GPIO_PIN_ASSIGNMENT;
+typedef struct _ATOM_VRAM_USAGE_BY_FIRMWARE_V1_5
+{
+  ATOM_COMMON_TABLE_HEADER sHeader;  
+  ATOM_FIRMWARE_VRAM_RESERVE_INFO_V1_5	asFirmwareVramReserveInfo[ATOM_MAX_FIRMWARE_VRAM_USAGE_INFO];
+}ATOM_VRAM_USAGE_BY_FIRMWARE_V1_5;
+
+/****************************************************************************/	
+// Structure used in GPIO_Pin_LUTTable
+/****************************************************************************/	
+typedef struct _ATOM_GPIO_PIN_ASSIGNMENT
+{
+  USHORT                   usGpioPin_AIndex;
+  UCHAR                    ucGpioPinBitShift;
+  UCHAR                    ucGPIO_ID;
+}ATOM_GPIO_PIN_ASSIGNMENT;
 
-typedef struct _ATOM_GPIO_PIN_LUT {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	ATOM_GPIO_PIN_ASSIGNMENT asGPIO_Pin[1];
-} ATOM_GPIO_PIN_LUT;
+typedef struct _ATOM_GPIO_PIN_LUT
+{
+  ATOM_COMMON_TABLE_HEADER  sHeader;
+  ATOM_GPIO_PIN_ASSIGNMENT	asGPIO_Pin[1];
+}ATOM_GPIO_PIN_LUT;
 
-/****************************************************************************/
-/*  Structure used in ComponentVideoInfoTable */
-/****************************************************************************/
+/****************************************************************************/	
+// Structure used in ComponentVideoInfoTable	
+/****************************************************************************/	
 #define GPIO_PIN_ACTIVE_HIGH          0x1
 
 #define MAX_SUPPORTED_CV_STANDARDS    5
 
-/*  definitions for ATOM_D_INFO.ucSettings */
-#define ATOM_GPIO_SETTINGS_BITSHIFT_MASK  0x1F	/*  [4:0] */
-#define ATOM_GPIO_SETTINGS_RESERVED_MASK  0x60	/*  [6:5] = must be zeroed out */
-#define ATOM_GPIO_SETTINGS_ACTIVE_MASK    0x80	/*  [7] */
+// definitions for ATOM_D_INFO.ucSettings
+#define ATOM_GPIO_SETTINGS_BITSHIFT_MASK  0x1F    // [4:0]
+#define ATOM_GPIO_SETTINGS_RESERVED_MASK  0x60    // [6:5] = must be zeroed out
+#define ATOM_GPIO_SETTINGS_ACTIVE_MASK    0x80    // [7]
 
-typedef struct _ATOM_GPIO_INFO {
-	USHORT usAOffset;
-	UCHAR ucSettings;
-	UCHAR ucReserved;
-} ATOM_GPIO_INFO;
+typedef struct _ATOM_GPIO_INFO
+{
+  USHORT  usAOffset;
+  UCHAR   ucSettings;
+  UCHAR   ucReserved;
+}ATOM_GPIO_INFO;
 
-/*  definitions for ATOM_COMPONENT_VIDEO_INFO.ucMiscInfo (bit vector) */
+// definitions for ATOM_COMPONENT_VIDEO_INFO.ucMiscInfo (bit vector)
 #define ATOM_CV_RESTRICT_FORMAT_SELECTION           0x2
 
-/*  definitions for ATOM_COMPONENT_VIDEO_INFO.uc480i/uc480p/uc720p/uc1080i */
-#define ATOM_GPIO_DEFAULT_MODE_EN                   0x80	/* [7]; */
-#define ATOM_GPIO_SETTING_PERMODE_MASK              0x7F	/* [6:0] */
-
-/*  definitions for ATOM_COMPONENT_VIDEO_INFO.ucLetterBoxMode */
-/* Line 3 out put 5V. */
-#define ATOM_CV_LINE3_ASPECTRATIO_16_9_GPIO_A       0x01	/* represent gpio 3 state for 16:9 */
-#define ATOM_CV_LINE3_ASPECTRATIO_16_9_GPIO_B       0x02	/* represent gpio 4 state for 16:9 */
-#define ATOM_CV_LINE3_ASPECTRATIO_16_9_GPIO_SHIFT   0x0
-
-/* Line 3 out put 2.2V */
-#define ATOM_CV_LINE3_ASPECTRATIO_4_3_LETBOX_GPIO_A 0x04	/* represent gpio 3 state for 4:3 Letter box */
-#define ATOM_CV_LINE3_ASPECTRATIO_4_3_LETBOX_GPIO_B 0x08	/* represent gpio 4 state for 4:3 Letter box */
-#define ATOM_CV_LINE3_ASPECTRATIO_4_3_LETBOX_GPIO_SHIFT 0x2
-
-/* Line 3 out put 0V */
-#define ATOM_CV_LINE3_ASPECTRATIO_4_3_GPIO_A        0x10	/* represent gpio 3 state for 4:3 */
-#define ATOM_CV_LINE3_ASPECTRATIO_4_3_GPIO_B        0x20	/* represent gpio 4 state for 4:3 */
-#define ATOM_CV_LINE3_ASPECTRATIO_4_3_GPIO_SHIFT    0x4
-
-#define ATOM_CV_LINE3_ASPECTRATIO_MASK              0x3F	/*  bit [5:0] */
-
-#define ATOM_CV_LINE3_ASPECTRATIO_EXIST             0x80	/* bit 7 */
-
-/* GPIO bit index in gpio setting per mode value, also represend the block no. in gpio blocks. */
-#define ATOM_GPIO_INDEX_LINE3_ASPECRATIO_GPIO_A   3	/* bit 3 in uc480i/uc480p/uc720p/uc1080i, which represend the default gpio bit setting for the mode. */
-#define ATOM_GPIO_INDEX_LINE3_ASPECRATIO_GPIO_B   4	/* bit 4 in uc480i/uc480p/uc720p/uc1080i, which represend the default gpio bit setting for the mode. */
-
-typedef struct _ATOM_COMPONENT_VIDEO_INFO {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	USHORT usMask_PinRegisterIndex;
-	USHORT usEN_PinRegisterIndex;
-	USHORT usY_PinRegisterIndex;
-	USHORT usA_PinRegisterIndex;
-	UCHAR ucBitShift;
-	UCHAR ucPinActiveState;	/* ucPinActiveState: Bit0=1 active high, =0 active low */
-	ATOM_DTD_FORMAT sReserved;	/*  must be zeroed out */
-	UCHAR ucMiscInfo;
-	UCHAR uc480i;
-	UCHAR uc480p;
-	UCHAR uc720p;
-	UCHAR uc1080i;
-	UCHAR ucLetterBoxMode;
-	UCHAR ucReserved[3];
-	UCHAR ucNumOfWbGpioBlocks;	/* For Component video D-Connector support. If zere, NTSC type connector */
-	ATOM_GPIO_INFO aWbGpioStateBlock[MAX_SUPPORTED_CV_STANDARDS];
-	ATOM_DTD_FORMAT aModeTimings[MAX_SUPPORTED_CV_STANDARDS];
-} ATOM_COMPONENT_VIDEO_INFO;
-
-/* ucTableFormatRevision=2 */
-/* ucTableContentRevision=1 */
-typedef struct _ATOM_COMPONENT_VIDEO_INFO_V21 {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	UCHAR ucMiscInfo;
-	UCHAR uc480i;
-	UCHAR uc480p;
-	UCHAR uc720p;
-	UCHAR uc1080i;
-	UCHAR ucReserved;
-	UCHAR ucLetterBoxMode;
-	UCHAR ucNumOfWbGpioBlocks;	/* For Component video D-Connector support. If zere, NTSC type connector */
-	ATOM_GPIO_INFO aWbGpioStateBlock[MAX_SUPPORTED_CV_STANDARDS];
-	ATOM_DTD_FORMAT aModeTimings[MAX_SUPPORTED_CV_STANDARDS];
-} ATOM_COMPONENT_VIDEO_INFO_V21;
+// definitions for ATOM_COMPONENT_VIDEO_INFO.uc480i/uc480p/uc720p/uc1080i
+#define ATOM_GPIO_DEFAULT_MODE_EN                   0x80 //[7];
+#define ATOM_GPIO_SETTING_PERMODE_MASK              0x7F //[6:0]
+
+// definitions for ATOM_COMPONENT_VIDEO_INFO.ucLetterBoxMode
+//Line 3 out put 5V.
+#define ATOM_CV_LINE3_ASPECTRATIO_16_9_GPIO_A       0x01     //represent gpio 3 state for 16:9
+#define ATOM_CV_LINE3_ASPECTRATIO_16_9_GPIO_B       0x02     //represent gpio 4 state for 16:9
+#define ATOM_CV_LINE3_ASPECTRATIO_16_9_GPIO_SHIFT   0x0   
+
+//Line 3 out put 2.2V              
+#define ATOM_CV_LINE3_ASPECTRATIO_4_3_LETBOX_GPIO_A 0x04     //represent gpio 3 state for 4:3 Letter box
+#define ATOM_CV_LINE3_ASPECTRATIO_4_3_LETBOX_GPIO_B 0x08     //represent gpio 4 state for 4:3 Letter box
+#define ATOM_CV_LINE3_ASPECTRATIO_4_3_LETBOX_GPIO_SHIFT 0x2     
+
+//Line 3 out put 0V
+#define ATOM_CV_LINE3_ASPECTRATIO_4_3_GPIO_A        0x10     //represent gpio 3 state for 4:3
+#define ATOM_CV_LINE3_ASPECTRATIO_4_3_GPIO_B        0x20     //represent gpio 4 state for 4:3
+#define ATOM_CV_LINE3_ASPECTRATIO_4_3_GPIO_SHIFT    0x4 
+
+#define ATOM_CV_LINE3_ASPECTRATIO_MASK              0x3F     // bit [5:0]
+
+#define ATOM_CV_LINE3_ASPECTRATIO_EXIST             0x80     //bit 7
+
+//GPIO bit index in gpio setting per mode value, also represend the block no. in gpio blocks.
+#define ATOM_GPIO_INDEX_LINE3_ASPECRATIO_GPIO_A   3   //bit 3 in uc480i/uc480p/uc720p/uc1080i, which represend the default gpio bit setting for the mode.
+#define ATOM_GPIO_INDEX_LINE3_ASPECRATIO_GPIO_B   4   //bit 4 in uc480i/uc480p/uc720p/uc1080i, which represend the default gpio bit setting for the mode.
+
+
+typedef struct _ATOM_COMPONENT_VIDEO_INFO
+{
+  ATOM_COMMON_TABLE_HEADER sHeader;
+  USHORT             usMask_PinRegisterIndex;
+  USHORT             usEN_PinRegisterIndex;
+  USHORT             usY_PinRegisterIndex;
+  USHORT             usA_PinRegisterIndex;
+  UCHAR              ucBitShift;
+  UCHAR              ucPinActiveState;  //ucPinActiveState: Bit0=1 active high, =0 active low
+  ATOM_DTD_FORMAT    sReserved;         // must be zeroed out
+  UCHAR              ucMiscInfo;
+  UCHAR              uc480i;
+  UCHAR              uc480p;
+  UCHAR              uc720p;
+  UCHAR              uc1080i;
+  UCHAR              ucLetterBoxMode;
+  UCHAR              ucReserved[3];
+  UCHAR              ucNumOfWbGpioBlocks; //For Component video D-Connector support. If zere, NTSC type connector
+  ATOM_GPIO_INFO     aWbGpioStateBlock[MAX_SUPPORTED_CV_STANDARDS];
+  ATOM_DTD_FORMAT    aModeTimings[MAX_SUPPORTED_CV_STANDARDS];
+}ATOM_COMPONENT_VIDEO_INFO;
+
+//ucTableFormatRevision=2
+//ucTableContentRevision=1
+typedef struct _ATOM_COMPONENT_VIDEO_INFO_V21
+{
+  ATOM_COMMON_TABLE_HEADER sHeader;
+  UCHAR              ucMiscInfo;
+  UCHAR              uc480i;
+  UCHAR              uc480p;
+  UCHAR              uc720p;
+  UCHAR              uc1080i;
+  UCHAR              ucReserved;
+  UCHAR              ucLetterBoxMode;
+  UCHAR              ucNumOfWbGpioBlocks; //For Component video D-Connector support. If zere, NTSC type connector
+  ATOM_GPIO_INFO     aWbGpioStateBlock[MAX_SUPPORTED_CV_STANDARDS];
+  ATOM_DTD_FORMAT    aModeTimings[MAX_SUPPORTED_CV_STANDARDS];
+}ATOM_COMPONENT_VIDEO_INFO_V21;
 
 #define ATOM_COMPONENT_VIDEO_INFO_LAST  ATOM_COMPONENT_VIDEO_INFO_V21
 
-/****************************************************************************/
-/*  Structure used in object_InfoTable */
-/****************************************************************************/
-typedef struct _ATOM_OBJECT_HEADER {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	USHORT usDeviceSupport;
-	USHORT usConnectorObjectTableOffset;
-	USHORT usRouterObjectTableOffset;
-	USHORT usEncoderObjectTableOffset;
-	USHORT usProtectionObjectTableOffset;	/* only available when Protection block is independent. */
-	USHORT usDisplayPathTableOffset;
-} ATOM_OBJECT_HEADER;
-
-typedef struct _ATOM_DISPLAY_OBJECT_PATH {
-	USHORT usDeviceTag;	/* supported device */
-	USHORT usSize;		/* the size of ATOM_DISPLAY_OBJECT_PATH */
-	USHORT usConnObjectId;	/* Connector Object ID */
-	USHORT usGPUObjectId;	/* GPU ID */
-	USHORT usGraphicObjIds[1];	/* 1st Encoder Obj source from GPU to last Graphic Obj destinate to connector. */
-} ATOM_DISPLAY_OBJECT_PATH;
-
-typedef struct _ATOM_DISPLAY_OBJECT_PATH_TABLE {
-	UCHAR ucNumOfDispPath;
-	UCHAR ucVersion;
-	UCHAR ucPadding[2];
-	ATOM_DISPLAY_OBJECT_PATH asDispPath[1];
-} ATOM_DISPLAY_OBJECT_PATH_TABLE;
-
-typedef struct _ATOM_OBJECT	/* each object has this structure */
-{
-	USHORT usObjectID;
-	USHORT usSrcDstTableOffset;
-	USHORT usRecordOffset;	/* this pointing to a bunch of records defined below */
-	USHORT usReserved;
-} ATOM_OBJECT;
-
-typedef struct _ATOM_OBJECT_TABLE	/* Above 4 object table offset pointing to a bunch of objects all have this structure */
-{
-	UCHAR ucNumberOfObjects;
-	UCHAR ucPadding[3];
-	ATOM_OBJECT asObjects[1];
-} ATOM_OBJECT_TABLE;
-
-typedef struct _ATOM_SRC_DST_TABLE_FOR_ONE_OBJECT	/* usSrcDstTableOffset pointing to this structure */
-{
-	UCHAR ucNumberOfSrc;
-	USHORT usSrcObjectID[1];
-	UCHAR ucNumberOfDst;
-	USHORT usDstObjectID[1];
-} ATOM_SRC_DST_TABLE_FOR_ONE_OBJECT;
-
-/* Related definitions, all records are differnt but they have a commond header */
-typedef struct _ATOM_COMMON_RECORD_HEADER {
-	UCHAR ucRecordType;	/* An emun to indicate the record type */
-	UCHAR ucRecordSize;	/* The size of the whole record in byte */
-} ATOM_COMMON_RECORD_HEADER;
-
-#define ATOM_I2C_RECORD_TYPE                           1
+/****************************************************************************/	
+// Structure used in object_InfoTable
+/****************************************************************************/	
+typedef struct _ATOM_OBJECT_HEADER
+{ 
+  ATOM_COMMON_TABLE_HEADER	sHeader;
+  USHORT                    usDeviceSupport;
+  USHORT                    usConnectorObjectTableOffset;
+  USHORT                    usRouterObjectTableOffset;
+  USHORT                    usEncoderObjectTableOffset;
+  USHORT                    usProtectionObjectTableOffset; //only available when Protection block is independent.
+  USHORT                    usDisplayPathTableOffset;
+}ATOM_OBJECT_HEADER;
+
+typedef struct _ATOM_OBJECT_HEADER_V3
+{ 
+  ATOM_COMMON_TABLE_HEADER	sHeader;
+  USHORT                    usDeviceSupport;
+  USHORT                    usConnectorObjectTableOffset;
+  USHORT                    usRouterObjectTableOffset;
+  USHORT                    usEncoderObjectTableOffset;
+  USHORT                    usProtectionObjectTableOffset; //only available when Protection block is independent.
+  USHORT                    usDisplayPathTableOffset;
+  USHORT                    usMiscObjectTableOffset;
+}ATOM_OBJECT_HEADER_V3;
+
+typedef struct  _ATOM_DISPLAY_OBJECT_PATH
+{
+  USHORT    usDeviceTag;                                   //supported device 
+  USHORT    usSize;                                        //the size of ATOM_DISPLAY_OBJECT_PATH
+  USHORT    usConnObjectId;                                //Connector Object ID 
+  USHORT    usGPUObjectId;                                 //GPU ID 
+  USHORT    usGraphicObjIds[1];                             //1st Encoder Obj source from GPU to last Graphic Obj destinate to connector.
+}ATOM_DISPLAY_OBJECT_PATH;
+
+typedef struct _ATOM_DISPLAY_OBJECT_PATH_TABLE
+{
+  UCHAR                           ucNumOfDispPath;
+  UCHAR                           ucVersion;
+  UCHAR                           ucPadding[2];
+  ATOM_DISPLAY_OBJECT_PATH        asDispPath[1];
+}ATOM_DISPLAY_OBJECT_PATH_TABLE;
+
+
+typedef struct _ATOM_OBJECT                                //each object has this structure    
+{
+  USHORT              usObjectID;
+  USHORT              usSrcDstTableOffset;
+  USHORT              usRecordOffset;                     //this pointing to a bunch of records defined below
+  USHORT              usReserved;
+}ATOM_OBJECT;
+
+typedef struct _ATOM_OBJECT_TABLE                         //Above 4 object table offset pointing to a bunch of objects all have this structure     
+{
+  UCHAR               ucNumberOfObjects;
+  UCHAR               ucPadding[3];
+  ATOM_OBJECT         asObjects[1];
+}ATOM_OBJECT_TABLE;
+
+typedef struct _ATOM_SRC_DST_TABLE_FOR_ONE_OBJECT         //usSrcDstTableOffset pointing to this structure
+{
+  UCHAR               ucNumberOfSrc;
+  USHORT              usSrcObjectID[1];
+  UCHAR               ucNumberOfDst;
+  USHORT              usDstObjectID[1];
+}ATOM_SRC_DST_TABLE_FOR_ONE_OBJECT;
+
+
+//Two definitions below are for OPM on MXM module designs
+
+#define EXT_HPDPIN_LUTINDEX_0                   0
+#define EXT_HPDPIN_LUTINDEX_1                   1
+#define EXT_HPDPIN_LUTINDEX_2                   2
+#define EXT_HPDPIN_LUTINDEX_3                   3
+#define EXT_HPDPIN_LUTINDEX_4                   4
+#define EXT_HPDPIN_LUTINDEX_5                   5
+#define EXT_HPDPIN_LUTINDEX_6                   6
+#define EXT_HPDPIN_LUTINDEX_7                   7
+#define MAX_NUMBER_OF_EXT_HPDPIN_LUT_ENTRIES   (EXT_HPDPIN_LUTINDEX_7+1)
+
+#define EXT_AUXDDC_LUTINDEX_0                   0
+#define EXT_AUXDDC_LUTINDEX_1                   1
+#define EXT_AUXDDC_LUTINDEX_2                   2
+#define EXT_AUXDDC_LUTINDEX_3                   3
+#define EXT_AUXDDC_LUTINDEX_4                   4
+#define EXT_AUXDDC_LUTINDEX_5                   5
+#define EXT_AUXDDC_LUTINDEX_6                   6
+#define EXT_AUXDDC_LUTINDEX_7                   7
+#define MAX_NUMBER_OF_EXT_AUXDDC_LUT_ENTRIES   (EXT_AUXDDC_LUTINDEX_7+1)
+
+typedef struct _EXT_DISPLAY_PATH
+{
+  USHORT  usDeviceTag;                    //A bit vector to show what devices are supported 
+  USHORT  usDeviceACPIEnum;               //16bit device ACPI id. 
+  USHORT  usDeviceConnector;              //A physical connector for displays to plug in, using object connector definitions
+  UCHAR   ucExtAUXDDCLutIndex;            //An index into external AUX/DDC channel LUT
+  UCHAR   ucExtHPDPINLutIndex;            //An index into external HPD pin LUT
+  USHORT  usExtEncoderObjId;              //external encoder object id
+  USHORT  usReserved[3]; 
+}EXT_DISPLAY_PATH;
+   
+#define NUMBER_OF_UCHAR_FOR_GUID          16
+#define MAX_NUMBER_OF_EXT_DISPLAY_PATH    7
+
+typedef  struct _ATOM_EXTERNAL_DISPLAY_CONNECTION_INFO
+{
+  ATOM_COMMON_TABLE_HEADER sHeader;
+  UCHAR                    ucGuid [NUMBER_OF_UCHAR_FOR_GUID];     // a GUID is a 16 byte long string
+  EXT_DISPLAY_PATH         sPath[MAX_NUMBER_OF_EXT_DISPLAY_PATH]; // total of fixed 7 entries.
+  UCHAR                    ucChecksum;                            // a  simple Checksum of the sum of whole structure equal to 0x0. 
+  UCHAR                    Reserved [7];                          // for potential expansion
+}ATOM_EXTERNAL_DISPLAY_CONNECTION_INFO;
+
+//Related definitions, all records are differnt but they have a commond header
+typedef struct _ATOM_COMMON_RECORD_HEADER
+{
+  UCHAR               ucRecordType;                      //An emun to indicate the record type
+  UCHAR               ucRecordSize;                      //The size of the whole record in byte
+}ATOM_COMMON_RECORD_HEADER;
+
+
+#define ATOM_I2C_RECORD_TYPE                           1         
 #define ATOM_HPD_INT_RECORD_TYPE                       2
 #define ATOM_OUTPUT_PROTECTION_RECORD_TYPE             3
 #define ATOM_CONNECTOR_DEVICE_TAG_RECORD_TYPE          4
-#define	ATOM_CONNECTOR_DVI_EXT_INPUT_RECORD_TYPE	     5	/* Obsolete, switch to use GPIO_CNTL_RECORD_TYPE */
-#define ATOM_ENCODER_FPGA_CONTROL_RECORD_TYPE          6	/* Obsolete, switch to use GPIO_CNTL_RECORD_TYPE */
+#define	ATOM_CONNECTOR_DVI_EXT_INPUT_RECORD_TYPE	     5 //Obsolete, switch to use GPIO_CNTL_RECORD_TYPE
+#define ATOM_ENCODER_FPGA_CONTROL_RECORD_TYPE          6 //Obsolete, switch to use GPIO_CNTL_RECORD_TYPE
 #define ATOM_CONNECTOR_CVTV_SHARE_DIN_RECORD_TYPE      7
-#define ATOM_JTAG_RECORD_TYPE                          8	/* Obsolete, switch to use GPIO_CNTL_RECORD_TYPE */
+#define ATOM_JTAG_RECORD_TYPE                          8 //Obsolete, switch to use GPIO_CNTL_RECORD_TYPE
 #define ATOM_OBJECT_GPIO_CNTL_RECORD_TYPE              9
 #define ATOM_ENCODER_DVO_CF_RECORD_TYPE               10
 #define ATOM_CONNECTOR_CF_RECORD_TYPE                 11
 #define	ATOM_CONNECTOR_HARDCODE_DTD_RECORD_TYPE	      12
 #define ATOM_CONNECTOR_PCIE_SUBCONNECTOR_RECORD_TYPE  13
-#define ATOM_ROUTER_DDC_PATH_SELECT_RECORD_TYPE				14
-#define ATOM_ROUTER_DATA_CLOCK_PATH_SELECT_RECORD_TYPE					15
-
-/* Must be updated when new record type is added,equal to that record definition! */
-#define ATOM_MAX_OBJECT_RECORD_NUMBER             ATOM_CONNECTOR_CF_RECORD_TYPE
-
-typedef struct _ATOM_I2C_RECORD {
-	ATOM_COMMON_RECORD_HEADER sheader;
-	ATOM_I2C_ID_CONFIG sucI2cId;
-	UCHAR ucI2CAddr;	/* The slave address, it's 0 when the record is attached to connector for DDC */
-} ATOM_I2C_RECORD;
-
-typedef struct _ATOM_HPD_INT_RECORD {
-	ATOM_COMMON_RECORD_HEADER sheader;
-	UCHAR ucHPDIntGPIOID;	/* Corresponding block in GPIO_PIN_INFO table gives the pin info */
-	UCHAR ucPluggged_PinState;
-} ATOM_HPD_INT_RECORD;
-
-typedef struct _ATOM_OUTPUT_PROTECTION_RECORD {
-	ATOM_COMMON_RECORD_HEADER sheader;
-	UCHAR ucProtectionFlag;
-	UCHAR ucReserved;
-} ATOM_OUTPUT_PROTECTION_RECORD;
-
-typedef struct _ATOM_CONNECTOR_DEVICE_TAG {
-	ULONG ulACPIDeviceEnum;	/* Reserved for now */
-	USHORT usDeviceID;	/* This Id is same as "ATOM_DEVICE_XXX_SUPPORT" */
-	USHORT usPadding;
-} ATOM_CONNECTOR_DEVICE_TAG;
-
-typedef struct _ATOM_CONNECTOR_DEVICE_TAG_RECORD {
-	ATOM_COMMON_RECORD_HEADER sheader;
-	UCHAR ucNumberOfDevice;
-	UCHAR ucReserved;
-	ATOM_CONNECTOR_DEVICE_TAG asDeviceTag[1];	/* This Id is same as "ATOM_DEVICE_XXX_SUPPORT", 1 is only for allocation */
-} ATOM_CONNECTOR_DEVICE_TAG_RECORD;
-
-typedef struct _ATOM_CONNECTOR_DVI_EXT_INPUT_RECORD {
-	ATOM_COMMON_RECORD_HEADER sheader;
-	UCHAR ucConfigGPIOID;
-	UCHAR ucConfigGPIOState;	/* Set to 1 when it's active high to enable external flow in */
-	UCHAR ucFlowinGPIPID;
-	UCHAR ucExtInGPIPID;
-} ATOM_CONNECTOR_DVI_EXT_INPUT_RECORD;
-
-typedef struct _ATOM_ENCODER_FPGA_CONTROL_RECORD {
-	ATOM_COMMON_RECORD_HEADER sheader;
-	UCHAR ucCTL1GPIO_ID;
-	UCHAR ucCTL1GPIOState;	/* Set to 1 when it's active high */
-	UCHAR ucCTL2GPIO_ID;
-	UCHAR ucCTL2GPIOState;	/* Set to 1 when it's active high */
-	UCHAR ucCTL3GPIO_ID;
-	UCHAR ucCTL3GPIOState;	/* Set to 1 when it's active high */
-	UCHAR ucCTLFPGA_IN_ID;
-	UCHAR ucPadding[3];
-} ATOM_ENCODER_FPGA_CONTROL_RECORD;
-
-typedef struct _ATOM_CONNECTOR_CVTV_SHARE_DIN_RECORD {
-	ATOM_COMMON_RECORD_HEADER sheader;
-	UCHAR ucGPIOID;		/* Corresponding block in GPIO_PIN_INFO table gives the pin info */
-	UCHAR ucTVActiveState;	/* Indicating when the pin==0 or 1 when TV is connected */
-} ATOM_CONNECTOR_CVTV_SHARE_DIN_RECORD;
-
-typedef struct _ATOM_JTAG_RECORD {
-	ATOM_COMMON_RECORD_HEADER sheader;
-	UCHAR ucTMSGPIO_ID;
-	UCHAR ucTMSGPIOState;	/* Set to 1 when it's active high */
-	UCHAR ucTCKGPIO_ID;
-	UCHAR ucTCKGPIOState;	/* Set to 1 when it's active high */
-	UCHAR ucTDOGPIO_ID;
-	UCHAR ucTDOGPIOState;	/* Set to 1 when it's active high */
-	UCHAR ucTDIGPIO_ID;
-	UCHAR ucTDIGPIOState;	/* Set to 1 when it's active high */
-	UCHAR ucPadding[2];
-} ATOM_JTAG_RECORD;
-
-/* The following generic object gpio pin control record type will replace JTAG_RECORD/FPGA_CONTROL_RECORD/DVI_EXT_INPUT_RECORD above gradually */
-typedef struct _ATOM_GPIO_PIN_CONTROL_PAIR {
-	UCHAR ucGPIOID;		/*  GPIO_ID, find the corresponding ID in GPIO_LUT table */
-	UCHAR ucGPIO_PinState;	/*  Pin state showing how to set-up the pin */
-} ATOM_GPIO_PIN_CONTROL_PAIR;
-
-typedef struct _ATOM_OBJECT_GPIO_CNTL_RECORD {
-	ATOM_COMMON_RECORD_HEADER sheader;
-	UCHAR ucFlags;		/*  Future expnadibility */
-	UCHAR ucNumberOfPins;	/*  Number of GPIO pins used to control the object */
-	ATOM_GPIO_PIN_CONTROL_PAIR asGpio[1];	/*  the real gpio pin pair determined by number of pins ucNumberOfPins */
-} ATOM_OBJECT_GPIO_CNTL_RECORD;
-
-/* Definitions for GPIO pin state */
+#define ATOM_ROUTER_DDC_PATH_SELECT_RECORD_TYPE	      14
+#define ATOM_ROUTER_DATA_CLOCK_PATH_SELECT_RECORD_TYPE	15
+#define ATOM_CONNECTOR_HPDPIN_LUT_RECORD_TYPE          16 //This is for the case when connectors are not known to object table
+#define ATOM_CONNECTOR_AUXDDC_LUT_RECORD_TYPE          17 //This is for the case when connectors are not known to object table
+#define ATOM_OBJECT_LINK_RECORD_TYPE                   18 //Once this record is present under one object, it indicats the oobject is linked to another obj described by the record
+#define ATOM_CONNECTOR_REMOTE_CAP_RECORD_TYPE          19
+
+
+//Must be updated when new record type is added,equal to that record definition!
+#define ATOM_MAX_OBJECT_RECORD_NUMBER             ATOM_CONNECTOR_REMOTE_CAP_RECORD_TYPE
+
+typedef struct  _ATOM_I2C_RECORD
+{
+  ATOM_COMMON_RECORD_HEADER   sheader;
+  ATOM_I2C_ID_CONFIG          sucI2cId; 
+  UCHAR                       ucI2CAddr;              //The slave address, it's 0 when the record is attached to connector for DDC
+}ATOM_I2C_RECORD;
+
+typedef struct  _ATOM_HPD_INT_RECORD
+{
+  ATOM_COMMON_RECORD_HEADER   sheader;
+  UCHAR                       ucHPDIntGPIOID;         //Corresponding block in GPIO_PIN_INFO table gives the pin info           
+  UCHAR                       ucPlugged_PinState;
+}ATOM_HPD_INT_RECORD;
+
+
+typedef struct  _ATOM_OUTPUT_PROTECTION_RECORD 
+{
+  ATOM_COMMON_RECORD_HEADER   sheader;
+  UCHAR                       ucProtectionFlag;
+  UCHAR                       ucReserved;
+}ATOM_OUTPUT_PROTECTION_RECORD;
+
+typedef struct  _ATOM_CONNECTOR_DEVICE_TAG
+{
+  ULONG                       ulACPIDeviceEnum;       //Reserved for now
+  USHORT                      usDeviceID;             //This Id is same as "ATOM_DEVICE_XXX_SUPPORT"
+  USHORT                      usPadding;
+}ATOM_CONNECTOR_DEVICE_TAG;
+
+typedef struct  _ATOM_CONNECTOR_DEVICE_TAG_RECORD
+{
+  ATOM_COMMON_RECORD_HEADER   sheader;
+  UCHAR                       ucNumberOfDevice;
+  UCHAR                       ucReserved;
+  ATOM_CONNECTOR_DEVICE_TAG   asDeviceTag[1];         //This Id is same as "ATOM_DEVICE_XXX_SUPPORT", 1 is only for allocation
+}ATOM_CONNECTOR_DEVICE_TAG_RECORD;
+
+
+typedef struct  _ATOM_CONNECTOR_DVI_EXT_INPUT_RECORD
+{
+  ATOM_COMMON_RECORD_HEADER   sheader;
+  UCHAR						            ucConfigGPIOID;
+  UCHAR						            ucConfigGPIOState;	    //Set to 1 when it's active high to enable external flow in
+  UCHAR                       ucFlowinGPIPID;
+  UCHAR                       ucExtInGPIPID;
+}ATOM_CONNECTOR_DVI_EXT_INPUT_RECORD;
+
+typedef struct  _ATOM_ENCODER_FPGA_CONTROL_RECORD
+{
+  ATOM_COMMON_RECORD_HEADER   sheader;
+  UCHAR                       ucCTL1GPIO_ID;
+  UCHAR                       ucCTL1GPIOState;        //Set to 1 when it's active high
+  UCHAR                       ucCTL2GPIO_ID;
+  UCHAR                       ucCTL2GPIOState;        //Set to 1 when it's active high
+  UCHAR                       ucCTL3GPIO_ID;
+  UCHAR                       ucCTL3GPIOState;        //Set to 1 when it's active high
+  UCHAR                       ucCTLFPGA_IN_ID;
+  UCHAR                       ucPadding[3];
+}ATOM_ENCODER_FPGA_CONTROL_RECORD;
+
+typedef struct  _ATOM_CONNECTOR_CVTV_SHARE_DIN_RECORD
+{
+  ATOM_COMMON_RECORD_HEADER   sheader;
+  UCHAR                       ucGPIOID;               //Corresponding block in GPIO_PIN_INFO table gives the pin info 
+  UCHAR                       ucTVActiveState;        //Indicating when the pin==0 or 1 when TV is connected
+}ATOM_CONNECTOR_CVTV_SHARE_DIN_RECORD;
+
+typedef struct  _ATOM_JTAG_RECORD
+{
+  ATOM_COMMON_RECORD_HEADER   sheader;
+  UCHAR                       ucTMSGPIO_ID;
+  UCHAR                       ucTMSGPIOState;         //Set to 1 when it's active high
+  UCHAR                       ucTCKGPIO_ID;
+  UCHAR                       ucTCKGPIOState;         //Set to 1 when it's active high
+  UCHAR                       ucTDOGPIO_ID;
+  UCHAR                       ucTDOGPIOState;         //Set to 1 when it's active high
+  UCHAR                       ucTDIGPIO_ID;
+  UCHAR                       ucTDIGPIOState;         //Set to 1 when it's active high
+  UCHAR                       ucPadding[2];
+}ATOM_JTAG_RECORD;
+
+
+//The following generic object gpio pin control record type will replace JTAG_RECORD/FPGA_CONTROL_RECORD/DVI_EXT_INPUT_RECORD above gradually
+typedef struct _ATOM_GPIO_PIN_CONTROL_PAIR
+{
+  UCHAR                       ucGPIOID;               // GPIO_ID, find the corresponding ID in GPIO_LUT table
+  UCHAR                       ucGPIO_PinState;        // Pin state showing how to set-up the pin
+}ATOM_GPIO_PIN_CONTROL_PAIR;
+
+typedef struct  _ATOM_OBJECT_GPIO_CNTL_RECORD
+{
+  ATOM_COMMON_RECORD_HEADER   sheader;
+  UCHAR                       ucFlags;                // Future expnadibility
+  UCHAR                       ucNumberOfPins;         // Number of GPIO pins used to control the object
+  ATOM_GPIO_PIN_CONTROL_PAIR  asGpio[1];              // the real gpio pin pair determined by number of pins ucNumberOfPins
+}ATOM_OBJECT_GPIO_CNTL_RECORD;
+
+//Definitions for GPIO pin state 
 #define GPIO_PIN_TYPE_INPUT             0x00
 #define GPIO_PIN_TYPE_OUTPUT            0x10
 #define GPIO_PIN_TYPE_HW_CONTROL        0x20
 
-/* For GPIO_PIN_TYPE_OUTPUT the following is defined */
+//For GPIO_PIN_TYPE_OUTPUT the following is defined 
 #define GPIO_PIN_OUTPUT_STATE_MASK      0x01
 #define GPIO_PIN_OUTPUT_STATE_SHIFT     0
 #define GPIO_PIN_STATE_ACTIVE_LOW       0x0
 #define GPIO_PIN_STATE_ACTIVE_HIGH      0x1
 
-typedef struct _ATOM_ENCODER_DVO_CF_RECORD {
-	ATOM_COMMON_RECORD_HEADER sheader;
-	ULONG ulStrengthControl;	/*  DVOA strength control for CF */
-	UCHAR ucPadding[2];
-} ATOM_ENCODER_DVO_CF_RECORD;
+// Indexes to GPIO array in GLSync record 
+#define ATOM_GPIO_INDEX_GLSYNC_REFCLK    0
+#define ATOM_GPIO_INDEX_GLSYNC_HSYNC     1
+#define ATOM_GPIO_INDEX_GLSYNC_VSYNC     2
+#define ATOM_GPIO_INDEX_GLSYNC_SWAP_REQ  3
+#define ATOM_GPIO_INDEX_GLSYNC_SWAP_GNT  4
+#define ATOM_GPIO_INDEX_GLSYNC_INTERRUPT 5
+#define ATOM_GPIO_INDEX_GLSYNC_V_RESET   6
+#define ATOM_GPIO_INDEX_GLSYNC_MAX       7
+
+typedef struct  _ATOM_ENCODER_DVO_CF_RECORD
+{
+  ATOM_COMMON_RECORD_HEADER   sheader;
+  ULONG                       ulStrengthControl;      // DVOA strength control for CF
+  UCHAR                       ucPadding[2];
+}ATOM_ENCODER_DVO_CF_RECORD;
 
-/*  value for ATOM_CONNECTOR_CF_RECORD.ucConnectedDvoBundle */
+// value for ATOM_CONNECTOR_CF_RECORD.ucConnectedDvoBundle
 #define ATOM_CONNECTOR_CF_RECORD_CONNECTED_UPPER12BITBUNDLEA   1
 #define ATOM_CONNECTOR_CF_RECORD_CONNECTED_LOWER12BITBUNDLEB   2
 
-typedef struct _ATOM_CONNECTOR_CF_RECORD {
-	ATOM_COMMON_RECORD_HEADER sheader;
-	USHORT usMaxPixClk;
-	UCHAR ucFlowCntlGpioId;
-	UCHAR ucSwapCntlGpioId;
-	UCHAR ucConnectedDvoBundle;
-	UCHAR ucPadding;
-} ATOM_CONNECTOR_CF_RECORD;
-
-typedef struct _ATOM_CONNECTOR_HARDCODE_DTD_RECORD {
-	ATOM_COMMON_RECORD_HEADER sheader;
-	ATOM_DTD_FORMAT asTiming;
-} ATOM_CONNECTOR_HARDCODE_DTD_RECORD;
-
-typedef struct _ATOM_CONNECTOR_PCIE_SUBCONNECTOR_RECORD {
-	ATOM_COMMON_RECORD_HEADER sheader;	/* ATOM_CONNECTOR_PCIE_SUBCONNECTOR_RECORD_TYPE */
-	UCHAR ucSubConnectorType;	/* CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_D|X_ID_DUAL_LINK_DVI_D|HDMI_TYPE_A */
-	UCHAR ucReserved;
-} ATOM_CONNECTOR_PCIE_SUBCONNECTOR_RECORD;
-
-typedef struct _ATOM_ROUTER_DDC_PATH_SELECT_RECORD {
-	ATOM_COMMON_RECORD_HEADER sheader;
-	UCHAR ucMuxType;	/* decide the number of ucMuxState, =0, no pin state, =1: single state with complement, >1: multiple state */
-	UCHAR ucMuxControlPin;
-	UCHAR ucMuxState[2];	/* for alligment purpose */
-} ATOM_ROUTER_DDC_PATH_SELECT_RECORD;
-
-typedef struct _ATOM_ROUTER_DATA_CLOCK_PATH_SELECT_RECORD {
-	ATOM_COMMON_RECORD_HEADER sheader;
-	UCHAR ucMuxType;
-	UCHAR ucMuxControlPin;
-	UCHAR ucMuxState[2];	/* for alligment purpose */
-} ATOM_ROUTER_DATA_CLOCK_PATH_SELECT_RECORD;
-
-/*  define ucMuxType */
+typedef struct  _ATOM_CONNECTOR_CF_RECORD
+{
+  ATOM_COMMON_RECORD_HEADER   sheader;
+  USHORT                      usMaxPixClk;
+  UCHAR                       ucFlowCntlGpioId;
+  UCHAR                       ucSwapCntlGpioId;
+  UCHAR                       ucConnectedDvoBundle;
+  UCHAR                       ucPadding;
+}ATOM_CONNECTOR_CF_RECORD;
+
+typedef struct  _ATOM_CONNECTOR_HARDCODE_DTD_RECORD
+{
+  ATOM_COMMON_RECORD_HEADER   sheader;
+	ATOM_DTD_FORMAT							asTiming;
+}ATOM_CONNECTOR_HARDCODE_DTD_RECORD;
+
+typedef struct _ATOM_CONNECTOR_PCIE_SUBCONNECTOR_RECORD
+{
+  ATOM_COMMON_RECORD_HEADER   sheader;                //ATOM_CONNECTOR_PCIE_SUBCONNECTOR_RECORD_TYPE
+  UCHAR                       ucSubConnectorType;     //CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_D|X_ID_DUAL_LINK_DVI_D|HDMI_TYPE_A
+  UCHAR                       ucReserved;
+}ATOM_CONNECTOR_PCIE_SUBCONNECTOR_RECORD;
+
+
+typedef struct _ATOM_ROUTER_DDC_PATH_SELECT_RECORD
+{
+	ATOM_COMMON_RECORD_HEADER   sheader;                
+	UCHAR												ucMuxType;							//decide the number of ucMuxState, =0, no pin state, =1: single state with complement, >1: multiple state
+	UCHAR												ucMuxControlPin;
+	UCHAR												ucMuxState[2];					//for alligment purpose
+}ATOM_ROUTER_DDC_PATH_SELECT_RECORD;
+
+typedef struct _ATOM_ROUTER_DATA_CLOCK_PATH_SELECT_RECORD
+{
+	ATOM_COMMON_RECORD_HEADER   sheader;                
+	UCHAR												ucMuxType;
+	UCHAR												ucMuxControlPin;
+	UCHAR												ucMuxState[2];					//for alligment purpose
+}ATOM_ROUTER_DATA_CLOCK_PATH_SELECT_RECORD;
+
+// define ucMuxType
 #define ATOM_ROUTER_MUX_PIN_STATE_MASK								0x0f
 #define ATOM_ROUTER_MUX_PIN_SINGLE_STATE_COMPLEMENT		0x01
 
-/****************************************************************************/
-/*  ASIC voltage data table */
-/****************************************************************************/
-typedef struct _ATOM_VOLTAGE_INFO_HEADER {
-	USHORT usVDDCBaseLevel;	/* In number of 50mv unit */
-	USHORT usReserved;	/* For possible extension table offset */
-	UCHAR ucNumOfVoltageEntries;
-	UCHAR ucBytesPerVoltageEntry;
-	UCHAR ucVoltageStep;	/* Indicating in how many mv increament is one step, 0.5mv unit */
-	UCHAR ucDefaultVoltageEntry;
-	UCHAR ucVoltageControlI2cLine;
-	UCHAR ucVoltageControlAddress;
-	UCHAR ucVoltageControlOffset;
-} ATOM_VOLTAGE_INFO_HEADER;
-
-typedef struct _ATOM_VOLTAGE_INFO {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	ATOM_VOLTAGE_INFO_HEADER viHeader;
-	UCHAR ucVoltageEntries[64];	/* 64 is for allocation, the actual number of entry is present at ucNumOfVoltageEntries*ucBytesPerVoltageEntry */
-} ATOM_VOLTAGE_INFO;
-
-typedef struct _ATOM_VOLTAGE_FORMULA {
-	USHORT usVoltageBaseLevel;	/*  In number of 1mv unit */
-	USHORT usVoltageStep;	/*  Indicating in how many mv increament is one step, 1mv unit */
-	UCHAR ucNumOfVoltageEntries;	/*  Number of Voltage Entry, which indicate max Voltage */
-	UCHAR ucFlag;		/*  bit0=0 :step is 1mv =1 0.5mv */
-	UCHAR ucBaseVID;	/*  if there is no lookup table, VID= BaseVID + ( Vol - BaseLevle ) /VoltageStep */
-	UCHAR ucReserved;
-	UCHAR ucVIDAdjustEntries[32];	/*  32 is for allocation, the actual number of entry is present at ucNumOfVoltageEntries */
-} ATOM_VOLTAGE_FORMULA;
-
-typedef struct _ATOM_VOLTAGE_CONTROL {
-	UCHAR ucVoltageControlId;	/* Indicate it is controlled by I2C or GPIO or HW state machine */
-	UCHAR ucVoltageControlI2cLine;
-	UCHAR ucVoltageControlAddress;
-	UCHAR ucVoltageControlOffset;
-	USHORT usGpioPin_AIndex;	/* GPIO_PAD register index */
-	UCHAR ucGpioPinBitShift[9];	/* at most 8 pin support 255 VIDs, termintate with 0xff */
-	UCHAR ucReserved;
-} ATOM_VOLTAGE_CONTROL;
-
-/*  Define ucVoltageControlId */
+typedef struct _ATOM_CONNECTOR_HPDPIN_LUT_RECORD     //record for ATOM_CONNECTOR_HPDPIN_LUT_RECORD_TYPE
+{
+  ATOM_COMMON_RECORD_HEADER   sheader;
+  UCHAR                       ucHPDPINMap[MAX_NUMBER_OF_EXT_HPDPIN_LUT_ENTRIES];  //An fixed size array which maps external pins to internal GPIO_PIN_INFO table 
+}ATOM_CONNECTOR_HPDPIN_LUT_RECORD;
+
+typedef struct _ATOM_CONNECTOR_AUXDDC_LUT_RECORD  //record for ATOM_CONNECTOR_AUXDDC_LUT_RECORD_TYPE
+{
+  ATOM_COMMON_RECORD_HEADER   sheader;
+  ATOM_I2C_ID_CONFIG          ucAUXDDCMap[MAX_NUMBER_OF_EXT_AUXDDC_LUT_ENTRIES];  //An fixed size array which maps external pins to internal DDC ID
+}ATOM_CONNECTOR_AUXDDC_LUT_RECORD;
+
+typedef struct _ATOM_OBJECT_LINK_RECORD
+{
+  ATOM_COMMON_RECORD_HEADER   sheader;
+  USHORT                      usObjectID;         //could be connector, encorder or other object in object.h
+}ATOM_OBJECT_LINK_RECORD;
+
+typedef struct _ATOM_CONNECTOR_REMOTE_CAP_RECORD
+{
+  ATOM_COMMON_RECORD_HEADER   sheader;
+  USHORT                      usReserved;
+}ATOM_CONNECTOR_REMOTE_CAP_RECORD;
+
+/****************************************************************************/	
+// ASIC voltage data table
+/****************************************************************************/	
+typedef struct  _ATOM_VOLTAGE_INFO_HEADER
+{
+   USHORT   usVDDCBaseLevel;                //In number of 50mv unit
+   USHORT   usReserved;                     //For possible extension table offset
+   UCHAR    ucNumOfVoltageEntries;
+   UCHAR    ucBytesPerVoltageEntry;
+   UCHAR    ucVoltageStep;                  //Indicating in how many mv increament is one step, 0.5mv unit
+   UCHAR    ucDefaultVoltageEntry;
+   UCHAR    ucVoltageControlI2cLine;
+   UCHAR    ucVoltageControlAddress;
+   UCHAR    ucVoltageControlOffset;
+}ATOM_VOLTAGE_INFO_HEADER;
+
+typedef struct  _ATOM_VOLTAGE_INFO
+{
+   ATOM_COMMON_TABLE_HEADER	sHeader; 
+   ATOM_VOLTAGE_INFO_HEADER viHeader;
+   UCHAR    ucVoltageEntries[64];            //64 is for allocation, the actual number of entry is present at ucNumOfVoltageEntries*ucBytesPerVoltageEntry
+}ATOM_VOLTAGE_INFO;
+
+
+typedef struct  _ATOM_VOLTAGE_FORMULA
+{
+   USHORT   usVoltageBaseLevel;             // In number of 1mv unit
+   USHORT   usVoltageStep;                  // Indicating in how many mv increament is one step, 1mv unit
+	 UCHAR		ucNumOfVoltageEntries;					// Number of Voltage Entry, which indicate max Voltage
+	 UCHAR		ucFlag;													// bit0=0 :step is 1mv =1 0.5mv
+	 UCHAR		ucBaseVID;											// if there is no lookup table, VID= BaseVID + ( Vol - BaseLevle ) /VoltageStep
+	 UCHAR		ucReserved;
+	 UCHAR		ucVIDAdjustEntries[32];					// 32 is for allocation, the actual number of entry is present at ucNumOfVoltageEntries
+}ATOM_VOLTAGE_FORMULA;
+
+typedef struct  _VOLTAGE_LUT_ENTRY
+{
+	 USHORT		usVoltageCode;									// The Voltage ID, either GPIO or I2C code
+	 USHORT		usVoltageValue;									// The corresponding Voltage Value, in mV
+}VOLTAGE_LUT_ENTRY;
+
+typedef struct  _ATOM_VOLTAGE_FORMULA_V2
+{
+	 UCHAR		ucNumOfVoltageEntries;					// Number of Voltage Entry, which indicate max Voltage
+	 UCHAR		ucReserved[3];
+	 VOLTAGE_LUT_ENTRY asVIDAdjustEntries[32];// 32 is for allocation, the actual number of entries is in ucNumOfVoltageEntries
+}ATOM_VOLTAGE_FORMULA_V2;
+
+typedef struct _ATOM_VOLTAGE_CONTROL
+{
+	UCHAR		 ucVoltageControlId;							//Indicate it is controlled by I2C or GPIO or HW state machine		
+  UCHAR    ucVoltageControlI2cLine;
+  UCHAR    ucVoltageControlAddress;
+  UCHAR    ucVoltageControlOffset;	 	
+  USHORT   usGpioPin_AIndex;								//GPIO_PAD register index
+  UCHAR    ucGpioPinBitShift[9];						//at most 8 pin support 255 VIDs, termintate with 0xff
+	UCHAR		 ucReserved;
+}ATOM_VOLTAGE_CONTROL;
+
+// Define ucVoltageControlId
 #define	VOLTAGE_CONTROLLED_BY_HW							0x00
 #define	VOLTAGE_CONTROLLED_BY_I2C_MASK				0x7F
 #define	VOLTAGE_CONTROLLED_BY_GPIO						0x80
-#define	VOLTAGE_CONTROL_ID_LM64								0x01	/* I2C control, used for R5xx Core Voltage */
-#define	VOLTAGE_CONTROL_ID_DAC								0x02	/* I2C control, used for R5xx/R6xx MVDDC,MVDDQ or VDDCI */
-#define	VOLTAGE_CONTROL_ID_VT116xM						0x03	/* I2C control, used for R6xx Core Voltage */
-#define VOLTAGE_CONTROL_ID_DS4402							0x04
-
-typedef struct _ATOM_VOLTAGE_OBJECT {
-	UCHAR ucVoltageType;	/* Indicate Voltage Source: VDDC, MVDDC, MVDDQ or MVDDCI */
-	UCHAR ucSize;		/* Size of Object */
-	ATOM_VOLTAGE_CONTROL asControl;	/* describ how to control */
-	ATOM_VOLTAGE_FORMULA asFormula;	/* Indicate How to convert real Voltage to VID */
-} ATOM_VOLTAGE_OBJECT;
-
-typedef struct _ATOM_VOLTAGE_OBJECT_INFO {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	ATOM_VOLTAGE_OBJECT asVoltageObj[3];	/* Info for Voltage control */
-} ATOM_VOLTAGE_OBJECT_INFO;
-
-typedef struct _ATOM_LEAKID_VOLTAGE {
-	UCHAR ucLeakageId;
-	UCHAR ucReserved;
-	USHORT usVoltage;
-} ATOM_LEAKID_VOLTAGE;
-
-typedef struct _ATOM_ASIC_PROFILE_VOLTAGE {
-	UCHAR ucProfileId;
-	UCHAR ucReserved;
-	USHORT usSize;
-	USHORT usEfuseSpareStartAddr;
-	USHORT usFuseIndex[8];	/* from LSB to MSB, Max 8bit,end of 0xffff if less than 8 efuse id, */
-	ATOM_LEAKID_VOLTAGE asLeakVol[2];	/* Leakid and relatd voltage */
-} ATOM_ASIC_PROFILE_VOLTAGE;
-
-/* ucProfileId */
-#define	ATOM_ASIC_PROFILE_ID_EFUSE_VOLTAGE			1
+#define	VOLTAGE_CONTROL_ID_LM64								0x01									//I2C control, used for R5xx Core Voltage
+#define	VOLTAGE_CONTROL_ID_DAC								0x02									//I2C control, used for R5xx/R6xx MVDDC,MVDDQ or VDDCI
+#define	VOLTAGE_CONTROL_ID_VT116xM						0x03									//I2C control, used for R6xx Core Voltage
+#define VOLTAGE_CONTROL_ID_DS4402							0x04									
+
+typedef struct  _ATOM_VOLTAGE_OBJECT
+{
+ 	 UCHAR		ucVoltageType;									//Indicate Voltage Source: VDDC, MVDDC, MVDDQ or MVDDCI	 
+	 UCHAR		ucSize;													//Size of Object	
+	 ATOM_VOLTAGE_CONTROL			asControl;			//describ how to control 	 
+ 	 ATOM_VOLTAGE_FORMULA			asFormula;			//Indicate How to convert real Voltage to VID 
+}ATOM_VOLTAGE_OBJECT;
+
+typedef struct  _ATOM_VOLTAGE_OBJECT_V2
+{
+ 	 UCHAR		ucVoltageType;									//Indicate Voltage Source: VDDC, MVDDC, MVDDQ or MVDDCI	 
+	 UCHAR		ucSize;													//Size of Object	
+	 ATOM_VOLTAGE_CONTROL			asControl;			//describ how to control 	 
+ 	 ATOM_VOLTAGE_FORMULA_V2	asFormula;			//Indicate How to convert real Voltage to VID 
+}ATOM_VOLTAGE_OBJECT_V2;
+
+typedef struct  _ATOM_VOLTAGE_OBJECT_INFO
+{
+   ATOM_COMMON_TABLE_HEADER	sHeader; 
+	 ATOM_VOLTAGE_OBJECT			asVoltageObj[3];	//Info for Voltage control	  	 
+}ATOM_VOLTAGE_OBJECT_INFO;
+
+typedef struct  _ATOM_VOLTAGE_OBJECT_INFO_V2
+{
+   ATOM_COMMON_TABLE_HEADER	sHeader; 
+	 ATOM_VOLTAGE_OBJECT_V2			asVoltageObj[3];	//Info for Voltage control	  	 
+}ATOM_VOLTAGE_OBJECT_INFO_V2;
+
+typedef struct  _ATOM_LEAKID_VOLTAGE
+{
+	UCHAR		ucLeakageId;
+	UCHAR		ucReserved;
+	USHORT	usVoltage;
+}ATOM_LEAKID_VOLTAGE;
+
+typedef struct  _ATOM_ASIC_PROFILE_VOLTAGE
+{
+	UCHAR		ucProfileId;
+	UCHAR		ucReserved;
+	USHORT	usSize;
+	USHORT	usEfuseSpareStartAddr;
+	USHORT	usFuseIndex[8];												//from LSB to MSB, Max 8bit,end of 0xffff if less than 8 efuse id, 
+	ATOM_LEAKID_VOLTAGE					asLeakVol[2];			//Leakid and relatd voltage
+}ATOM_ASIC_PROFILE_VOLTAGE;
+
+//ucProfileId
+#define	ATOM_ASIC_PROFILE_ID_EFUSE_VOLTAGE			1		
 #define	ATOM_ASIC_PROFILE_ID_EFUSE_PERFORMANCE_VOLTAGE			1
 #define	ATOM_ASIC_PROFILE_ID_EFUSE_THERMAL_VOLTAGE					2
 
-typedef struct _ATOM_ASIC_PROFILING_INFO {
-	ATOM_COMMON_TABLE_HEADER asHeader;
-	ATOM_ASIC_PROFILE_VOLTAGE asVoltage;
-} ATOM_ASIC_PROFILING_INFO;
-
-typedef struct _ATOM_POWER_SOURCE_OBJECT {
-	UCHAR ucPwrSrcId;	/*  Power source */
-	UCHAR ucPwrSensorType;	/*  GPIO, I2C or none */
-	UCHAR ucPwrSensId;	/*  if GPIO detect, it is GPIO id,  if I2C detect, it is I2C id */
-	UCHAR ucPwrSensSlaveAddr;	/*  Slave address if I2C detect */
-	UCHAR ucPwrSensRegIndex;	/*  I2C register Index if I2C detect */
-	UCHAR ucPwrSensRegBitMask;	/*  detect which bit is used if I2C detect */
-	UCHAR ucPwrSensActiveState;	/*  high active or low active */
-	UCHAR ucReserve[3];	/*  reserve */
-	USHORT usSensPwr;	/*  in unit of watt */
-} ATOM_POWER_SOURCE_OBJECT;
-
-typedef struct _ATOM_POWER_SOURCE_INFO {
-	ATOM_COMMON_TABLE_HEADER asHeader;
-	UCHAR asPwrbehave[16];
-	ATOM_POWER_SOURCE_OBJECT asPwrObj[1];
-} ATOM_POWER_SOURCE_INFO;
-
-/* Define ucPwrSrcId */
+typedef struct  _ATOM_ASIC_PROFILING_INFO
+{
+  ATOM_COMMON_TABLE_HEADER			asHeader; 
+	ATOM_ASIC_PROFILE_VOLTAGE			asVoltage;
+}ATOM_ASIC_PROFILING_INFO;
+
+typedef struct _ATOM_POWER_SOURCE_OBJECT
+{
+	UCHAR	ucPwrSrcId;													// Power source
+	UCHAR	ucPwrSensorType;										// GPIO, I2C or none
+	UCHAR	ucPwrSensId;											  // if GPIO detect, it is GPIO id,  if I2C detect, it is I2C id
+	UCHAR	ucPwrSensSlaveAddr;									// Slave address if I2C detect
+	UCHAR ucPwrSensRegIndex;									// I2C register Index if I2C detect
+	UCHAR ucPwrSensRegBitMask;								// detect which bit is used if I2C detect
+	UCHAR	ucPwrSensActiveState;								// high active or low active
+	UCHAR	ucReserve[3];												// reserve		
+	USHORT usSensPwr;													// in unit of watt
+}ATOM_POWER_SOURCE_OBJECT;
+
+typedef struct _ATOM_POWER_SOURCE_INFO
+{
+		ATOM_COMMON_TABLE_HEADER		asHeader;
+		UCHAR												asPwrbehave[16];
+		ATOM_POWER_SOURCE_OBJECT		asPwrObj[1];
+}ATOM_POWER_SOURCE_INFO;
+
+
+//Define ucPwrSrcId
 #define POWERSOURCE_PCIE_ID1						0x00
 #define POWERSOURCE_6PIN_CONNECTOR_ID1	0x01
 #define POWERSOURCE_8PIN_CONNECTOR_ID1	0x02
 #define POWERSOURCE_6PIN_CONNECTOR_ID2	0x04
 #define POWERSOURCE_8PIN_CONNECTOR_ID2	0x08
 
-/* define ucPwrSensorId */
+//define ucPwrSensorId
 #define POWER_SENSOR_ALWAYS							0x00
 #define POWER_SENSOR_GPIO								0x01
 #define POWER_SENSOR_I2C								0x02
 
+typedef struct _ATOM_INTEGRATED_SYSTEM_INFO_V6
+{
+  ATOM_COMMON_TABLE_HEADER   sHeader;
+  ULONG  ulBootUpEngineClock;
+  ULONG  ulDentistVCOFreq;          
+  ULONG  ulBootUpUMAClock;          
+  ULONG  ulReserved1[8];            
+  ULONG  ulBootUpReqDisplayVector;
+  ULONG  ulOtherDisplayMisc;
+  ULONG  ulGPUCapInfo;
+  ULONG  ulReserved2[3];            
+  ULONG  ulSystemConfig;            
+  ULONG  ulCPUCapInfo;              
+  USHORT usMaxNBVoltage;  
+  USHORT usMinNBVoltage;  
+  USHORT usBootUpNBVoltage;         
+  USHORT usExtDispConnInfoOffset;  
+  UCHAR  ucHtcTmpLmt;   
+  UCHAR  ucTjOffset;    
+  UCHAR  ucMemoryType;  
+  UCHAR  ucUMAChannelNumber;
+  ULONG  ulCSR_M3_ARB_CNTL_DEFAULT[10];  
+  ULONG  ulCSR_M3_ARB_CNTL_UVD[10]; 
+  ULONG  ulCSR_M3_ARB_CNTL_FS3D[10];
+  ULONG  ulReserved3[42]; 
+  ATOM_EXTERNAL_DISPLAY_CONNECTION_INFO sExtDispConnInfo;   
+}ATOM_INTEGRATED_SYSTEM_INFO_V6;   
+
+/**********************************************************************************************************************
+// ATOM_INTEGRATED_SYSTEM_INFO_V6 Description
+//ulBootUpEngineClock:              VBIOS bootup Engine clock frequency, in 10kHz unit. 
+//ulDentistVCOFreq:                 Dentist VCO clock in 10kHz unit. 
+//ulBootUpUMAClock:                 System memory boot up clock frequency in 10Khz unit. 
+//ulReserved1[8]                    Reserved by now, must be 0x0. 
+//ulBootUpReqDisplayVector	        VBIOS boot up display IDs
+//                                  ATOM_DEVICE_CRT1_SUPPORT                  0x0001
+//                                  ATOM_DEVICE_CRT2_SUPPORT                  0x0010
+//                                  ATOM_DEVICE_DFP1_SUPPORT                  0x0008 
+//                                  ATOM_DEVICE_DFP6_SUPPORT                  0x0040 
+//                                  ATOM_DEVICE_DFP2_SUPPORT                  0x0080       
+//                                  ATOM_DEVICE_DFP3_SUPPORT                  0x0200       
+//                                  ATOM_DEVICE_DFP4_SUPPORT                  0x0400        
+//                                  ATOM_DEVICE_DFP5_SUPPORT                  0x0800
+//                                  ATOM_DEVICE_LCD1_SUPPORT                  0x0002
+//ulOtherDisplayMisc      	        Other display related flags, not defined yet. 
+//ulGPUCapInfo                      TBD
+//ulReserved2[3]                    must be 0x0 for the reserved.
+//ulSystemConfig                    TBD
+//ulCPUCapInfo                      TBD
+//usMaxNBVoltage                    High NB voltage in unit of mv, calculated using current VDDNB (D24F2xDC) and VDDNB offset fuse. 
+//usMinNBVoltage                    Low NB voltage in unit of mv, calculated using current VDDNB (D24F2xDC) and VDDNB offset fuse.
+//usBootUpNBVoltage                 Boot up NB voltage in unit of mv.
+//ucHtcTmpLmt                       Bit [22:16] of D24F3x64 Thermal Control (HTC) Register.
+//ucTjOffset                        Bit [28:22] of D24F3xE4 Thermtrip Status Register,may not be needed.
+//ucMemoryType                      [3:0]=1:DDR1;=2:DDR2;=3:DDR3.[7:4] is reserved.
+//ucUMAChannelNumber      	        System memory channel numbers. 
+//usExtDispConnectionInfoOffset     ATOM_EXTERNAL_DISPLAY_CONNECTION_INFO offset relative to beginning of this table. 
+//ulCSR_M3_ARB_CNTL_DEFAULT[10]     Arrays with values for CSR M3 arbiter for default
+//ulCSR_M3_ARB_CNTL_UVD[10]         Arrays with values for CSR M3 arbiter for UVD playback.
+//ulCSR_M3_ARB_CNTL_FS3D[10]        Arrays with values for CSR M3 arbiter for Full Screen 3D applications.
+**********************************************************************************************************************/
+
 /**************************************************************************/
-/*  This portion is only used when ext thermal chip or engine/memory clock SS chip is populated on a design */
-/* Memory SS Info Table */
-/* Define Memory Clock SS chip ID */
+// This portion is only used when ext thermal chip or engine/memory clock SS chip is populated on a design
+//Memory SS Info Table
+//Define Memory Clock SS chip ID
 #define ICS91719  1
 #define ICS91720  2
 
-/* Define one structure to inform SW a "block of data" writing to external SS chip via I2C protocol */
-typedef struct _ATOM_I2C_DATA_RECORD {
-	UCHAR ucNunberOfBytes;	/* Indicates how many bytes SW needs to write to the external ASIC for one block, besides to "Start" and "Stop" */
-	UCHAR ucI2CData[1];	/* I2C data in bytes, should be less than 16 bytes usually */
-} ATOM_I2C_DATA_RECORD;
-
-/* Define one structure to inform SW how many blocks of data writing to external SS chip via I2C protocol, in addition to other information */
-typedef struct _ATOM_I2C_DEVICE_SETUP_INFO {
-	ATOM_I2C_ID_CONFIG_ACCESS sucI2cId;	/* I2C line and HW/SW assisted cap. */
-	UCHAR ucSSChipID;	/* SS chip being used */
-	UCHAR ucSSChipSlaveAddr;	/* Slave Address to set up this SS chip */
-	UCHAR ucNumOfI2CDataRecords;	/* number of data block */
-	ATOM_I2C_DATA_RECORD asI2CData[1];
-} ATOM_I2C_DEVICE_SETUP_INFO;
-
-/* ========================================================================================== */
-typedef struct _ATOM_ASIC_MVDD_INFO {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	ATOM_I2C_DEVICE_SETUP_INFO asI2CSetup[1];
-} ATOM_ASIC_MVDD_INFO;
-
-/* ========================================================================================== */
+//Define one structure to inform SW a "block of data" writing to external SS chip via I2C protocol
+typedef struct _ATOM_I2C_DATA_RECORD
+{
+  UCHAR         ucNunberOfBytes;                                              //Indicates how many bytes SW needs to write to the external ASIC for one block, besides to "Start" and "Stop"
+  UCHAR         ucI2CData[1];                                                 //I2C data in bytes, should be less than 16 bytes usually
+}ATOM_I2C_DATA_RECORD;
+
+
+//Define one structure to inform SW how many blocks of data writing to external SS chip via I2C protocol, in addition to other information
+typedef struct _ATOM_I2C_DEVICE_SETUP_INFO
+{
+  ATOM_I2C_ID_CONFIG_ACCESS       sucI2cId;               //I2C line and HW/SW assisted cap.
+  UCHAR		                        ucSSChipID;             //SS chip being used
+  UCHAR		                        ucSSChipSlaveAddr;      //Slave Address to set up this SS chip
+  UCHAR                           ucNumOfI2CDataRecords;  //number of data block
+  ATOM_I2C_DATA_RECORD            asI2CData[1];  
+}ATOM_I2C_DEVICE_SETUP_INFO;
+
+//==========================================================================================
+typedef struct  _ATOM_ASIC_MVDD_INFO
+{
+  ATOM_COMMON_TABLE_HEADER	      sHeader; 
+  ATOM_I2C_DEVICE_SETUP_INFO      asI2CSetup[1];
+}ATOM_ASIC_MVDD_INFO;
+
+//==========================================================================================
 #define ATOM_MCLK_SS_INFO         ATOM_ASIC_MVDD_INFO
 
-/* ========================================================================================== */
+//==========================================================================================
 /**************************************************************************/
 
-typedef struct _ATOM_ASIC_SS_ASSIGNMENT {
-	ULONG ulTargetClockRange;	/* Clock Out frequence (VCO ), in unit of 10Khz */
-	USHORT usSpreadSpectrumPercentage;	/* in unit of 0.01% */
-	USHORT usSpreadRateInKhz;	/* in unit of kHz, modulation freq */
-	UCHAR ucClockIndication;	/* Indicate which clock source needs SS */
-	UCHAR ucSpreadSpectrumMode;	/* Bit1=0 Down Spread,=1 Center Spread. */
-	UCHAR ucReserved[2];
-} ATOM_ASIC_SS_ASSIGNMENT;
-
-/* Define ucSpreadSpectrumType */
+typedef struct _ATOM_ASIC_SS_ASSIGNMENT
+{
+	ULONG								ulTargetClockRange;						//Clock Out frequence (VCO ), in unit of 10Khz
+  USHORT              usSpreadSpectrumPercentage;		//in unit of 0.01%
+	USHORT							usSpreadRateInKhz;						//in unit of kHz, modulation freq
+  UCHAR               ucClockIndication;					  //Indicate which clock source needs SS
+	UCHAR								ucSpreadSpectrumMode;					//Bit1=0 Down Spread,=1 Center Spread.
+	UCHAR								ucReserved[2];
+}ATOM_ASIC_SS_ASSIGNMENT;
+
+//Define ucClockIndication, SW uses the IDs below to search if the SS is requried/enabled on a clock branch/signal type.
+//SS is not required or enabled if a match is not found.
 #define ASIC_INTERNAL_MEMORY_SS			1
 #define ASIC_INTERNAL_ENGINE_SS			2
-#define ASIC_INTERNAL_UVD_SS				3
+#define ASIC_INTERNAL_UVD_SS        3
+#define ASIC_INTERNAL_SS_ON_TMDS    4
+#define ASIC_INTERNAL_SS_ON_HDMI    5
+#define ASIC_INTERNAL_SS_ON_LVDS    6
+#define ASIC_INTERNAL_SS_ON_DP      7
+#define ASIC_INTERNAL_SS_ON_DCPLL   8
+
+typedef struct _ATOM_ASIC_SS_ASSIGNMENT_V2
+{
+	ULONG								ulTargetClockRange;						//For mem/engine/uvd, Clock Out frequence (VCO ), in unit of 10Khz
+                                                    //For TMDS/HDMI/LVDS, it is pixel clock , for DP, it is link clock ( 27000 or 16200 )
+  USHORT              usSpreadSpectrumPercentage;		//in unit of 0.01%
+	USHORT							usSpreadRateIn10Hz;						//in unit of 10Hz, modulation freq
+  UCHAR               ucClockIndication;					  //Indicate which clock source needs SS
+	UCHAR								ucSpreadSpectrumMode;					//Bit0=0 Down Spread,=1 Center Spread, bit1=0: internal SS bit1=1: external SS
+	UCHAR								ucReserved[2];
+}ATOM_ASIC_SS_ASSIGNMENT_V2;
+
+//ucSpreadSpectrumMode
+//#define ATOM_SS_DOWN_SPREAD_MODE_MASK          0x00000000
+//#define ATOM_SS_DOWN_SPREAD_MODE               0x00000000
+//#define ATOM_SS_CENTRE_SPREAD_MODE_MASK        0x00000001
+//#define ATOM_SS_CENTRE_SPREAD_MODE             0x00000001
+//#define ATOM_INTERNAL_SS_MASK                  0x00000000
+//#define ATOM_EXTERNAL_SS_MASK                  0x00000002
+
+typedef struct _ATOM_ASIC_INTERNAL_SS_INFO
+{
+  ATOM_COMMON_TABLE_HEADER	      sHeader; 
+  ATOM_ASIC_SS_ASSIGNMENT		      asSpreadSpectrum[4];
+}ATOM_ASIC_INTERNAL_SS_INFO;
 
-typedef struct _ATOM_ASIC_INTERNAL_SS_INFO {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	ATOM_ASIC_SS_ASSIGNMENT asSpreadSpectrum[4];
-} ATOM_ASIC_INTERNAL_SS_INFO;
+typedef struct _ATOM_ASIC_INTERNAL_SS_INFO_V2
+{
+  ATOM_COMMON_TABLE_HEADER	      sHeader; 
+  ATOM_ASIC_SS_ASSIGNMENT_V2		  asSpreadSpectrum[1];      //this is point only. 
+}ATOM_ASIC_INTERNAL_SS_INFO_V2;
 
-/* ==============================Scratch Pad Definition Portion=============================== */
+typedef struct _ATOM_ASIC_SS_ASSIGNMENT_V3
+{
+	ULONG								ulTargetClockRange;						//For mem/engine/uvd, Clock Out frequence (VCO ), in unit of 10Khz
+                                                    //For TMDS/HDMI/LVDS, it is pixel clock , for DP, it is link clock ( 27000 or 16200 )
+  USHORT              usSpreadSpectrumPercentage;		//in unit of 0.01%
+	USHORT							usSpreadRateIn10Hz;						//in unit of 10Hz, modulation freq
+  UCHAR               ucClockIndication;					  //Indicate which clock source needs SS
+	UCHAR								ucSpreadSpectrumMode;					//Bit0=0 Down Spread,=1 Center Spread, bit1=0: internal SS bit1=1: external SS
+	UCHAR								ucReserved[2];
+}ATOM_ASIC_SS_ASSIGNMENT_V3;
+
+typedef struct _ATOM_ASIC_INTERNAL_SS_INFO_V3
+{
+  ATOM_COMMON_TABLE_HEADER	      sHeader; 
+  ATOM_ASIC_SS_ASSIGNMENT_V3		  asSpreadSpectrum[1];      //this is pointer only. 
+}ATOM_ASIC_INTERNAL_SS_INFO_V3;
+
+
+//==============================Scratch Pad Definition Portion===============================
 #define ATOM_DEVICE_CONNECT_INFO_DEF  0
 #define ATOM_ROM_LOCATION_DEF         1
 #define ATOM_TV_STANDARD_DEF          2
@@ -2995,7 +3852,8 @@ typedef struct _ATOM_ASIC_INTERNAL_SS_INFO {
 #define ATOM_I2C_CHANNEL_STATUS_DEF   8
 #define ATOM_I2C_CHANNEL_STATUS1_DEF  9
 
-/*  BIOS_0_SCRATCH Definition */
+
+// BIOS_0_SCRATCH Definition 
 #define ATOM_S0_CRT1_MONO               0x00000001L
 #define ATOM_S0_CRT1_COLOR              0x00000002L
 #define ATOM_S0_CRT1_MASK               (ATOM_S0_CRT1_MONO+ATOM_S0_CRT1_COLOR)
@@ -3008,6 +3866,7 @@ typedef struct _ATOM_ASIC_INTERNAL_SS_INFO {
 #define ATOM_S0_CV_DIN_A                0x00000020L
 #define ATOM_S0_CV_MASK_A               (ATOM_S0_CV_A+ATOM_S0_CV_DIN_A)
 
+
 #define ATOM_S0_CRT2_MONO               0x00000100L
 #define ATOM_S0_CRT2_COLOR              0x00000200L
 #define ATOM_S0_CRT2_MASK               (ATOM_S0_CRT2_MONO+ATOM_S0_CRT2_COLOR)
@@ -3025,28 +3884,27 @@ typedef struct _ATOM_ASIC_INTERNAL_SS_INFO {
 #define ATOM_S0_DFP2                    0x00020000L
 #define ATOM_S0_LCD1                    0x00040000L
 #define ATOM_S0_LCD2                    0x00080000L
-#define ATOM_S0_TV2                     0x00100000L
-#define ATOM_S0_DFP3			0x00200000L
-#define ATOM_S0_DFP4			0x00400000L
-#define ATOM_S0_DFP5			0x00800000L
+#define ATOM_S0_DFP6                    0x00100000L
+#define ATOM_S0_DFP3                    0x00200000L
+#define ATOM_S0_DFP4                    0x00400000L
+#define ATOM_S0_DFP5                    0x00800000L
 
-#define ATOM_S0_DFP_MASK \
-	(ATOM_S0_DFP1 | ATOM_S0_DFP2 | ATOM_S0_DFP3 | ATOM_S0_DFP4 | ATOM_S0_DFP5)
+#define ATOM_S0_DFP_MASK                ATOM_S0_DFP1 | ATOM_S0_DFP2 | ATOM_S0_DFP3 | ATOM_S0_DFP4 | ATOM_S0_DFP5 | ATOM_S0_DFP6
 
-#define ATOM_S0_FAD_REGISTER_BUG        0x02000000L	/*  If set, indicates we are running a PCIE asic with */
-						    /*  the FAD/HDP reg access bug.  Bit is read by DAL */
+#define ATOM_S0_FAD_REGISTER_BUG        0x02000000L // If set, indicates we are running a PCIE asic with 
+                                                    // the FAD/HDP reg access bug.  Bit is read by DAL, this is obsolete from RV5xx
 
 #define ATOM_S0_THERMAL_STATE_MASK      0x1C000000L
 #define ATOM_S0_THERMAL_STATE_SHIFT     26
 
 #define ATOM_S0_SYSTEM_POWER_STATE_MASK 0xE0000000L
-#define ATOM_S0_SYSTEM_POWER_STATE_SHIFT 29
+#define ATOM_S0_SYSTEM_POWER_STATE_SHIFT 29 
 
 #define ATOM_S0_SYSTEM_POWER_STATE_VALUE_AC     1
 #define ATOM_S0_SYSTEM_POWER_STATE_VALUE_DC     2
 #define ATOM_S0_SYSTEM_POWER_STATE_VALUE_LITEAC 3
 
-/* Byte aligned defintion for BIOS usage */
+//Byte aligned defintion for BIOS usage
 #define ATOM_S0_CRT1_MONOb0             0x01
 #define ATOM_S0_CRT1_COLORb0            0x02
 #define ATOM_S0_CRT1_MASKb0             (ATOM_S0_CRT1_MONOb0+ATOM_S0_CRT1_COLORb0)
@@ -3076,8 +3934,11 @@ typedef struct _ATOM_ASIC_INTERNAL_SS_INFO {
 #define ATOM_S0_DFP2b2                  0x02
 #define ATOM_S0_LCD1b2                  0x04
 #define ATOM_S0_LCD2b2                  0x08
-#define ATOM_S0_TV2b2                   0x10
-#define ATOM_S0_DFP3b2									0x20
+#define ATOM_S0_DFP6b2                  0x10
+#define ATOM_S0_DFP3b2                  0x20
+#define ATOM_S0_DFP4b2                  0x40
+#define ATOM_S0_DFP5b2                  0x80
+
 
 #define ATOM_S0_THERMAL_STATE_MASKb3    0x1C
 #define ATOM_S0_THERMAL_STATE_SHIFTb3   2
@@ -3085,43 +3946,20 @@ typedef struct _ATOM_ASIC_INTERNAL_SS_INFO {
 #define ATOM_S0_SYSTEM_POWER_STATE_MASKb3 0xE0
 #define ATOM_S0_LCD1_SHIFT              18
 
-/*  BIOS_1_SCRATCH Definition */
+// BIOS_1_SCRATCH Definition
 #define ATOM_S1_ROM_LOCATION_MASK       0x0000FFFFL
 #define ATOM_S1_PCI_BUS_DEV_MASK        0xFFFF0000L
 
-/*       BIOS_2_SCRATCH Definition */
+//	BIOS_2_SCRATCH Definition
 #define ATOM_S2_TV1_STANDARD_MASK       0x0000000FL
 #define ATOM_S2_CURRENT_BL_LEVEL_MASK   0x0000FF00L
 #define ATOM_S2_CURRENT_BL_LEVEL_SHIFT  8
 
-#define ATOM_S2_CRT1_DPMS_STATE         0x00010000L
-#define ATOM_S2_LCD1_DPMS_STATE	        0x00020000L
-#define ATOM_S2_TV1_DPMS_STATE          0x00040000L
-#define ATOM_S2_DFP1_DPMS_STATE         0x00080000L
-#define ATOM_S2_CRT2_DPMS_STATE         0x00100000L
-#define ATOM_S2_LCD2_DPMS_STATE         0x00200000L
-#define ATOM_S2_TV2_DPMS_STATE          0x00400000L
-#define ATOM_S2_DFP2_DPMS_STATE         0x00800000L
-#define ATOM_S2_CV_DPMS_STATE           0x01000000L
-#define ATOM_S2_DFP3_DPMS_STATE					0x02000000L
-#define ATOM_S2_DFP4_DPMS_STATE					0x04000000L
-#define ATOM_S2_DFP5_DPMS_STATE					0x08000000L
-
-#define ATOM_S2_DFP_DPM_STATE \
-	(ATOM_S2_DFP1_DPMS_STATE | ATOM_S2_DFP2_DPMS_STATE | \
-	 ATOM_S2_DFP3_DPMS_STATE | ATOM_S2_DFP4_DPMS_STATE | \
-	 ATOM_S2_DFP5_DPMS_STATE)
-
-#define ATOM_S2_DEVICE_DPMS_STATE \
-	(ATOM_S2_CRT1_DPMS_STATE + ATOM_S2_LCD1_DPMS_STATE + \
-	 ATOM_S2_TV1_DPMS_STATE + ATOM_S2_DFP_DPMS_STATE + \
-	 ATOM_S2_CRT2_DPMS_STATE + ATOM_S2_LCD2_DPMS_STATE + \
-	 ATOM_S2_TV2_DPMS_STATE + ATOM_S2_CV_DPMS_STATE)
-
 #define ATOM_S2_FORCEDLOWPWRMODE_STATE_MASK       0x0C000000L
 #define ATOM_S2_FORCEDLOWPWRMODE_STATE_MASK_SHIFT 26
 #define ATOM_S2_FORCEDLOWPWRMODE_STATE_CHANGE     0x10000000L
 
+#define ATOM_S2_DEVICE_DPMS_STATE       0x00010000L
 #define ATOM_S2_VRI_BRIGHT_ENABLE       0x20000000L
 
 #define ATOM_S2_DISPLAY_ROTATION_0_DEGREE     0x0
@@ -3131,21 +3969,11 @@ typedef struct _ATOM_ASIC_INTERNAL_SS_INFO {
 #define ATOM_S2_DISPLAY_ROTATION_DEGREE_SHIFT 30
 #define ATOM_S2_DISPLAY_ROTATION_ANGLE_MASK   0xC0000000L
 
-/* Byte aligned defintion for BIOS usage */
+
+//Byte aligned defintion for BIOS usage
 #define ATOM_S2_TV1_STANDARD_MASKb0     0x0F
 #define ATOM_S2_CURRENT_BL_LEVEL_MASKb1 0xFF
-#define ATOM_S2_CRT1_DPMS_STATEb2       0x01
-#define ATOM_S2_LCD1_DPMS_STATEb2       0x02
-#define ATOM_S2_TV1_DPMS_STATEb2        0x04
-#define ATOM_S2_DFP1_DPMS_STATEb2       0x08
-#define ATOM_S2_CRT2_DPMS_STATEb2       0x10
-#define ATOM_S2_LCD2_DPMS_STATEb2       0x20
-#define ATOM_S2_TV2_DPMS_STATEb2        0x40
-#define ATOM_S2_DFP2_DPMS_STATEb2       0x80
-#define ATOM_S2_CV_DPMS_STATEb3         0x01
-#define ATOM_S2_DFP3_DPMS_STATEb3				0x02
-#define ATOM_S2_DFP4_DPMS_STATEb3				0x04
-#define ATOM_S2_DFP5_DPMS_STATEb3				0x08
+#define ATOM_S2_DEVICE_DPMS_STATEb2     0x01
 
 #define ATOM_S2_DEVICE_DPMS_MASKw1      0x3FF
 #define ATOM_S2_FORCEDLOWPWRMODE_STATE_MASKb3     0x0C
@@ -3153,21 +3981,22 @@ typedef struct _ATOM_ASIC_INTERNAL_SS_INFO {
 #define ATOM_S2_VRI_BRIGHT_ENABLEb3     0x20
 #define ATOM_S2_ROTATION_STATE_MASKb3   0xC0
 
-/*  BIOS_3_SCRATCH Definition */
+
+// BIOS_3_SCRATCH Definition
 #define ATOM_S3_CRT1_ACTIVE             0x00000001L
 #define ATOM_S3_LCD1_ACTIVE             0x00000002L
 #define ATOM_S3_TV1_ACTIVE              0x00000004L
 #define ATOM_S3_DFP1_ACTIVE             0x00000008L
 #define ATOM_S3_CRT2_ACTIVE             0x00000010L
 #define ATOM_S3_LCD2_ACTIVE             0x00000020L
-#define ATOM_S3_TV2_ACTIVE              0x00000040L
+#define ATOM_S3_DFP6_ACTIVE             0x00000040L
 #define ATOM_S3_DFP2_ACTIVE             0x00000080L
 #define ATOM_S3_CV_ACTIVE               0x00000100L
 #define ATOM_S3_DFP3_ACTIVE							0x00000200L
 #define ATOM_S3_DFP4_ACTIVE							0x00000400L
 #define ATOM_S3_DFP5_ACTIVE							0x00000800L
 
-#define ATOM_S3_DEVICE_ACTIVE_MASK      0x000003FFL
+#define ATOM_S3_DEVICE_ACTIVE_MASK      0x00000FFFL
 
 #define ATOM_S3_LCD_FULLEXPANSION_ACTIVE         0x00001000L
 #define ATOM_S3_LCD_EXPANSION_ASPEC_RATIO_ACTIVE 0x00002000L
@@ -3178,7 +4007,7 @@ typedef struct _ATOM_ASIC_INTERNAL_SS_INFO {
 #define ATOM_S3_DFP1_CRTC_ACTIVE        0x00080000L
 #define ATOM_S3_CRT2_CRTC_ACTIVE        0x00100000L
 #define ATOM_S3_LCD2_CRTC_ACTIVE        0x00200000L
-#define ATOM_S3_TV2_CRTC_ACTIVE         0x00400000L
+#define ATOM_S3_DFP6_CRTC_ACTIVE        0x00400000L
 #define ATOM_S3_DFP2_CRTC_ACTIVE        0x00800000L
 #define ATOM_S3_CV_CRTC_ACTIVE          0x01000000L
 #define ATOM_S3_DFP3_CRTC_ACTIVE				0x02000000L
@@ -3187,17 +4016,18 @@ typedef struct _ATOM_ASIC_INTERNAL_SS_INFO {
 
 #define ATOM_S3_DEVICE_CRTC_ACTIVE_MASK 0x0FFF0000L
 #define ATOM_S3_ASIC_GUI_ENGINE_HUNG    0x20000000L
+//Below two definitions are not supported in pplib, but in the old powerplay in DAL
 #define ATOM_S3_ALLOW_FAST_PWR_SWITCH   0x40000000L
 #define ATOM_S3_RQST_GPU_USE_MIN_PWR    0x80000000L
 
-/* Byte aligned defintion for BIOS usage */
+//Byte aligned defintion for BIOS usage
 #define ATOM_S3_CRT1_ACTIVEb0           0x01
 #define ATOM_S3_LCD1_ACTIVEb0           0x02
 #define ATOM_S3_TV1_ACTIVEb0            0x04
 #define ATOM_S3_DFP1_ACTIVEb0           0x08
 #define ATOM_S3_CRT2_ACTIVEb0           0x10
 #define ATOM_S3_LCD2_ACTIVEb0           0x20
-#define ATOM_S3_TV2_ACTIVEb0            0x40
+#define ATOM_S3_DFP6_ACTIVEb0           0x40
 #define ATOM_S3_DFP2_ACTIVEb0           0x80
 #define ATOM_S3_CV_ACTIVEb1             0x01
 #define ATOM_S3_DFP3_ACTIVEb1						0x02
@@ -3212,7 +4042,7 @@ typedef struct _ATOM_ASIC_INTERNAL_SS_INFO {
 #define ATOM_S3_DFP1_CRTC_ACTIVEb2      0x08
 #define ATOM_S3_CRT2_CRTC_ACTIVEb2      0x10
 #define ATOM_S3_LCD2_CRTC_ACTIVEb2      0x20
-#define ATOM_S3_TV2_CRTC_ACTIVEb2       0x40
+#define ATOM_S3_DFP6_CRTC_ACTIVEb2      0x40
 #define ATOM_S3_DFP2_CRTC_ACTIVEb2      0x80
 #define ATOM_S3_CV_CRTC_ACTIVEb3        0x01
 #define ATOM_S3_DFP3_CRTC_ACTIVEb3			0x02
@@ -3221,35 +4051,31 @@ typedef struct _ATOM_ASIC_INTERNAL_SS_INFO {
 
 #define ATOM_S3_ACTIVE_CRTC2w1          0xFFF
 
-#define ATOM_S3_ASIC_GUI_ENGINE_HUNGb3	0x20
-#define ATOM_S3_ALLOW_FAST_PWR_SWITCHb3 0x40
-#define ATOM_S3_RQST_GPU_USE_MIN_PWRb3  0x80
-
-/*  BIOS_4_SCRATCH Definition */
+// BIOS_4_SCRATCH Definition
 #define ATOM_S4_LCD1_PANEL_ID_MASK      0x000000FFL
 #define ATOM_S4_LCD1_REFRESH_MASK       0x0000FF00L
 #define ATOM_S4_LCD1_REFRESH_SHIFT      8
 
-/* Byte aligned defintion for BIOS usage */
+//Byte aligned defintion for BIOS usage
 #define ATOM_S4_LCD1_PANEL_ID_MASKb0	  0x0FF
 #define ATOM_S4_LCD1_REFRESH_MASKb1		  ATOM_S4_LCD1_PANEL_ID_MASKb0
 #define ATOM_S4_VRAM_INFO_MASKb2        ATOM_S4_LCD1_PANEL_ID_MASKb0
 
-/*  BIOS_5_SCRATCH Definition, BIOS_5_SCRATCH is used by Firmware only !!!! */
+// BIOS_5_SCRATCH Definition, BIOS_5_SCRATCH is used by Firmware only !!!!
 #define ATOM_S5_DOS_REQ_CRT1b0          0x01
 #define ATOM_S5_DOS_REQ_LCD1b0          0x02
 #define ATOM_S5_DOS_REQ_TV1b0           0x04
 #define ATOM_S5_DOS_REQ_DFP1b0          0x08
 #define ATOM_S5_DOS_REQ_CRT2b0          0x10
 #define ATOM_S5_DOS_REQ_LCD2b0          0x20
-#define ATOM_S5_DOS_REQ_TV2b0           0x40
+#define ATOM_S5_DOS_REQ_DFP6b0          0x40
 #define ATOM_S5_DOS_REQ_DFP2b0          0x80
 #define ATOM_S5_DOS_REQ_CVb1            0x01
 #define ATOM_S5_DOS_REQ_DFP3b1					0x02
 #define ATOM_S5_DOS_REQ_DFP4b1					0x04
 #define ATOM_S5_DOS_REQ_DFP5b1					0x08
 
-#define ATOM_S5_DOS_REQ_DEVICEw0        0x03FF
+#define ATOM_S5_DOS_REQ_DEVICEw0        0x0FFF
 
 #define ATOM_S5_DOS_REQ_CRT1            0x0001
 #define ATOM_S5_DOS_REQ_LCD1            0x0002
@@ -3257,22 +4083,21 @@ typedef struct _ATOM_ASIC_INTERNAL_SS_INFO {
 #define ATOM_S5_DOS_REQ_DFP1            0x0008
 #define ATOM_S5_DOS_REQ_CRT2            0x0010
 #define ATOM_S5_DOS_REQ_LCD2            0x0020
-#define ATOM_S5_DOS_REQ_TV2             0x0040
+#define ATOM_S5_DOS_REQ_DFP6            0x0040
 #define ATOM_S5_DOS_REQ_DFP2            0x0080
 #define ATOM_S5_DOS_REQ_CV              0x0100
-#define ATOM_S5_DOS_REQ_DFP3						0x0200
-#define ATOM_S5_DOS_REQ_DFP4						0x0400
-#define ATOM_S5_DOS_REQ_DFP5						0x0800
+#define ATOM_S5_DOS_REQ_DFP3            0x0200
+#define ATOM_S5_DOS_REQ_DFP4            0x0400
+#define ATOM_S5_DOS_REQ_DFP5            0x0800
 
 #define ATOM_S5_DOS_FORCE_CRT1b2        ATOM_S5_DOS_REQ_CRT1b0
 #define ATOM_S5_DOS_FORCE_TV1b2         ATOM_S5_DOS_REQ_TV1b0
 #define ATOM_S5_DOS_FORCE_CRT2b2        ATOM_S5_DOS_REQ_CRT2b0
 #define ATOM_S5_DOS_FORCE_CVb3          ATOM_S5_DOS_REQ_CVb1
-#define ATOM_S5_DOS_FORCE_DEVICEw1 \
-	(ATOM_S5_DOS_FORCE_CRT1b2 + ATOM_S5_DOS_FORCE_TV1b2 + \
-	 ATOM_S5_DOS_FORCE_CRT2b2 + (ATOM_S5_DOS_FORCE_CVb3 << 8))
+#define ATOM_S5_DOS_FORCE_DEVICEw1      (ATOM_S5_DOS_FORCE_CRT1b2+ATOM_S5_DOS_FORCE_TV1b2+ATOM_S5_DOS_FORCE_CRT2b2+\
+                                        (ATOM_S5_DOS_FORCE_CVb3<<8))
 
-/*  BIOS_6_SCRATCH Definition */
+// BIOS_6_SCRATCH Definition
 #define ATOM_S6_DEVICE_CHANGE           0x00000001L
 #define ATOM_S6_SCALER_CHANGE           0x00000002L
 #define ATOM_S6_LID_CHANGE              0x00000004L
@@ -3285,11 +4110,11 @@ typedef struct _ATOM_ASIC_INTERNAL_SS_INFO {
 #define ATOM_S6_HW_I2C_BUSY_STATE       0x00000200L
 #define ATOM_S6_THERMAL_STATE_CHANGE    0x00000400L
 #define ATOM_S6_INTERRUPT_SET_BY_BIOS   0x00000800L
-#define ATOM_S6_REQ_LCD_EXPANSION_FULL         0x00001000L	/* Normal expansion Request bit for LCD */
-#define ATOM_S6_REQ_LCD_EXPANSION_ASPEC_RATIO  0x00002000L	/* Aspect ratio expansion Request bit for LCD */
+#define ATOM_S6_REQ_LCD_EXPANSION_FULL         0x00001000L //Normal expansion Request bit for LCD
+#define ATOM_S6_REQ_LCD_EXPANSION_ASPEC_RATIO  0x00002000L //Aspect ratio expansion Request bit for LCD
 
-#define ATOM_S6_DISPLAY_STATE_CHANGE    0x00004000L	/* This bit is recycled when ATOM_BIOS_INFO_BIOS_SCRATCH6_SCL2_REDEFINE is set,previously it's SCL2_H_expansion */
-#define ATOM_S6_I2C_STATE_CHANGE        0x00008000L	/* This bit is recycled,when ATOM_BIOS_INFO_BIOS_SCRATCH6_SCL2_REDEFINE is set,previously it's SCL2_V_expansion */
+#define ATOM_S6_DISPLAY_STATE_CHANGE    0x00004000L        //This bit is recycled when ATOM_BIOS_INFO_BIOS_SCRATCH6_SCL2_REDEFINE is set,previously it's SCL2_H_expansion
+#define ATOM_S6_I2C_STATE_CHANGE        0x00008000L        //This bit is recycled,when ATOM_BIOS_INFO_BIOS_SCRATCH6_SCL2_REDEFINE is set,previously it's SCL2_V_expansion
 
 #define ATOM_S6_ACC_REQ_CRT1            0x00010000L
 #define ATOM_S6_ACC_REQ_LCD1            0x00020000L
@@ -3297,7 +4122,7 @@ typedef struct _ATOM_ASIC_INTERNAL_SS_INFO {
 #define ATOM_S6_ACC_REQ_DFP1            0x00080000L
 #define ATOM_S6_ACC_REQ_CRT2            0x00100000L
 #define ATOM_S6_ACC_REQ_LCD2            0x00200000L
-#define ATOM_S6_ACC_REQ_TV2             0x00400000L
+#define ATOM_S6_ACC_REQ_DFP6            0x00400000L
 #define ATOM_S6_ACC_REQ_DFP2            0x00800000L
 #define ATOM_S6_ACC_REQ_CV              0x01000000L
 #define ATOM_S6_ACC_REQ_DFP3						0x02000000L
@@ -3310,7 +4135,7 @@ typedef struct _ATOM_ASIC_INTERNAL_SS_INFO {
 #define ATOM_S6_VRI_BRIGHTNESS_CHANGE       0x40000000L
 #define ATOM_S6_CONFIG_DISPLAY_CHANGE_MASK  0x80000000L
 
-/* Byte aligned defintion for BIOS usage */
+//Byte aligned defintion for BIOS usage
 #define ATOM_S6_DEVICE_CHANGEb0         0x01
 #define ATOM_S6_SCALER_CHANGEb0         0x02
 #define ATOM_S6_LID_CHANGEb0            0x04
@@ -3320,11 +4145,11 @@ typedef struct _ATOM_ASIC_INTERNAL_SS_INFO {
 #define ATOM_S6_LID_STATEb0             0x40
 #define ATOM_S6_DOCK_STATEb0            0x80
 #define ATOM_S6_CRITICAL_STATEb1        0x01
-#define ATOM_S6_HW_I2C_BUSY_STATEb1     0x02
+#define ATOM_S6_HW_I2C_BUSY_STATEb1     0x02  
 #define ATOM_S6_THERMAL_STATE_CHANGEb1  0x04
 #define ATOM_S6_INTERRUPT_SET_BY_BIOSb1 0x08
-#define ATOM_S6_REQ_LCD_EXPANSION_FULLb1        0x10
-#define ATOM_S6_REQ_LCD_EXPANSION_ASPEC_RATIOb1 0x20
+#define ATOM_S6_REQ_LCD_EXPANSION_FULLb1        0x10    
+#define ATOM_S6_REQ_LCD_EXPANSION_ASPEC_RATIOb1 0x20 
 
 #define ATOM_S6_ACC_REQ_CRT1b2          0x01
 #define ATOM_S6_ACC_REQ_LCD1b2          0x02
@@ -3332,12 +4157,12 @@ typedef struct _ATOM_ASIC_INTERNAL_SS_INFO {
 #define ATOM_S6_ACC_REQ_DFP1b2          0x08
 #define ATOM_S6_ACC_REQ_CRT2b2          0x10
 #define ATOM_S6_ACC_REQ_LCD2b2          0x20
-#define ATOM_S6_ACC_REQ_TV2b2           0x40
+#define ATOM_S6_ACC_REQ_DFP6b2          0x40
 #define ATOM_S6_ACC_REQ_DFP2b2          0x80
 #define ATOM_S6_ACC_REQ_CVb3            0x01
-#define ATOM_S6_ACC_REQ_DFP3b3					0x02
-#define ATOM_S6_ACC_REQ_DFP4b3					0x04
-#define ATOM_S6_ACC_REQ_DFP5b3					0x08
+#define ATOM_S6_ACC_REQ_DFP3b3          0x02
+#define ATOM_S6_ACC_REQ_DFP4b3          0x04
+#define ATOM_S6_ACC_REQ_DFP5b3          0x08
 
 #define ATOM_S6_ACC_REQ_DEVICEw1        ATOM_S5_DOS_REQ_DEVICEw0
 #define ATOM_S6_SYSTEM_POWER_MODE_CHANGEb3 0x10
@@ -3366,7 +4191,7 @@ typedef struct _ATOM_ASIC_INTERNAL_SS_INFO {
 #define ATOM_S6_VRI_BRIGHTNESS_CHANGE_SHIFT     30
 #define ATOM_S6_CONFIG_DISPLAY_CHANGE_SHIFT     31
 
-/*  BIOS_7_SCRATCH Definition, BIOS_7_SCRATCH is used by Firmware only !!!! */
+// BIOS_7_SCRATCH Definition, BIOS_7_SCRATCH is used by Firmware only !!!!
 #define ATOM_S7_DOS_MODE_TYPEb0             0x03
 #define ATOM_S7_DOS_MODE_VGAb0              0x00
 #define ATOM_S7_DOS_MODE_VESAb0             0x01
@@ -3378,220 +4203,194 @@ typedef struct _ATOM_ASIC_INTERNAL_SS_INFO {
 
 #define ATOM_S7_DOS_8BIT_DAC_EN_SHIFT       8
 
-/*  BIOS_8_SCRATCH Definition */
+// BIOS_8_SCRATCH Definition
 #define ATOM_S8_I2C_CHANNEL_BUSY_MASK       0x00000FFFF
-#define ATOM_S8_I2C_HW_ENGINE_BUSY_MASK     0x0FFFF0000
+#define ATOM_S8_I2C_HW_ENGINE_BUSY_MASK     0x0FFFF0000   
 
 #define ATOM_S8_I2C_CHANNEL_BUSY_SHIFT      0
 #define ATOM_S8_I2C_ENGINE_BUSY_SHIFT       16
 
-/*  BIOS_9_SCRATCH Definition */
-#ifndef ATOM_S9_I2C_CHANNEL_COMPLETED_MASK
+// BIOS_9_SCRATCH Definition
+#ifndef ATOM_S9_I2C_CHANNEL_COMPLETED_MASK 
 #define ATOM_S9_I2C_CHANNEL_COMPLETED_MASK  0x0000FFFF
 #endif
-#ifndef ATOM_S9_I2C_CHANNEL_ABORTED_MASK
+#ifndef ATOM_S9_I2C_CHANNEL_ABORTED_MASK  
 #define ATOM_S9_I2C_CHANNEL_ABORTED_MASK    0xFFFF0000
 #endif
-#ifndef ATOM_S9_I2C_CHANNEL_COMPLETED_SHIFT
+#ifndef ATOM_S9_I2C_CHANNEL_COMPLETED_SHIFT 
 #define ATOM_S9_I2C_CHANNEL_COMPLETED_SHIFT 0
 #endif
-#ifndef ATOM_S9_I2C_CHANNEL_ABORTED_SHIFT
+#ifndef ATOM_S9_I2C_CHANNEL_ABORTED_SHIFT   
 #define ATOM_S9_I2C_CHANNEL_ABORTED_SHIFT   16
 #endif
 
+ 
 #define ATOM_FLAG_SET                         0x20
 #define ATOM_FLAG_CLEAR                       0
-#define CLEAR_ATOM_S6_ACC_MODE \
-	((ATOM_ACC_CHANGE_INFO_DEF << 8) | \
-	 ATOM_S6_ACC_MODE_SHIFT | ATOM_FLAG_CLEAR)
-#define SET_ATOM_S6_DEVICE_CHANGE \
-	((ATOM_ACC_CHANGE_INFO_DEF << 8) | \
-	 ATOM_S6_DEVICE_CHANGE_SHIFT | ATOM_FLAG_SET)
-#define SET_ATOM_S6_VRI_BRIGHTNESS_CHANGE \
-	((ATOM_ACC_CHANGE_INFO_DEF << 8) | \
-	 ATOM_S6_VRI_BRIGHTNESS_CHANGE_SHIFT | ATOM_FLAG_SET)
-#define SET_ATOM_S6_SCALER_CHANGE \
-	((ATOM_ACC_CHANGE_INFO_DEF << 8) | \
-	 ATOM_S6_SCALER_CHANGE_SHIFT | ATOM_FLAG_SET)
-#define SET_ATOM_S6_LID_CHANGE \
-	((ATOM_ACC_CHANGE_INFO_DEF << 8) | \
-	 ATOM_S6_LID_CHANGE_SHIFT | ATOM_FLAG_SET)
-
-#define SET_ATOM_S6_LID_STATE \
-	((ATOM_ACC_CHANGE_INFO_DEF << 8) |\
-	 ATOM_S6_LID_STATE_SHIFT | ATOM_FLAG_SET)
-#define CLEAR_ATOM_S6_LID_STATE \
-	((ATOM_ACC_CHANGE_INFO_DEF << 8) | \
-	 ATOM_S6_LID_STATE_SHIFT | ATOM_FLAG_CLEAR)
-
-#define SET_ATOM_S6_DOCK_CHANGE \
-	((ATOM_ACC_CHANGE_INFO_DEF << 8)| \
-	 ATOM_S6_DOCKING_CHANGE_SHIFT | ATOM_FLAG_SET)
-#define SET_ATOM_S6_DOCK_STATE \
-	((ATOM_ACC_CHANGE_INFO_DEF << 8) | \
-	 ATOM_S6_DOCK_STATE_SHIFT | ATOM_FLAG_SET)
-#define CLEAR_ATOM_S6_DOCK_STATE \
-	((ATOM_ACC_CHANGE_INFO_DEF << 8) | \
-	 ATOM_S6_DOCK_STATE_SHIFT | ATOM_FLAG_CLEAR)
-
-#define SET_ATOM_S6_THERMAL_STATE_CHANGE \
-	((ATOM_ACC_CHANGE_INFO_DEF << 8) | \
-	 ATOM_S6_THERMAL_STATE_CHANGE_SHIFT | ATOM_FLAG_SET)
-#define SET_ATOM_S6_SYSTEM_POWER_MODE_CHANGE \
-	((ATOM_ACC_CHANGE_INFO_DEF << 8) | \
-	 ATOM_S6_SYSTEM_POWER_MODE_CHANGE_SHIFT | ATOM_FLAG_SET)
-#define SET_ATOM_S6_INTERRUPT_SET_BY_BIOS \
-	((ATOM_ACC_CHANGE_INFO_DEF << 8) | \
-	 ATOM_S6_INTERRUPT_SET_BY_BIOS_SHIFT | ATOM_FLAG_SET)
-
-#define SET_ATOM_S6_CRITICAL_STATE \
-	((ATOM_ACC_CHANGE_INFO_DEF << 8) | \
-	 ATOM_S6_CRITICAL_STATE_SHIFT | ATOM_FLAG_SET)
-#define CLEAR_ATOM_S6_CRITICAL_STATE \
-	((ATOM_ACC_CHANGE_INFO_DEF << 8) | \
-	 ATOM_S6_CRITICAL_STATE_SHIFT | ATOM_FLAG_CLEAR)
-
-#define SET_ATOM_S6_REQ_SCALER \
-	((ATOM_ACC_CHANGE_INFO_DEF << 8) | \
-	 ATOM_S6_REQ_SCALER_SHIFT | ATOM_FLAG_SET)
-#define CLEAR_ATOM_S6_REQ_SCALER \
-	((ATOM_ACC_CHANGE_INFO_DEF << 8) | \
-	 ATOM_S6_REQ_SCALER_SHIFT | ATOM_FLAG_CLEAR )
-
-#define SET_ATOM_S6_REQ_SCALER_ARATIO \
-	((ATOM_ACC_CHANGE_INFO_DEF << 8) | \
-	 ATOM_S6_REQ_SCALER_ARATIO_SHIFT | ATOM_FLAG_SET )
-#define CLEAR_ATOM_S6_REQ_SCALER_ARATIO \
-	((ATOM_ACC_CHANGE_INFO_DEF << 8) | \
-	 ATOM_S6_REQ_SCALER_ARATIO_SHIFT | ATOM_FLAG_CLEAR )
-
-#define SET_ATOM_S6_I2C_STATE_CHANGE \
-	((ATOM_ACC_CHANGE_INFO_DEF << 8) | \
-	 ATOM_S6_I2C_STATE_CHANGE_SHIFT | ATOM_FLAG_SET )
-
-#define SET_ATOM_S6_DISPLAY_STATE_CHANGE \
-	((ATOM_ACC_CHANGE_INFO_DEF << 8) | \
-	 ATOM_S6_DISPLAY_STATE_CHANGE_SHIFT | ATOM_FLAG_SET )
-
-#define SET_ATOM_S6_DEVICE_RECONFIG \
-	((ATOM_ACC_CHANGE_INFO_DEF << 8) | \
-	 ATOM_S6_CONFIG_DISPLAY_CHANGE_SHIFT | ATOM_FLAG_SET)
-#define CLEAR_ATOM_S0_LCD1 \
-	((ATOM_DEVICE_CONNECT_INFO_DEF << 8 ) | \
-	 ATOM_S0_LCD1_SHIFT | ATOM_FLAG_CLEAR )
-#define SET_ATOM_S7_DOS_8BIT_DAC_EN \
-	((ATOM_DOS_MODE_INFO_DEF << 8) | \
-	 ATOM_S7_DOS_8BIT_DAC_EN_SHIFT | ATOM_FLAG_SET )
-#define CLEAR_ATOM_S7_DOS_8BIT_DAC_EN \
-	((ATOM_DOS_MODE_INFO_DEF << 8) | \
-	 ATOM_S7_DOS_8BIT_DAC_EN_SHIFT | ATOM_FLAG_CLEAR )
+#define CLEAR_ATOM_S6_ACC_MODE                ((ATOM_ACC_CHANGE_INFO_DEF << 8 )|ATOM_S6_ACC_MODE_SHIFT | ATOM_FLAG_CLEAR)
+#define SET_ATOM_S6_DEVICE_CHANGE             ((ATOM_ACC_CHANGE_INFO_DEF << 8 )|ATOM_S6_DEVICE_CHANGE_SHIFT | ATOM_FLAG_SET)
+#define SET_ATOM_S6_VRI_BRIGHTNESS_CHANGE     ((ATOM_ACC_CHANGE_INFO_DEF << 8 )|ATOM_S6_VRI_BRIGHTNESS_CHANGE_SHIFT | ATOM_FLAG_SET)
+#define SET_ATOM_S6_SCALER_CHANGE             ((ATOM_ACC_CHANGE_INFO_DEF << 8 )|ATOM_S6_SCALER_CHANGE_SHIFT | ATOM_FLAG_SET)
+#define SET_ATOM_S6_LID_CHANGE                ((ATOM_ACC_CHANGE_INFO_DEF << 8 )|ATOM_S6_LID_CHANGE_SHIFT | ATOM_FLAG_SET)
 
-/****************************************************************************/
-/* Portion II: Definitinos only used in Driver */
+#define SET_ATOM_S6_LID_STATE                 ((ATOM_ACC_CHANGE_INFO_DEF << 8 )|ATOM_S6_LID_STATE_SHIFT | ATOM_FLAG_SET)
+#define CLEAR_ATOM_S6_LID_STATE               ((ATOM_ACC_CHANGE_INFO_DEF << 8 )|ATOM_S6_LID_STATE_SHIFT | ATOM_FLAG_CLEAR)
+
+#define SET_ATOM_S6_DOCK_CHANGE			          ((ATOM_ACC_CHANGE_INFO_DEF << 8 )|ATOM_S6_DOCKING_CHANGE_SHIFT | ATOM_FLAG_SET)
+#define SET_ATOM_S6_DOCK_STATE                ((ATOM_ACC_CHANGE_INFO_DEF << 8 )|ATOM_S6_DOCK_STATE_SHIFT | ATOM_FLAG_SET)
+#define CLEAR_ATOM_S6_DOCK_STATE              ((ATOM_ACC_CHANGE_INFO_DEF << 8 )|ATOM_S6_DOCK_STATE_SHIFT | ATOM_FLAG_CLEAR)
+
+#define SET_ATOM_S6_THERMAL_STATE_CHANGE      ((ATOM_ACC_CHANGE_INFO_DEF << 8 )|ATOM_S6_THERMAL_STATE_CHANGE_SHIFT | ATOM_FLAG_SET)
+#define SET_ATOM_S6_SYSTEM_POWER_MODE_CHANGE  ((ATOM_ACC_CHANGE_INFO_DEF << 8 )|ATOM_S6_SYSTEM_POWER_MODE_CHANGE_SHIFT | ATOM_FLAG_SET)
+#define SET_ATOM_S6_INTERRUPT_SET_BY_BIOS     ((ATOM_ACC_CHANGE_INFO_DEF << 8 )|ATOM_S6_INTERRUPT_SET_BY_BIOS_SHIFT | ATOM_FLAG_SET)
+
+#define SET_ATOM_S6_CRITICAL_STATE            ((ATOM_ACC_CHANGE_INFO_DEF << 8 )|ATOM_S6_CRITICAL_STATE_SHIFT | ATOM_FLAG_SET)
+#define CLEAR_ATOM_S6_CRITICAL_STATE          ((ATOM_ACC_CHANGE_INFO_DEF << 8 )|ATOM_S6_CRITICAL_STATE_SHIFT | ATOM_FLAG_CLEAR)
+
+#define SET_ATOM_S6_REQ_SCALER                ((ATOM_ACC_CHANGE_INFO_DEF << 8 )|ATOM_S6_REQ_SCALER_SHIFT | ATOM_FLAG_SET)  
+#define CLEAR_ATOM_S6_REQ_SCALER              ((ATOM_ACC_CHANGE_INFO_DEF << 8 )|ATOM_S6_REQ_SCALER_SHIFT | ATOM_FLAG_CLEAR )
+
+#define SET_ATOM_S6_REQ_SCALER_ARATIO         ((ATOM_ACC_CHANGE_INFO_DEF << 8 )|ATOM_S6_REQ_SCALER_ARATIO_SHIFT | ATOM_FLAG_SET )
+#define CLEAR_ATOM_S6_REQ_SCALER_ARATIO       ((ATOM_ACC_CHANGE_INFO_DEF << 8 )|ATOM_S6_REQ_SCALER_ARATIO_SHIFT | ATOM_FLAG_CLEAR )
+
+#define SET_ATOM_S6_I2C_STATE_CHANGE          ((ATOM_ACC_CHANGE_INFO_DEF << 8 )|ATOM_S6_I2C_STATE_CHANGE_SHIFT | ATOM_FLAG_SET )
+
+#define SET_ATOM_S6_DISPLAY_STATE_CHANGE      ((ATOM_ACC_CHANGE_INFO_DEF << 8 )|ATOM_S6_DISPLAY_STATE_CHANGE_SHIFT | ATOM_FLAG_SET )
+
+#define SET_ATOM_S6_DEVICE_RECONFIG           ((ATOM_ACC_CHANGE_INFO_DEF << 8 )|ATOM_S6_CONFIG_DISPLAY_CHANGE_SHIFT | ATOM_FLAG_SET)
+#define CLEAR_ATOM_S0_LCD1                    ((ATOM_DEVICE_CONNECT_INFO_DEF << 8 )|  ATOM_S0_LCD1_SHIFT | ATOM_FLAG_CLEAR )
+#define SET_ATOM_S7_DOS_8BIT_DAC_EN           ((ATOM_DOS_MODE_INFO_DEF << 8 )|ATOM_S7_DOS_8BIT_DAC_EN_SHIFT | ATOM_FLAG_SET )
+#define CLEAR_ATOM_S7_DOS_8BIT_DAC_EN         ((ATOM_DOS_MODE_INFO_DEF << 8 )|ATOM_S7_DOS_8BIT_DAC_EN_SHIFT | ATOM_FLAG_CLEAR )
+
+/****************************************************************************/	
+//Portion II: Definitinos only used in Driver
 /****************************************************************************/
 
-/*  Macros used by driver */
+// Macros used by driver
+#ifdef __cplusplus
+#define GetIndexIntoMasterTable(MasterOrData, FieldName) ((reinterpret_cast<char*>(&(static_cast<ATOM_MASTER_LIST_OF_##MasterOrData##_TABLES*>(0))->FieldName)-static_cast<char*>(0))/sizeof(USHORT))
 
-#define	GetIndexIntoMasterTable(MasterOrData, FieldName) (((char *)(&((ATOM_MASTER_LIST_OF_##MasterOrData##_TABLES *)0)->FieldName)-(char *)0)/sizeof(USHORT))
+#define GET_COMMAND_TABLE_COMMANDSET_REVISION(TABLE_HEADER_OFFSET) (((static_cast<ATOM_COMMON_TABLE_HEADER*>(TABLE_HEADER_OFFSET))->ucTableFormatRevision )&0x3F)
+#define GET_COMMAND_TABLE_PARAMETER_REVISION(TABLE_HEADER_OFFSET)  (((static_cast<ATOM_COMMON_TABLE_HEADER*>(TABLE_HEADER_OFFSET))->ucTableContentRevision)&0x3F)
+#else // not __cplusplus
+#define	GetIndexIntoMasterTable(MasterOrData, FieldName) (((char*)(&((ATOM_MASTER_LIST_OF_##MasterOrData##_TABLES*)0)->FieldName)-(char*)0)/sizeof(USHORT))
 
 #define GET_COMMAND_TABLE_COMMANDSET_REVISION(TABLE_HEADER_OFFSET) ((((ATOM_COMMON_TABLE_HEADER*)TABLE_HEADER_OFFSET)->ucTableFormatRevision)&0x3F)
 #define GET_COMMAND_TABLE_PARAMETER_REVISION(TABLE_HEADER_OFFSET)  ((((ATOM_COMMON_TABLE_HEADER*)TABLE_HEADER_OFFSET)->ucTableContentRevision)&0x3F)
+#endif // __cplusplus
 
 #define GET_DATA_TABLE_MAJOR_REVISION GET_COMMAND_TABLE_COMMANDSET_REVISION
 #define GET_DATA_TABLE_MINOR_REVISION GET_COMMAND_TABLE_PARAMETER_REVISION
 
-/****************************************************************************/
-/* Portion III: Definitinos only used in VBIOS */
+/****************************************************************************/	
+//Portion III: Definitinos only used in VBIOS
 /****************************************************************************/
 #define ATOM_DAC_SRC					0x80
 #define ATOM_SRC_DAC1					0
 #define ATOM_SRC_DAC2					0x80
 
-#ifdef	UEFI_BUILD
-#define	USHORT	UTEMP
-#endif
-
-typedef struct _MEMORY_PLLINIT_PARAMETERS {
-	ULONG ulTargetMemoryClock;	/* In 10Khz unit */
-	UCHAR ucAction;		/* not define yet */
-	UCHAR ucFbDiv_Hi;	/* Fbdiv Hi byte */
-	UCHAR ucFbDiv;		/* FB value */
-	UCHAR ucPostDiv;	/* Post div */
-} MEMORY_PLLINIT_PARAMETERS;
+typedef struct _MEMORY_PLLINIT_PARAMETERS
+{
+  ULONG ulTargetMemoryClock; //In 10Khz unit
+  UCHAR   ucAction;					 //not define yet
+  UCHAR   ucFbDiv_Hi;				 //Fbdiv Hi byte
+  UCHAR   ucFbDiv;					 //FB value
+  UCHAR   ucPostDiv;				 //Post div
+}MEMORY_PLLINIT_PARAMETERS;
 
 #define MEMORY_PLLINIT_PS_ALLOCATION  MEMORY_PLLINIT_PARAMETERS
 
-#define	GPIO_PIN_WRITE													0x01
+
+#define	GPIO_PIN_WRITE													0x01			
 #define	GPIO_PIN_READ														0x00
 
-typedef struct _GPIO_PIN_CONTROL_PARAMETERS {
-	UCHAR ucGPIO_ID;	/* return value, read from GPIO pins */
-	UCHAR ucGPIOBitShift;	/* define which bit in uGPIOBitVal need to be update */
-	UCHAR ucGPIOBitVal;	/* Set/Reset corresponding bit defined in ucGPIOBitMask */
-	UCHAR ucAction;		/* =GPIO_PIN_WRITE: Read; =GPIO_PIN_READ: Write */
-} GPIO_PIN_CONTROL_PARAMETERS;
-
-typedef struct _ENABLE_SCALER_PARAMETERS {
-	UCHAR ucScaler;		/*  ATOM_SCALER1, ATOM_SCALER2 */
-	UCHAR ucEnable;		/*  ATOM_SCALER_DISABLE or ATOM_SCALER_CENTER or ATOM_SCALER_EXPANSION */
-	UCHAR ucTVStandard;	/*  */
-	UCHAR ucPadding[1];
-} ENABLE_SCALER_PARAMETERS;
-#define ENABLE_SCALER_PS_ALLOCATION ENABLE_SCALER_PARAMETERS
-
-/* ucEnable: */
+typedef struct  _GPIO_PIN_CONTROL_PARAMETERS
+{
+  UCHAR ucGPIO_ID;           //return value, read from GPIO pins
+  UCHAR ucGPIOBitShift;	     //define which bit in uGPIOBitVal need to be update 
+	UCHAR ucGPIOBitVal;		     //Set/Reset corresponding bit defined in ucGPIOBitMask
+  UCHAR ucAction;				     //=GPIO_PIN_WRITE: Read; =GPIO_PIN_READ: Write
+}GPIO_PIN_CONTROL_PARAMETERS;
+
+typedef struct _ENABLE_SCALER_PARAMETERS
+{
+  UCHAR ucScaler;            // ATOM_SCALER1, ATOM_SCALER2
+  UCHAR ucEnable;            // ATOM_SCALER_DISABLE or ATOM_SCALER_CENTER or ATOM_SCALER_EXPANSION
+  UCHAR ucTVStandard;        // 
+  UCHAR ucPadding[1];
+}ENABLE_SCALER_PARAMETERS; 
+#define ENABLE_SCALER_PS_ALLOCATION ENABLE_SCALER_PARAMETERS 
+
+//ucEnable:
 #define SCALER_BYPASS_AUTO_CENTER_NO_REPLICATION    0
 #define SCALER_BYPASS_AUTO_CENTER_AUTO_REPLICATION  1
 #define SCALER_ENABLE_2TAP_ALPHA_MODE               2
 #define SCALER_ENABLE_MULTITAP_MODE                 3
 
-typedef struct _ENABLE_HARDWARE_ICON_CURSOR_PARAMETERS {
-	ULONG usHWIconHorzVertPosn;	/*  Hardware Icon Vertical position */
-	UCHAR ucHWIconVertOffset;	/*  Hardware Icon Vertical offset */
-	UCHAR ucHWIconHorzOffset;	/*  Hardware Icon Horizontal offset */
-	UCHAR ucSelection;	/*  ATOM_CURSOR1 or ATOM_ICON1 or ATOM_CURSOR2 or ATOM_ICON2 */
-	UCHAR ucEnable;		/*  ATOM_ENABLE or ATOM_DISABLE */
-} ENABLE_HARDWARE_ICON_CURSOR_PARAMETERS;
-
-typedef struct _ENABLE_HARDWARE_ICON_CURSOR_PS_ALLOCATION {
-	ENABLE_HARDWARE_ICON_CURSOR_PARAMETERS sEnableIcon;
-	ENABLE_CRTC_PARAMETERS sReserved;
-} ENABLE_HARDWARE_ICON_CURSOR_PS_ALLOCATION;
-
-typedef struct _ENABLE_GRAPH_SURFACE_PARAMETERS {
-	USHORT usHight;		/*  Image Hight */
-	USHORT usWidth;		/*  Image Width */
-	UCHAR ucSurface;	/*  Surface 1 or 2 */
-	UCHAR ucPadding[3];
-} ENABLE_GRAPH_SURFACE_PARAMETERS;
-
-typedef struct _ENABLE_GRAPH_SURFACE_PARAMETERS_V1_2 {
-	USHORT usHight;		/*  Image Hight */
-	USHORT usWidth;		/*  Image Width */
-	UCHAR ucSurface;	/*  Surface 1 or 2 */
-	UCHAR ucEnable;		/*  ATOM_ENABLE or ATOM_DISABLE */
-	UCHAR ucPadding[2];
-} ENABLE_GRAPH_SURFACE_PARAMETERS_V1_2;
-
-typedef struct _ENABLE_GRAPH_SURFACE_PS_ALLOCATION {
-	ENABLE_GRAPH_SURFACE_PARAMETERS sSetSurface;
-	ENABLE_YUV_PS_ALLOCATION sReserved;	/*  Don't set this one */
-} ENABLE_GRAPH_SURFACE_PS_ALLOCATION;
-
-typedef struct _MEMORY_CLEAN_UP_PARAMETERS {
-	USHORT usMemoryStart;	/* in 8Kb boundry, offset from memory base address */
-	USHORT usMemorySize;	/* 8Kb blocks aligned */
-} MEMORY_CLEAN_UP_PARAMETERS;
+typedef struct _ENABLE_HARDWARE_ICON_CURSOR_PARAMETERS
+{
+  ULONG  usHWIconHorzVertPosn;        // Hardware Icon Vertical position
+  UCHAR  ucHWIconVertOffset;          // Hardware Icon Vertical offset
+  UCHAR  ucHWIconHorzOffset;          // Hardware Icon Horizontal offset
+  UCHAR  ucSelection;                 // ATOM_CURSOR1 or ATOM_ICON1 or ATOM_CURSOR2 or ATOM_ICON2
+  UCHAR  ucEnable;                    // ATOM_ENABLE or ATOM_DISABLE
+}ENABLE_HARDWARE_ICON_CURSOR_PARAMETERS;
+
+typedef struct _ENABLE_HARDWARE_ICON_CURSOR_PS_ALLOCATION
+{
+  ENABLE_HARDWARE_ICON_CURSOR_PARAMETERS  sEnableIcon;
+  ENABLE_CRTC_PARAMETERS                  sReserved;  
+}ENABLE_HARDWARE_ICON_CURSOR_PS_ALLOCATION;
+
+typedef struct _ENABLE_GRAPH_SURFACE_PARAMETERS
+{
+  USHORT usHight;                     // Image Hight
+  USHORT usWidth;                     // Image Width
+  UCHAR  ucSurface;                   // Surface 1 or 2	
+  UCHAR  ucPadding[3];
+}ENABLE_GRAPH_SURFACE_PARAMETERS;
+
+typedef struct _ENABLE_GRAPH_SURFACE_PARAMETERS_V1_2
+{
+  USHORT usHight;                     // Image Hight
+  USHORT usWidth;                     // Image Width
+  UCHAR  ucSurface;                   // Surface 1 or 2
+  UCHAR  ucEnable;                    // ATOM_ENABLE or ATOM_DISABLE
+  UCHAR  ucPadding[2];
+}ENABLE_GRAPH_SURFACE_PARAMETERS_V1_2;
+
+typedef struct _ENABLE_GRAPH_SURFACE_PARAMETERS_V1_3
+{
+  USHORT usHight;                     // Image Hight
+  USHORT usWidth;                     // Image Width
+  UCHAR  ucSurface;                   // Surface 1 or 2
+  UCHAR  ucEnable;                    // ATOM_ENABLE or ATOM_DISABLE
+  USHORT usDeviceId;                  // Active Device Id for this surface. If no device, set to 0. 
+}ENABLE_GRAPH_SURFACE_PARAMETERS_V1_3;
+
+typedef struct _ENABLE_GRAPH_SURFACE_PS_ALLOCATION
+{
+  ENABLE_GRAPH_SURFACE_PARAMETERS sSetSurface;          
+  ENABLE_YUV_PS_ALLOCATION        sReserved; // Don't set this one
+}ENABLE_GRAPH_SURFACE_PS_ALLOCATION;
+
+typedef struct _MEMORY_CLEAN_UP_PARAMETERS
+{
+  USHORT  usMemoryStart;                //in 8Kb boundry, offset from memory base address
+  USHORT  usMemorySize;                 //8Kb blocks aligned
+}MEMORY_CLEAN_UP_PARAMETERS;
 #define MEMORY_CLEAN_UP_PS_ALLOCATION MEMORY_CLEAN_UP_PARAMETERS
 
-typedef struct _GET_DISPLAY_SURFACE_SIZE_PARAMETERS {
-	USHORT usX_Size;	/* When use as input parameter, usX_Size indicates which CRTC */
-	USHORT usY_Size;
-} GET_DISPLAY_SURFACE_SIZE_PARAMETERS;
+typedef struct  _GET_DISPLAY_SURFACE_SIZE_PARAMETERS
+{
+  USHORT  usX_Size;                     //When use as input parameter, usX_Size indicates which CRTC                 
+  USHORT  usY_Size;
+}GET_DISPLAY_SURFACE_SIZE_PARAMETERS; 
 
-typedef struct _INDIRECT_IO_ACCESS {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	UCHAR IOAccessSequence[256];
+typedef struct _INDIRECT_IO_ACCESS
+{
+  ATOM_COMMON_TABLE_HEADER sHeader;  
+  UCHAR                    IOAccessSequence[256];
 } INDIRECT_IO_ACCESS;
 
 #define INDIRECT_READ              0x00
@@ -3615,93 +4414,108 @@ typedef struct _INDIRECT_IO_ACCESS {
 #define INDIRECT_IO_NBMISC_READ    INDIRECT_IO_NBMISC | INDIRECT_READ
 #define INDIRECT_IO_NBMISC_WRITE   INDIRECT_IO_NBMISC | INDIRECT_WRITE
 
-typedef struct _ATOM_OEM_INFO {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	ATOM_I2C_ID_CONFIG_ACCESS sucI2cId;
-} ATOM_OEM_INFO;
-
-typedef struct _ATOM_TV_MODE {
-	UCHAR ucVMode_Num;	/* Video mode number */
-	UCHAR ucTV_Mode_Num;	/* Internal TV mode number */
-} ATOM_TV_MODE;
-
-typedef struct _ATOM_BIOS_INT_TVSTD_MODE {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	USHORT usTV_Mode_LUT_Offset;	/*  Pointer to standard to internal number conversion table */
-	USHORT usTV_FIFO_Offset;	/*  Pointer to FIFO entry table */
-	USHORT usNTSC_Tbl_Offset;	/*  Pointer to SDTV_Mode_NTSC table */
-	USHORT usPAL_Tbl_Offset;	/*  Pointer to SDTV_Mode_PAL table */
-	USHORT usCV_Tbl_Offset;	/*  Pointer to SDTV_Mode_PAL table */
-} ATOM_BIOS_INT_TVSTD_MODE;
-
-typedef struct _ATOM_TV_MODE_SCALER_PTR {
-	USHORT ucFilter0_Offset;	/* Pointer to filter format 0 coefficients */
-	USHORT usFilter1_Offset;	/* Pointer to filter format 0 coefficients */
-	UCHAR ucTV_Mode_Num;
-} ATOM_TV_MODE_SCALER_PTR;
-
-typedef struct _ATOM_STANDARD_VESA_TIMING {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	ATOM_DTD_FORMAT aModeTimings[16];	/*  16 is not the real array number, just for initial allocation */
-} ATOM_STANDARD_VESA_TIMING;
-
-typedef struct _ATOM_STD_FORMAT {
-	USHORT usSTD_HDisp;
-	USHORT usSTD_VDisp;
-	USHORT usSTD_RefreshRate;
-	USHORT usReserved;
-} ATOM_STD_FORMAT;
-
-typedef struct _ATOM_VESA_TO_EXTENDED_MODE {
-	USHORT usVESA_ModeNumber;
-	USHORT usExtendedModeNumber;
-} ATOM_VESA_TO_EXTENDED_MODE;
-
-typedef struct _ATOM_VESA_TO_INTENAL_MODE_LUT {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	ATOM_VESA_TO_EXTENDED_MODE asVESA_ToExtendedModeInfo[76];
-} ATOM_VESA_TO_INTENAL_MODE_LUT;
+typedef struct _ATOM_OEM_INFO
+{ 
+  ATOM_COMMON_TABLE_HEADER	sHeader;
+  ATOM_I2C_ID_CONFIG_ACCESS sucI2cId;
+}ATOM_OEM_INFO;
+
+typedef struct _ATOM_TV_MODE
+{
+   UCHAR	ucVMode_Num;			  //Video mode number
+   UCHAR	ucTV_Mode_Num;			//Internal TV mode number
+}ATOM_TV_MODE;
+
+typedef struct _ATOM_BIOS_INT_TVSTD_MODE
+{
+  ATOM_COMMON_TABLE_HEADER sHeader;  
+   USHORT	usTV_Mode_LUT_Offset;	// Pointer to standard to internal number conversion table
+   USHORT	usTV_FIFO_Offset;		  // Pointer to FIFO entry table
+   USHORT	usNTSC_Tbl_Offset;		// Pointer to SDTV_Mode_NTSC table
+   USHORT	usPAL_Tbl_Offset;		  // Pointer to SDTV_Mode_PAL table 
+   USHORT	usCV_Tbl_Offset;		  // Pointer to SDTV_Mode_PAL table 
+}ATOM_BIOS_INT_TVSTD_MODE;
+
+
+typedef struct _ATOM_TV_MODE_SCALER_PTR
+{
+   USHORT	ucFilter0_Offset;		//Pointer to filter format 0 coefficients
+   USHORT	usFilter1_Offset;		//Pointer to filter format 0 coefficients
+   UCHAR	ucTV_Mode_Num;
+}ATOM_TV_MODE_SCALER_PTR;
+
+typedef struct _ATOM_STANDARD_VESA_TIMING
+{
+  ATOM_COMMON_TABLE_HEADER sHeader;  
+  ATOM_DTD_FORMAT 				 aModeTimings[16];      // 16 is not the real array number, just for initial allocation
+}ATOM_STANDARD_VESA_TIMING;
+
+
+typedef struct _ATOM_STD_FORMAT
+{ 
+  USHORT    usSTD_HDisp;
+  USHORT    usSTD_VDisp;
+  USHORT    usSTD_RefreshRate;
+  USHORT    usReserved;
+}ATOM_STD_FORMAT;
+
+typedef struct _ATOM_VESA_TO_EXTENDED_MODE
+{
+  USHORT  usVESA_ModeNumber;
+  USHORT  usExtendedModeNumber;
+}ATOM_VESA_TO_EXTENDED_MODE;
+
+typedef struct _ATOM_VESA_TO_INTENAL_MODE_LUT
+{ 
+  ATOM_COMMON_TABLE_HEADER   sHeader;  
+  ATOM_VESA_TO_EXTENDED_MODE asVESA_ToExtendedModeInfo[76];
+}ATOM_VESA_TO_INTENAL_MODE_LUT;
 
 /*************** ATOM Memory Related Data Structure ***********************/
-typedef struct _ATOM_MEMORY_VENDOR_BLOCK {
-	UCHAR ucMemoryType;
-	UCHAR ucMemoryVendor;
-	UCHAR ucAdjMCId;
-	UCHAR ucDynClkId;
-	ULONG ulDllResetClkRange;
-} ATOM_MEMORY_VENDOR_BLOCK;
-
-typedef struct _ATOM_MEMORY_SETTING_ID_CONFIG {
+typedef struct _ATOM_MEMORY_VENDOR_BLOCK{
+	UCHAR												ucMemoryType;
+	UCHAR												ucMemoryVendor;
+	UCHAR												ucAdjMCId;
+	UCHAR												ucDynClkId;
+	ULONG												ulDllResetClkRange;
+}ATOM_MEMORY_VENDOR_BLOCK;
+
+
+typedef struct _ATOM_MEMORY_SETTING_ID_CONFIG{
 #if ATOM_BIG_ENDIAN
-	ULONG ucMemBlkId:8;
-	ULONG ulMemClockRange:24;
+	ULONG												ucMemBlkId:8;
+	ULONG												ulMemClockRange:24;
 #else
-	ULONG ulMemClockRange:24;
-	ULONG ucMemBlkId:8;
+	ULONG												ulMemClockRange:24;
+	ULONG												ucMemBlkId:8;
 #endif
-} ATOM_MEMORY_SETTING_ID_CONFIG;
-
-typedef union _ATOM_MEMORY_SETTING_ID_CONFIG_ACCESS {
-	ATOM_MEMORY_SETTING_ID_CONFIG slAccess;
-	ULONG ulAccess;
-} ATOM_MEMORY_SETTING_ID_CONFIG_ACCESS;
-
-typedef struct _ATOM_MEMORY_SETTING_DATA_BLOCK {
-	ATOM_MEMORY_SETTING_ID_CONFIG_ACCESS ulMemoryID;
-	ULONG aulMemData[1];
-} ATOM_MEMORY_SETTING_DATA_BLOCK;
-
-typedef struct _ATOM_INIT_REG_INDEX_FORMAT {
-	USHORT usRegIndex;	/*  MC register index */
-	UCHAR ucPreRegDataLength;	/*  offset in ATOM_INIT_REG_DATA_BLOCK.saRegDataBuf */
-} ATOM_INIT_REG_INDEX_FORMAT;
-
-typedef struct _ATOM_INIT_REG_BLOCK {
-	USHORT usRegIndexTblSize;	/* size of asRegIndexBuf */
-	USHORT usRegDataBlkSize;	/* size of ATOM_MEMORY_SETTING_DATA_BLOCK */
-	ATOM_INIT_REG_INDEX_FORMAT asRegIndexBuf[1];
-	ATOM_MEMORY_SETTING_DATA_BLOCK asRegDataBuf[1];
-} ATOM_INIT_REG_BLOCK;
+}ATOM_MEMORY_SETTING_ID_CONFIG;
+
+typedef union _ATOM_MEMORY_SETTING_ID_CONFIG_ACCESS
+{
+  ATOM_MEMORY_SETTING_ID_CONFIG slAccess;
+  ULONG                         ulAccess;
+}ATOM_MEMORY_SETTING_ID_CONFIG_ACCESS;
+
+
+typedef struct _ATOM_MEMORY_SETTING_DATA_BLOCK{
+	ATOM_MEMORY_SETTING_ID_CONFIG_ACCESS			ulMemoryID;
+	ULONG															        aulMemData[1];
+}ATOM_MEMORY_SETTING_DATA_BLOCK;
+
+
+typedef struct _ATOM_INIT_REG_INDEX_FORMAT{
+	 USHORT											usRegIndex;                                     // MC register index
+	 UCHAR											ucPreRegDataLength;                             // offset in ATOM_INIT_REG_DATA_BLOCK.saRegDataBuf
+}ATOM_INIT_REG_INDEX_FORMAT;
+
+
+typedef struct _ATOM_INIT_REG_BLOCK{
+	USHORT													usRegIndexTblSize;													//size of asRegIndexBuf
+	USHORT													usRegDataBlkSize;														//size of ATOM_MEMORY_SETTING_DATA_BLOCK
+	ATOM_INIT_REG_INDEX_FORMAT			asRegIndexBuf[1];
+	ATOM_MEMORY_SETTING_DATA_BLOCK	asRegDataBuf[1];
+}ATOM_INIT_REG_BLOCK;
 
 #define END_OF_REG_INDEX_BLOCK  0x0ffff
 #define END_OF_REG_DATA_BLOCK   0x00000000
@@ -3716,16 +4530,19 @@ typedef struct _ATOM_INIT_REG_BLOCK {
 #define INDEX_ACCESS_RANGE_END		    (INDEX_ACCESS_RANGE_BEGIN + 1)
 #define VALUE_INDEX_ACCESS_SINGLE	    (INDEX_ACCESS_RANGE_END + 1)
 
-typedef struct _ATOM_MC_INIT_PARAM_TABLE {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	USHORT usAdjustARB_SEQDataOffset;
-	USHORT usMCInitMemTypeTblOffset;
-	USHORT usMCInitCommonTblOffset;
-	USHORT usMCInitPowerDownTblOffset;
-	ULONG ulARB_SEQDataBuf[32];
-	ATOM_INIT_REG_BLOCK asMCInitMemType;
-	ATOM_INIT_REG_BLOCK asMCInitCommon;
-} ATOM_MC_INIT_PARAM_TABLE;
+
+typedef struct _ATOM_MC_INIT_PARAM_TABLE
+{ 
+  ATOM_COMMON_TABLE_HEADER		sHeader;
+  USHORT											usAdjustARB_SEQDataOffset;
+  USHORT											usMCInitMemTypeTblOffset;
+  USHORT											usMCInitCommonTblOffset;
+  USHORT											usMCInitPowerDownTblOffset;
+	ULONG												ulARB_SEQDataBuf[32];
+	ATOM_INIT_REG_BLOCK					asMCInitMemType;
+	ATOM_INIT_REG_BLOCK					asMCInitCommon;
+}ATOM_MC_INIT_PARAM_TABLE;
+
 
 #define _4Mx16              0x2
 #define _4Mx32              0x3
@@ -3751,221 +4568,272 @@ typedef struct _ATOM_MC_INIT_PARAM_TABLE {
 
 #define QIMONDA             INFINEON
 #define PROMOS              MOSEL
+#define KRETON              INFINEON
 
-/* ///////////Support for GDDR5 MC uCode to reside in upper 64K of ROM///////////// */
+/////////////Support for GDDR5 MC uCode to reside in upper 64K of ROM/////////////
 
 #define UCODE_ROM_START_ADDRESS		0x1c000
-#define	UCODE_SIGNATURE			0x4375434d	/*  'MCuC' - MC uCode */
-
-/* uCode block header for reference */
-
-typedef struct _MCuCodeHeader {
-	ULONG ulSignature;
-	UCHAR ucRevision;
-	UCHAR ucChecksum;
-	UCHAR ucReserved1;
-	UCHAR ucReserved2;
-	USHORT usParametersLength;
-	USHORT usUCodeLength;
-	USHORT usReserved1;
-	USHORT usReserved2;
+#define	UCODE_SIGNATURE			0x4375434d // 'MCuC' - MC uCode
+
+//uCode block header for reference
+
+typedef struct _MCuCodeHeader
+{
+  ULONG  ulSignature;
+  UCHAR  ucRevision;
+  UCHAR  ucChecksum;
+  UCHAR  ucReserved1;
+  UCHAR  ucReserved2;
+  USHORT usParametersLength;
+  USHORT usUCodeLength;
+  USHORT usReserved1;
+  USHORT usReserved2;
 } MCuCodeHeader;
 
-/* //////////////////////////////////////////////////////////////////////////////// */
+//////////////////////////////////////////////////////////////////////////////////
 
 #define ATOM_MAX_NUMBER_OF_VRAM_MODULE	16
 
 #define ATOM_VRAM_MODULE_MEMORY_VENDOR_ID_MASK	0xF
-typedef struct _ATOM_VRAM_MODULE_V1 {
-	ULONG ulReserved;
-	USHORT usEMRSValue;
-	USHORT usMRSValue;
-	USHORT usReserved;
-	UCHAR ucExtMemoryID;	/*  An external indicator (by hardcode, callback or pin) to tell what is the current memory module */
-	UCHAR ucMemoryType;	/*  [7:4]=0x1:DDR1;=0x2:DDR2;=0x3:DDR3;=0x4:DDR4;[3:0] reserved; */
-	UCHAR ucMemoryVenderID;	/*  Predefined,never change across designs or memory type/vender */
-	UCHAR ucMemoryDeviceCfg;	/*  [7:4]=0x0:4M;=0x1:8M;=0x2:16M;0x3:32M....[3:0]=0x0:x4;=0x1:x8;=0x2:x16;=0x3:x32... */
-	UCHAR ucRow;		/*  Number of Row,in power of 2; */
-	UCHAR ucColumn;		/*  Number of Column,in power of 2; */
-	UCHAR ucBank;		/*  Nunber of Bank; */
-	UCHAR ucRank;		/*  Number of Rank, in power of 2 */
-	UCHAR ucChannelNum;	/*  Number of channel; */
-	UCHAR ucChannelConfig;	/*  [3:0]=Indication of what channel combination;[4:7]=Channel bit width, in number of 2 */
-	UCHAR ucDefaultMVDDQ_ID;	/*  Default MVDDQ setting for this memory block, ID linking to MVDDQ info table to find real set-up data; */
-	UCHAR ucDefaultMVDDC_ID;	/*  Default MVDDC setting for this memory block, ID linking to MVDDC info table to find real set-up data; */
-	UCHAR ucReserved[2];
-} ATOM_VRAM_MODULE_V1;
-
-typedef struct _ATOM_VRAM_MODULE_V2 {
-	ULONG ulReserved;
-	ULONG ulFlags;		/*  To enable/disable functionalities based on memory type */
-	ULONG ulEngineClock;	/*  Override of default engine clock for particular memory type */
-	ULONG ulMemoryClock;	/*  Override of default memory clock for particular memory type */
-	USHORT usEMRS2Value;	/*  EMRS2 Value is used for GDDR2 and GDDR4 memory type */
-	USHORT usEMRS3Value;	/*  EMRS3 Value is used for GDDR2 and GDDR4 memory type */
-	USHORT usEMRSValue;
-	USHORT usMRSValue;
-	USHORT usReserved;
-	UCHAR ucExtMemoryID;	/*  An external indicator (by hardcode, callback or pin) to tell what is the current memory module */
-	UCHAR ucMemoryType;	/*  [7:4]=0x1:DDR1;=0x2:DDR2;=0x3:DDR3;=0x4:DDR4;[3:0] - must not be used for now; */
-	UCHAR ucMemoryVenderID;	/*  Predefined,never change across designs or memory type/vender. If not predefined, vendor detection table gets executed */
-	UCHAR ucMemoryDeviceCfg;	/*  [7:4]=0x0:4M;=0x1:8M;=0x2:16M;0x3:32M....[3:0]=0x0:x4;=0x1:x8;=0x2:x16;=0x3:x32... */
-	UCHAR ucRow;		/*  Number of Row,in power of 2; */
-	UCHAR ucColumn;		/*  Number of Column,in power of 2; */
-	UCHAR ucBank;		/*  Nunber of Bank; */
-	UCHAR ucRank;		/*  Number of Rank, in power of 2 */
-	UCHAR ucChannelNum;	/*  Number of channel; */
-	UCHAR ucChannelConfig;	/*  [3:0]=Indication of what channel combination;[4:7]=Channel bit width, in number of 2 */
-	UCHAR ucDefaultMVDDQ_ID;	/*  Default MVDDQ setting for this memory block, ID linking to MVDDQ info table to find real set-up data; */
-	UCHAR ucDefaultMVDDC_ID;	/*  Default MVDDC setting for this memory block, ID linking to MVDDC info table to find real set-up data; */
-	UCHAR ucRefreshRateFactor;
-	UCHAR ucReserved[3];
-} ATOM_VRAM_MODULE_V2;
-
-typedef struct _ATOM_MEMORY_TIMING_FORMAT {
-	ULONG ulClkRange;	/*  memory clock in 10kHz unit, when target memory clock is below this clock, use this memory timing */
-	union {
-		USHORT usMRS;	/*  mode register */
-		USHORT usDDR3_MR0;
-	};
-	union {
-		USHORT usEMRS;	/*  extended mode register */
-		USHORT usDDR3_MR1;
-	};
-	UCHAR ucCL;		/*  CAS latency */
-	UCHAR ucWL;		/*  WRITE Latency */
-	UCHAR uctRAS;		/*  tRAS */
-	UCHAR uctRC;		/*  tRC */
-	UCHAR uctRFC;		/*  tRFC */
-	UCHAR uctRCDR;		/*  tRCDR */
-	UCHAR uctRCDW;		/*  tRCDW */
-	UCHAR uctRP;		/*  tRP */
-	UCHAR uctRRD;		/*  tRRD */
-	UCHAR uctWR;		/*  tWR */
-	UCHAR uctWTR;		/*  tWTR */
-	UCHAR uctPDIX;		/*  tPDIX */
-	UCHAR uctFAW;		/*  tFAW */
-	UCHAR uctAOND;		/*  tAOND */
-	union {
-		struct {
-			UCHAR ucflag;	/*  flag to control memory timing calculation. bit0= control EMRS2 Infineon */
-			UCHAR ucReserved;
-		};
-		USHORT usDDR3_MR2;
-	};
-} ATOM_MEMORY_TIMING_FORMAT;
-
-typedef struct _ATOM_MEMORY_TIMING_FORMAT_V1 {
-	ULONG ulClkRange;	/*  memory clock in 10kHz unit, when target memory clock is below this clock, use this memory timing */
-	USHORT usMRS;		/*  mode register */
-	USHORT usEMRS;		/*  extended mode register */
-	UCHAR ucCL;		/*  CAS latency */
-	UCHAR ucWL;		/*  WRITE Latency */
-	UCHAR uctRAS;		/*  tRAS */
-	UCHAR uctRC;		/*  tRC */
-	UCHAR uctRFC;		/*  tRFC */
-	UCHAR uctRCDR;		/*  tRCDR */
-	UCHAR uctRCDW;		/*  tRCDW */
-	UCHAR uctRP;		/*  tRP */
-	UCHAR uctRRD;		/*  tRRD */
-	UCHAR uctWR;		/*  tWR */
-	UCHAR uctWTR;		/*  tWTR */
-	UCHAR uctPDIX;		/*  tPDIX */
-	UCHAR uctFAW;		/*  tFAW */
-	UCHAR uctAOND;		/*  tAOND */
-	UCHAR ucflag;		/*  flag to control memory timing calculation. bit0= control EMRS2 Infineon */
-/* ///////////////////////GDDR parameters/////////////////////////////////// */
-	UCHAR uctCCDL;		/*  */
-	UCHAR uctCRCRL;		/*  */
-	UCHAR uctCRCWL;		/*  */
-	UCHAR uctCKE;		/*  */
-	UCHAR uctCKRSE;		/*  */
-	UCHAR uctCKRSX;		/*  */
-	UCHAR uctFAW32;		/*  */
-	UCHAR ucReserved1;	/*  */
-	UCHAR ucReserved2;	/*  */
-	UCHAR ucTerminator;
-} ATOM_MEMORY_TIMING_FORMAT_V1;
-
-typedef struct _ATOM_MEMORY_FORMAT {
-	ULONG ulDllDisClock;	/*  memory DLL will be disable when target memory clock is below this clock */
-	union {
-		USHORT usEMRS2Value;	/*  EMRS2 Value is used for GDDR2 and GDDR4 memory type */
-		USHORT usDDR3_Reserved;	/*  Not used for DDR3 memory */
-	};
-	union {
-		USHORT usEMRS3Value;	/*  EMRS3 Value is used for GDDR2 and GDDR4 memory type */
-		USHORT usDDR3_MR3;	/*  Used for DDR3 memory */
-	};
-	UCHAR ucMemoryType;	/*  [7:4]=0x1:DDR1;=0x2:DDR2;=0x3:DDR3;=0x4:DDR4;[3:0] - must not be used for now; */
-	UCHAR ucMemoryVenderID;	/*  Predefined,never change across designs or memory type/vender. If not predefined, vendor detection table gets executed */
-	UCHAR ucRow;		/*  Number of Row,in power of 2; */
-	UCHAR ucColumn;		/*  Number of Column,in power of 2; */
-	UCHAR ucBank;		/*  Nunber of Bank; */
-	UCHAR ucRank;		/*  Number of Rank, in power of 2 */
-	UCHAR ucBurstSize;	/*  burst size, 0= burst size=4  1= burst size=8 */
-	UCHAR ucDllDisBit;	/*  position of DLL Enable/Disable bit in EMRS ( Extended Mode Register ) */
-	UCHAR ucRefreshRateFactor;	/*  memory refresh rate in unit of ms */
-	UCHAR ucDensity;	/*  _8Mx32, _16Mx32, _16Mx16, _32Mx16 */
-	UCHAR ucPreamble;	/* [7:4] Write Preamble, [3:0] Read Preamble */
-	UCHAR ucMemAttrib;	/*  Memory Device Addribute, like RDBI/WDBI etc */
-	ATOM_MEMORY_TIMING_FORMAT asMemTiming[5];	/* Memory Timing block sort from lower clock to higher clock */
-} ATOM_MEMORY_FORMAT;
-
-typedef struct _ATOM_VRAM_MODULE_V3 {
-	ULONG ulChannelMapCfg;	/*  board dependent paramenter:Channel combination */
-	USHORT usSize;		/*  size of ATOM_VRAM_MODULE_V3 */
-	USHORT usDefaultMVDDQ;	/*  board dependent parameter:Default Memory Core Voltage */
-	USHORT usDefaultMVDDC;	/*  board dependent parameter:Default Memory IO Voltage */
-	UCHAR ucExtMemoryID;	/*  An external indicator (by hardcode, callback or pin) to tell what is the current memory module */
-	UCHAR ucChannelNum;	/*  board dependent parameter:Number of channel; */
-	UCHAR ucChannelSize;	/*  board dependent parameter:32bit or 64bit */
-	UCHAR ucVREFI;		/*  board dependnt parameter: EXT or INT +160mv to -140mv */
-	UCHAR ucNPL_RT;		/*  board dependent parameter:NPL round trip delay, used for calculate memory timing parameters */
-	UCHAR ucFlag;		/*  To enable/disable functionalities based on memory type */
-	ATOM_MEMORY_FORMAT asMemory;	/*  describ all of video memory parameters from memory spec */
-} ATOM_VRAM_MODULE_V3;
-
-/* ATOM_VRAM_MODULE_V3.ucNPL_RT */
+typedef struct _ATOM_VRAM_MODULE_V1
+{
+  ULONG                      ulReserved;
+  USHORT                     usEMRSValue;  
+  USHORT                     usMRSValue;
+  USHORT                     usReserved;
+  UCHAR                      ucExtMemoryID;     // An external indicator (by hardcode, callback or pin) to tell what is the current memory module
+  UCHAR                      ucMemoryType;      // [7:4]=0x1:DDR1;=0x2:DDR2;=0x3:DDR3;=0x4:DDR4;[3:0] reserved;
+  UCHAR                      ucMemoryVenderID;  // Predefined,never change across designs or memory type/vender 
+  UCHAR                      ucMemoryDeviceCfg; // [7:4]=0x0:4M;=0x1:8M;=0x2:16M;0x3:32M....[3:0]=0x0:x4;=0x1:x8;=0x2:x16;=0x3:x32...
+  UCHAR                      ucRow;             // Number of Row,in power of 2;
+  UCHAR                      ucColumn;          // Number of Column,in power of 2;
+  UCHAR                      ucBank;            // Nunber of Bank;
+  UCHAR                      ucRank;            // Number of Rank, in power of 2
+  UCHAR                      ucChannelNum;      // Number of channel;
+  UCHAR                      ucChannelConfig;   // [3:0]=Indication of what channel combination;[4:7]=Channel bit width, in number of 2
+  UCHAR                      ucDefaultMVDDQ_ID; // Default MVDDQ setting for this memory block, ID linking to MVDDQ info table to find real set-up data;
+  UCHAR                      ucDefaultMVDDC_ID; // Default MVDDC setting for this memory block, ID linking to MVDDC info table to find real set-up data;
+  UCHAR                      ucReserved[2];
+}ATOM_VRAM_MODULE_V1;
+
+
+typedef struct _ATOM_VRAM_MODULE_V2
+{
+  ULONG                      ulReserved;
+  ULONG                      ulFlags;     			// To enable/disable functionalities based on memory type
+  ULONG                      ulEngineClock;     // Override of default engine clock for particular memory type
+  ULONG                      ulMemoryClock;     // Override of default memory clock for particular memory type
+  USHORT                     usEMRS2Value;      // EMRS2 Value is used for GDDR2 and GDDR4 memory type
+  USHORT                     usEMRS3Value;      // EMRS3 Value is used for GDDR2 and GDDR4 memory type
+  USHORT                     usEMRSValue;  
+  USHORT                     usMRSValue;
+  USHORT                     usReserved;
+  UCHAR                      ucExtMemoryID;     // An external indicator (by hardcode, callback or pin) to tell what is the current memory module
+  UCHAR                      ucMemoryType;      // [7:4]=0x1:DDR1;=0x2:DDR2;=0x3:DDR3;=0x4:DDR4;[3:0] - must not be used for now;
+  UCHAR                      ucMemoryVenderID;  // Predefined,never change across designs or memory type/vender. If not predefined, vendor detection table gets executed
+  UCHAR                      ucMemoryDeviceCfg; // [7:4]=0x0:4M;=0x1:8M;=0x2:16M;0x3:32M....[3:0]=0x0:x4;=0x1:x8;=0x2:x16;=0x3:x32...
+  UCHAR                      ucRow;             // Number of Row,in power of 2;
+  UCHAR                      ucColumn;          // Number of Column,in power of 2;
+  UCHAR                      ucBank;            // Nunber of Bank;
+  UCHAR                      ucRank;            // Number of Rank, in power of 2
+  UCHAR                      ucChannelNum;      // Number of channel;
+  UCHAR                      ucChannelConfig;   // [3:0]=Indication of what channel combination;[4:7]=Channel bit width, in number of 2
+  UCHAR                      ucDefaultMVDDQ_ID; // Default MVDDQ setting for this memory block, ID linking to MVDDQ info table to find real set-up data;
+  UCHAR                      ucDefaultMVDDC_ID; // Default MVDDC setting for this memory block, ID linking to MVDDC info table to find real set-up data;
+  UCHAR                      ucRefreshRateFactor;
+  UCHAR                      ucReserved[3];
+}ATOM_VRAM_MODULE_V2;
+
+
+typedef	struct _ATOM_MEMORY_TIMING_FORMAT
+{
+	ULONG											 ulClkRange;				// memory clock in 10kHz unit, when target memory clock is below this clock, use this memory timing 	
+  union{
+	  USHORT										 usMRS;							// mode register						
+    USHORT                     usDDR3_MR0;
+  };
+  union{
+	  USHORT										 usEMRS;						// extended mode register
+    USHORT                     usDDR3_MR1;
+  };
+	UCHAR											 ucCL;							// CAS latency
+	UCHAR											 ucWL;							// WRITE Latency				
+	UCHAR											 uctRAS;						// tRAS
+	UCHAR											 uctRC;							// tRC	
+	UCHAR											 uctRFC;						// tRFC
+	UCHAR											 uctRCDR;						// tRCDR	
+	UCHAR											 uctRCDW;						// tRCDW
+	UCHAR											 uctRP;							// tRP
+	UCHAR											 uctRRD;						// tRRD	
+	UCHAR											 uctWR;							// tWR
+	UCHAR											 uctWTR;						// tWTR
+	UCHAR											 uctPDIX;						// tPDIX
+	UCHAR											 uctFAW;						// tFAW
+	UCHAR											 uctAOND;						// tAOND
+  union 
+  {
+    struct {
+	    UCHAR											 ucflag;						// flag to control memory timing calculation. bit0= control EMRS2 Infineon 
+	    UCHAR											 ucReserved;						
+    };
+    USHORT                   usDDR3_MR2;
+  };
+}ATOM_MEMORY_TIMING_FORMAT;
+
+
+typedef	struct _ATOM_MEMORY_TIMING_FORMAT_V1
+{
+	ULONG											 ulClkRange;				// memory clock in 10kHz unit, when target memory clock is below this clock, use this memory timing 	
+	USHORT										 usMRS;							// mode register						
+	USHORT										 usEMRS;						// extended mode register
+	UCHAR											 ucCL;							// CAS latency
+	UCHAR											 ucWL;							// WRITE Latency				
+	UCHAR											 uctRAS;						// tRAS
+	UCHAR											 uctRC;							// tRC	
+	UCHAR											 uctRFC;						// tRFC
+	UCHAR											 uctRCDR;						// tRCDR	
+	UCHAR											 uctRCDW;						// tRCDW
+	UCHAR											 uctRP;							// tRP
+	UCHAR											 uctRRD;						// tRRD	
+	UCHAR											 uctWR;							// tWR
+	UCHAR											 uctWTR;						// tWTR
+	UCHAR											 uctPDIX;						// tPDIX
+	UCHAR											 uctFAW;						// tFAW
+	UCHAR											 uctAOND;						// tAOND
+	UCHAR											 ucflag;						// flag to control memory timing calculation. bit0= control EMRS2 Infineon 
+////////////////////////////////////GDDR parameters///////////////////////////////////
+	UCHAR											 uctCCDL;						// 
+	UCHAR											 uctCRCRL;						// 
+	UCHAR											 uctCRCWL;						// 
+	UCHAR											 uctCKE;						// 
+	UCHAR											 uctCKRSE;						// 
+	UCHAR											 uctCKRSX;						// 
+	UCHAR											 uctFAW32;						// 
+	UCHAR											 ucMR5lo;					// 
+	UCHAR											 ucMR5hi;					// 
+	UCHAR											 ucTerminator;
+}ATOM_MEMORY_TIMING_FORMAT_V1;
+
+typedef	struct _ATOM_MEMORY_TIMING_FORMAT_V2
+{
+	ULONG											 ulClkRange;				// memory clock in 10kHz unit, when target memory clock is below this clock, use this memory timing 	
+	USHORT										 usMRS;							// mode register						
+	USHORT										 usEMRS;						// extended mode register
+	UCHAR											 ucCL;							// CAS latency
+	UCHAR											 ucWL;							// WRITE Latency				
+	UCHAR											 uctRAS;						// tRAS
+	UCHAR											 uctRC;							// tRC	
+	UCHAR											 uctRFC;						// tRFC
+	UCHAR											 uctRCDR;						// tRCDR	
+	UCHAR											 uctRCDW;						// tRCDW
+	UCHAR											 uctRP;							// tRP
+	UCHAR											 uctRRD;						// tRRD	
+	UCHAR											 uctWR;							// tWR
+	UCHAR											 uctWTR;						// tWTR
+	UCHAR											 uctPDIX;						// tPDIX
+	UCHAR											 uctFAW;						// tFAW
+	UCHAR											 uctAOND;						// tAOND
+	UCHAR											 ucflag;						// flag to control memory timing calculation. bit0= control EMRS2 Infineon 
+////////////////////////////////////GDDR parameters///////////////////////////////////
+	UCHAR											 uctCCDL;						// 
+	UCHAR											 uctCRCRL;						// 
+	UCHAR											 uctCRCWL;						// 
+	UCHAR											 uctCKE;						// 
+	UCHAR											 uctCKRSE;						// 
+	UCHAR											 uctCKRSX;						// 
+	UCHAR											 uctFAW32;						// 
+	UCHAR											 ucMR4lo;					// 
+	UCHAR											 ucMR4hi;					// 
+	UCHAR											 ucMR5lo;					// 
+	UCHAR											 ucMR5hi;					// 
+	UCHAR											 ucTerminator;
+	UCHAR											 ucReserved;	
+}ATOM_MEMORY_TIMING_FORMAT_V2;
+
+typedef	struct _ATOM_MEMORY_FORMAT
+{
+	ULONG											 ulDllDisClock;			// memory DLL will be disable when target memory clock is below this clock
+  union{
+    USHORT                     usEMRS2Value;      // EMRS2 Value is used for GDDR2 and GDDR4 memory type
+    USHORT                     usDDR3_Reserved;   // Not used for DDR3 memory
+  };
+  union{
+    USHORT                     usEMRS3Value;      // EMRS3 Value is used for GDDR2 and GDDR4 memory type
+    USHORT                     usDDR3_MR3;        // Used for DDR3 memory
+  };
+  UCHAR                      ucMemoryType;      // [7:4]=0x1:DDR1;=0x2:DDR2;=0x3:DDR3;=0x4:DDR4;[3:0] - must not be used for now;
+  UCHAR                      ucMemoryVenderID;  // Predefined,never change across designs or memory type/vender. If not predefined, vendor detection table gets executed
+  UCHAR                      ucRow;             // Number of Row,in power of 2;
+  UCHAR                      ucColumn;          // Number of Column,in power of 2;
+  UCHAR                      ucBank;            // Nunber of Bank;
+  UCHAR                      ucRank;            // Number of Rank, in power of 2
+	UCHAR											 ucBurstSize;				// burst size, 0= burst size=4  1= burst size=8
+  UCHAR                      ucDllDisBit;				// position of DLL Enable/Disable bit in EMRS ( Extended Mode Register )
+  UCHAR                      ucRefreshRateFactor;	// memory refresh rate in unit of ms	
+	UCHAR											 ucDensity;					// _8Mx32, _16Mx32, _16Mx16, _32Mx16
+	UCHAR											 ucPreamble;				//[7:4] Write Preamble, [3:0] Read Preamble
+  UCHAR											 ucMemAttrib;				// Memory Device Addribute, like RDBI/WDBI etc
+	ATOM_MEMORY_TIMING_FORMAT	 asMemTiming[5];		//Memory Timing block sort from lower clock to higher clock
+}ATOM_MEMORY_FORMAT;
+
+
+typedef struct _ATOM_VRAM_MODULE_V3
+{
+	ULONG											 ulChannelMapCfg;		// board dependent paramenter:Channel combination
+	USHORT										 usSize;						// size of ATOM_VRAM_MODULE_V3
+  USHORT                     usDefaultMVDDQ;		// board dependent parameter:Default Memory Core Voltage
+  USHORT                     usDefaultMVDDC;		// board dependent parameter:Default Memory IO Voltage
+	UCHAR                      ucExtMemoryID;     // An external indicator (by hardcode, callback or pin) to tell what is the current memory module
+  UCHAR                      ucChannelNum;      // board dependent parameter:Number of channel;
+	UCHAR											 ucChannelSize;			// board dependent parameter:32bit or 64bit	
+	UCHAR											 ucVREFI;						// board dependnt parameter: EXT or INT +160mv to -140mv
+	UCHAR											 ucNPL_RT;					// board dependent parameter:NPL round trip delay, used for calculate memory timing parameters
+	UCHAR											 ucFlag;						// To enable/disable functionalities based on memory type
+	ATOM_MEMORY_FORMAT				 asMemory;					// describ all of video memory parameters from memory spec
+}ATOM_VRAM_MODULE_V3;
+
+
+//ATOM_VRAM_MODULE_V3.ucNPL_RT
 #define NPL_RT_MASK															0x0f
 #define BATTERY_ODT_MASK												0xc0
 
 #define ATOM_VRAM_MODULE		 ATOM_VRAM_MODULE_V3
 
-typedef struct _ATOM_VRAM_MODULE_V4 {
-	ULONG ulChannelMapCfg;	/*  board dependent parameter: Channel combination */
-	USHORT usModuleSize;	/*  size of ATOM_VRAM_MODULE_V4, make it easy for VBIOS to look for next entry of VRAM_MODULE */
-	USHORT usPrivateReserved;	/*  BIOS internal reserved space to optimize code size, updated by the compiler, shouldn't be modified manually!! */
-	/*  MC_ARB_RAMCFG (includes NOOFBANK,NOOFRANKS,NOOFROWS,NOOFCOLS) */
-	USHORT usReserved;
-	UCHAR ucExtMemoryID;	/*  An external indicator (by hardcode, callback or pin) to tell what is the current memory module */
-	UCHAR ucMemoryType;	/*  [7:4]=0x1:DDR1;=0x2:DDR2;=0x3:DDR3;=0x4:DDR4; 0x5:DDR5 [3:0] - Must be 0x0 for now; */
-	UCHAR ucChannelNum;	/*  Number of channels present in this module config */
-	UCHAR ucChannelWidth;	/*  0 - 32 bits; 1 - 64 bits */
-	UCHAR ucDensity;	/*  _8Mx32, _16Mx32, _16Mx16, _32Mx16 */
-	UCHAR ucFlag;		/*  To enable/disable functionalities based on memory type */
-	UCHAR ucMisc;		/*  bit0: 0 - single rank; 1 - dual rank;   bit2: 0 - burstlength 4, 1 - burstlength 8 */
-	UCHAR ucVREFI;		/*  board dependent parameter */
-	UCHAR ucNPL_RT;		/*  board dependent parameter:NPL round trip delay, used for calculate memory timing parameters */
-	UCHAR ucPreamble;	/*  [7:4] Write Preamble, [3:0] Read Preamble */
-	UCHAR ucMemorySize;	/*  BIOS internal reserved space to optimize code size, updated by the compiler, shouldn't be modified manually!! */
-	/*  Total memory size in unit of 16MB for CONFIG_MEMSIZE - bit[23:0] zeros */
-	UCHAR ucReserved[3];
-
-/* compare with V3, we flat the struct by merging ATOM_MEMORY_FORMAT (as is) into V4 as the same level */
-	union {
-		USHORT usEMRS2Value;	/*  EMRS2 Value is used for GDDR2 and GDDR4 memory type */
-		USHORT usDDR3_Reserved;
-	};
-	union {
-		USHORT usEMRS3Value;	/*  EMRS3 Value is used for GDDR2 and GDDR4 memory type */
-		USHORT usDDR3_MR3;	/*  Used for DDR3 memory */
-	};
-	UCHAR ucMemoryVenderID;	/*  Predefined, If not predefined, vendor detection table gets executed */
-	UCHAR ucRefreshRateFactor;	/*  [1:0]=RefreshFactor (00=8ms, 01=16ms, 10=32ms,11=64ms) */
-	UCHAR ucReserved2[2];
-	ATOM_MEMORY_TIMING_FORMAT asMemTiming[5];	/* Memory Timing block sort from lower clock to higher clock */
-} ATOM_VRAM_MODULE_V4;
+typedef struct _ATOM_VRAM_MODULE_V4
+{
+  ULONG	  ulChannelMapCfg;	                // board dependent parameter: Channel combination
+  USHORT  usModuleSize;                     // size of ATOM_VRAM_MODULE_V4, make it easy for VBIOS to look for next entry of VRAM_MODULE
+  USHORT  usPrivateReserved;                // BIOS internal reserved space to optimize code size, updated by the compiler, shouldn't be modified manually!!
+                                            // MC_ARB_RAMCFG (includes NOOFBANK,NOOFRANKS,NOOFROWS,NOOFCOLS)
+  USHORT  usReserved;
+  UCHAR   ucExtMemoryID;    		            // An external indicator (by hardcode, callback or pin) to tell what is the current memory module
+  UCHAR   ucMemoryType;                     // [7:4]=0x1:DDR1;=0x2:DDR2;=0x3:DDR3;=0x4:DDR4; 0x5:DDR5 [3:0] - Must be 0x0 for now;
+  UCHAR   ucChannelNum;                     // Number of channels present in this module config
+  UCHAR   ucChannelWidth;                   // 0 - 32 bits; 1 - 64 bits
+	UCHAR   ucDensity;                        // _8Mx32, _16Mx32, _16Mx16, _32Mx16
+	UCHAR	  ucFlag;						                // To enable/disable functionalities based on memory type
+	UCHAR	  ucMisc;						                // bit0: 0 - single rank; 1 - dual rank;   bit2: 0 - burstlength 4, 1 - burstlength 8
+  UCHAR		ucVREFI;                          // board dependent parameter
+  UCHAR   ucNPL_RT;                         // board dependent parameter:NPL round trip delay, used for calculate memory timing parameters
+  UCHAR		ucPreamble;                       // [7:4] Write Preamble, [3:0] Read Preamble
+  UCHAR   ucMemorySize;                     // BIOS internal reserved space to optimize code size, updated by the compiler, shouldn't be modified manually!!
+                                            // Total memory size in unit of 16MB for CONFIG_MEMSIZE - bit[23:0] zeros
+  UCHAR   ucReserved[3];
+
+//compare with V3, we flat the struct by merging ATOM_MEMORY_FORMAT (as is) into V4 as the same level
+  union{
+    USHORT	usEMRS2Value;                   // EMRS2 Value is used for GDDR2 and GDDR4 memory type
+    USHORT  usDDR3_Reserved;
+  };
+  union{
+    USHORT	usEMRS3Value;                   // EMRS3 Value is used for GDDR2 and GDDR4 memory type
+    USHORT  usDDR3_MR3;                     // Used for DDR3 memory
+  };  
+  UCHAR   ucMemoryVenderID;  		            // Predefined, If not predefined, vendor detection table gets executed
+  UCHAR	  ucRefreshRateFactor;              // [1:0]=RefreshFactor (00=8ms, 01=16ms, 10=32ms,11=64ms)
+  UCHAR   ucReserved2[2];
+  ATOM_MEMORY_TIMING_FORMAT  asMemTiming[5];//Memory Timing block sort from lower clock to higher clock
+}ATOM_VRAM_MODULE_V4;
 
 #define VRAM_MODULE_V4_MISC_RANK_MASK       0x3
 #define VRAM_MODULE_V4_MISC_DUAL_RANK       0x1
@@ -3973,96 +4841,139 @@ typedef struct _ATOM_VRAM_MODULE_V4 {
 #define VRAM_MODULE_V4_MISC_BL8             0x4
 #define VRAM_MODULE_V4_MISC_DUAL_CS         0x10
 
-typedef struct _ATOM_VRAM_MODULE_V5 {
-	ULONG ulChannelMapCfg;	/*  board dependent parameter: Channel combination */
-	USHORT usModuleSize;	/*  size of ATOM_VRAM_MODULE_V4, make it easy for VBIOS to look for next entry of VRAM_MODULE */
-	USHORT usPrivateReserved;	/*  BIOS internal reserved space to optimize code size, updated by the compiler, shouldn't be modified manually!! */
-	/*  MC_ARB_RAMCFG (includes NOOFBANK,NOOFRANKS,NOOFROWS,NOOFCOLS) */
-	USHORT usReserved;
-	UCHAR ucExtMemoryID;	/*  An external indicator (by hardcode, callback or pin) to tell what is the current memory module */
-	UCHAR ucMemoryType;	/*  [7:4]=0x1:DDR1;=0x2:DDR2;=0x3:DDR3;=0x4:DDR4; 0x5:DDR5 [3:0] - Must be 0x0 for now; */
-	UCHAR ucChannelNum;	/*  Number of channels present in this module config */
-	UCHAR ucChannelWidth;	/*  0 - 32 bits; 1 - 64 bits */
-	UCHAR ucDensity;	/*  _8Mx32, _16Mx32, _16Mx16, _32Mx16 */
-	UCHAR ucFlag;		/*  To enable/disable functionalities based on memory type */
-	UCHAR ucMisc;		/*  bit0: 0 - single rank; 1 - dual rank;   bit2: 0 - burstlength 4, 1 - burstlength 8 */
-	UCHAR ucVREFI;		/*  board dependent parameter */
-	UCHAR ucNPL_RT;		/*  board dependent parameter:NPL round trip delay, used for calculate memory timing parameters */
-	UCHAR ucPreamble;	/*  [7:4] Write Preamble, [3:0] Read Preamble */
-	UCHAR ucMemorySize;	/*  BIOS internal reserved space to optimize code size, updated by the compiler, shouldn't be modified manually!! */
-	/*  Total memory size in unit of 16MB for CONFIG_MEMSIZE - bit[23:0] zeros */
-	UCHAR ucReserved[3];
+typedef struct _ATOM_VRAM_MODULE_V5
+{
+  ULONG	  ulChannelMapCfg;	                // board dependent parameter: Channel combination
+  USHORT  usModuleSize;                     // size of ATOM_VRAM_MODULE_V4, make it easy for VBIOS to look for next entry of VRAM_MODULE
+  USHORT  usPrivateReserved;                // BIOS internal reserved space to optimize code size, updated by the compiler, shouldn't be modified manually!!
+                                            // MC_ARB_RAMCFG (includes NOOFBANK,NOOFRANKS,NOOFROWS,NOOFCOLS)
+  USHORT  usReserved;
+  UCHAR   ucExtMemoryID;    		            // An external indicator (by hardcode, callback or pin) to tell what is the current memory module
+  UCHAR   ucMemoryType;                     // [7:4]=0x1:DDR1;=0x2:DDR2;=0x3:DDR3;=0x4:DDR4; 0x5:DDR5 [3:0] - Must be 0x0 for now;
+  UCHAR   ucChannelNum;                     // Number of channels present in this module config
+  UCHAR   ucChannelWidth;                   // 0 - 32 bits; 1 - 64 bits
+	UCHAR   ucDensity;                        // _8Mx32, _16Mx32, _16Mx16, _32Mx16
+	UCHAR	  ucFlag;						                // To enable/disable functionalities based on memory type
+	UCHAR	  ucMisc;						                // bit0: 0 - single rank; 1 - dual rank;   bit2: 0 - burstlength 4, 1 - burstlength 8
+  UCHAR		ucVREFI;                          // board dependent parameter
+  UCHAR   ucNPL_RT;                         // board dependent parameter:NPL round trip delay, used for calculate memory timing parameters
+  UCHAR		ucPreamble;                       // [7:4] Write Preamble, [3:0] Read Preamble
+  UCHAR   ucMemorySize;                     // BIOS internal reserved space to optimize code size, updated by the compiler, shouldn't be modified manually!!
+                                            // Total memory size in unit of 16MB for CONFIG_MEMSIZE - bit[23:0] zeros
+  UCHAR   ucReserved[3];
+
+//compare with V3, we flat the struct by merging ATOM_MEMORY_FORMAT (as is) into V4 as the same level
+  USHORT	usEMRS2Value;      		            // EMRS2 Value is used for GDDR2 and GDDR4 memory type
+  USHORT	usEMRS3Value;      		            // EMRS3 Value is used for GDDR2 and GDDR4 memory type
+  UCHAR   ucMemoryVenderID;  		            // Predefined, If not predefined, vendor detection table gets executed
+  UCHAR	  ucRefreshRateFactor;              // [1:0]=RefreshFactor (00=8ms, 01=16ms, 10=32ms,11=64ms)
+  UCHAR	  ucFIFODepth;			                // FIFO depth supposes to be detected during vendor detection, but if we dont do vendor detection we have to hardcode FIFO Depth
+  UCHAR   ucCDR_Bandwidth;		   // [0:3]=Read CDR bandwidth, [4:7] - Write CDR Bandwidth
+  ATOM_MEMORY_TIMING_FORMAT_V1  asMemTiming[5];//Memory Timing block sort from lower clock to higher clock
+}ATOM_VRAM_MODULE_V5;
+
+typedef struct _ATOM_VRAM_MODULE_V6
+{
+  ULONG	  ulChannelMapCfg;	                // board dependent parameter: Channel combination
+  USHORT  usModuleSize;                     // size of ATOM_VRAM_MODULE_V4, make it easy for VBIOS to look for next entry of VRAM_MODULE
+  USHORT  usPrivateReserved;                // BIOS internal reserved space to optimize code size, updated by the compiler, shouldn't be modified manually!!
+                                            // MC_ARB_RAMCFG (includes NOOFBANK,NOOFRANKS,NOOFROWS,NOOFCOLS)
+  USHORT  usReserved;
+  UCHAR   ucExtMemoryID;    		            // An external indicator (by hardcode, callback or pin) to tell what is the current memory module
+  UCHAR   ucMemoryType;                     // [7:4]=0x1:DDR1;=0x2:DDR2;=0x3:DDR3;=0x4:DDR4; 0x5:DDR5 [3:0] - Must be 0x0 for now;
+  UCHAR   ucChannelNum;                     // Number of channels present in this module config
+  UCHAR   ucChannelWidth;                   // 0 - 32 bits; 1 - 64 bits
+	UCHAR   ucDensity;                        // _8Mx32, _16Mx32, _16Mx16, _32Mx16
+	UCHAR	  ucFlag;						                // To enable/disable functionalities based on memory type
+	UCHAR	  ucMisc;						                // bit0: 0 - single rank; 1 - dual rank;   bit2: 0 - burstlength 4, 1 - burstlength 8
+  UCHAR		ucVREFI;                          // board dependent parameter
+  UCHAR   ucNPL_RT;                         // board dependent parameter:NPL round trip delay, used for calculate memory timing parameters
+  UCHAR		ucPreamble;                       // [7:4] Write Preamble, [3:0] Read Preamble
+  UCHAR   ucMemorySize;                     // BIOS internal reserved space to optimize code size, updated by the compiler, shouldn't be modified manually!!
+                                            // Total memory size in unit of 16MB for CONFIG_MEMSIZE - bit[23:0] zeros
+  UCHAR   ucReserved[3];
+
+//compare with V3, we flat the struct by merging ATOM_MEMORY_FORMAT (as is) into V4 as the same level
+  USHORT	usEMRS2Value;      		            // EMRS2 Value is used for GDDR2 and GDDR4 memory type
+  USHORT	usEMRS3Value;      		            // EMRS3 Value is used for GDDR2 and GDDR4 memory type
+  UCHAR   ucMemoryVenderID;  		            // Predefined, If not predefined, vendor detection table gets executed
+  UCHAR	  ucRefreshRateFactor;              // [1:0]=RefreshFactor (00=8ms, 01=16ms, 10=32ms,11=64ms)
+  UCHAR	  ucFIFODepth;			                // FIFO depth supposes to be detected during vendor detection, but if we dont do vendor detection we have to hardcode FIFO Depth
+  UCHAR   ucCDR_Bandwidth;		   // [0:3]=Read CDR bandwidth, [4:7] - Write CDR Bandwidth
+  ATOM_MEMORY_TIMING_FORMAT_V2  asMemTiming[5];//Memory Timing block sort from lower clock to higher clock
+}ATOM_VRAM_MODULE_V6;
+
+
+
+typedef struct _ATOM_VRAM_INFO_V2
+{
+  ATOM_COMMON_TABLE_HEADER   sHeader;
+  UCHAR                      ucNumOfVRAMModule;
+  ATOM_VRAM_MODULE           aVramInfo[ATOM_MAX_NUMBER_OF_VRAM_MODULE];      // just for allocation, real number of blocks is in ucNumOfVRAMModule;
+}ATOM_VRAM_INFO_V2;
 
-/* compare with V3, we flat the struct by merging ATOM_MEMORY_FORMAT (as is) into V4 as the same level */
-	USHORT usEMRS2Value;	/*  EMRS2 Value is used for GDDR2 and GDDR4 memory type */
-	USHORT usEMRS3Value;	/*  EMRS3 Value is used for GDDR2 and GDDR4 memory type */
-	UCHAR ucMemoryVenderID;	/*  Predefined, If not predefined, vendor detection table gets executed */
-	UCHAR ucRefreshRateFactor;	/*  [1:0]=RefreshFactor (00=8ms, 01=16ms, 10=32ms,11=64ms) */
-	UCHAR ucFIFODepth;	/*  FIFO depth supposes to be detected during vendor detection, but if we dont do vendor detection we have to hardcode FIFO Depth */
-	UCHAR ucCDR_Bandwidth;	/*  [0:3]=Read CDR bandwidth, [4:7] - Write CDR Bandwidth */
-	ATOM_MEMORY_TIMING_FORMAT_V1 asMemTiming[5];	/* Memory Timing block sort from lower clock to higher clock */
-} ATOM_VRAM_MODULE_V5;
-
-typedef struct _ATOM_VRAM_INFO_V2 {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	UCHAR ucNumOfVRAMModule;
-	ATOM_VRAM_MODULE aVramInfo[ATOM_MAX_NUMBER_OF_VRAM_MODULE];	/*  just for allocation, real number of blocks is in ucNumOfVRAMModule; */
-} ATOM_VRAM_INFO_V2;
-
-typedef struct _ATOM_VRAM_INFO_V3 {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	USHORT usMemAdjustTblOffset;	/*  offset of ATOM_INIT_REG_BLOCK structure for memory vendor specific MC adjust setting */
-	USHORT usMemClkPatchTblOffset;	/*      offset of ATOM_INIT_REG_BLOCK structure for memory clock specific MC setting */
-	USHORT usRerseved;
-	UCHAR aVID_PinsShift[9];	/*  8 bit strap maximum+terminator */
-	UCHAR ucNumOfVRAMModule;
-	ATOM_VRAM_MODULE aVramInfo[ATOM_MAX_NUMBER_OF_VRAM_MODULE];	/*  just for allocation, real number of blocks is in ucNumOfVRAMModule; */
-	ATOM_INIT_REG_BLOCK asMemPatch;	/*  for allocation */
-	/*      ATOM_INIT_REG_BLOCK                              aMemAdjust; */
-} ATOM_VRAM_INFO_V3;
+typedef struct _ATOM_VRAM_INFO_V3
+{
+  ATOM_COMMON_TABLE_HEADER   sHeader;
+	USHORT										 usMemAdjustTblOffset;													 // offset of ATOM_INIT_REG_BLOCK structure for memory vendor specific MC adjust setting
+	USHORT										 usMemClkPatchTblOffset;												 //	offset of ATOM_INIT_REG_BLOCK structure for memory clock specific MC setting
+	USHORT										 usRerseved;
+	UCHAR           	         aVID_PinsShift[9];															 // 8 bit strap maximum+terminator
+  UCHAR                      ucNumOfVRAMModule;
+  ATOM_VRAM_MODULE		       aVramInfo[ATOM_MAX_NUMBER_OF_VRAM_MODULE];      // just for allocation, real number of blocks is in ucNumOfVRAMModule;
+	ATOM_INIT_REG_BLOCK				 asMemPatch;																		 // for allocation
+																																						 //	ATOM_INIT_REG_BLOCK				 aMemAdjust;
+}ATOM_VRAM_INFO_V3;
 
 #define	ATOM_VRAM_INFO_LAST	     ATOM_VRAM_INFO_V3
 
-typedef struct _ATOM_VRAM_INFO_V4 {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	USHORT usMemAdjustTblOffset;	/*  offset of ATOM_INIT_REG_BLOCK structure for memory vendor specific MC adjust setting */
-	USHORT usMemClkPatchTblOffset;	/*      offset of ATOM_INIT_REG_BLOCK structure for memory clock specific MC setting */
-	USHORT usRerseved;
-	UCHAR ucMemDQ7_0ByteRemap;	/*  DQ line byte remap, =0: Memory Data line BYTE0, =1: BYTE1, =2: BYTE2, =3: BYTE3 */
-	ULONG ulMemDQ7_0BitRemap;	/*  each DQ line ( 7~0) use 3bits, like: DQ0=Bit[2:0], DQ1:[5:3], ... DQ7:[23:21] */
-	UCHAR ucReservde[4];
-	UCHAR ucNumOfVRAMModule;
-	ATOM_VRAM_MODULE_V4 aVramInfo[ATOM_MAX_NUMBER_OF_VRAM_MODULE];	/*  just for allocation, real number of blocks is in ucNumOfVRAMModule; */
-	ATOM_INIT_REG_BLOCK asMemPatch;	/*  for allocation */
-	/*      ATOM_INIT_REG_BLOCK                              aMemAdjust; */
-} ATOM_VRAM_INFO_V4;
-
-typedef struct _ATOM_VRAM_GPIO_DETECTION_INFO {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	UCHAR aVID_PinsShift[9];	/* 8 bit strap maximum+terminator */
-} ATOM_VRAM_GPIO_DETECTION_INFO;
-
-typedef struct _ATOM_MEMORY_TRAINING_INFO {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	UCHAR ucTrainingLoop;
-	UCHAR ucReserved[3];
-	ATOM_INIT_REG_BLOCK asMemTrainingSetting;
-} ATOM_MEMORY_TRAINING_INFO;
-
-typedef struct SW_I2C_CNTL_DATA_PARAMETERS {
-	UCHAR ucControl;
-	UCHAR ucData;
-	UCHAR ucSatus;
-	UCHAR ucTemp;
+typedef struct _ATOM_VRAM_INFO_V4
+{
+  ATOM_COMMON_TABLE_HEADER   sHeader;
+	USHORT										 usMemAdjustTblOffset;													 // offset of ATOM_INIT_REG_BLOCK structure for memory vendor specific MC adjust setting
+	USHORT										 usMemClkPatchTblOffset;												 //	offset of ATOM_INIT_REG_BLOCK structure for memory clock specific MC setting
+	USHORT										 usRerseved;
+	UCHAR           	         ucMemDQ7_0ByteRemap;													   // DQ line byte remap, =0: Memory Data line BYTE0, =1: BYTE1, =2: BYTE2, =3: BYTE3
+  ULONG                      ulMemDQ7_0BitRemap;                             // each DQ line ( 7~0) use 3bits, like: DQ0=Bit[2:0], DQ1:[5:3], ... DQ7:[23:21]
+  UCHAR                      ucReservde[4]; 
+  UCHAR                      ucNumOfVRAMModule;
+  ATOM_VRAM_MODULE_V4		     aVramInfo[ATOM_MAX_NUMBER_OF_VRAM_MODULE];      // just for allocation, real number of blocks is in ucNumOfVRAMModule;
+	ATOM_INIT_REG_BLOCK				 asMemPatch;																		 // for allocation
+																																						 //	ATOM_INIT_REG_BLOCK				 aMemAdjust;
+}ATOM_VRAM_INFO_V4;
+
+typedef struct _ATOM_VRAM_GPIO_DETECTION_INFO
+{
+  ATOM_COMMON_TABLE_HEADER   sHeader;
+  UCHAR           	         aVID_PinsShift[9];   //8 bit strap maximum+terminator
+}ATOM_VRAM_GPIO_DETECTION_INFO;
+
+
+typedef struct _ATOM_MEMORY_TRAINING_INFO
+{
+	ATOM_COMMON_TABLE_HEADER   sHeader;
+	UCHAR											 ucTrainingLoop;
+	UCHAR											 ucReserved[3];
+	ATOM_INIT_REG_BLOCK				 asMemTrainingSetting;
+}ATOM_MEMORY_TRAINING_INFO;
+
+
+typedef struct SW_I2C_CNTL_DATA_PARAMETERS
+{
+  UCHAR    ucControl;
+  UCHAR    ucData; 
+  UCHAR    ucSatus; 
+  UCHAR    ucTemp; 
 } SW_I2C_CNTL_DATA_PARAMETERS;
 
 #define SW_I2C_CNTL_DATA_PS_ALLOCATION  SW_I2C_CNTL_DATA_PARAMETERS
 
-typedef struct _SW_I2C_IO_DATA_PARAMETERS {
-	USHORT GPIO_Info;
-	UCHAR ucAct;
-	UCHAR ucData;
-} SW_I2C_IO_DATA_PARAMETERS;
+typedef struct _SW_I2C_IO_DATA_PARAMETERS
+{                               
+  USHORT   GPIO_Info;
+  UCHAR    ucAct; 
+  UCHAR    ucData; 
+ } SW_I2C_IO_DATA_PARAMETERS;
 
 #define SW_I2C_IO_DATA_PS_ALLOCATION  SW_I2C_IO_DATA_PARAMETERS
 
@@ -4087,127 +4998,136 @@ typedef struct _SW_I2C_IO_DATA_PARAMETERS {
 #define SW_I2C_CNTL_CLOSE     5
 #define SW_I2C_CNTL_WRITE1BIT 6
 
-/* ==============================VESA definition Portion=============================== */
+//==============================VESA definition Portion===============================
 #define VESA_OEM_PRODUCT_REV			            '01.00'
-#define VESA_MODE_ATTRIBUTE_MODE_SUPPORT	     0xBB	/* refer to VBE spec p.32, no TTY support */
+#define VESA_MODE_ATTRIBUTE_MODE_SUPPORT	     0xBB	//refer to VBE spec p.32, no TTY support
 #define VESA_MODE_WIN_ATTRIBUTE						     7
 #define VESA_WIN_SIZE											     64
 
-typedef struct _PTR_32_BIT_STRUCTURE {
-	USHORT Offset16;
-	USHORT Segment16;
+typedef struct _PTR_32_BIT_STRUCTURE
+{
+	USHORT	Offset16;			
+	USHORT	Segment16;				
 } PTR_32_BIT_STRUCTURE;
 
-typedef union _PTR_32_BIT_UNION {
-	PTR_32_BIT_STRUCTURE SegmentOffset;
-	ULONG Ptr32_Bit;
+typedef union _PTR_32_BIT_UNION
+{
+	PTR_32_BIT_STRUCTURE	SegmentOffset;
+	ULONG					        Ptr32_Bit;
 } PTR_32_BIT_UNION;
 
-typedef struct _VBE_1_2_INFO_BLOCK_UPDATABLE {
-	UCHAR VbeSignature[4];
-	USHORT VbeVersion;
-	PTR_32_BIT_UNION OemStringPtr;
-	UCHAR Capabilities[4];
-	PTR_32_BIT_UNION VideoModePtr;
-	USHORT TotalMemory;
+typedef struct _VBE_1_2_INFO_BLOCK_UPDATABLE
+{
+	UCHAR				      VbeSignature[4];
+	USHORT				    VbeVersion;
+	PTR_32_BIT_UNION	OemStringPtr;
+	UCHAR				      Capabilities[4];
+	PTR_32_BIT_UNION	VideoModePtr;
+	USHORT				    TotalMemory;
 } VBE_1_2_INFO_BLOCK_UPDATABLE;
 
-typedef struct _VBE_2_0_INFO_BLOCK_UPDATABLE {
-	VBE_1_2_INFO_BLOCK_UPDATABLE CommonBlock;
-	USHORT OemSoftRev;
-	PTR_32_BIT_UNION OemVendorNamePtr;
-	PTR_32_BIT_UNION OemProductNamePtr;
-	PTR_32_BIT_UNION OemProductRevPtr;
+
+typedef struct _VBE_2_0_INFO_BLOCK_UPDATABLE
+{
+	VBE_1_2_INFO_BLOCK_UPDATABLE	CommonBlock;
+	USHORT							    OemSoftRev;
+	PTR_32_BIT_UNION				OemVendorNamePtr;
+	PTR_32_BIT_UNION				OemProductNamePtr;
+	PTR_32_BIT_UNION				OemProductRevPtr;
 } VBE_2_0_INFO_BLOCK_UPDATABLE;
 
-typedef union _VBE_VERSION_UNION {
-	VBE_2_0_INFO_BLOCK_UPDATABLE VBE_2_0_InfoBlock;
-	VBE_1_2_INFO_BLOCK_UPDATABLE VBE_1_2_InfoBlock;
+typedef union _VBE_VERSION_UNION
+{
+	VBE_2_0_INFO_BLOCK_UPDATABLE	VBE_2_0_InfoBlock;
+	VBE_1_2_INFO_BLOCK_UPDATABLE	VBE_1_2_InfoBlock;
 } VBE_VERSION_UNION;
 
-typedef struct _VBE_INFO_BLOCK {
-	VBE_VERSION_UNION UpdatableVBE_Info;
-	UCHAR Reserved[222];
-	UCHAR OemData[256];
+typedef struct _VBE_INFO_BLOCK
+{
+	VBE_VERSION_UNION			UpdatableVBE_Info;
+	UCHAR						      Reserved[222];
+	UCHAR						      OemData[256];
 } VBE_INFO_BLOCK;
 
-typedef struct _VBE_FP_INFO {
-	USHORT HSize;
-	USHORT VSize;
-	USHORT FPType;
-	UCHAR RedBPP;
-	UCHAR GreenBPP;
-	UCHAR BlueBPP;
-	UCHAR ReservedBPP;
-	ULONG RsvdOffScrnMemSize;
-	ULONG RsvdOffScrnMEmPtr;
-	UCHAR Reserved[14];
+typedef struct _VBE_FP_INFO
+{
+  USHORT	HSize;
+	USHORT	VSize;
+	USHORT	FPType;
+	UCHAR		RedBPP;
+	UCHAR		GreenBPP;
+	UCHAR		BlueBPP;
+	UCHAR		ReservedBPP;
+	ULONG		RsvdOffScrnMemSize;
+	ULONG		RsvdOffScrnMEmPtr;
+	UCHAR		Reserved[14];
 } VBE_FP_INFO;
 
-typedef struct _VESA_MODE_INFO_BLOCK {
-/*  Mandatory information for all VBE revisions */
-	USHORT ModeAttributes;	/*                  dw      ?       ; mode attributes */
-	UCHAR WinAAttributes;	/*                    db      ?       ; window A attributes */
-	UCHAR WinBAttributes;	/*                    db      ?       ; window B attributes */
-	USHORT WinGranularity;	/*                    dw      ?       ; window granularity */
-	USHORT WinSize;		/*                    dw      ?       ; window size */
-	USHORT WinASegment;	/*                    dw      ?       ; window A start segment */
-	USHORT WinBSegment;	/*                    dw      ?       ; window B start segment */
-	ULONG WinFuncPtr;	/*                    dd      ?       ; real mode pointer to window function */
-	USHORT BytesPerScanLine;	/*                    dw      ?       ; bytes per scan line */
-
-/* ; Mandatory information for VBE 1.2 and above */
-	USHORT XResolution;	/*                         dw      ?       ; horizontal resolution in pixels or characters */
-	USHORT YResolution;	/*                   dw      ?       ; vertical resolution in pixels or characters */
-	UCHAR XCharSize;	/*                   db      ?       ; character cell width in pixels */
-	UCHAR YCharSize;	/*                   db      ?       ; character cell height in pixels */
-	UCHAR NumberOfPlanes;	/*                   db      ?       ; number of memory planes */
-	UCHAR BitsPerPixel;	/*                   db      ?       ; bits per pixel */
-	UCHAR NumberOfBanks;	/*                   db      ?       ; number of banks */
-	UCHAR MemoryModel;	/*                   db      ?       ; memory model type */
-	UCHAR BankSize;		/*                   db      ?       ; bank size in KB */
-	UCHAR NumberOfImagePages;	/*            db    ?       ; number of images */
-	UCHAR ReservedForPageFunction;	/* db  1       ; reserved for page function */
-
-/* ; Direct Color fields(required for direct/6 and YUV/7 memory models) */
-	UCHAR RedMaskSize;	/*           db      ?       ; size of direct color red mask in bits */
-	UCHAR RedFieldPosition;	/*           db      ?       ; bit position of lsb of red mask */
-	UCHAR GreenMaskSize;	/*           db      ?       ; size of direct color green mask in bits */
-	UCHAR GreenFieldPosition;	/*           db      ?       ; bit position of lsb of green mask */
-	UCHAR BlueMaskSize;	/*           db      ?       ; size of direct color blue mask in bits */
-	UCHAR BlueFieldPosition;	/*           db      ?       ; bit position of lsb of blue mask */
-	UCHAR RsvdMaskSize;	/*           db      ?       ; size of direct color reserved mask in bits */
-	UCHAR RsvdFieldPosition;	/*           db      ?       ; bit position of lsb of reserved mask */
-	UCHAR DirectColorModeInfo;	/*           db      ?       ; direct color mode attributes */
-
-/* ; Mandatory information for VBE 2.0 and above */
-	ULONG PhysBasePtr;	/*           dd      ?       ; physical address for flat memory frame buffer */
-	ULONG Reserved_1;	/*           dd      0       ; reserved - always set to 0 */
-	USHORT Reserved_2;	/*     dw    0       ; reserved - always set to 0 */
-
-/* ; Mandatory information for VBE 3.0 and above */
-	USHORT LinBytesPerScanLine;	/*         dw      ?       ; bytes per scan line for linear modes */
-	UCHAR BnkNumberOfImagePages;	/*         db      ?       ; number of images for banked modes */
-	UCHAR LinNumberOfImagPages;	/*         db      ?       ; number of images for linear modes */
-	UCHAR LinRedMaskSize;	/*         db      ?       ; size of direct color red mask(linear modes) */
-	UCHAR LinRedFieldPosition;	/*         db      ?       ; bit position of lsb of red mask(linear modes) */
-	UCHAR LinGreenMaskSize;	/*         db      ?       ; size of direct color green mask(linear modes) */
-	UCHAR LinGreenFieldPosition;	/*         db      ?       ; bit position of lsb of green mask(linear modes) */
-	UCHAR LinBlueMaskSize;	/*         db      ?       ; size of direct color blue mask(linear modes) */
-	UCHAR LinBlueFieldPosition;	/*         db      ?       ; bit position of lsb of blue mask(linear modes) */
-	UCHAR LinRsvdMaskSize;	/*         db      ?       ; size of direct color reserved mask(linear modes) */
-	UCHAR LinRsvdFieldPosition;	/*         db      ?       ; bit position of lsb of reserved mask(linear modes) */
-	ULONG MaxPixelClock;	/*         dd      ?       ; maximum pixel clock(in Hz) for graphics mode */
-	UCHAR Reserved;		/*         db      190 dup (0) */
+typedef struct _VESA_MODE_INFO_BLOCK
+{
+// Mandatory information for all VBE revisions
+  USHORT    ModeAttributes;  //			dw	?	; mode attributes
+	UCHAR     WinAAttributes;  //			db	?	; window A attributes
+	UCHAR     WinBAttributes;  //			db	?	; window B attributes
+	USHORT    WinGranularity;  //			dw	?	; window granularity
+	USHORT    WinSize;         //			dw	?	; window size
+	USHORT    WinASegment;     //			dw	?	; window A start segment
+	USHORT    WinBSegment;     //			dw	?	; window B start segment
+	ULONG     WinFuncPtr;      //			dd	?	; real mode pointer to window function
+	USHORT    BytesPerScanLine;//			dw	?	; bytes per scan line
+
+//; Mandatory information for VBE 1.2 and above
+  USHORT    XResolution;      //			dw	?	; horizontal resolution in pixels or characters
+	USHORT    YResolution;      //			dw	?	; vertical resolution in pixels or characters
+	UCHAR     XCharSize;        //			db	?	; character cell width in pixels
+	UCHAR     YCharSize;        //			db	?	; character cell height in pixels
+	UCHAR     NumberOfPlanes;   //			db	?	; number of memory planes
+	UCHAR     BitsPerPixel;     //			db	?	; bits per pixel
+	UCHAR     NumberOfBanks;    //			db	?	; number of banks
+	UCHAR     MemoryModel;      //			db	?	; memory model type
+	UCHAR     BankSize;         //			db	?	; bank size in KB
+	UCHAR     NumberOfImagePages;//		  db	?	; number of images
+	UCHAR     ReservedForPageFunction;//db	1	; reserved for page function
+
+//; Direct Color fields(required for direct/6 and YUV/7 memory models)
+	UCHAR			RedMaskSize;        //		db	?	; size of direct color red mask in bits
+	UCHAR			RedFieldPosition;   //		db	?	; bit position of lsb of red mask
+	UCHAR			GreenMaskSize;      //		db	?	; size of direct color green mask in bits
+	UCHAR			GreenFieldPosition; //		db	?	; bit position of lsb of green mask
+	UCHAR			BlueMaskSize;       //		db	?	; size of direct color blue mask in bits
+	UCHAR			BlueFieldPosition;  //		db	?	; bit position of lsb of blue mask
+	UCHAR			RsvdMaskSize;       //		db	?	; size of direct color reserved mask in bits
+	UCHAR			RsvdFieldPosition;  //		db	?	; bit position of lsb of reserved mask
+	UCHAR			DirectColorModeInfo;//		db	?	; direct color mode attributes
+
+//; Mandatory information for VBE 2.0 and above
+	ULONG			PhysBasePtr;        //		dd	?	; physical address for flat memory frame buffer
+	ULONG			Reserved_1;         //		dd	0	; reserved - always set to 0
+	USHORT		Reserved_2;         //	  dw	0	; reserved - always set to 0
+
+//; Mandatory information for VBE 3.0 and above
+	USHORT		LinBytesPerScanLine;  //	dw	?	; bytes per scan line for linear modes
+	UCHAR			BnkNumberOfImagePages;//	db	?	; number of images for banked modes
+	UCHAR			LinNumberOfImagPages; //	db	?	; number of images for linear modes
+	UCHAR			LinRedMaskSize;       //	db	?	; size of direct color red mask(linear modes)
+	UCHAR			LinRedFieldPosition;  //	db	?	; bit position of lsb of red mask(linear modes)
+	UCHAR			LinGreenMaskSize;     //	db	?	; size of direct color green mask(linear modes)
+	UCHAR			LinGreenFieldPosition;//	db	?	; bit position of lsb of green mask(linear modes)
+	UCHAR			LinBlueMaskSize;      //	db	?	; size of direct color blue mask(linear modes)
+	UCHAR			LinBlueFieldPosition; //	db	?	; bit position of lsb of blue mask(linear modes)
+	UCHAR			LinRsvdMaskSize;      //	db	?	; size of direct color reserved mask(linear modes)
+	UCHAR			LinRsvdFieldPosition; //	db	?	; bit position of lsb of reserved mask(linear modes)
+	ULONG			MaxPixelClock;        //	dd	?	; maximum pixel clock(in Hz) for graphics mode
+	UCHAR			Reserved;             //	db	190 dup (0)
 } VESA_MODE_INFO_BLOCK;
 
-/*  BIOS function CALLS */
-#define ATOM_BIOS_EXTENDED_FUNCTION_CODE        0xA0	/*  ATI Extended Function code */
+// BIOS function CALLS
+#define ATOM_BIOS_EXTENDED_FUNCTION_CODE        0xA0	        // ATI Extended Function code
 #define ATOM_BIOS_FUNCTION_COP_MODE             0x00
 #define ATOM_BIOS_FUNCTION_SHORT_QUERY1         0x04
 #define ATOM_BIOS_FUNCTION_SHORT_QUERY2         0x05
 #define ATOM_BIOS_FUNCTION_SHORT_QUERY3         0x06
-#define ATOM_BIOS_FUNCTION_GET_DDC              0x0B
+#define ATOM_BIOS_FUNCTION_GET_DDC              0x0B   
 #define ATOM_BIOS_FUNCTION_ASIC_DSTATE          0x0E
 #define ATOM_BIOS_FUNCTION_DEBUG_PLAY           0x0F
 #define ATOM_BIOS_FUNCTION_STV_STD              0x16
@@ -4217,100 +5137,135 @@ typedef struct _VESA_MODE_INFO_BLOCK {
 #define ATOM_BIOS_FUNCTION_PANEL_CONTROL        0x82
 #define ATOM_BIOS_FUNCTION_OLD_DEVICE_DET       0x83
 #define ATOM_BIOS_FUNCTION_OLD_DEVICE_SWITCH    0x84
-#define ATOM_BIOS_FUNCTION_HW_ICON              0x8A
+#define ATOM_BIOS_FUNCTION_HW_ICON              0x8A 
 #define ATOM_BIOS_FUNCTION_SET_CMOS             0x8B
-#define SUB_FUNCTION_UPDATE_DISPLAY_INFO        0x8000	/*  Sub function 80 */
-#define SUB_FUNCTION_UPDATE_EXPANSION_INFO      0x8100	/*  Sub function 80 */
+#define SUB_FUNCTION_UPDATE_DISPLAY_INFO        0x8000          // Sub function 80
+#define SUB_FUNCTION_UPDATE_EXPANSION_INFO      0x8100          // Sub function 80
 
 #define ATOM_BIOS_FUNCTION_DISPLAY_INFO         0x8D
 #define ATOM_BIOS_FUNCTION_DEVICE_ON_OFF        0x8E
-#define ATOM_BIOS_FUNCTION_VIDEO_STATE          0x8F
-#define ATOM_SUB_FUNCTION_GET_CRITICAL_STATE    0x0300	/*  Sub function 03 */
-#define ATOM_SUB_FUNCTION_GET_LIDSTATE          0x0700	/*  Sub function 7 */
-#define ATOM_SUB_FUNCTION_THERMAL_STATE_NOTICE  0x1400	/*  Notify caller the current thermal state */
-#define ATOM_SUB_FUNCTION_CRITICAL_STATE_NOTICE 0x8300	/*  Notify caller the current critical state */
-#define ATOM_SUB_FUNCTION_SET_LIDSTATE          0x8500	/*  Sub function 85 */
-#define ATOM_SUB_FUNCTION_GET_REQ_DISPLAY_FROM_SBIOS_MODE 0x8900	/*  Sub function 89 */
-#define ATOM_SUB_FUNCTION_INFORM_ADC_SUPPORT    0x9400	/*  Notify caller that ADC is supported */
-
-#define ATOM_BIOS_FUNCTION_VESA_DPMS            0x4F10	/*  Set DPMS */
-#define ATOM_SUB_FUNCTION_SET_DPMS              0x0001	/*  BL: Sub function 01 */
-#define ATOM_SUB_FUNCTION_GET_DPMS              0x0002	/*  BL: Sub function 02 */
-#define ATOM_PARAMETER_VESA_DPMS_ON             0x0000	/*  BH Parameter for DPMS ON. */
-#define ATOM_PARAMETER_VESA_DPMS_STANDBY        0x0100	/*  BH Parameter for DPMS STANDBY */
-#define ATOM_PARAMETER_VESA_DPMS_SUSPEND        0x0200	/*  BH Parameter for DPMS SUSPEND */
-#define ATOM_PARAMETER_VESA_DPMS_OFF            0x0400	/*  BH Parameter for DPMS OFF */
-#define ATOM_PARAMETER_VESA_DPMS_REDUCE_ON      0x0800	/*  BH Parameter for DPMS REDUCE ON (NOT SUPPORTED) */
+#define ATOM_BIOS_FUNCTION_VIDEO_STATE          0x8F 
+#define ATOM_SUB_FUNCTION_GET_CRITICAL_STATE    0x0300          // Sub function 03  
+#define ATOM_SUB_FUNCTION_GET_LIDSTATE          0x0700          // Sub function 7
+#define ATOM_SUB_FUNCTION_THERMAL_STATE_NOTICE  0x1400          // Notify caller the current thermal state
+#define ATOM_SUB_FUNCTION_CRITICAL_STATE_NOTICE 0x8300          // Notify caller the current critical state
+#define ATOM_SUB_FUNCTION_SET_LIDSTATE          0x8500          // Sub function 85
+#define ATOM_SUB_FUNCTION_GET_REQ_DISPLAY_FROM_SBIOS_MODE 0x8900// Sub function 89
+#define ATOM_SUB_FUNCTION_INFORM_ADC_SUPPORT    0x9400          // Notify caller that ADC is supported
+     
+
+#define ATOM_BIOS_FUNCTION_VESA_DPMS            0x4F10          // Set DPMS 
+#define ATOM_SUB_FUNCTION_SET_DPMS              0x0001          // BL: Sub function 01 
+#define ATOM_SUB_FUNCTION_GET_DPMS              0x0002          // BL: Sub function 02 
+#define ATOM_PARAMETER_VESA_DPMS_ON             0x0000          // BH Parameter for DPMS ON.  
+#define ATOM_PARAMETER_VESA_DPMS_STANDBY        0x0100          // BH Parameter for DPMS STANDBY  
+#define ATOM_PARAMETER_VESA_DPMS_SUSPEND        0x0200          // BH Parameter for DPMS SUSPEND
+#define ATOM_PARAMETER_VESA_DPMS_OFF            0x0400          // BH Parameter for DPMS OFF
+#define ATOM_PARAMETER_VESA_DPMS_REDUCE_ON      0x0800          // BH Parameter for DPMS REDUCE ON (NOT SUPPORTED)
 
 #define ATOM_BIOS_RETURN_CODE_MASK              0x0000FF00L
 #define ATOM_BIOS_REG_HIGH_MASK                 0x0000FF00L
 #define ATOM_BIOS_REG_LOW_MASK                  0x000000FFL
 
-/*  structure used for VBIOS only */
+// structure used for VBIOS only
 
-/* DispOutInfoTable */
-typedef struct _ASIC_TRANSMITTER_INFO {
+//DispOutInfoTable
+typedef struct _ASIC_TRANSMITTER_INFO
+{
 	USHORT usTransmitterObjId;
 	USHORT usSupportDevice;
-	UCHAR ucTransmitterCmdTblId;
-	UCHAR ucConfig;
-	UCHAR ucEncoderID;	/* available 1st encoder ( default ) */
-	UCHAR ucOptionEncoderID;	/* available 2nd encoder ( optional ) */
-	UCHAR uc2ndEncoderID;
-	UCHAR ucReserved;
-} ASIC_TRANSMITTER_INFO;
-
-typedef struct _ASIC_ENCODER_INFO {
+  UCHAR  ucTransmitterCmdTblId;
+	UCHAR  ucConfig;
+	UCHAR  ucEncoderID;					 //available 1st encoder ( default )
+	UCHAR  ucOptionEncoderID;    //available 2nd encoder ( optional )
+	UCHAR  uc2ndEncoderID;
+	UCHAR  ucReserved;
+}ASIC_TRANSMITTER_INFO;
+
+typedef struct _ASIC_ENCODER_INFO
+{
 	UCHAR ucEncoderID;
 	UCHAR ucEncoderConfig;
-	USHORT usEncoderCmdTblId;
-} ASIC_ENCODER_INFO;
+  USHORT usEncoderCmdTblId;
+}ASIC_ENCODER_INFO;
 
-typedef struct _ATOM_DISP_OUT_INFO {
-	ATOM_COMMON_TABLE_HEADER sHeader;
+typedef struct _ATOM_DISP_OUT_INFO
+{
+  ATOM_COMMON_TABLE_HEADER sHeader;  
+	USHORT ptrTransmitterInfo;
+	USHORT ptrEncoderInfo;
+	ASIC_TRANSMITTER_INFO  asTransmitterInfo[1];
+	ASIC_ENCODER_INFO      asEncoderInfo[1];
+}ATOM_DISP_OUT_INFO;
+
+typedef struct _ATOM_DISP_OUT_INFO_V2
+{
+  ATOM_COMMON_TABLE_HEADER sHeader;  
 	USHORT ptrTransmitterInfo;
 	USHORT ptrEncoderInfo;
-	ASIC_TRANSMITTER_INFO asTransmitterInfo[1];
-	ASIC_ENCODER_INFO asEncoderInfo[1];
-} ATOM_DISP_OUT_INFO;
+  USHORT ptrMainCallParserFar;                  // direct address of main parser call in VBIOS binary. 
+	ASIC_TRANSMITTER_INFO  asTransmitterInfo[1];
+	ASIC_ENCODER_INFO      asEncoderInfo[1];
+}ATOM_DISP_OUT_INFO_V2;
 
-/*  DispDevicePriorityInfo */
-typedef struct _ATOM_DISPLAY_DEVICE_PRIORITY_INFO {
-	ATOM_COMMON_TABLE_HEADER sHeader;
+// DispDevicePriorityInfo
+typedef struct _ATOM_DISPLAY_DEVICE_PRIORITY_INFO
+{
+  ATOM_COMMON_TABLE_HEADER sHeader;  
 	USHORT asDevicePriority[16];
-} ATOM_DISPLAY_DEVICE_PRIORITY_INFO;
-
-/* ProcessAuxChannelTransactionTable */
-typedef struct _PROCESS_AUX_CHANNEL_TRANSACTION_PARAMETERS {
-	USHORT lpAuxRequest;
-	USHORT lpDataOut;
-	UCHAR ucChannelID;
-	union {
-		UCHAR ucReplyStatus;
-		UCHAR ucDelay;
+}ATOM_DISPLAY_DEVICE_PRIORITY_INFO;
+
+//ProcessAuxChannelTransactionTable
+typedef struct _PROCESS_AUX_CHANNEL_TRANSACTION_PARAMETERS
+{
+	USHORT	lpAuxRequest;
+	USHORT  lpDataOut;
+	UCHAR		ucChannelID;
+	union
+	{
+  UCHAR   ucReplyStatus;
+	UCHAR   ucDelay;
+	};
+  UCHAR   ucDataOutLen;
+	UCHAR   ucReserved;
+}PROCESS_AUX_CHANNEL_TRANSACTION_PARAMETERS;
+
+//ProcessAuxChannelTransactionTable
+typedef struct _PROCESS_AUX_CHANNEL_TRANSACTION_PARAMETERS_V2
+{
+	USHORT	lpAuxRequest;
+	USHORT  lpDataOut;
+	UCHAR		ucChannelID;
+	union
+	{
+  UCHAR   ucReplyStatus;
+	UCHAR   ucDelay;
 	};
-	UCHAR ucDataOutLen;
-	UCHAR ucReserved;
-} PROCESS_AUX_CHANNEL_TRANSACTION_PARAMETERS;
+  UCHAR   ucDataOutLen;
+	UCHAR   ucHPD_ID;                                       //=0: HPD1, =1: HPD2, =2: HPD3, =3: HPD4, =4: HPD5, =5: HPD6
+}PROCESS_AUX_CHANNEL_TRANSACTION_PARAMETERS_V2;
 
 #define PROCESS_AUX_CHANNEL_TRANSACTION_PS_ALLOCATION			PROCESS_AUX_CHANNEL_TRANSACTION_PARAMETERS
 
-/* GetSinkType */
+//GetSinkType
 
-typedef struct _DP_ENCODER_SERVICE_PARAMETERS {
+typedef struct _DP_ENCODER_SERVICE_PARAMETERS
+{
 	USHORT ucLinkClock;
-	union {
-		UCHAR ucConfig;	/*  for DP training command */
-		UCHAR ucI2cId;	/*  use for GET_SINK_TYPE command */
+	union 
+	{
+	UCHAR ucConfig;				// for DP training command
+	UCHAR ucI2cId;				// use for GET_SINK_TYPE command
 	};
 	UCHAR ucAction;
 	UCHAR ucStatus;
 	UCHAR ucLaneNum;
 	UCHAR ucReserved[2];
-} DP_ENCODER_SERVICE_PARAMETERS;
+}DP_ENCODER_SERVICE_PARAMETERS;
 
-/*  ucAction */
+// ucAction
 #define ATOM_DP_ACTION_GET_SINK_TYPE							0x01
+/* obselete */
 #define ATOM_DP_ACTION_TRAINING_START							0x02
 #define ATOM_DP_ACTION_TRAINING_COMPLETE					0x03
 #define ATOM_DP_ACTION_TRAINING_PATTERN_SEL				0x04
@@ -4318,7 +5273,7 @@ typedef struct _DP_ENCODER_SERVICE_PARAMETERS {
 #define ATOM_DP_ACTION_GET_VSWING_PREEMP					0x06
 #define ATOM_DP_ACTION_BLANKING                   0x07
 
-/*  ucConfig */
+// ucConfig
 #define ATOM_DP_CONFIG_ENCODER_SEL_MASK						0x03
 #define ATOM_DP_CONFIG_DIG1_ENCODER								0x00
 #define ATOM_DP_CONFIG_DIG2_ENCODER								0x01
@@ -4326,14 +5281,14 @@ typedef struct _DP_ENCODER_SERVICE_PARAMETERS {
 #define ATOM_DP_CONFIG_LINK_SEL_MASK							0x04
 #define ATOM_DP_CONFIG_LINK_A											0x00
 #define ATOM_DP_CONFIG_LINK_B											0x04
-
+/* /obselete */
 #define DP_ENCODER_SERVICE_PS_ALLOCATION				WRITE_ONE_BYTE_HW_I2C_DATA_PARAMETERS
 
-/*  DP_TRAINING_TABLE */
-#define DPCD_SET_LINKRATE_LANENUM_PATTERN1_TBL_ADDR				ATOM_DP_TRAINING_TBL_ADDR
+// DP_TRAINING_TABLE
+#define DPCD_SET_LINKRATE_LANENUM_PATTERN1_TBL_ADDR				ATOM_DP_TRAINING_TBL_ADDR		
 #define DPCD_SET_SS_CNTL_TBL_ADDR													(ATOM_DP_TRAINING_TBL_ADDR + 8 )
-#define DPCD_SET_LANE_VSWING_PREEMP_TBL_ADDR							(ATOM_DP_TRAINING_TBL_ADDR + 16)
-#define DPCD_SET_TRAINING_PATTERN0_TBL_ADDR								(ATOM_DP_TRAINING_TBL_ADDR + 24)
+#define DPCD_SET_LANE_VSWING_PREEMP_TBL_ADDR							(ATOM_DP_TRAINING_TBL_ADDR + 16 )
+#define DPCD_SET_TRAINING_PATTERN0_TBL_ADDR								(ATOM_DP_TRAINING_TBL_ADDR + 24 )
 #define DPCD_SET_TRAINING_PATTERN2_TBL_ADDR								(ATOM_DP_TRAINING_TBL_ADDR + 32)
 #define DPCD_GET_LINKRATE_LANENUM_SS_TBL_ADDR							(ATOM_DP_TRAINING_TBL_ADDR + 40)
 #define	DPCD_GET_LANE_STATUS_ADJUST_TBL_ADDR							(ATOM_DP_TRAINING_TBL_ADDR + 48)
@@ -4341,183 +5296,241 @@ typedef struct _DP_ENCODER_SERVICE_PARAMETERS {
 #define DP_I2C_AUX_DDC_WRITE_TBL_ADDR											(ATOM_DP_TRAINING_TBL_ADDR + 64)
 #define DP_I2C_AUX_DDC_READ_START_TBL_ADDR								(ATOM_DP_TRAINING_TBL_ADDR + 72)
 #define DP_I2C_AUX_DDC_READ_TBL_ADDR											(ATOM_DP_TRAINING_TBL_ADDR + 76)
-#define DP_I2C_AUX_DDC_READ_END_TBL_ADDR									(ATOM_DP_TRAINING_TBL_ADDR + 80)
+#define DP_I2C_AUX_DDC_WRITE_END_TBL_ADDR                 (ATOM_DP_TRAINING_TBL_ADDR + 80) 
+#define DP_I2C_AUX_DDC_READ_END_TBL_ADDR									(ATOM_DP_TRAINING_TBL_ADDR + 84)
 
-typedef struct _PROCESS_I2C_CHANNEL_TRANSACTION_PARAMETERS {
-	UCHAR ucI2CSpeed;
-	union {
-		UCHAR ucRegIndex;
-		UCHAR ucStatus;
+typedef struct _PROCESS_I2C_CHANNEL_TRANSACTION_PARAMETERS
+{
+	UCHAR   ucI2CSpeed;
+ 	union
+	{
+   UCHAR ucRegIndex;
+   UCHAR ucStatus;
 	};
-	USHORT lpI2CDataOut;
-	UCHAR ucFlag;
-	UCHAR ucTransBytes;
-	UCHAR ucSlaveAddr;
-	UCHAR ucLineNumber;
-} PROCESS_I2C_CHANNEL_TRANSACTION_PARAMETERS;
+	USHORT  lpI2CDataOut;
+  UCHAR   ucFlag;               
+  UCHAR   ucTransBytes;
+  UCHAR   ucSlaveAddr;
+  UCHAR   ucLineNumber;
+}PROCESS_I2C_CHANNEL_TRANSACTION_PARAMETERS;
 
 #define PROCESS_I2C_CHANNEL_TRANSACTION_PS_ALLOCATION       PROCESS_I2C_CHANNEL_TRANSACTION_PARAMETERS
 
-/* ucFlag */
+//ucFlag
 #define HW_I2C_WRITE        1
 #define HW_I2C_READ         0
+#define I2C_2BYTE_ADDR      0x02
 
+typedef struct _SET_HWBLOCK_INSTANCE_PARAMETER_V2
+{
+   UCHAR ucHWBlkInst;                // HW block instance, 0, 1, 2, ...
+   UCHAR ucReserved[3]; 
+}SET_HWBLOCK_INSTANCE_PARAMETER_V2;
+
+#define HWBLKINST_INSTANCE_MASK       0x07
+#define HWBLKINST_HWBLK_MASK          0xF0
+#define HWBLKINST_HWBLK_SHIFT         0x04
+
+//ucHWBlock
+#define SELECT_DISP_ENGINE            0
+#define SELECT_DISP_PLL               1
+#define SELECT_DCIO_UNIPHY_LINK0      2
+#define SELECT_DCIO_UNIPHY_LINK1      3
+#define SELECT_DCIO_IMPCAL            4
+#define SELECT_DCIO_DIG               6
+#define SELECT_CRTC_PIXEL_RATE        7
+
+/****************************************************************************/	
+//Portion VI: Definitinos for vbios MC scratch registers that driver used
 /****************************************************************************/
-/* Portion VI: Definitinos being oboselete */
+
+#define MC_MISC0__MEMORY_TYPE_MASK    0xF0000000
+#define MC_MISC0__MEMORY_TYPE__GDDR1  0x10000000
+#define MC_MISC0__MEMORY_TYPE__DDR2   0x20000000
+#define MC_MISC0__MEMORY_TYPE__GDDR3  0x30000000
+#define MC_MISC0__MEMORY_TYPE__GDDR4  0x40000000
+#define MC_MISC0__MEMORY_TYPE__GDDR5  0x50000000
+#define MC_MISC0__MEMORY_TYPE__DDR3   0xB0000000
+
+/****************************************************************************/	
+//Portion VI: Definitinos being oboselete
 /****************************************************************************/
 
-/* ========================================================================================== */
-/* Remove the definitions below when driver is ready! */
-typedef struct _ATOM_DAC_INFO {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	USHORT usMaxFrequency;	/*  in 10kHz unit */
-	USHORT usReserved;
-} ATOM_DAC_INFO;
-
-typedef struct _COMPASSIONATE_DATA {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-
-	/* ==============================  DAC1 portion */
-	UCHAR ucDAC1_BG_Adjustment;
-	UCHAR ucDAC1_DAC_Adjustment;
-	USHORT usDAC1_FORCE_Data;
-	/* ==============================  DAC2 portion */
-	UCHAR ucDAC2_CRT2_BG_Adjustment;
-	UCHAR ucDAC2_CRT2_DAC_Adjustment;
-	USHORT usDAC2_CRT2_FORCE_Data;
-	USHORT usDAC2_CRT2_MUX_RegisterIndex;
-	UCHAR ucDAC2_CRT2_MUX_RegisterInfo;	/* Bit[4:0]=Bit position,Bit[7]=1:Active High;=0 Active Low */
-	UCHAR ucDAC2_NTSC_BG_Adjustment;
-	UCHAR ucDAC2_NTSC_DAC_Adjustment;
-	USHORT usDAC2_TV1_FORCE_Data;
-	USHORT usDAC2_TV1_MUX_RegisterIndex;
-	UCHAR ucDAC2_TV1_MUX_RegisterInfo;	/* Bit[4:0]=Bit position,Bit[7]=1:Active High;=0 Active Low */
-	UCHAR ucDAC2_CV_BG_Adjustment;
-	UCHAR ucDAC2_CV_DAC_Adjustment;
-	USHORT usDAC2_CV_FORCE_Data;
-	USHORT usDAC2_CV_MUX_RegisterIndex;
-	UCHAR ucDAC2_CV_MUX_RegisterInfo;	/* Bit[4:0]=Bit position,Bit[7]=1:Active High;=0 Active Low */
-	UCHAR ucDAC2_PAL_BG_Adjustment;
-	UCHAR ucDAC2_PAL_DAC_Adjustment;
-	USHORT usDAC2_TV2_FORCE_Data;
-} COMPASSIONATE_DATA;
+//==========================================================================================
+//Remove the definitions below when driver is ready!
+typedef struct _ATOM_DAC_INFO
+{
+  ATOM_COMMON_TABLE_HEADER sHeader;  
+  USHORT                   usMaxFrequency;      // in 10kHz unit
+  USHORT                   usReserved;
+}ATOM_DAC_INFO;
+
+
+typedef struct  _COMPASSIONATE_DATA           
+{
+  ATOM_COMMON_TABLE_HEADER sHeader; 
+
+  //==============================  DAC1 portion
+  UCHAR   ucDAC1_BG_Adjustment;
+  UCHAR   ucDAC1_DAC_Adjustment;
+  USHORT  usDAC1_FORCE_Data;
+  //==============================  DAC2 portion
+  UCHAR   ucDAC2_CRT2_BG_Adjustment;
+  UCHAR   ucDAC2_CRT2_DAC_Adjustment;
+  USHORT  usDAC2_CRT2_FORCE_Data;
+  USHORT  usDAC2_CRT2_MUX_RegisterIndex;
+  UCHAR   ucDAC2_CRT2_MUX_RegisterInfo;     //Bit[4:0]=Bit position,Bit[7]=1:Active High;=0 Active Low
+  UCHAR   ucDAC2_NTSC_BG_Adjustment;
+  UCHAR   ucDAC2_NTSC_DAC_Adjustment;
+  USHORT  usDAC2_TV1_FORCE_Data;
+  USHORT  usDAC2_TV1_MUX_RegisterIndex;
+  UCHAR   ucDAC2_TV1_MUX_RegisterInfo;      //Bit[4:0]=Bit position,Bit[7]=1:Active High;=0 Active Low
+  UCHAR   ucDAC2_CV_BG_Adjustment;
+  UCHAR   ucDAC2_CV_DAC_Adjustment;
+  USHORT  usDAC2_CV_FORCE_Data;
+  USHORT  usDAC2_CV_MUX_RegisterIndex;
+  UCHAR   ucDAC2_CV_MUX_RegisterInfo;       //Bit[4:0]=Bit position,Bit[7]=1:Active High;=0 Active Low
+  UCHAR   ucDAC2_PAL_BG_Adjustment;
+  UCHAR   ucDAC2_PAL_DAC_Adjustment;
+  USHORT  usDAC2_TV2_FORCE_Data;
+}COMPASSIONATE_DATA;
 
 /****************************Supported Device Info Table Definitions**********************/
-/*   ucConnectInfo: */
-/*     [7:4] - connector type */
-/*       = 1   - VGA connector */
-/*       = 2   - DVI-I */
-/*       = 3   - DVI-D */
-/*       = 4   - DVI-A */
-/*       = 5   - SVIDEO */
-/*       = 6   - COMPOSITE */
-/*       = 7   - LVDS */
-/*       = 8   - DIGITAL LINK */
-/*       = 9   - SCART */
-/*       = 0xA - HDMI_type A */
-/*       = 0xB - HDMI_type B */
-/*       = 0xE - Special case1 (DVI+DIN) */
-/*       Others=TBD */
-/*     [3:0] - DAC Associated */
-/*       = 0   - no DAC */
-/*       = 1   - DACA */
-/*       = 2   - DACB */
-/*       = 3   - External DAC */
-/*       Others=TBD */
-/*  */
-
-typedef struct _ATOM_CONNECTOR_INFO {
+//  ucConnectInfo:
+//    [7:4] - connector type
+//      = 1   - VGA connector   
+//      = 2   - DVI-I
+//      = 3   - DVI-D
+//      = 4   - DVI-A
+//      = 5   - SVIDEO
+//      = 6   - COMPOSITE
+//      = 7   - LVDS
+//      = 8   - DIGITAL LINK
+//      = 9   - SCART
+//      = 0xA - HDMI_type A
+//      = 0xB - HDMI_type B
+//      = 0xE - Special case1 (DVI+DIN)
+//      Others=TBD
+//    [3:0] - DAC Associated
+//      = 0   - no DAC
+//      = 1   - DACA
+//      = 2   - DACB
+//      = 3   - External DAC
+//      Others=TBD
+//    
+
+typedef struct _ATOM_CONNECTOR_INFO
+{
 #if ATOM_BIG_ENDIAN
-	UCHAR bfConnectorType:4;
-	UCHAR bfAssociatedDAC:4;
+  UCHAR   bfConnectorType:4;
+  UCHAR   bfAssociatedDAC:4;
 #else
-	UCHAR bfAssociatedDAC:4;
-	UCHAR bfConnectorType:4;
+  UCHAR   bfAssociatedDAC:4;
+  UCHAR   bfConnectorType:4;
 #endif
-} ATOM_CONNECTOR_INFO;
+}ATOM_CONNECTOR_INFO;
 
-typedef union _ATOM_CONNECTOR_INFO_ACCESS {
-	ATOM_CONNECTOR_INFO sbfAccess;
-	UCHAR ucAccess;
-} ATOM_CONNECTOR_INFO_ACCESS;
+typedef union _ATOM_CONNECTOR_INFO_ACCESS
+{
+  ATOM_CONNECTOR_INFO sbfAccess;
+  UCHAR               ucAccess;
+}ATOM_CONNECTOR_INFO_ACCESS;
+
+typedef struct _ATOM_CONNECTOR_INFO_I2C
+{
+  ATOM_CONNECTOR_INFO_ACCESS sucConnectorInfo;
+  ATOM_I2C_ID_CONFIG_ACCESS  sucI2cId;
+}ATOM_CONNECTOR_INFO_I2C;
 
-typedef struct _ATOM_CONNECTOR_INFO_I2C {
-	ATOM_CONNECTOR_INFO_ACCESS sucConnectorInfo;
-	ATOM_I2C_ID_CONFIG_ACCESS sucI2cId;
-} ATOM_CONNECTOR_INFO_I2C;
 
-typedef struct _ATOM_SUPPORTED_DEVICES_INFO {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	USHORT usDeviceSupport;
-	ATOM_CONNECTOR_INFO_I2C asConnInfo[ATOM_MAX_SUPPORTED_DEVICE_INFO];
-} ATOM_SUPPORTED_DEVICES_INFO;
+typedef struct _ATOM_SUPPORTED_DEVICES_INFO
+{ 
+  ATOM_COMMON_TABLE_HEADER	sHeader;
+  USHORT                    usDeviceSupport;
+  ATOM_CONNECTOR_INFO_I2C   asConnInfo[ATOM_MAX_SUPPORTED_DEVICE_INFO];
+}ATOM_SUPPORTED_DEVICES_INFO;
 
 #define NO_INT_SRC_MAPPED       0xFF
 
-typedef struct _ATOM_CONNECTOR_INC_SRC_BITMAP {
-	UCHAR ucIntSrcBitmap;
-} ATOM_CONNECTOR_INC_SRC_BITMAP;
-
-typedef struct _ATOM_SUPPORTED_DEVICES_INFO_2 {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	USHORT usDeviceSupport;
-	ATOM_CONNECTOR_INFO_I2C asConnInfo[ATOM_MAX_SUPPORTED_DEVICE_INFO_2];
-	ATOM_CONNECTOR_INC_SRC_BITMAP
-	    asIntSrcInfo[ATOM_MAX_SUPPORTED_DEVICE_INFO_2];
-} ATOM_SUPPORTED_DEVICES_INFO_2;
-
-typedef struct _ATOM_SUPPORTED_DEVICES_INFO_2d1 {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	USHORT usDeviceSupport;
-	ATOM_CONNECTOR_INFO_I2C asConnInfo[ATOM_MAX_SUPPORTED_DEVICE];
-	ATOM_CONNECTOR_INC_SRC_BITMAP asIntSrcInfo[ATOM_MAX_SUPPORTED_DEVICE];
-} ATOM_SUPPORTED_DEVICES_INFO_2d1;
+typedef struct _ATOM_CONNECTOR_INC_SRC_BITMAP
+{
+  UCHAR   ucIntSrcBitmap;
+}ATOM_CONNECTOR_INC_SRC_BITMAP;
+
+typedef struct _ATOM_SUPPORTED_DEVICES_INFO_2
+{ 
+  ATOM_COMMON_TABLE_HEADER      sHeader;
+  USHORT                        usDeviceSupport;
+  ATOM_CONNECTOR_INFO_I2C       asConnInfo[ATOM_MAX_SUPPORTED_DEVICE_INFO_2];
+  ATOM_CONNECTOR_INC_SRC_BITMAP asIntSrcInfo[ATOM_MAX_SUPPORTED_DEVICE_INFO_2];
+}ATOM_SUPPORTED_DEVICES_INFO_2;
+
+typedef struct _ATOM_SUPPORTED_DEVICES_INFO_2d1
+{ 
+  ATOM_COMMON_TABLE_HEADER      sHeader;
+  USHORT                        usDeviceSupport;
+  ATOM_CONNECTOR_INFO_I2C       asConnInfo[ATOM_MAX_SUPPORTED_DEVICE];
+  ATOM_CONNECTOR_INC_SRC_BITMAP asIntSrcInfo[ATOM_MAX_SUPPORTED_DEVICE];
+}ATOM_SUPPORTED_DEVICES_INFO_2d1;
 
 #define ATOM_SUPPORTED_DEVICES_INFO_LAST ATOM_SUPPORTED_DEVICES_INFO_2d1
 
-typedef struct _ATOM_MISC_CONTROL_INFO {
-	USHORT usFrequency;
-	UCHAR ucPLL_ChargePump;	/*  PLL charge-pump gain control */
-	UCHAR ucPLL_DutyCycle;	/*  PLL duty cycle control */
-	UCHAR ucPLL_VCO_Gain;	/*  PLL VCO gain control */
-	UCHAR ucPLL_VoltageSwing;	/*  PLL driver voltage swing control */
-} ATOM_MISC_CONTROL_INFO;
+
+
+typedef struct _ATOM_MISC_CONTROL_INFO
+{
+   USHORT usFrequency;
+   UCHAR  ucPLL_ChargePump;				                // PLL charge-pump gain control
+   UCHAR  ucPLL_DutyCycle;				                // PLL duty cycle control
+   UCHAR  ucPLL_VCO_Gain;				                  // PLL VCO gain control
+   UCHAR  ucPLL_VoltageSwing;			                // PLL driver voltage swing control
+}ATOM_MISC_CONTROL_INFO;  
+
 
 #define ATOM_MAX_MISC_INFO       4
 
-typedef struct _ATOM_TMDS_INFO {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	USHORT usMaxFrequency;	/*  in 10Khz */
-	ATOM_MISC_CONTROL_INFO asMiscInfo[ATOM_MAX_MISC_INFO];
-} ATOM_TMDS_INFO;
+typedef struct _ATOM_TMDS_INFO
+{
+  ATOM_COMMON_TABLE_HEADER sHeader;  
+  USHORT							usMaxFrequency;             // in 10Khz
+  ATOM_MISC_CONTROL_INFO				asMiscInfo[ATOM_MAX_MISC_INFO];
+}ATOM_TMDS_INFO;
+
 
-typedef struct _ATOM_ENCODER_ANALOG_ATTRIBUTE {
-	UCHAR ucTVStandard;	/* Same as TV standards defined above, */
-	UCHAR ucPadding[1];
-} ATOM_ENCODER_ANALOG_ATTRIBUTE;
+typedef struct _ATOM_ENCODER_ANALOG_ATTRIBUTE
+{
+  UCHAR ucTVStandard;     //Same as TV standards defined above, 
+  UCHAR ucPadding[1];
+}ATOM_ENCODER_ANALOG_ATTRIBUTE;
 
-typedef struct _ATOM_ENCODER_DIGITAL_ATTRIBUTE {
-	UCHAR ucAttribute;	/* Same as other digital encoder attributes defined above */
-	UCHAR ucPadding[1];
-} ATOM_ENCODER_DIGITAL_ATTRIBUTE;
+typedef struct _ATOM_ENCODER_DIGITAL_ATTRIBUTE
+{
+  UCHAR ucAttribute;      //Same as other digital encoder attributes defined above
+  UCHAR ucPadding[1];		
+}ATOM_ENCODER_DIGITAL_ATTRIBUTE;
 
-typedef union _ATOM_ENCODER_ATTRIBUTE {
-	ATOM_ENCODER_ANALOG_ATTRIBUTE sAlgAttrib;
-	ATOM_ENCODER_DIGITAL_ATTRIBUTE sDigAttrib;
-} ATOM_ENCODER_ATTRIBUTE;
+typedef union _ATOM_ENCODER_ATTRIBUTE
+{
+  ATOM_ENCODER_ANALOG_ATTRIBUTE sAlgAttrib;
+  ATOM_ENCODER_DIGITAL_ATTRIBUTE sDigAttrib;
+}ATOM_ENCODER_ATTRIBUTE;
 
-typedef struct _DVO_ENCODER_CONTROL_PARAMETERS {
-	USHORT usPixelClock;
-	USHORT usEncoderID;
-	UCHAR ucDeviceType;	/* Use ATOM_DEVICE_xxx1_Index to indicate device type only. */
-	UCHAR ucAction;		/* ATOM_ENABLE/ATOM_DISABLE/ATOM_HPD_INIT */
-	ATOM_ENCODER_ATTRIBUTE usDevAttr;
-} DVO_ENCODER_CONTROL_PARAMETERS;
 
-typedef struct _DVO_ENCODER_CONTROL_PS_ALLOCATION {
-	DVO_ENCODER_CONTROL_PARAMETERS sDVOEncoder;
-	WRITE_ONE_BYTE_HW_I2C_DATA_PS_ALLOCATION sReserved;	/* Caller doesn't need to init this portion */
-} DVO_ENCODER_CONTROL_PS_ALLOCATION;
+typedef struct _DVO_ENCODER_CONTROL_PARAMETERS
+{
+  USHORT usPixelClock; 
+  USHORT usEncoderID; 
+  UCHAR  ucDeviceType;												//Use ATOM_DEVICE_xxx1_Index to indicate device type only.	
+  UCHAR  ucAction;														//ATOM_ENABLE/ATOM_DISABLE/ATOM_HPD_INIT
+  ATOM_ENCODER_ATTRIBUTE usDevAttr;     		
+}DVO_ENCODER_CONTROL_PARAMETERS;
+
+typedef struct _DVO_ENCODER_CONTROL_PS_ALLOCATION
+{                               
+  DVO_ENCODER_CONTROL_PARAMETERS    sDVOEncoder;
+  WRITE_ONE_BYTE_HW_I2C_DATA_PS_ALLOCATION      sReserved;     //Caller doesn't need to init this portion
+}DVO_ENCODER_CONTROL_PS_ALLOCATION;
+
 
 #define ATOM_XTMDS_ASIC_SI164_ID        1
 #define ATOM_XTMDS_ASIC_SI178_ID        2
@@ -4526,27 +5539,30 @@ typedef struct _DVO_ENCODER_CONTROL_PS_ALLOCATION {
 #define ATOM_XTMDS_SUPPORTED_DUALLINK   0x00000002
 #define ATOM_XTMDS_MVPU_FPGA            0x00000004
 
-typedef struct _ATOM_XTMDS_INFO {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	USHORT usSingleLinkMaxFrequency;
-	ATOM_I2C_ID_CONFIG_ACCESS sucI2cId;	/* Point the ID on which I2C is used to control external chip */
-	UCHAR ucXtransimitterID;
-	UCHAR ucSupportedLink;	/*  Bit field, bit0=1, single link supported;bit1=1,dual link supported */
-	UCHAR ucSequnceAlterID;	/*  Even with the same external TMDS asic, it's possible that the program seqence alters */
-	/*  due to design. This ID is used to alert driver that the sequence is not "standard"! */
-	UCHAR ucMasterAddress;	/*  Address to control Master xTMDS Chip */
-	UCHAR ucSlaveAddress;	/*  Address to control Slave xTMDS Chip */
-} ATOM_XTMDS_INFO;
-
-typedef struct _DFP_DPMS_STATUS_CHANGE_PARAMETERS {
-	UCHAR ucEnable;		/*  ATOM_ENABLE=On or ATOM_DISABLE=Off */
-	UCHAR ucDevice;		/*  ATOM_DEVICE_DFP1_INDEX.... */
-	UCHAR ucPadding[2];
-} DFP_DPMS_STATUS_CHANGE_PARAMETERS;
+                           
+typedef struct _ATOM_XTMDS_INFO
+{
+  ATOM_COMMON_TABLE_HEADER   sHeader;  
+  USHORT                     usSingleLinkMaxFrequency; 
+  ATOM_I2C_ID_CONFIG_ACCESS  sucI2cId;           //Point the ID on which I2C is used to control external chip
+  UCHAR                      ucXtransimitterID;          
+  UCHAR                      ucSupportedLink;    // Bit field, bit0=1, single link supported;bit1=1,dual link supported
+  UCHAR                      ucSequnceAlterID;   // Even with the same external TMDS asic, it's possible that the program seqence alters 
+                                                 // due to design. This ID is used to alert driver that the sequence is not "standard"!              
+  UCHAR                      ucMasterAddress;    // Address to control Master xTMDS Chip
+  UCHAR                      ucSlaveAddress;     // Address to control Slave xTMDS Chip
+}ATOM_XTMDS_INFO;
+
+typedef struct _DFP_DPMS_STATUS_CHANGE_PARAMETERS
+{  
+  UCHAR ucEnable;                     // ATOM_ENABLE=On or ATOM_DISABLE=Off
+  UCHAR ucDevice;                     // ATOM_DEVICE_DFP1_INDEX....
+  UCHAR ucPadding[2];             
+}DFP_DPMS_STATUS_CHANGE_PARAMETERS;
 
 /****************************Legacy Power Play Table Definitions **********************/
 
-/* Definitions for ulPowerPlayMiscInfo */
+//Definitions for ulPowerPlayMiscInfo
 #define ATOM_PM_MISCINFO_SPLIT_CLOCK                     0x00000000L
 #define ATOM_PM_MISCINFO_USING_MCLK_SRC                  0x00000001L
 #define ATOM_PM_MISCINFO_USING_SCLK_SRC                  0x00000002L
@@ -4558,8 +5574,8 @@ typedef struct _DFP_DPMS_STATUS_CHANGE_PARAMETERS {
 
 #define ATOM_PM_MISCINFO_ENGINE_CLOCK_CONTRL_EN          0x00000020L
 #define ATOM_PM_MISCINFO_MEMORY_CLOCK_CONTRL_EN          0x00000040L
-#define ATOM_PM_MISCINFO_PROGRAM_VOLTAGE                 0x00000080L	/* When this bit set, ucVoltageDropIndex is not an index for GPIO pin, but a voltage ID that SW needs program */
-
+#define ATOM_PM_MISCINFO_PROGRAM_VOLTAGE                 0x00000080L  //When this bit set, ucVoltageDropIndex is not an index for GPIO pin, but a voltage ID that SW needs program  
+ 
 #define ATOM_PM_MISCINFO_ASIC_REDUCED_SPEED_SCLK_EN      0x00000100L
 #define ATOM_PM_MISCINFO_ASIC_DYNAMIC_VOLTAGE_EN         0x00000200L
 #define ATOM_PM_MISCINFO_ASIC_SLEEP_MODE_EN              0x00000400L
@@ -4569,22 +5585,22 @@ typedef struct _DFP_DPMS_STATUS_CHANGE_PARAMETERS {
 #define ATOM_PM_MISCINFO_LOW_LCD_REFRESH_RATE            0x00004000L
 
 #define ATOM_PM_MISCINFO_DRIVER_DEFAULT_MODE             0x00008000L
-#define ATOM_PM_MISCINFO_OVER_CLOCK_MODE                 0x00010000L
+#define ATOM_PM_MISCINFO_OVER_CLOCK_MODE                 0x00010000L 
 #define ATOM_PM_MISCINFO_OVER_DRIVE_MODE                 0x00020000L
 #define ATOM_PM_MISCINFO_POWER_SAVING_MODE               0x00040000L
 #define ATOM_PM_MISCINFO_THERMAL_DIODE_MODE              0x00080000L
 
-#define ATOM_PM_MISCINFO_FRAME_MODULATION_MASK           0x00300000L	/* 0-FM Disable, 1-2 level FM, 2-4 level FM, 3-Reserved */
-#define ATOM_PM_MISCINFO_FRAME_MODULATION_SHIFT          20
+#define ATOM_PM_MISCINFO_FRAME_MODULATION_MASK           0x00300000L  //0-FM Disable, 1-2 level FM, 2-4 level FM, 3-Reserved
+#define ATOM_PM_MISCINFO_FRAME_MODULATION_SHIFT          20 
 
 #define ATOM_PM_MISCINFO_DYN_CLK_3D_IDLE                 0x00400000L
 #define ATOM_PM_MISCINFO_DYNAMIC_CLOCK_DIVIDER_BY_2      0x00800000L
 #define ATOM_PM_MISCINFO_DYNAMIC_CLOCK_DIVIDER_BY_4      0x01000000L
-#define ATOM_PM_MISCINFO_DYNAMIC_HDP_BLOCK_EN            0x02000000L	/* When set, Dynamic */
-#define ATOM_PM_MISCINFO_DYNAMIC_MC_HOST_BLOCK_EN        0x04000000L	/* When set, Dynamic */
-#define ATOM_PM_MISCINFO_3D_ACCELERATION_EN              0x08000000L	/* When set, This mode is for acceleated 3D mode */
+#define ATOM_PM_MISCINFO_DYNAMIC_HDP_BLOCK_EN            0x02000000L  //When set, Dynamic 
+#define ATOM_PM_MISCINFO_DYNAMIC_MC_HOST_BLOCK_EN        0x04000000L  //When set, Dynamic
+#define ATOM_PM_MISCINFO_3D_ACCELERATION_EN              0x08000000L  //When set, This mode is for acceleated 3D mode
 
-#define ATOM_PM_MISCINFO_POWERPLAY_SETTINGS_GROUP_MASK   0x70000000L	/* 1-Optimal Battery Life Group, 2-High Battery, 3-Balanced, 4-High Performance, 5- Optimal Performance (Default state with Default clocks) */
+#define ATOM_PM_MISCINFO_POWERPLAY_SETTINGS_GROUP_MASK   0x70000000L  //1-Optimal Battery Life Group, 2-High Battery, 3-Balanced, 4-High Performance, 5- Optimal Performance (Default state with Default clocks) 
 #define ATOM_PM_MISCINFO_POWERPLAY_SETTINGS_GROUP_SHIFT  28
 #define ATOM_PM_MISCINFO_ENABLE_BACK_BIAS                0x80000000L
 
@@ -4594,55 +5610,59 @@ typedef struct _DFP_DPMS_STATUS_CHANGE_PARAMETERS {
 #define ATOM_PM_MISCINFO2_FS3D_OVERDRIVE_INFO            0x00000008L
 #define ATOM_PM_MISCINFO2_FORCEDLOWPWR_MODE              0x00000010L
 #define ATOM_PM_MISCINFO2_VDDCI_DYNAMIC_VOLTAGE_EN       0x00000020L
-#define ATOM_PM_MISCINFO2_VIDEO_PLAYBACK_CAPABLE         0x00000040L	/* If this bit is set in multi-pp mode, then driver will pack up one with the minior power consumption. */
-								      /* If it's not set in any pp mode, driver will use its default logic to pick a pp mode in video playback */
+#define ATOM_PM_MISCINFO2_VIDEO_PLAYBACK_CAPABLE         0x00000040L  //If this bit is set in multi-pp mode, then driver will pack up one with the minior power consumption. 
+                                                                      //If it's not set in any pp mode, driver will use its default logic to pick a pp mode in video playback
 #define ATOM_PM_MISCINFO2_NOT_VALID_ON_DC                0x00000080L
 #define ATOM_PM_MISCINFO2_STUTTER_MODE_EN                0x00000100L
-#define ATOM_PM_MISCINFO2_UVD_SUPPORT_MODE               0x00000200L
-
-/* ucTableFormatRevision=1 */
-/* ucTableContentRevision=1 */
-typedef struct _ATOM_POWERMODE_INFO {
-	ULONG ulMiscInfo;	/* The power level should be arranged in ascending order */
-	ULONG ulReserved1;	/*  must set to 0 */
-	ULONG ulReserved2;	/*  must set to 0 */
-	USHORT usEngineClock;
-	USHORT usMemoryClock;
-	UCHAR ucVoltageDropIndex;	/*  index to GPIO table */
-	UCHAR ucSelectedPanel_RefreshRate;	/*  panel refresh rate */
-	UCHAR ucMinTemperature;
-	UCHAR ucMaxTemperature;
-	UCHAR ucNumPciELanes;	/*  number of PCIE lanes */
-} ATOM_POWERMODE_INFO;
-
-/* ucTableFormatRevision=2 */
-/* ucTableContentRevision=1 */
-typedef struct _ATOM_POWERMODE_INFO_V2 {
-	ULONG ulMiscInfo;	/* The power level should be arranged in ascending order */
-	ULONG ulMiscInfo2;
-	ULONG ulEngineClock;
-	ULONG ulMemoryClock;
-	UCHAR ucVoltageDropIndex;	/*  index to GPIO table */
-	UCHAR ucSelectedPanel_RefreshRate;	/*  panel refresh rate */
-	UCHAR ucMinTemperature;
-	UCHAR ucMaxTemperature;
-	UCHAR ucNumPciELanes;	/*  number of PCIE lanes */
-} ATOM_POWERMODE_INFO_V2;
-
-/* ucTableFormatRevision=2 */
-/* ucTableContentRevision=2 */
-typedef struct _ATOM_POWERMODE_INFO_V3 {
-	ULONG ulMiscInfo;	/* The power level should be arranged in ascending order */
-	ULONG ulMiscInfo2;
-	ULONG ulEngineClock;
-	ULONG ulMemoryClock;
-	UCHAR ucVoltageDropIndex;	/*  index to Core (VDDC) votage table */
-	UCHAR ucSelectedPanel_RefreshRate;	/*  panel refresh rate */
-	UCHAR ucMinTemperature;
-	UCHAR ucMaxTemperature;
-	UCHAR ucNumPciELanes;	/*  number of PCIE lanes */
-	UCHAR ucVDDCI_VoltageDropIndex;	/*  index to VDDCI votage table */
-} ATOM_POWERMODE_INFO_V3;
+#define ATOM_PM_MISCINFO2_UVD_SUPPORT_MODE               0x00000200L 
+
+//ucTableFormatRevision=1
+//ucTableContentRevision=1
+typedef struct  _ATOM_POWERMODE_INFO
+{
+  ULONG     ulMiscInfo;                 //The power level should be arranged in ascending order
+  ULONG     ulReserved1;                // must set to 0
+  ULONG     ulReserved2;                // must set to 0
+  USHORT    usEngineClock;
+  USHORT    usMemoryClock;
+  UCHAR     ucVoltageDropIndex;         // index to GPIO table
+  UCHAR     ucSelectedPanel_RefreshRate;// panel refresh rate
+  UCHAR     ucMinTemperature;
+  UCHAR     ucMaxTemperature;
+  UCHAR     ucNumPciELanes;             // number of PCIE lanes
+}ATOM_POWERMODE_INFO;
+
+//ucTableFormatRevision=2
+//ucTableContentRevision=1
+typedef struct  _ATOM_POWERMODE_INFO_V2
+{
+  ULONG     ulMiscInfo;                 //The power level should be arranged in ascending order
+  ULONG     ulMiscInfo2;                
+  ULONG     ulEngineClock;                
+  ULONG     ulMemoryClock;
+  UCHAR     ucVoltageDropIndex;         // index to GPIO table
+  UCHAR     ucSelectedPanel_RefreshRate;// panel refresh rate
+  UCHAR     ucMinTemperature;
+  UCHAR     ucMaxTemperature;
+  UCHAR     ucNumPciELanes;             // number of PCIE lanes
+}ATOM_POWERMODE_INFO_V2;
+
+//ucTableFormatRevision=2
+//ucTableContentRevision=2
+typedef struct  _ATOM_POWERMODE_INFO_V3
+{
+  ULONG     ulMiscInfo;                 //The power level should be arranged in ascending order
+  ULONG     ulMiscInfo2;                
+  ULONG     ulEngineClock;                
+  ULONG     ulMemoryClock;
+  UCHAR     ucVoltageDropIndex;         // index to Core (VDDC) votage table
+  UCHAR     ucSelectedPanel_RefreshRate;// panel refresh rate
+  UCHAR     ucMinTemperature;
+  UCHAR     ucMaxTemperature;
+  UCHAR     ucNumPciELanes;             // number of PCIE lanes
+  UCHAR     ucVDDCI_VoltageDropIndex;   // index to VDDCI votage table
+}ATOM_POWERMODE_INFO_V3;
+
 
 #define ATOM_MAX_NUMBEROF_POWER_BLOCK  8
 
@@ -4655,59 +5675,264 @@ typedef struct _ATOM_POWERMODE_INFO_V3 {
 #define ATOM_PP_OVERDRIVE_THERMALCONTROLLER_MUA6649   0x04
 #define ATOM_PP_OVERDRIVE_THERMALCONTROLLER_LM64      0x05
 #define ATOM_PP_OVERDRIVE_THERMALCONTROLLER_F75375    0x06
-#define ATOM_PP_OVERDRIVE_THERMALCONTROLLER_ASC7512   0x07	/*  Andigilog */
-
-typedef struct _ATOM_POWERPLAY_INFO {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	UCHAR ucOverdriveThermalController;
-	UCHAR ucOverdriveI2cLine;
-	UCHAR ucOverdriveIntBitmap;
-	UCHAR ucOverdriveControllerAddress;
-	UCHAR ucSizeOfPowerModeEntry;
-	UCHAR ucNumOfPowerModeEntries;
-	ATOM_POWERMODE_INFO asPowerPlayInfo[ATOM_MAX_NUMBEROF_POWER_BLOCK];
-} ATOM_POWERPLAY_INFO;
-
-typedef struct _ATOM_POWERPLAY_INFO_V2 {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	UCHAR ucOverdriveThermalController;
-	UCHAR ucOverdriveI2cLine;
-	UCHAR ucOverdriveIntBitmap;
-	UCHAR ucOverdriveControllerAddress;
-	UCHAR ucSizeOfPowerModeEntry;
-	UCHAR ucNumOfPowerModeEntries;
-	ATOM_POWERMODE_INFO_V2 asPowerPlayInfo[ATOM_MAX_NUMBEROF_POWER_BLOCK];
-} ATOM_POWERPLAY_INFO_V2;
-
-typedef struct _ATOM_POWERPLAY_INFO_V3 {
-	ATOM_COMMON_TABLE_HEADER sHeader;
-	UCHAR ucOverdriveThermalController;
-	UCHAR ucOverdriveI2cLine;
-	UCHAR ucOverdriveIntBitmap;
-	UCHAR ucOverdriveControllerAddress;
-	UCHAR ucSizeOfPowerModeEntry;
-	UCHAR ucNumOfPowerModeEntries;
-	ATOM_POWERMODE_INFO_V3 asPowerPlayInfo[ATOM_MAX_NUMBEROF_POWER_BLOCK];
-} ATOM_POWERPLAY_INFO_V3;
+#define ATOM_PP_OVERDRIVE_THERMALCONTROLLER_ASC7512   0x07	// Andigilog
+
+
+typedef struct  _ATOM_POWERPLAY_INFO
+{
+  ATOM_COMMON_TABLE_HEADER	sHeader; 
+  UCHAR    ucOverdriveThermalController;
+  UCHAR    ucOverdriveI2cLine;
+  UCHAR    ucOverdriveIntBitmap;
+  UCHAR    ucOverdriveControllerAddress;
+  UCHAR    ucSizeOfPowerModeEntry;
+  UCHAR    ucNumOfPowerModeEntries;
+  ATOM_POWERMODE_INFO asPowerPlayInfo[ATOM_MAX_NUMBEROF_POWER_BLOCK];
+}ATOM_POWERPLAY_INFO;
+
+typedef struct  _ATOM_POWERPLAY_INFO_V2
+{
+  ATOM_COMMON_TABLE_HEADER	sHeader; 
+  UCHAR    ucOverdriveThermalController;
+  UCHAR    ucOverdriveI2cLine;
+  UCHAR    ucOverdriveIntBitmap;
+  UCHAR    ucOverdriveControllerAddress;
+  UCHAR    ucSizeOfPowerModeEntry;
+  UCHAR    ucNumOfPowerModeEntries;
+  ATOM_POWERMODE_INFO_V2 asPowerPlayInfo[ATOM_MAX_NUMBEROF_POWER_BLOCK];
+}ATOM_POWERPLAY_INFO_V2;
+  
+typedef struct  _ATOM_POWERPLAY_INFO_V3
+{
+  ATOM_COMMON_TABLE_HEADER	sHeader; 
+  UCHAR    ucOverdriveThermalController;
+  UCHAR    ucOverdriveI2cLine;
+  UCHAR    ucOverdriveIntBitmap;
+  UCHAR    ucOverdriveControllerAddress;
+  UCHAR    ucSizeOfPowerModeEntry;
+  UCHAR    ucNumOfPowerModeEntries;
+  ATOM_POWERMODE_INFO_V3 asPowerPlayInfo[ATOM_MAX_NUMBEROF_POWER_BLOCK];
+}ATOM_POWERPLAY_INFO_V3;
+
+/* New PPlib */
+/**************************************************************************/
+typedef struct _ATOM_PPLIB_THERMALCONTROLLER
+
+{
+    UCHAR ucType;           // one of ATOM_PP_THERMALCONTROLLER_*
+    UCHAR ucI2cLine;        // as interpreted by DAL I2C
+    UCHAR ucI2cAddress;
+    UCHAR ucFanParameters;  // Fan Control Parameters.
+    UCHAR ucFanMinRPM;      // Fan Minimum RPM (hundreds) -- for display purposes only.
+    UCHAR ucFanMaxRPM;      // Fan Maximum RPM (hundreds) -- for display purposes only.
+    UCHAR ucReserved;       // ----
+    UCHAR ucFlags;          // to be defined
+} ATOM_PPLIB_THERMALCONTROLLER;
+
+#define ATOM_PP_FANPARAMETERS_TACHOMETER_PULSES_PER_REVOLUTION_MASK 0x0f
+#define ATOM_PP_FANPARAMETERS_NOFAN                                 0x80    // No fan is connected to this controller.
+
+#define ATOM_PP_THERMALCONTROLLER_NONE      0
+#define ATOM_PP_THERMALCONTROLLER_LM63      1  // Not used by PPLib
+#define ATOM_PP_THERMALCONTROLLER_ADM1032   2  // Not used by PPLib
+#define ATOM_PP_THERMALCONTROLLER_ADM1030   3  // Not used by PPLib
+#define ATOM_PP_THERMALCONTROLLER_MUA6649   4  // Not used by PPLib
+#define ATOM_PP_THERMALCONTROLLER_LM64      5
+#define ATOM_PP_THERMALCONTROLLER_F75375    6  // Not used by PPLib
+#define ATOM_PP_THERMALCONTROLLER_RV6xx     7
+#define ATOM_PP_THERMALCONTROLLER_RV770     8
+#define ATOM_PP_THERMALCONTROLLER_ADT7473   9
+
+typedef struct _ATOM_PPLIB_STATE
+{
+    UCHAR ucNonClockStateIndex;
+    UCHAR ucClockStateIndices[1]; // variable-sized
+} ATOM_PPLIB_STATE;
+
+//// ATOM_PPLIB_POWERPLAYTABLE::ulPlatformCaps
+#define ATOM_PP_PLATFORM_CAP_BACKBIAS 1
+#define ATOM_PP_PLATFORM_CAP_POWERPLAY 2
+#define ATOM_PP_PLATFORM_CAP_SBIOSPOWERSOURCE 4
+#define ATOM_PP_PLATFORM_CAP_ASPM_L0s 8
+#define ATOM_PP_PLATFORM_CAP_ASPM_L1 16
+#define ATOM_PP_PLATFORM_CAP_HARDWAREDC 32
+#define ATOM_PP_PLATFORM_CAP_GEMINIPRIMARY 64
+#define ATOM_PP_PLATFORM_CAP_STEPVDDC 128
+#define ATOM_PP_PLATFORM_CAP_VOLTAGECONTROL 256
+#define ATOM_PP_PLATFORM_CAP_SIDEPORTCONTROL 512
+#define ATOM_PP_PLATFORM_CAP_TURNOFFPLL_ASPML1 1024
+#define ATOM_PP_PLATFORM_CAP_HTLINKCONTROL 2048
+
+typedef struct _ATOM_PPLIB_POWERPLAYTABLE
+{
+      ATOM_COMMON_TABLE_HEADER sHeader;
+
+      UCHAR ucDataRevision;
+
+      UCHAR ucNumStates;
+      UCHAR ucStateEntrySize;
+      UCHAR ucClockInfoSize;
+      UCHAR ucNonClockSize;
+
+      // offset from start of this table to array of ucNumStates ATOM_PPLIB_STATE structures
+      USHORT usStateArrayOffset;
+
+      // offset from start of this table to array of ASIC-specific structures,
+      // currently ATOM_PPLIB_CLOCK_INFO.
+      USHORT usClockInfoArrayOffset;
+
+      // offset from start of this table to array of ATOM_PPLIB_NONCLOCK_INFO
+      USHORT usNonClockInfoArrayOffset;
+
+      USHORT usBackbiasTime;    // in microseconds
+      USHORT usVoltageTime;     // in microseconds
+      USHORT usTableSize;       //the size of this structure, or the extended structure
+
+      ULONG ulPlatformCaps;            // See ATOM_PPLIB_CAPS_*
+
+      ATOM_PPLIB_THERMALCONTROLLER    sThermalController;
+
+      USHORT usBootClockInfoOffset;
+      USHORT usBootNonClockInfoOffset;
+
+} ATOM_PPLIB_POWERPLAYTABLE;
+
+//// ATOM_PPLIB_NONCLOCK_INFO::usClassification
+#define ATOM_PPLIB_CLASSIFICATION_UI_MASK          0x0007
+#define ATOM_PPLIB_CLASSIFICATION_UI_SHIFT         0
+#define ATOM_PPLIB_CLASSIFICATION_UI_NONE          0
+#define ATOM_PPLIB_CLASSIFICATION_UI_BATTERY       1
+#define ATOM_PPLIB_CLASSIFICATION_UI_BALANCED      3
+#define ATOM_PPLIB_CLASSIFICATION_UI_PERFORMANCE   5
+// 2, 4, 6, 7 are reserved
+
+#define ATOM_PPLIB_CLASSIFICATION_BOOT                   0x0008
+#define ATOM_PPLIB_CLASSIFICATION_THERMAL                0x0010
+#define ATOM_PPLIB_CLASSIFICATION_LIMITEDPOWERSOURCE     0x0020
+#define ATOM_PPLIB_CLASSIFICATION_REST                   0x0040
+#define ATOM_PPLIB_CLASSIFICATION_FORCED                 0x0080
+#define ATOM_PPLIB_CLASSIFICATION_3DPERFORMANCE          0x0100
+#define ATOM_PPLIB_CLASSIFICATION_OVERDRIVETEMPLATE      0x0200
+#define ATOM_PPLIB_CLASSIFICATION_UVDSTATE               0x0400
+#define ATOM_PPLIB_CLASSIFICATION_3DLOW                  0x0800
+#define ATOM_PPLIB_CLASSIFICATION_ACPI                   0x1000
+// remaining 3 bits are reserved
+
+//// ATOM_PPLIB_NONCLOCK_INFO::ulCapsAndSettings
+#define ATOM_PPLIB_SINGLE_DISPLAY_ONLY           0x00000001
+#define ATOM_PPLIB_SUPPORTS_VIDEO_PLAYBACK         0x00000002
+
+// 0 is 2.5Gb/s, 1 is 5Gb/s
+#define ATOM_PPLIB_PCIE_LINK_SPEED_MASK            0x00000004
+#define ATOM_PPLIB_PCIE_LINK_SPEED_SHIFT           2
+
+// lanes - 1: 1, 2, 4, 8, 12, 16 permitted by PCIE spec
+#define ATOM_PPLIB_PCIE_LINK_WIDTH_MASK            0x000000F8
+#define ATOM_PPLIB_PCIE_LINK_WIDTH_SHIFT           3
+
+// lookup into reduced refresh-rate table
+#define ATOM_PPLIB_LIMITED_REFRESHRATE_VALUE_MASK  0x00000F00
+#define ATOM_PPLIB_LIMITED_REFRESHRATE_VALUE_SHIFT 8
+
+#define ATOM_PPLIB_LIMITED_REFRESHRATE_UNLIMITED    0
+#define ATOM_PPLIB_LIMITED_REFRESHRATE_50HZ         1
+// 2-15 TBD as needed.
+
+#define ATOM_PPLIB_SOFTWARE_DISABLE_LOADBALANCING        0x00001000
+#define ATOM_PPLIB_SOFTWARE_ENABLE_SLEEP_FOR_TIMESTAMPS  0x00002000
+#define ATOM_PPLIB_ENABLE_VARIBRIGHT                     0x00008000
+
+#define ATOM_PPLIB_DISALLOW_ON_DC                       0x00004000
+
+// Contained in an array starting at the offset
+// in ATOM_PPLIB_POWERPLAYTABLE::usNonClockInfoArrayOffset.
+// referenced from ATOM_PPLIB_STATE_INFO::ucNonClockStateIndex
+typedef struct _ATOM_PPLIB_NONCLOCK_INFO
+{
+      USHORT usClassification;
+      UCHAR  ucMinTemperature;
+      UCHAR  ucMaxTemperature;
+      ULONG  ulCapsAndSettings;
+      UCHAR  ucRequiredPower;
+      UCHAR  ucUnused1[3];
+} ATOM_PPLIB_NONCLOCK_INFO;
+
+// Contained in an array starting at the offset
+// in ATOM_PPLIB_POWERPLAYTABLE::usClockInfoArrayOffset.
+// referenced from ATOM_PPLIB_STATE::ucClockStateIndices
+typedef struct _ATOM_PPLIB_R600_CLOCK_INFO
+{
+      USHORT usEngineClockLow;
+      UCHAR ucEngineClockHigh;
+
+      USHORT usMemoryClockLow;
+      UCHAR ucMemoryClockHigh;
+
+      USHORT usVDDC;
+      USHORT usUnused1;
+      USHORT usUnused2;
+
+      ULONG ulFlags; // ATOM_PPLIB_R600_FLAGS_*
+
+} ATOM_PPLIB_R600_CLOCK_INFO;
+
+// ulFlags in ATOM_PPLIB_R600_CLOCK_INFO
+#define ATOM_PPLIB_R600_FLAGS_PCIEGEN2          1
+#define ATOM_PPLIB_R600_FLAGS_UVDSAFE           2
+#define ATOM_PPLIB_R600_FLAGS_BACKBIASENABLE    4
+#define ATOM_PPLIB_R600_FLAGS_MEMORY_ODT_OFF    8
+#define ATOM_PPLIB_R600_FLAGS_MEMORY_DLL_OFF    16
+
+typedef struct _ATOM_PPLIB_RS780_CLOCK_INFO
+
+{
+      USHORT usLowEngineClockLow;         // Low Engine clock in MHz (the same way as on the R600).
+      UCHAR  ucLowEngineClockHigh;
+      USHORT usHighEngineClockLow;        // High Engine clock in MHz.
+      UCHAR  ucHighEngineClockHigh;
+      USHORT usMemoryClockLow;            // For now one of the ATOM_PPLIB_RS780_SPMCLK_XXXX constants.
+      UCHAR  ucMemoryClockHigh;           // Currentyl unused.
+      UCHAR  ucPadding;                   // For proper alignment and size.
+      USHORT usVDDC;                      // For the 780, use: None, Low, High, Variable
+      UCHAR  ucMaxHTLinkWidth;            // From SBIOS - {2, 4, 8, 16}
+      UCHAR  ucMinHTLinkWidth;            // From SBIOS - {2, 4, 8, 16}. Effective only if CDLW enabled. Minimum down stream width could be bigger as display BW requriement.
+      USHORT usHTLinkFreq;                // See definition ATOM_PPLIB_RS780_HTLINKFREQ_xxx or in MHz(>=200).
+      ULONG  ulFlags; 
+} ATOM_PPLIB_RS780_CLOCK_INFO;
+
+#define ATOM_PPLIB_RS780_VOLTAGE_NONE       0 
+#define ATOM_PPLIB_RS780_VOLTAGE_LOW        1 
+#define ATOM_PPLIB_RS780_VOLTAGE_HIGH       2 
+#define ATOM_PPLIB_RS780_VOLTAGE_VARIABLE   3 
+
+#define ATOM_PPLIB_RS780_SPMCLK_NONE        0   // We cannot change the side port memory clock, leave it as it is.
+#define ATOM_PPLIB_RS780_SPMCLK_LOW         1
+#define ATOM_PPLIB_RS780_SPMCLK_HIGH        2
+
+#define ATOM_PPLIB_RS780_HTLINKFREQ_NONE       0 
+#define ATOM_PPLIB_RS780_HTLINKFREQ_LOW        1 
+#define ATOM_PPLIB_RS780_HTLINKFREQ_HIGH       2 
 
 /**************************************************************************/
 
-/*  Following definitions are for compatiblity issue in different SW components. */
+
+// Following definitions are for compatiblity issue in different SW components. 
 #define ATOM_MASTER_DATA_TABLE_REVISION   0x01
-#define Object_Info												Object_Header
+#define Object_Info												Object_Header			
 #define	AdjustARB_SEQ											MC_InitParameter
 #define	VRAM_GPIO_DetectionInfo						VoltageObjectInfo
-#define	ASIC_VDDCI_Info                   ASIC_ProfilingInfo
+#define	ASIC_VDDCI_Info                   ASIC_ProfilingInfo														
 #define ASIC_MVDDQ_Info										MemoryTrainingInfo
-#define SS_Info                           PPLL_SS_Info
+#define SS_Info                           PPLL_SS_Info                      
 #define ASIC_MVDDC_Info                   ASIC_InternalSS_Info
 #define DispDevicePriorityInfo						SaveRestoreInfo
 #define DispOutInfo												TV_VideoMode
 
+
 #define ATOM_ENCODER_OBJECT_TABLE         ATOM_OBJECT_TABLE
 #define ATOM_CONNECTOR_OBJECT_TABLE       ATOM_OBJECT_TABLE
 
-/* New device naming, remove them when both DAL/VBIOS is ready */
+//New device naming, remove them when both DAL/VBIOS is ready
 #define DFP2I_OUTPUT_CONTROL_PARAMETERS    CRT1_OUTPUT_CONTROL_PARAMETERS
 #define DFP2I_OUTPUT_CONTROL_PS_ALLOCATION DFP2I_OUTPUT_CONTROL_PARAMETERS
 
@@ -4722,7 +5947,7 @@ typedef struct _ATOM_POWERPLAY_INFO_V3 {
 
 #define ATOM_DEVICE_DFP1I_INDEX            ATOM_DEVICE_DFP1_INDEX
 #define ATOM_DEVICE_DFP1X_INDEX            ATOM_DEVICE_DFP2_INDEX
-
+ 
 #define ATOM_DEVICE_DFP2I_INDEX            0x00000009
 #define ATOM_DEVICE_DFP2I_SUPPORT          (0x1L << ATOM_DEVICE_DFP2I_INDEX)
 
@@ -4740,7 +5965,7 @@ typedef struct _ATOM_POWERPLAY_INFO_V3 {
 
 #define ATOM_S3_DFP2I_ACTIVEb1             0x02
 
-#define ATOM_S3_DFP1I_ACTIVE               ATOM_S3_DFP1_ACTIVE
+#define ATOM_S3_DFP1I_ACTIVE               ATOM_S3_DFP1_ACTIVE 
 #define ATOM_S3_DFP1X_ACTIVE               ATOM_S3_DFP2_ACTIVE
 
 #define ATOM_S3_DFP2I_ACTIVE               0x00000200L
@@ -4759,14 +5984,14 @@ typedef struct _ATOM_POWERPLAY_INFO_V3 {
 #define ATOM_S6_ACC_REQ_DFP2Ib3            0x02
 #define ATOM_S6_ACC_REQ_DFP2I              0x02000000L
 
-#define TMDS1XEncoderControl               DVOEncoderControl
+#define TMDS1XEncoderControl               DVOEncoderControl           
 #define DFP1XOutputControl                 DVOOutputControl
 
 #define ExternalDFPOutputControl           DFP1XOutputControl
 #define EnableExternalTMDS_Encoder         TMDS1XEncoderControl
 
 #define DFP1IOutputControl                 TMDSAOutputControl
-#define DFP2IOutputControl                 LVTMAOutputControl
+#define DFP2IOutputControl                 LVTMAOutputControl      
 
 #define DAC1_ENCODER_CONTROL_PARAMETERS    DAC_ENCODER_CONTROL_PARAMETERS
 #define DAC1_ENCODER_CONTROL_PS_ALLOCATION DAC_ENCODER_CONTROL_PS_ALLOCATION
@@ -4775,7 +6000,7 @@ typedef struct _ATOM_POWERPLAY_INFO_V3 {
 #define DAC2_ENCODER_CONTROL_PS_ALLOCATION DAC_ENCODER_CONTROL_PS_ALLOCATION
 
 #define ucDac1Standard  ucDacStandard
-#define ucDac2Standard  ucDacStandard
+#define ucDac2Standard  ucDacStandard  
 
 #define TMDS1EncoderControl TMDSAEncoderControl
 #define TMDS2EncoderControl LVTMAEncoderControl
@@ -4785,12 +6010,56 @@ typedef struct _ATOM_POWERPLAY_INFO_V3 {
 #define CRT1OutputControl   DAC1OutputControl
 #define CRT2OutputControl   DAC2OutputControl
 
-/* These two lines will be removed for sure in a few days, will follow up with Michael V. */
+//These two lines will be removed for sure in a few days, will follow up with Michael V.
 #define EnableLVDS_SS   EnableSpreadSpectrumOnPPLL
-#define ENABLE_LVDS_SS_PARAMETERS_V3  ENABLE_SPREAD_SPECTRUM_ON_PPLL
+#define ENABLE_LVDS_SS_PARAMETERS_V3  ENABLE_SPREAD_SPECTRUM_ON_PPLL  
+
+//#define ATOM_S2_CRT1_DPMS_STATE         0x00010000L
+//#define ATOM_S2_LCD1_DPMS_STATE	        ATOM_S2_CRT1_DPMS_STATE
+//#define ATOM_S2_TV1_DPMS_STATE          ATOM_S2_CRT1_DPMS_STATE
+//#define ATOM_S2_DFP1_DPMS_STATE         ATOM_S2_CRT1_DPMS_STATE
+//#define ATOM_S2_CRT2_DPMS_STATE         ATOM_S2_CRT1_DPMS_STATE
+
+#define ATOM_S6_ACC_REQ_TV2             0x00400000L
+#define ATOM_DEVICE_TV2_INDEX           0x00000006
+#define ATOM_DEVICE_TV2_SUPPORT         (0x1L << ATOM_DEVICE_TV2_INDEX)
+#define ATOM_S0_TV2                     0x00100000L
+#define ATOM_S3_TV2_ACTIVE              ATOM_S3_DFP6_ACTIVE
+#define ATOM_S3_TV2_CRTC_ACTIVE         ATOM_S3_DFP6_CRTC_ACTIVE
+
+//
+#define ATOM_S2_CRT1_DPMS_STATE         0x00010000L
+#define ATOM_S2_LCD1_DPMS_STATE	        0x00020000L
+#define ATOM_S2_TV1_DPMS_STATE          0x00040000L
+#define ATOM_S2_DFP1_DPMS_STATE         0x00080000L
+#define ATOM_S2_CRT2_DPMS_STATE         0x00100000L
+#define ATOM_S2_LCD2_DPMS_STATE         0x00200000L
+#define ATOM_S2_TV2_DPMS_STATE          0x00400000L
+#define ATOM_S2_DFP2_DPMS_STATE         0x00800000L
+#define ATOM_S2_CV_DPMS_STATE           0x01000000L
+#define ATOM_S2_DFP3_DPMS_STATE					0x02000000L
+#define ATOM_S2_DFP4_DPMS_STATE					0x04000000L
+#define ATOM_S2_DFP5_DPMS_STATE					0x08000000L
+
+#define ATOM_S2_CRT1_DPMS_STATEb2       0x01
+#define ATOM_S2_LCD1_DPMS_STATEb2       0x02
+#define ATOM_S2_TV1_DPMS_STATEb2        0x04
+#define ATOM_S2_DFP1_DPMS_STATEb2       0x08
+#define ATOM_S2_CRT2_DPMS_STATEb2       0x10
+#define ATOM_S2_LCD2_DPMS_STATEb2       0x20
+#define ATOM_S2_TV2_DPMS_STATEb2        0x40
+#define ATOM_S2_DFP2_DPMS_STATEb2       0x80
+#define ATOM_S2_CV_DPMS_STATEb3         0x01
+#define ATOM_S2_DFP3_DPMS_STATEb3				0x02
+#define ATOM_S2_DFP4_DPMS_STATEb3				0x04
+#define ATOM_S2_DFP5_DPMS_STATEb3				0x08
+
+#define ATOM_S3_ASIC_GUI_ENGINE_HUNGb3	0x20
+#define ATOM_S3_ALLOW_FAST_PWR_SWITCHb3 0x40
+#define ATOM_S3_RQST_GPU_USE_MIN_PWRb3  0x80
 
 /*********************************************************************************/
 
-#pragma pack()			/*  BIOS data must use byte aligment */
+#pragma pack() // BIOS data must use byte aligment
 
 #endif /* _ATOMBIOS_H */
diff --git a/drivers/gpu/drm/radeon/atombios_crtc.c b/drivers/gpu/drm/radeon/atombios_crtc.c
index 6a015929deee..dd9fdf560611 100644
--- a/drivers/gpu/drm/radeon/atombios_crtc.c
+++ b/drivers/gpu/drm/radeon/atombios_crtc.c
@@ -31,10 +31,6 @@
 #include "atom.h"
 #include "atom-bits.h"
 
-/* evil but including atombios.h is much worse */
-bool radeon_atom_get_tv_timings(struct radeon_device *rdev, int index,
-				SET_CRTC_TIMING_PARAMETERS_PS_ALLOCATION *crtc_timing,
-				int32_t *pixel_clock);
 static void atombios_overscan_setup(struct drm_crtc *crtc,
 				    struct drm_display_mode *mode,
 				    struct drm_display_mode *adjusted_mode)
@@ -245,162 +241,432 @@ void atombios_crtc_dpms(struct drm_crtc *crtc, int mode)
 {
 	struct drm_device *dev = crtc->dev;
 	struct radeon_device *rdev = dev->dev_private;
+	struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc);
 
 	switch (mode) {
 	case DRM_MODE_DPMS_ON:
+		atombios_enable_crtc(crtc, ATOM_ENABLE);
 		if (ASIC_IS_DCE3(rdev))
-			atombios_enable_crtc_memreq(crtc, 1);
-		atombios_enable_crtc(crtc, 1);
-		atombios_blank_crtc(crtc, 0);
+			atombios_enable_crtc_memreq(crtc, ATOM_ENABLE);
+		atombios_blank_crtc(crtc, ATOM_DISABLE);
+		/* XXX re-enable when interrupt support is added */
+		if (!ASIC_IS_DCE4(rdev))
+			drm_vblank_post_modeset(dev, radeon_crtc->crtc_id);
+		radeon_crtc_load_lut(crtc);
 		break;
 	case DRM_MODE_DPMS_STANDBY:
 	case DRM_MODE_DPMS_SUSPEND:
 	case DRM_MODE_DPMS_OFF:
-		atombios_blank_crtc(crtc, 1);
-		atombios_enable_crtc(crtc, 0);
+		/* XXX re-enable when interrupt support is added */
+		if (!ASIC_IS_DCE4(rdev))
+			drm_vblank_pre_modeset(dev, radeon_crtc->crtc_id);
+		atombios_blank_crtc(crtc, ATOM_ENABLE);
 		if (ASIC_IS_DCE3(rdev))
-			atombios_enable_crtc_memreq(crtc, 0);
+			atombios_enable_crtc_memreq(crtc, ATOM_DISABLE);
+		atombios_enable_crtc(crtc, ATOM_DISABLE);
 		break;
 	}
-
-	if (mode != DRM_MODE_DPMS_OFF) {
-		radeon_crtc_load_lut(crtc);
-	}
 }
 
 static void
 atombios_set_crtc_dtd_timing(struct drm_crtc *crtc,
-			     SET_CRTC_USING_DTD_TIMING_PARAMETERS * crtc_param)
+			     struct drm_display_mode *mode)
 {
+	struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc);
 	struct drm_device *dev = crtc->dev;
 	struct radeon_device *rdev = dev->dev_private;
-	SET_CRTC_USING_DTD_TIMING_PARAMETERS conv_param;
+	SET_CRTC_USING_DTD_TIMING_PARAMETERS args;
 	int index = GetIndexIntoMasterTable(COMMAND, SetCRTC_UsingDTDTiming);
+	u16 misc = 0;
+
+	memset(&args, 0, sizeof(args));
+	args.usH_Size = cpu_to_le16(mode->crtc_hdisplay);
+	args.usH_Blanking_Time =
+		cpu_to_le16(mode->crtc_hblank_end - mode->crtc_hdisplay);
+	args.usV_Size = cpu_to_le16(mode->crtc_vdisplay);
+	args.usV_Blanking_Time =
+	    cpu_to_le16(mode->crtc_vblank_end - mode->crtc_vdisplay);
+	args.usH_SyncOffset =
+		cpu_to_le16(mode->crtc_hsync_start - mode->crtc_hdisplay);
+	args.usH_SyncWidth =
+		cpu_to_le16(mode->crtc_hsync_end - mode->crtc_hsync_start);
+	args.usV_SyncOffset =
+		cpu_to_le16(mode->crtc_vsync_start - mode->crtc_vdisplay);
+	args.usV_SyncWidth =
+		cpu_to_le16(mode->crtc_vsync_end - mode->crtc_vsync_start);
+	/*args.ucH_Border = mode->hborder;*/
+	/*args.ucV_Border = mode->vborder;*/
+
+	if (mode->flags & DRM_MODE_FLAG_NVSYNC)
+		misc |= ATOM_VSYNC_POLARITY;
+	if (mode->flags & DRM_MODE_FLAG_NHSYNC)
+		misc |= ATOM_HSYNC_POLARITY;
+	if (mode->flags & DRM_MODE_FLAG_CSYNC)
+		misc |= ATOM_COMPOSITESYNC;
+	if (mode->flags & DRM_MODE_FLAG_INTERLACE)
+		misc |= ATOM_INTERLACE;
+	if (mode->flags & DRM_MODE_FLAG_DBLSCAN)
+		misc |= ATOM_DOUBLE_CLOCK_MODE;
+
+	args.susModeMiscInfo.usAccess = cpu_to_le16(misc);
+	args.ucCRTC = radeon_crtc->crtc_id;
 
-	conv_param.usH_Size = cpu_to_le16(crtc_param->usH_Size);
-	conv_param.usH_Blanking_Time =
-	    cpu_to_le16(crtc_param->usH_Blanking_Time);
-	conv_param.usV_Size = cpu_to_le16(crtc_param->usV_Size);
-	conv_param.usV_Blanking_Time =
-	    cpu_to_le16(crtc_param->usV_Blanking_Time);
-	conv_param.usH_SyncOffset = cpu_to_le16(crtc_param->usH_SyncOffset);
-	conv_param.usH_SyncWidth = cpu_to_le16(crtc_param->usH_SyncWidth);
-	conv_param.usV_SyncOffset = cpu_to_le16(crtc_param->usV_SyncOffset);
-	conv_param.usV_SyncWidth = cpu_to_le16(crtc_param->usV_SyncWidth);
-	conv_param.susModeMiscInfo.usAccess =
-	    cpu_to_le16(crtc_param->susModeMiscInfo.usAccess);
-	conv_param.ucCRTC = crtc_param->ucCRTC;
-
-	printk("executing set crtc dtd timing\n");
-	atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&conv_param);
+	atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args);
 }
 
-void atombios_crtc_set_timing(struct drm_crtc *crtc,
-			      SET_CRTC_TIMING_PARAMETERS_PS_ALLOCATION *
-			      crtc_param)
+static void atombios_crtc_set_timing(struct drm_crtc *crtc,
+				     struct drm_display_mode *mode)
 {
+	struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc);
 	struct drm_device *dev = crtc->dev;
 	struct radeon_device *rdev = dev->dev_private;
-	SET_CRTC_TIMING_PARAMETERS_PS_ALLOCATION conv_param;
+	SET_CRTC_TIMING_PARAMETERS_PS_ALLOCATION args;
 	int index = GetIndexIntoMasterTable(COMMAND, SetCRTC_Timing);
+	u16 misc = 0;
+
+	memset(&args, 0, sizeof(args));
+	args.usH_Total = cpu_to_le16(mode->crtc_htotal);
+	args.usH_Disp = cpu_to_le16(mode->crtc_hdisplay);
+	args.usH_SyncStart = cpu_to_le16(mode->crtc_hsync_start);
+	args.usH_SyncWidth =
+		cpu_to_le16(mode->crtc_hsync_end - mode->crtc_hsync_start);
+	args.usV_Total = cpu_to_le16(mode->crtc_vtotal);
+	args.usV_Disp = cpu_to_le16(mode->crtc_vdisplay);
+	args.usV_SyncStart = cpu_to_le16(mode->crtc_vsync_start);
+	args.usV_SyncWidth =
+		cpu_to_le16(mode->crtc_vsync_end - mode->crtc_vsync_start);
+
+	if (mode->flags & DRM_MODE_FLAG_NVSYNC)
+		misc |= ATOM_VSYNC_POLARITY;
+	if (mode->flags & DRM_MODE_FLAG_NHSYNC)
+		misc |= ATOM_HSYNC_POLARITY;
+	if (mode->flags & DRM_MODE_FLAG_CSYNC)
+		misc |= ATOM_COMPOSITESYNC;
+	if (mode->flags & DRM_MODE_FLAG_INTERLACE)
+		misc |= ATOM_INTERLACE;
+	if (mode->flags & DRM_MODE_FLAG_DBLSCAN)
+		misc |= ATOM_DOUBLE_CLOCK_MODE;
+
+	args.susModeMiscInfo.usAccess = cpu_to_le16(misc);
+	args.ucCRTC = radeon_crtc->crtc_id;
 
-	conv_param.usH_Total = cpu_to_le16(crtc_param->usH_Total);
-	conv_param.usH_Disp = cpu_to_le16(crtc_param->usH_Disp);
-	conv_param.usH_SyncStart = cpu_to_le16(crtc_param->usH_SyncStart);
-	conv_param.usH_SyncWidth = cpu_to_le16(crtc_param->usH_SyncWidth);
-	conv_param.usV_Total = cpu_to_le16(crtc_param->usV_Total);
-	conv_param.usV_Disp = cpu_to_le16(crtc_param->usV_Disp);
-	conv_param.usV_SyncStart = cpu_to_le16(crtc_param->usV_SyncStart);
-	conv_param.usV_SyncWidth = cpu_to_le16(crtc_param->usV_SyncWidth);
-	conv_param.susModeMiscInfo.usAccess =
-	    cpu_to_le16(crtc_param->susModeMiscInfo.usAccess);
-	conv_param.ucCRTC = crtc_param->ucCRTC;
-	conv_param.ucOverscanRight = crtc_param->ucOverscanRight;
-	conv_param.ucOverscanLeft = crtc_param->ucOverscanLeft;
-	conv_param.ucOverscanBottom = crtc_param->ucOverscanBottom;
-	conv_param.ucOverscanTop = crtc_param->ucOverscanTop;
-	conv_param.ucReserved = crtc_param->ucReserved;
-
-	printk("executing set crtc timing\n");
-	atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&conv_param);
+	atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args);
 }
 
-void atombios_crtc_set_pll(struct drm_crtc *crtc, struct drm_display_mode *mode)
+union atom_enable_ss {
+	ENABLE_LVDS_SS_PARAMETERS legacy;
+	ENABLE_SPREAD_SPECTRUM_ON_PPLL_PS_ALLOCATION v1;
+};
+
+static void atombios_set_ss(struct drm_crtc *crtc, int enable)
 {
 	struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc);
 	struct drm_device *dev = crtc->dev;
 	struct radeon_device *rdev = dev->dev_private;
 	struct drm_encoder *encoder = NULL;
 	struct radeon_encoder *radeon_encoder = NULL;
-	uint8_t frev, crev;
-	int index = GetIndexIntoMasterTable(COMMAND, SetPixelClock);
-	SET_PIXEL_CLOCK_PS_ALLOCATION args;
-	PIXEL_CLOCK_PARAMETERS *spc1_ptr;
-	PIXEL_CLOCK_PARAMETERS_V2 *spc2_ptr;
-	PIXEL_CLOCK_PARAMETERS_V3 *spc3_ptr;
-	uint32_t sclock = mode->clock;
-	uint32_t ref_div = 0, fb_div = 0, frac_fb_div = 0, post_div = 0;
-	struct radeon_pll *pll;
-	int pll_flags = 0;
+	struct radeon_encoder_atom_dig *dig = NULL;
+	int index = GetIndexIntoMasterTable(COMMAND, EnableSpreadSpectrumOnPPLL);
+	union atom_enable_ss args;
+	uint16_t percentage = 0;
+	uint8_t type = 0, step = 0, delay = 0, range = 0;
+
+	/* XXX add ss support for DCE4 */
+	if (ASIC_IS_DCE4(rdev))
+		return;
+
+	list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) {
+		if (encoder->crtc == crtc) {
+			radeon_encoder = to_radeon_encoder(encoder);
+			/* only enable spread spectrum on LVDS */
+			if (radeon_encoder->devices & (ATOM_DEVICE_LCD_SUPPORT)) {
+				dig = radeon_encoder->enc_priv;
+				if (dig && dig->ss) {
+					percentage = dig->ss->percentage;
+					type = dig->ss->type;
+					step = dig->ss->step;
+					delay = dig->ss->delay;
+					range = dig->ss->range;
+				} else if (enable)
+					return;
+			} else if (enable)
+				return;
+			break;
+		}
+	}
+
+	if (!radeon_encoder)
+		return;
 
 	memset(&args, 0, sizeof(args));
+	if (ASIC_IS_AVIVO(rdev)) {
+		args.v1.usSpreadSpectrumPercentage = cpu_to_le16(percentage);
+		args.v1.ucSpreadSpectrumType = type;
+		args.v1.ucSpreadSpectrumStep = step;
+		args.v1.ucSpreadSpectrumDelay = delay;
+		args.v1.ucSpreadSpectrumRange = range;
+		args.v1.ucPpll = radeon_crtc->crtc_id ? ATOM_PPLL2 : ATOM_PPLL1;
+		args.v1.ucEnable = enable;
+	} else {
+		args.legacy.usSpreadSpectrumPercentage = cpu_to_le16(percentage);
+		args.legacy.ucSpreadSpectrumType = type;
+		args.legacy.ucSpreadSpectrumStepSize_Delay = (step & 3) << 2;
+		args.legacy.ucSpreadSpectrumStepSize_Delay |= (delay & 7) << 4;
+		args.legacy.ucEnable = enable;
+	}
+	atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args);
+}
+
+union adjust_pixel_clock {
+	ADJUST_DISPLAY_PLL_PS_ALLOCATION v1;
+	ADJUST_DISPLAY_PLL_PS_ALLOCATION_V3 v3;
+};
+
+static u32 atombios_adjust_pll(struct drm_crtc *crtc,
+			       struct drm_display_mode *mode,
+			       struct radeon_pll *pll)
+{
+	struct drm_device *dev = crtc->dev;
+	struct radeon_device *rdev = dev->dev_private;
+	struct drm_encoder *encoder = NULL;
+	struct radeon_encoder *radeon_encoder = NULL;
+	u32 adjusted_clock = mode->clock;
+	int encoder_mode = 0;
 
+	/* reset the pll flags */
+	pll->flags = 0;
+
+	/* select the PLL algo */
 	if (ASIC_IS_AVIVO(rdev)) {
-		uint32_t ss_cntl;
+		if (radeon_new_pll == 0)
+			pll->algo = PLL_ALGO_LEGACY;
+		else
+			pll->algo = PLL_ALGO_NEW;
+	} else {
+		if (radeon_new_pll == 1)
+			pll->algo = PLL_ALGO_NEW;
+		else
+			pll->algo = PLL_ALGO_LEGACY;
+	}
 
+	if (ASIC_IS_AVIVO(rdev)) {
 		if ((rdev->family == CHIP_RS600) ||
 		    (rdev->family == CHIP_RS690) ||
 		    (rdev->family == CHIP_RS740))
-			pll_flags |= (RADEON_PLL_USE_FRAC_FB_DIV |
-				      RADEON_PLL_PREFER_CLOSEST_LOWER);
+			pll->flags |= (RADEON_PLL_USE_FRAC_FB_DIV |
+				       RADEON_PLL_PREFER_CLOSEST_LOWER);
 
 		if (ASIC_IS_DCE32(rdev) && mode->clock > 200000)	/* range limits??? */
-			pll_flags |= RADEON_PLL_PREFER_HIGH_FB_DIV;
+			pll->flags |= RADEON_PLL_PREFER_HIGH_FB_DIV;
 		else
-			pll_flags |= RADEON_PLL_PREFER_LOW_REF_DIV;
-
-		/* disable spread spectrum clocking for now -- thanks Hedy Lamarr */
-		if (radeon_crtc->crtc_id == 0) {
-			ss_cntl = RREG32(AVIVO_P1PLL_INT_SS_CNTL);
-			WREG32(AVIVO_P1PLL_INT_SS_CNTL, ss_cntl & ~1);
-		} else {
-			ss_cntl = RREG32(AVIVO_P2PLL_INT_SS_CNTL);
-			WREG32(AVIVO_P2PLL_INT_SS_CNTL, ss_cntl & ~1);
-		}
+			pll->flags |= RADEON_PLL_PREFER_LOW_REF_DIV;
 	} else {
-		pll_flags |= RADEON_PLL_LEGACY;
+		pll->flags |= RADEON_PLL_LEGACY;
 
 		if (mode->clock > 200000)	/* range limits??? */
-			pll_flags |= RADEON_PLL_PREFER_HIGH_FB_DIV;
+			pll->flags |= RADEON_PLL_PREFER_HIGH_FB_DIV;
 		else
-			pll_flags |= RADEON_PLL_PREFER_LOW_REF_DIV;
+			pll->flags |= RADEON_PLL_PREFER_LOW_REF_DIV;
 
 	}
 
 	list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) {
 		if (encoder->crtc == crtc) {
-			if (!ASIC_IS_AVIVO(rdev)) {
-				if (encoder->encoder_type !=
-				    DRM_MODE_ENCODER_DAC)
-					pll_flags |= RADEON_PLL_NO_ODD_POST_DIV;
-				if (!ASIC_IS_AVIVO(rdev)
-				    && (encoder->encoder_type ==
-					DRM_MODE_ENCODER_LVDS))
-					pll_flags |= RADEON_PLL_USE_REF_DIV;
+			radeon_encoder = to_radeon_encoder(encoder);
+			encoder_mode = atombios_get_encoder_mode(encoder);
+			if (ASIC_IS_AVIVO(rdev)) {
+				/* DVO wants 2x pixel clock if the DVO chip is in 12 bit mode */
+				if (radeon_encoder->encoder_id == ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DVO1)
+					adjusted_clock = mode->clock * 2;
+				/* LVDS PLL quirks */
+				if (encoder->encoder_type == DRM_MODE_ENCODER_LVDS) {
+					struct radeon_encoder_atom_dig *dig = radeon_encoder->enc_priv;
+					pll->algo = dig->pll_algo;
+				}
+			} else {
+				if (encoder->encoder_type != DRM_MODE_ENCODER_DAC)
+					pll->flags |= RADEON_PLL_NO_ODD_POST_DIV;
+				if (encoder->encoder_type == DRM_MODE_ENCODER_LVDS)
+					pll->flags |= RADEON_PLL_USE_REF_DIV;
+			}
+			break;
+		}
+	}
+
+	/* DCE3+ has an AdjustDisplayPll that will adjust the pixel clock
+	 * accordingly based on the encoder/transmitter to work around
+	 * special hw requirements.
+	 */
+	if (ASIC_IS_DCE3(rdev)) {
+		union adjust_pixel_clock args;
+		u8 frev, crev;
+		int index;
+
+		index = GetIndexIntoMasterTable(COMMAND, AdjustDisplayPll);
+		atom_parse_cmd_header(rdev->mode_info.atom_context, index, &frev,
+				      &crev);
+
+		memset(&args, 0, sizeof(args));
+
+		switch (frev) {
+		case 1:
+			switch (crev) {
+			case 1:
+			case 2:
+				args.v1.usPixelClock = cpu_to_le16(mode->clock / 10);
+				args.v1.ucTransmitterID = radeon_encoder->encoder_id;
+				args.v1.ucEncodeMode = encoder_mode;
+
+				atom_execute_table(rdev->mode_info.atom_context,
+						   index, (uint32_t *)&args);
+				adjusted_clock = le16_to_cpu(args.v1.usPixelClock) * 10;
+				break;
+			case 3:
+				args.v3.sInput.usPixelClock = cpu_to_le16(mode->clock / 10);
+				args.v3.sInput.ucTransmitterID = radeon_encoder->encoder_id;
+				args.v3.sInput.ucEncodeMode = encoder_mode;
+				args.v3.sInput.ucDispPllConfig = 0;
+				if (radeon_encoder->devices & (ATOM_DEVICE_DFP_SUPPORT)) {
+					struct radeon_encoder_atom_dig *dig = radeon_encoder->enc_priv;
+
+					if (encoder_mode == ATOM_ENCODER_MODE_DP)
+						args.v3.sInput.ucDispPllConfig |=
+							DISPPLL_CONFIG_COHERENT_MODE;
+					else {
+						if (dig->coherent_mode)
+							args.v3.sInput.ucDispPllConfig |=
+								DISPPLL_CONFIG_COHERENT_MODE;
+						if (mode->clock > 165000)
+							args.v3.sInput.ucDispPllConfig |=
+								DISPPLL_CONFIG_DUAL_LINK;
+					}
+				} else if (radeon_encoder->devices & (ATOM_DEVICE_LCD_SUPPORT)) {
+					/* may want to enable SS on DP/eDP eventually */
+					args.v3.sInput.ucDispPllConfig |=
+						DISPPLL_CONFIG_SS_ENABLE;
+					if (mode->clock > 165000)
+						args.v3.sInput.ucDispPllConfig |=
+							DISPPLL_CONFIG_DUAL_LINK;
+				}
+				atom_execute_table(rdev->mode_info.atom_context,
+						   index, (uint32_t *)&args);
+				adjusted_clock = le32_to_cpu(args.v3.sOutput.ulDispPllFreq) * 10;
+				if (args.v3.sOutput.ucRefDiv) {
+					pll->flags |= RADEON_PLL_USE_REF_DIV;
+					pll->reference_div = args.v3.sOutput.ucRefDiv;
+				}
+				if (args.v3.sOutput.ucPostDiv) {
+					pll->flags |= RADEON_PLL_USE_POST_DIV;
+					pll->post_div = args.v3.sOutput.ucPostDiv;
+				}
+				break;
+			default:
+				DRM_ERROR("Unknown table version %d %d\n", frev, crev);
+				return adjusted_clock;
 			}
+			break;
+		default:
+			DRM_ERROR("Unknown table version %d %d\n", frev, crev);
+			return adjusted_clock;
+		}
+	}
+	return adjusted_clock;
+}
+
+union set_pixel_clock {
+	SET_PIXEL_CLOCK_PS_ALLOCATION base;
+	PIXEL_CLOCK_PARAMETERS v1;
+	PIXEL_CLOCK_PARAMETERS_V2 v2;
+	PIXEL_CLOCK_PARAMETERS_V3 v3;
+	PIXEL_CLOCK_PARAMETERS_V5 v5;
+};
+
+static void atombios_crtc_set_dcpll(struct drm_crtc *crtc)
+{
+	struct drm_device *dev = crtc->dev;
+	struct radeon_device *rdev = dev->dev_private;
+	u8 frev, crev;
+	int index;
+	union set_pixel_clock args;
+
+	memset(&args, 0, sizeof(args));
+
+	index = GetIndexIntoMasterTable(COMMAND, SetPixelClock);
+	atom_parse_cmd_header(rdev->mode_info.atom_context, index, &frev,
+			      &crev);
+
+	switch (frev) {
+	case 1:
+		switch (crev) {
+		case 5:
+			/* if the default dcpll clock is specified,
+			 * SetPixelClock provides the dividers
+			 */
+			args.v5.ucCRTC = ATOM_CRTC_INVALID;
+			args.v5.usPixelClock = rdev->clock.default_dispclk;
+			args.v5.ucPpll = ATOM_DCPLL;
+			break;
+		default:
+			DRM_ERROR("Unknown table version %d %d\n", frev, crev);
+			return;
+		}
+		break;
+	default:
+		DRM_ERROR("Unknown table version %d %d\n", frev, crev);
+		return;
+	}
+	atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args);
+}
+
+static void atombios_crtc_set_pll(struct drm_crtc *crtc, struct drm_display_mode *mode)
+{
+	struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc);
+	struct drm_device *dev = crtc->dev;
+	struct radeon_device *rdev = dev->dev_private;
+	struct drm_encoder *encoder = NULL;
+	struct radeon_encoder *radeon_encoder = NULL;
+	u8 frev, crev;
+	int index;
+	union set_pixel_clock args;
+	u32 pll_clock = mode->clock;
+	u32 ref_div = 0, fb_div = 0, frac_fb_div = 0, post_div = 0;
+	struct radeon_pll *pll;
+	u32 adjusted_clock;
+	int encoder_mode = 0;
+
+	memset(&args, 0, sizeof(args));
+
+	list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) {
+		if (encoder->crtc == crtc) {
 			radeon_encoder = to_radeon_encoder(encoder);
+			encoder_mode = atombios_get_encoder_mode(encoder);
 			break;
 		}
 	}
 
-	if (radeon_crtc->crtc_id == 0)
+	if (!radeon_encoder)
+		return;
+
+	switch (radeon_crtc->pll_id) {
+	case ATOM_PPLL1:
 		pll = &rdev->clock.p1pll;
-	else
+		break;
+	case ATOM_PPLL2:
 		pll = &rdev->clock.p2pll;
+		break;
+	case ATOM_DCPLL:
+	case ATOM_PPLL_INVALID:
+		pll = &rdev->clock.dcpll;
+		break;
+	}
+
+	/* adjust pixel clock as needed */
+	adjusted_clock = atombios_adjust_pll(crtc, mode, pll);
 
-	radeon_compute_pll(pll, mode->clock, &sclock, &fb_div, &frac_fb_div,
-			   &ref_div, &post_div, pll_flags);
+	radeon_compute_pll(pll, adjusted_clock, &pll_clock, &fb_div, &frac_fb_div,
+			   &ref_div, &post_div);
 
+	index = GetIndexIntoMasterTable(COMMAND, SetPixelClock);
 	atom_parse_cmd_header(rdev->mode_info.atom_context, index, &frev,
 			      &crev);
 
@@ -408,46 +674,47 @@ void atombios_crtc_set_pll(struct drm_crtc *crtc, struct drm_display_mode *mode)
 	case 1:
 		switch (crev) {
 		case 1:
-			spc1_ptr = (PIXEL_CLOCK_PARAMETERS *) & args.sPCLKInput;
-			spc1_ptr->usPixelClock = cpu_to_le16(sclock);
-			spc1_ptr->usRefDiv = cpu_to_le16(ref_div);
-			spc1_ptr->usFbDiv = cpu_to_le16(fb_div);
-			spc1_ptr->ucFracFbDiv = frac_fb_div;
-			spc1_ptr->ucPostDiv = post_div;
-			spc1_ptr->ucPpll =
-			    radeon_crtc->crtc_id ? ATOM_PPLL2 : ATOM_PPLL1;
-			spc1_ptr->ucCRTC = radeon_crtc->crtc_id;
-			spc1_ptr->ucRefDivSrc = 1;
+			args.v1.usPixelClock = cpu_to_le16(mode->clock / 10);
+			args.v1.usRefDiv = cpu_to_le16(ref_div);
+			args.v1.usFbDiv = cpu_to_le16(fb_div);
+			args.v1.ucFracFbDiv = frac_fb_div;
+			args.v1.ucPostDiv = post_div;
+			args.v1.ucPpll = radeon_crtc->pll_id;
+			args.v1.ucCRTC = radeon_crtc->crtc_id;
+			args.v1.ucRefDivSrc = 1;
 			break;
 		case 2:
-			spc2_ptr =
-			    (PIXEL_CLOCK_PARAMETERS_V2 *) & args.sPCLKInput;
-			spc2_ptr->usPixelClock = cpu_to_le16(sclock);
-			spc2_ptr->usRefDiv = cpu_to_le16(ref_div);
-			spc2_ptr->usFbDiv = cpu_to_le16(fb_div);
-			spc2_ptr->ucFracFbDiv = frac_fb_div;
-			spc2_ptr->ucPostDiv = post_div;
-			spc2_ptr->ucPpll =
-			    radeon_crtc->crtc_id ? ATOM_PPLL2 : ATOM_PPLL1;
-			spc2_ptr->ucCRTC = radeon_crtc->crtc_id;
-			spc2_ptr->ucRefDivSrc = 1;
+			args.v2.usPixelClock = cpu_to_le16(mode->clock / 10);
+			args.v2.usRefDiv = cpu_to_le16(ref_div);
+			args.v2.usFbDiv = cpu_to_le16(fb_div);
+			args.v2.ucFracFbDiv = frac_fb_div;
+			args.v2.ucPostDiv = post_div;
+			args.v2.ucPpll = radeon_crtc->pll_id;
+			args.v2.ucCRTC = radeon_crtc->crtc_id;
+			args.v2.ucRefDivSrc = 1;
 			break;
 		case 3:
-			if (!encoder)
-				return;
-			spc3_ptr =
-			    (PIXEL_CLOCK_PARAMETERS_V3 *) & args.sPCLKInput;
-			spc3_ptr->usPixelClock = cpu_to_le16(sclock);
-			spc3_ptr->usRefDiv = cpu_to_le16(ref_div);
-			spc3_ptr->usFbDiv = cpu_to_le16(fb_div);
-			spc3_ptr->ucFracFbDiv = frac_fb_div;
-			spc3_ptr->ucPostDiv = post_div;
-			spc3_ptr->ucPpll =
-			    radeon_crtc->crtc_id ? ATOM_PPLL2 : ATOM_PPLL1;
-			spc3_ptr->ucMiscInfo = (radeon_crtc->crtc_id << 2);
-			spc3_ptr->ucTransmitterId = radeon_encoder->encoder_id;
-			spc3_ptr->ucEncoderMode =
-			    atombios_get_encoder_mode(encoder);
+			args.v3.usPixelClock = cpu_to_le16(mode->clock / 10);
+			args.v3.usRefDiv = cpu_to_le16(ref_div);
+			args.v3.usFbDiv = cpu_to_le16(fb_div);
+			args.v3.ucFracFbDiv = frac_fb_div;
+			args.v3.ucPostDiv = post_div;
+			args.v3.ucPpll = radeon_crtc->pll_id;
+			args.v3.ucMiscInfo = (radeon_crtc->pll_id << 2);
+			args.v3.ucTransmitterId = radeon_encoder->encoder_id;
+			args.v3.ucEncoderMode = encoder_mode;
+			break;
+		case 5:
+			args.v5.ucCRTC = radeon_crtc->crtc_id;
+			args.v5.usPixelClock = cpu_to_le16(mode->clock / 10);
+			args.v5.ucRefDiv = ref_div;
+			args.v5.usFbDiv = cpu_to_le16(fb_div);
+			args.v5.ulFbDivDecFrac = cpu_to_le32(frac_fb_div * 100000);
+			args.v5.ucPostDiv = post_div;
+			args.v5.ucMiscInfo = 0; /* HDMI depth, etc. */
+			args.v5.ucTransmitterID = radeon_encoder->encoder_id;
+			args.v5.ucEncoderMode = encoder_mode;
+			args.v5.ucPpll = radeon_crtc->pll_id;
 			break;
 		default:
 			DRM_ERROR("Unknown table version %d %d\n", frev, crev);
@@ -459,33 +726,177 @@ void atombios_crtc_set_pll(struct drm_crtc *crtc, struct drm_display_mode *mode)
 		return;
 	}
 
-	printk("executing set pll\n");
 	atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args);
 }
 
-int atombios_crtc_set_base(struct drm_crtc *crtc, int x, int y,
-			   struct drm_framebuffer *old_fb)
+static int evergreen_crtc_set_base(struct drm_crtc *crtc, int x, int y,
+				   struct drm_framebuffer *old_fb)
 {
 	struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc);
 	struct drm_device *dev = crtc->dev;
 	struct radeon_device *rdev = dev->dev_private;
 	struct radeon_framebuffer *radeon_fb;
 	struct drm_gem_object *obj;
-	struct drm_radeon_gem_object *obj_priv;
+	struct radeon_bo *rbo;
 	uint64_t fb_location;
 	uint32_t fb_format, fb_pitch_pixels, tiling_flags;
+	int r;
 
-	if (!crtc->fb)
-		return -EINVAL;
+	/* no fb bound */
+	if (!crtc->fb) {
+		DRM_DEBUG("No FB bound\n");
+		return 0;
+	}
 
 	radeon_fb = to_radeon_framebuffer(crtc->fb);
 
+	/* Pin framebuffer & get tilling informations */
 	obj = radeon_fb->obj;
-	obj_priv = obj->driver_private;
+	rbo = obj->driver_private;
+	r = radeon_bo_reserve(rbo, false);
+	if (unlikely(r != 0))
+		return r;
+	r = radeon_bo_pin(rbo, RADEON_GEM_DOMAIN_VRAM, &fb_location);
+	if (unlikely(r != 0)) {
+		radeon_bo_unreserve(rbo);
+		return -EINVAL;
+	}
+	radeon_bo_get_tiling_flags(rbo, &tiling_flags, NULL);
+	radeon_bo_unreserve(rbo);
 
-	if (radeon_gem_object_pin(obj, RADEON_GEM_DOMAIN_VRAM, &fb_location)) {
+	switch (crtc->fb->bits_per_pixel) {
+	case 8:
+		fb_format = (EVERGREEN_GRPH_DEPTH(EVERGREEN_GRPH_DEPTH_8BPP) |
+			     EVERGREEN_GRPH_FORMAT(EVERGREEN_GRPH_FORMAT_INDEXED));
+		break;
+	case 15:
+		fb_format = (EVERGREEN_GRPH_DEPTH(EVERGREEN_GRPH_DEPTH_16BPP) |
+			     EVERGREEN_GRPH_FORMAT(EVERGREEN_GRPH_FORMAT_ARGB1555));
+		break;
+	case 16:
+		fb_format = (EVERGREEN_GRPH_DEPTH(EVERGREEN_GRPH_DEPTH_16BPP) |
+			     EVERGREEN_GRPH_FORMAT(EVERGREEN_GRPH_FORMAT_ARGB565));
+		break;
+	case 24:
+	case 32:
+		fb_format = (EVERGREEN_GRPH_DEPTH(EVERGREEN_GRPH_DEPTH_32BPP) |
+			     EVERGREEN_GRPH_FORMAT(EVERGREEN_GRPH_FORMAT_ARGB8888));
+		break;
+	default:
+		DRM_ERROR("Unsupported screen depth %d\n",
+			  crtc->fb->bits_per_pixel);
+		return -EINVAL;
+	}
+
+	switch (radeon_crtc->crtc_id) {
+	case 0:
+		WREG32(AVIVO_D1VGA_CONTROL, 0);
+		break;
+	case 1:
+		WREG32(AVIVO_D2VGA_CONTROL, 0);
+		break;
+	case 2:
+		WREG32(EVERGREEN_D3VGA_CONTROL, 0);
+		break;
+	case 3:
+		WREG32(EVERGREEN_D4VGA_CONTROL, 0);
+		break;
+	case 4:
+		WREG32(EVERGREEN_D5VGA_CONTROL, 0);
+		break;
+	case 5:
+		WREG32(EVERGREEN_D6VGA_CONTROL, 0);
+		break;
+	default:
+		break;
+	}
+
+	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS_HIGH + radeon_crtc->crtc_offset,
+	       upper_32_bits(fb_location));
+	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS_HIGH + radeon_crtc->crtc_offset,
+	       upper_32_bits(fb_location));
+	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS + radeon_crtc->crtc_offset,
+	       (u32)fb_location & EVERGREEN_GRPH_SURFACE_ADDRESS_MASK);
+	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS + radeon_crtc->crtc_offset,
+	       (u32) fb_location & EVERGREEN_GRPH_SURFACE_ADDRESS_MASK);
+	WREG32(EVERGREEN_GRPH_CONTROL + radeon_crtc->crtc_offset, fb_format);
+
+	WREG32(EVERGREEN_GRPH_SURFACE_OFFSET_X + radeon_crtc->crtc_offset, 0);
+	WREG32(EVERGREEN_GRPH_SURFACE_OFFSET_Y + radeon_crtc->crtc_offset, 0);
+	WREG32(EVERGREEN_GRPH_X_START + radeon_crtc->crtc_offset, 0);
+	WREG32(EVERGREEN_GRPH_Y_START + radeon_crtc->crtc_offset, 0);
+	WREG32(EVERGREEN_GRPH_X_END + radeon_crtc->crtc_offset, crtc->fb->width);
+	WREG32(EVERGREEN_GRPH_Y_END + radeon_crtc->crtc_offset, crtc->fb->height);
+
+	fb_pitch_pixels = crtc->fb->pitch / (crtc->fb->bits_per_pixel / 8);
+	WREG32(EVERGREEN_GRPH_PITCH + radeon_crtc->crtc_offset, fb_pitch_pixels);
+	WREG32(EVERGREEN_GRPH_ENABLE + radeon_crtc->crtc_offset, 1);
+
+	WREG32(EVERGREEN_DESKTOP_HEIGHT + radeon_crtc->crtc_offset,
+	       crtc->mode.vdisplay);
+	x &= ~3;
+	y &= ~1;
+	WREG32(EVERGREEN_VIEWPORT_START + radeon_crtc->crtc_offset,
+	       (x << 16) | y);
+	WREG32(EVERGREEN_VIEWPORT_SIZE + radeon_crtc->crtc_offset,
+	       (crtc->mode.hdisplay << 16) | crtc->mode.vdisplay);
+
+	if (crtc->mode.flags & DRM_MODE_FLAG_INTERLACE)
+		WREG32(EVERGREEN_DATA_FORMAT + radeon_crtc->crtc_offset,
+		       EVERGREEN_INTERLEAVE_EN);
+	else
+		WREG32(EVERGREEN_DATA_FORMAT + radeon_crtc->crtc_offset, 0);
+
+	if (old_fb && old_fb != crtc->fb) {
+		radeon_fb = to_radeon_framebuffer(old_fb);
+		rbo = radeon_fb->obj->driver_private;
+		r = radeon_bo_reserve(rbo, false);
+		if (unlikely(r != 0))
+			return r;
+		radeon_bo_unpin(rbo);
+		radeon_bo_unreserve(rbo);
+	}
+
+	/* Bytes per pixel may have changed */
+	radeon_bandwidth_update(rdev);
+
+	return 0;
+}
+
+static int avivo_crtc_set_base(struct drm_crtc *crtc, int x, int y,
+			       struct drm_framebuffer *old_fb)
+{
+	struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc);
+	struct drm_device *dev = crtc->dev;
+	struct radeon_device *rdev = dev->dev_private;
+	struct radeon_framebuffer *radeon_fb;
+	struct drm_gem_object *obj;
+	struct radeon_bo *rbo;
+	uint64_t fb_location;
+	uint32_t fb_format, fb_pitch_pixels, tiling_flags;
+	int r;
+
+	/* no fb bound */
+	if (!crtc->fb) {
+		DRM_DEBUG("No FB bound\n");
+		return 0;
+	}
+
+	radeon_fb = to_radeon_framebuffer(crtc->fb);
+
+	/* Pin framebuffer & get tilling informations */
+	obj = radeon_fb->obj;
+	rbo = obj->driver_private;
+	r = radeon_bo_reserve(rbo, false);
+	if (unlikely(r != 0))
+		return r;
+	r = radeon_bo_pin(rbo, RADEON_GEM_DOMAIN_VRAM, &fb_location);
+	if (unlikely(r != 0)) {
+		radeon_bo_unreserve(rbo);
 		return -EINVAL;
 	}
+	radeon_bo_get_tiling_flags(rbo, &tiling_flags, NULL);
+	radeon_bo_unreserve(rbo);
 
 	switch (crtc->fb->bits_per_pixel) {
 	case 8:
@@ -515,8 +926,6 @@ int atombios_crtc_set_base(struct drm_crtc *crtc, int x, int y,
 		return -EINVAL;
 	}
 
-	radeon_object_get_tiling_flags(obj->driver_private,
-				       &tiling_flags, NULL);
 	if (tiling_flags & RADEON_TILING_MACRO)
 		fb_format |= AVIVO_D1GRPH_MACRO_ADDRESS_MODE;
 
@@ -527,6 +936,16 @@ int atombios_crtc_set_base(struct drm_crtc *crtc, int x, int y,
 		WREG32(AVIVO_D1VGA_CONTROL, 0);
 	else
 		WREG32(AVIVO_D2VGA_CONTROL, 0);
+
+	if (rdev->family >= CHIP_RV770) {
+		if (radeon_crtc->crtc_id) {
+			WREG32(R700_D2GRPH_PRIMARY_SURFACE_ADDRESS_HIGH, 0);
+			WREG32(R700_D2GRPH_SECONDARY_SURFACE_ADDRESS_HIGH, 0);
+		} else {
+			WREG32(R700_D1GRPH_PRIMARY_SURFACE_ADDRESS_HIGH, 0);
+			WREG32(R700_D1GRPH_SECONDARY_SURFACE_ADDRESS_HIGH, 0);
+		}
+	}
 	WREG32(AVIVO_D1GRPH_PRIMARY_SURFACE_ADDRESS + radeon_crtc->crtc_offset,
 	       (u32) fb_location);
 	WREG32(AVIVO_D1GRPH_SECONDARY_SURFACE_ADDRESS +
@@ -561,147 +980,132 @@ int atombios_crtc_set_base(struct drm_crtc *crtc, int x, int y,
 
 	if (old_fb && old_fb != crtc->fb) {
 		radeon_fb = to_radeon_framebuffer(old_fb);
-		radeon_gem_object_unpin(radeon_fb->obj);
+		rbo = radeon_fb->obj->driver_private;
+		r = radeon_bo_reserve(rbo, false);
+		if (unlikely(r != 0))
+			return r;
+		radeon_bo_unpin(rbo);
+		radeon_bo_unreserve(rbo);
 	}
+
+	/* Bytes per pixel may have changed */
+	radeon_bandwidth_update(rdev);
+
 	return 0;
 }
 
-int atombios_crtc_mode_set(struct drm_crtc *crtc,
-			   struct drm_display_mode *mode,
-			   struct drm_display_mode *adjusted_mode,
-			   int x, int y, struct drm_framebuffer *old_fb)
+int atombios_crtc_set_base(struct drm_crtc *crtc, int x, int y,
+			   struct drm_framebuffer *old_fb)
 {
-	struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc);
 	struct drm_device *dev = crtc->dev;
 	struct radeon_device *rdev = dev->dev_private;
-	struct drm_encoder *encoder;
-	SET_CRTC_TIMING_PARAMETERS_PS_ALLOCATION crtc_timing;
-	int need_tv_timings = 0;
-	bool ret;
 
-	/* TODO color tiling */
-	memset(&crtc_timing, 0, sizeof(crtc_timing));
+	if (ASIC_IS_DCE4(rdev))
+		return evergreen_crtc_set_base(crtc, x, y, old_fb);
+	else if (ASIC_IS_AVIVO(rdev))
+		return avivo_crtc_set_base(crtc, x, y, old_fb);
+	else
+		return radeon_crtc_set_base(crtc, x, y, old_fb);
+}
 
-	list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) {
-		/* find tv std */
-		if (encoder->crtc == crtc) {
-			struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
+/* properly set additional regs when using atombios */
+static void radeon_legacy_atom_fixup(struct drm_crtc *crtc)
+{
+	struct drm_device *dev = crtc->dev;
+	struct radeon_device *rdev = dev->dev_private;
+	struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc);
+	u32 disp_merge_cntl;
 
-			if (radeon_encoder->active_device & ATOM_DEVICE_TV_SUPPORT) {
-				struct radeon_encoder_atom_dac *tv_dac = radeon_encoder->enc_priv;
-				if (tv_dac) {
-					if (tv_dac->tv_std == TV_STD_NTSC ||
-					    tv_dac->tv_std == TV_STD_NTSC_J ||
-					    tv_dac->tv_std == TV_STD_PAL_M)
-						need_tv_timings = 1;
-					else
-						need_tv_timings = 2;
-					break;
+	switch (radeon_crtc->crtc_id) {
+	case 0:
+		disp_merge_cntl = RREG32(RADEON_DISP_MERGE_CNTL);
+		disp_merge_cntl &= ~RADEON_DISP_RGB_OFFSET_EN;
+		WREG32(RADEON_DISP_MERGE_CNTL, disp_merge_cntl);
+		break;
+	case 1:
+		disp_merge_cntl = RREG32(RADEON_DISP2_MERGE_CNTL);
+		disp_merge_cntl &= ~RADEON_DISP2_RGB_OFFSET_EN;
+		WREG32(RADEON_DISP2_MERGE_CNTL, disp_merge_cntl);
+		WREG32(RADEON_FP_H2_SYNC_STRT_WID,   RREG32(RADEON_CRTC2_H_SYNC_STRT_WID));
+		WREG32(RADEON_FP_V2_SYNC_STRT_WID,   RREG32(RADEON_CRTC2_V_SYNC_STRT_WID));
+		break;
+	}
+}
+
+static int radeon_atom_pick_pll(struct drm_crtc *crtc)
+{
+	struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc);
+	struct drm_device *dev = crtc->dev;
+	struct radeon_device *rdev = dev->dev_private;
+	struct drm_encoder *test_encoder;
+	struct drm_crtc *test_crtc;
+	uint32_t pll_in_use = 0;
+
+	if (ASIC_IS_DCE4(rdev)) {
+		/* if crtc is driving DP and we have an ext clock, use that */
+		list_for_each_entry(test_encoder, &dev->mode_config.encoder_list, head) {
+			if (test_encoder->crtc && (test_encoder->crtc == crtc)) {
+				if (atombios_get_encoder_mode(test_encoder) == ATOM_ENCODER_MODE_DP) {
+					if (rdev->clock.dp_extclk)
+						return ATOM_PPLL_INVALID;
 				}
 			}
 		}
-	}
 
-	crtc_timing.ucCRTC = radeon_crtc->crtc_id;
-	if (need_tv_timings) {
-		ret = radeon_atom_get_tv_timings(rdev, need_tv_timings - 1,
-						 &crtc_timing, &adjusted_mode->clock);
-		if (ret == false)
-			need_tv_timings = 0;
-	}
-
-	if (!need_tv_timings) {
-		crtc_timing.usH_Total = adjusted_mode->crtc_htotal;
-		crtc_timing.usH_Disp = adjusted_mode->crtc_hdisplay;
-		crtc_timing.usH_SyncStart = adjusted_mode->crtc_hsync_start;
-		crtc_timing.usH_SyncWidth =
-			adjusted_mode->crtc_hsync_end - adjusted_mode->crtc_hsync_start;
+		/* otherwise, pick one of the plls */
+		list_for_each_entry(test_crtc, &dev->mode_config.crtc_list, head) {
+			struct radeon_crtc *radeon_test_crtc;
 
-		crtc_timing.usV_Total = adjusted_mode->crtc_vtotal;
-		crtc_timing.usV_Disp = adjusted_mode->crtc_vdisplay;
-		crtc_timing.usV_SyncStart = adjusted_mode->crtc_vsync_start;
-		crtc_timing.usV_SyncWidth =
-			adjusted_mode->crtc_vsync_end - adjusted_mode->crtc_vsync_start;
+			if (crtc == test_crtc)
+				continue;
 
-		if (adjusted_mode->flags & DRM_MODE_FLAG_NVSYNC)
-			crtc_timing.susModeMiscInfo.usAccess |= ATOM_VSYNC_POLARITY;
+			radeon_test_crtc = to_radeon_crtc(test_crtc);
+			if ((radeon_test_crtc->pll_id >= ATOM_PPLL1) &&
+			    (radeon_test_crtc->pll_id <= ATOM_PPLL2))
+				pll_in_use |= (1 << radeon_test_crtc->pll_id);
+		}
+		if (!(pll_in_use & 1))
+			return ATOM_PPLL1;
+		return ATOM_PPLL2;
+	} else
+		return radeon_crtc->crtc_id;
 
-		if (adjusted_mode->flags & DRM_MODE_FLAG_NHSYNC)
-			crtc_timing.susModeMiscInfo.usAccess |= ATOM_HSYNC_POLARITY;
+}
 
-		if (adjusted_mode->flags & DRM_MODE_FLAG_CSYNC)
-			crtc_timing.susModeMiscInfo.usAccess |= ATOM_COMPOSITESYNC;
+int atombios_crtc_mode_set(struct drm_crtc *crtc,
+			   struct drm_display_mode *mode,
+			   struct drm_display_mode *adjusted_mode,
+			   int x, int y, struct drm_framebuffer *old_fb)
+{
+	struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc);
+	struct drm_device *dev = crtc->dev;
+	struct radeon_device *rdev = dev->dev_private;
 
-		if (adjusted_mode->flags & DRM_MODE_FLAG_INTERLACE)
-			crtc_timing.susModeMiscInfo.usAccess |= ATOM_INTERLACE;
+	/* TODO color tiling */
 
-		if (adjusted_mode->flags & DRM_MODE_FLAG_DBLSCAN)
-			crtc_timing.susModeMiscInfo.usAccess |= ATOM_DOUBLE_CLOCK_MODE;
-	}
+	/* pick pll */
+	radeon_crtc->pll_id = radeon_atom_pick_pll(crtc);
 
+	atombios_set_ss(crtc, 0);
+	/* always set DCPLL */
+	if (ASIC_IS_DCE4(rdev))
+		atombios_crtc_set_dcpll(crtc);
 	atombios_crtc_set_pll(crtc, adjusted_mode);
-	atombios_crtc_set_timing(crtc, &crtc_timing);
+	atombios_set_ss(crtc, 1);
 
-	if (ASIC_IS_AVIVO(rdev))
-		atombios_crtc_set_base(crtc, x, y, old_fb);
+	if (ASIC_IS_DCE4(rdev))
+		atombios_set_crtc_dtd_timing(crtc, adjusted_mode);
+	else if (ASIC_IS_AVIVO(rdev))
+		atombios_crtc_set_timing(crtc, adjusted_mode);
 	else {
-		if (radeon_crtc->crtc_id == 0) {
-			SET_CRTC_USING_DTD_TIMING_PARAMETERS crtc_dtd_timing;
-			memset(&crtc_dtd_timing, 0, sizeof(crtc_dtd_timing));
-
-			/* setup FP shadow regs on R4xx */
-			crtc_dtd_timing.ucCRTC = radeon_crtc->crtc_id;
-			crtc_dtd_timing.usH_Size = adjusted_mode->crtc_hdisplay;
-			crtc_dtd_timing.usV_Size = adjusted_mode->crtc_vdisplay;
-			crtc_dtd_timing.usH_Blanking_Time =
-			    adjusted_mode->crtc_hblank_end -
-			    adjusted_mode->crtc_hdisplay;
-			crtc_dtd_timing.usV_Blanking_Time =
-			    adjusted_mode->crtc_vblank_end -
-			    adjusted_mode->crtc_vdisplay;
-			crtc_dtd_timing.usH_SyncOffset =
-			    adjusted_mode->crtc_hsync_start -
-			    adjusted_mode->crtc_hdisplay;
-			crtc_dtd_timing.usV_SyncOffset =
-			    adjusted_mode->crtc_vsync_start -
-			    adjusted_mode->crtc_vdisplay;
-			crtc_dtd_timing.usH_SyncWidth =
-			    adjusted_mode->crtc_hsync_end -
-			    adjusted_mode->crtc_hsync_start;
-			crtc_dtd_timing.usV_SyncWidth =
-			    adjusted_mode->crtc_vsync_end -
-			    adjusted_mode->crtc_vsync_start;
-			/* crtc_dtd_timing.ucH_Border = adjusted_mode->crtc_hborder; */
-			/* crtc_dtd_timing.ucV_Border = adjusted_mode->crtc_vborder; */
-
-			if (adjusted_mode->flags & DRM_MODE_FLAG_NVSYNC)
-				crtc_dtd_timing.susModeMiscInfo.usAccess |=
-				    ATOM_VSYNC_POLARITY;
-
-			if (adjusted_mode->flags & DRM_MODE_FLAG_NHSYNC)
-				crtc_dtd_timing.susModeMiscInfo.usAccess |=
-				    ATOM_HSYNC_POLARITY;
-
-			if (adjusted_mode->flags & DRM_MODE_FLAG_CSYNC)
-				crtc_dtd_timing.susModeMiscInfo.usAccess |=
-				    ATOM_COMPOSITESYNC;
-
-			if (adjusted_mode->flags & DRM_MODE_FLAG_INTERLACE)
-				crtc_dtd_timing.susModeMiscInfo.usAccess |=
-				    ATOM_INTERLACE;
-
-			if (adjusted_mode->flags & DRM_MODE_FLAG_DBLSCAN)
-				crtc_dtd_timing.susModeMiscInfo.usAccess |=
-				    ATOM_DOUBLE_CLOCK_MODE;
-
-			atombios_set_crtc_dtd_timing(crtc, &crtc_dtd_timing);
-		}
-		radeon_crtc_set_base(crtc, x, y, old_fb);
-		radeon_legacy_atom_set_surface(crtc);
+		atombios_crtc_set_timing(crtc, adjusted_mode);
+		if (radeon_crtc->crtc_id == 0)
+			atombios_set_crtc_dtd_timing(crtc, adjusted_mode);
+		radeon_legacy_atom_fixup(crtc);
 	}
+	atombios_crtc_set_base(crtc, x, y, old_fb);
 	atombios_overscan_setup(crtc, mode, adjusted_mode);
 	atombios_scaler_setup(crtc);
-	radeon_bandwidth_update(rdev);
 	return 0;
 }
 
@@ -716,14 +1120,14 @@ static bool atombios_crtc_mode_fixup(struct drm_crtc *crtc,
 
 static void atombios_crtc_prepare(struct drm_crtc *crtc)
 {
+	atombios_lock_crtc(crtc, ATOM_ENABLE);
 	atombios_crtc_dpms(crtc, DRM_MODE_DPMS_OFF);
-	atombios_lock_crtc(crtc, 1);
 }
 
 static void atombios_crtc_commit(struct drm_crtc *crtc)
 {
 	atombios_crtc_dpms(crtc, DRM_MODE_DPMS_ON);
-	atombios_lock_crtc(crtc, 0);
+	atombios_lock_crtc(crtc, ATOM_DISABLE);
 }
 
 static const struct drm_crtc_helper_funcs atombios_helper_funcs = {
@@ -733,13 +1137,43 @@ static const struct drm_crtc_helper_funcs atombios_helper_funcs = {
 	.mode_set_base = atombios_crtc_set_base,
 	.prepare = atombios_crtc_prepare,
 	.commit = atombios_crtc_commit,
+	.load_lut = radeon_crtc_load_lut,
 };
 
 void radeon_atombios_init_crtc(struct drm_device *dev,
 			       struct radeon_crtc *radeon_crtc)
 {
-	if (radeon_crtc->crtc_id == 1)
-		radeon_crtc->crtc_offset =
-		    AVIVO_D2CRTC_H_TOTAL - AVIVO_D1CRTC_H_TOTAL;
+	struct radeon_device *rdev = dev->dev_private;
+
+	if (ASIC_IS_DCE4(rdev)) {
+		switch (radeon_crtc->crtc_id) {
+		case 0:
+		default:
+			radeon_crtc->crtc_offset = EVERGREEN_CRTC0_REGISTER_OFFSET;
+			break;
+		case 1:
+			radeon_crtc->crtc_offset = EVERGREEN_CRTC1_REGISTER_OFFSET;
+			break;
+		case 2:
+			radeon_crtc->crtc_offset = EVERGREEN_CRTC2_REGISTER_OFFSET;
+			break;
+		case 3:
+			radeon_crtc->crtc_offset = EVERGREEN_CRTC3_REGISTER_OFFSET;
+			break;
+		case 4:
+			radeon_crtc->crtc_offset = EVERGREEN_CRTC4_REGISTER_OFFSET;
+			break;
+		case 5:
+			radeon_crtc->crtc_offset = EVERGREEN_CRTC5_REGISTER_OFFSET;
+			break;
+		}
+	} else {
+		if (radeon_crtc->crtc_id == 1)
+			radeon_crtc->crtc_offset =
+				AVIVO_D2CRTC_H_TOTAL - AVIVO_D1CRTC_H_TOTAL;
+		else
+			radeon_crtc->crtc_offset = 0;
+	}
+	radeon_crtc->pll_id = -1;
 	drm_crtc_helper_add(&radeon_crtc->base, &atombios_helper_funcs);
 }
diff --git a/drivers/gpu/drm/radeon/atombios_dp.c b/drivers/gpu/drm/radeon/atombios_dp.c
new file mode 100644
index 000000000000..8a133bda00a2
--- /dev/null
+++ b/drivers/gpu/drm/radeon/atombios_dp.c
@@ -0,0 +1,809 @@
+/*
+ * Copyright 2007-8 Advanced Micro Devices, Inc.
+ * Copyright 2008 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Dave Airlie
+ *          Alex Deucher
+ */
+#include "drmP.h"
+#include "radeon_drm.h"
+#include "radeon.h"
+
+#include "atom.h"
+#include "atom-bits.h"
+#include "drm_dp_helper.h"
+
+/* move these to drm_dp_helper.c/h */
+#define DP_LINK_CONFIGURATION_SIZE 9
+#define DP_LINK_STATUS_SIZE	   6
+#define DP_DPCD_SIZE	           8
+
+static char *voltage_names[] = {
+        "0.4V", "0.6V", "0.8V", "1.2V"
+};
+static char *pre_emph_names[] = {
+        "0dB", "3.5dB", "6dB", "9.5dB"
+};
+
+static const int dp_clocks[] = {
+	54000,  /* 1 lane, 1.62 Ghz */
+	90000,  /* 1 lane, 2.70 Ghz */
+	108000, /* 2 lane, 1.62 Ghz */
+	180000, /* 2 lane, 2.70 Ghz */
+	216000, /* 4 lane, 1.62 Ghz */
+	360000, /* 4 lane, 2.70 Ghz */
+};
+
+static const int num_dp_clocks = sizeof(dp_clocks) / sizeof(int);
+
+/* common helper functions */
+static int dp_lanes_for_mode_clock(u8 dpcd[DP_DPCD_SIZE], int mode_clock)
+{
+	int i;
+	u8 max_link_bw;
+	u8 max_lane_count;
+
+	if (!dpcd)
+		return 0;
+
+	max_link_bw = dpcd[DP_MAX_LINK_RATE];
+	max_lane_count = dpcd[DP_MAX_LANE_COUNT] & DP_MAX_LANE_COUNT_MASK;
+
+	switch (max_link_bw) {
+	case DP_LINK_BW_1_62:
+	default:
+		for (i = 0; i < num_dp_clocks; i++) {
+			if (i % 2)
+				continue;
+			switch (max_lane_count) {
+			case 1:
+				if (i > 1)
+					return 0;
+				break;
+			case 2:
+				if (i > 3)
+					return 0;
+				break;
+			case 4:
+			default:
+				break;
+			}
+			if (dp_clocks[i] > mode_clock) {
+				if (i < 2)
+					return 1;
+				else if (i < 4)
+					return 2;
+				else
+					return 4;
+			}
+		}
+		break;
+	case DP_LINK_BW_2_7:
+		for (i = 0; i < num_dp_clocks; i++) {
+			switch (max_lane_count) {
+			case 1:
+				if (i > 1)
+					return 0;
+				break;
+			case 2:
+				if (i > 3)
+					return 0;
+				break;
+			case 4:
+			default:
+				break;
+			}
+			if (dp_clocks[i] > mode_clock) {
+				if (i < 2)
+					return 1;
+				else if (i < 4)
+					return 2;
+				else
+					return 4;
+			}
+		}
+		break;
+	}
+
+	return 0;
+}
+
+static int dp_link_clock_for_mode_clock(u8 dpcd[DP_DPCD_SIZE], int mode_clock)
+{
+	int i;
+	u8 max_link_bw;
+	u8 max_lane_count;
+
+	if (!dpcd)
+		return 0;
+
+	max_link_bw = dpcd[DP_MAX_LINK_RATE];
+	max_lane_count = dpcd[DP_MAX_LANE_COUNT] & DP_MAX_LANE_COUNT_MASK;
+
+	switch (max_link_bw) {
+	case DP_LINK_BW_1_62:
+	default:
+		for (i = 0; i < num_dp_clocks; i++) {
+			if (i % 2)
+				continue;
+			switch (max_lane_count) {
+			case 1:
+				if (i > 1)
+					return 0;
+				break;
+			case 2:
+				if (i > 3)
+					return 0;
+				break;
+			case 4:
+			default:
+				break;
+			}
+			if (dp_clocks[i] > mode_clock)
+				return 162000;
+		}
+		break;
+	case DP_LINK_BW_2_7:
+		for (i = 0; i < num_dp_clocks; i++) {
+			switch (max_lane_count) {
+			case 1:
+				if (i > 1)
+					return 0;
+				break;
+			case 2:
+				if (i > 3)
+					return 0;
+				break;
+			case 4:
+			default:
+				break;
+			}
+			if (dp_clocks[i] > mode_clock)
+				return (i % 2) ? 270000 : 162000;
+		}
+	}
+
+	return 0;
+}
+
+int dp_mode_valid(u8 dpcd[DP_DPCD_SIZE], int mode_clock)
+{
+	int lanes = dp_lanes_for_mode_clock(dpcd, mode_clock);
+	int bw = dp_lanes_for_mode_clock(dpcd, mode_clock);
+
+	if ((lanes == 0) || (bw == 0))
+		return MODE_CLOCK_HIGH;
+
+	return MODE_OK;
+}
+
+static u8 dp_link_status(u8 link_status[DP_LINK_STATUS_SIZE], int r)
+{
+	return link_status[r - DP_LANE0_1_STATUS];
+}
+
+static u8 dp_get_lane_status(u8 link_status[DP_LINK_STATUS_SIZE],
+			     int lane)
+{
+	int i = DP_LANE0_1_STATUS + (lane >> 1);
+	int s = (lane & 1) * 4;
+	u8 l = dp_link_status(link_status, i);
+	return (l >> s) & 0xf;
+}
+
+static bool dp_clock_recovery_ok(u8 link_status[DP_LINK_STATUS_SIZE],
+				 int lane_count)
+{
+	int lane;
+	u8 lane_status;
+
+	for (lane = 0; lane < lane_count; lane++) {
+		lane_status = dp_get_lane_status(link_status, lane);
+		if ((lane_status & DP_LANE_CR_DONE) == 0)
+			return false;
+	}
+	return true;
+}
+
+static bool dp_channel_eq_ok(u8 link_status[DP_LINK_STATUS_SIZE],
+			     int lane_count)
+{
+	u8 lane_align;
+	u8 lane_status;
+	int lane;
+
+	lane_align = dp_link_status(link_status,
+				    DP_LANE_ALIGN_STATUS_UPDATED);
+	if ((lane_align & DP_INTERLANE_ALIGN_DONE) == 0)
+		return false;
+	for (lane = 0; lane < lane_count; lane++) {
+		lane_status = dp_get_lane_status(link_status, lane);
+		if ((lane_status & DP_CHANNEL_EQ_BITS) != DP_CHANNEL_EQ_BITS)
+			return false;
+	}
+	return true;
+}
+
+static u8 dp_get_adjust_request_voltage(uint8_t link_status[DP_LINK_STATUS_SIZE],
+					int lane)
+
+{
+	int i = DP_ADJUST_REQUEST_LANE0_1 + (lane >> 1);
+	int s = ((lane & 1) ?
+		 DP_ADJUST_VOLTAGE_SWING_LANE1_SHIFT :
+		 DP_ADJUST_VOLTAGE_SWING_LANE0_SHIFT);
+	u8 l = dp_link_status(link_status, i);
+
+	return ((l >> s) & 0x3) << DP_TRAIN_VOLTAGE_SWING_SHIFT;
+}
+
+static u8 dp_get_adjust_request_pre_emphasis(uint8_t link_status[DP_LINK_STATUS_SIZE],
+					     int lane)
+{
+	int i = DP_ADJUST_REQUEST_LANE0_1 + (lane >> 1);
+	int s = ((lane & 1) ?
+		 DP_ADJUST_PRE_EMPHASIS_LANE1_SHIFT :
+		 DP_ADJUST_PRE_EMPHASIS_LANE0_SHIFT);
+	u8 l = dp_link_status(link_status, i);
+
+	return ((l >> s) & 0x3) << DP_TRAIN_PRE_EMPHASIS_SHIFT;
+}
+
+/* XXX fix me -- chip specific */
+#define DP_VOLTAGE_MAX         DP_TRAIN_VOLTAGE_SWING_1200
+static u8 dp_pre_emphasis_max(u8 voltage_swing)
+{
+	switch (voltage_swing & DP_TRAIN_VOLTAGE_SWING_MASK) {
+	case DP_TRAIN_VOLTAGE_SWING_400:
+		return DP_TRAIN_PRE_EMPHASIS_6;
+	case DP_TRAIN_VOLTAGE_SWING_600:
+		return DP_TRAIN_PRE_EMPHASIS_6;
+	case DP_TRAIN_VOLTAGE_SWING_800:
+		return DP_TRAIN_PRE_EMPHASIS_3_5;
+	case DP_TRAIN_VOLTAGE_SWING_1200:
+	default:
+		return DP_TRAIN_PRE_EMPHASIS_0;
+	}
+}
+
+static void dp_get_adjust_train(u8 link_status[DP_LINK_STATUS_SIZE],
+				int lane_count,
+				u8 train_set[4])
+{
+	u8 v = 0;
+	u8 p = 0;
+	int lane;
+
+	for (lane = 0; lane < lane_count; lane++) {
+		u8 this_v = dp_get_adjust_request_voltage(link_status, lane);
+		u8 this_p = dp_get_adjust_request_pre_emphasis(link_status, lane);
+
+		DRM_DEBUG("requested signal parameters: lane %d voltage %s pre_emph %s\n",
+			  lane,
+			  voltage_names[this_v >> DP_TRAIN_VOLTAGE_SWING_SHIFT],
+			  pre_emph_names[this_p >> DP_TRAIN_PRE_EMPHASIS_SHIFT]);
+
+		if (this_v > v)
+			v = this_v;
+		if (this_p > p)
+			p = this_p;
+	}
+
+	if (v >= DP_VOLTAGE_MAX)
+		v = DP_VOLTAGE_MAX | DP_TRAIN_MAX_SWING_REACHED;
+
+	if (p >= dp_pre_emphasis_max(v))
+		p = dp_pre_emphasis_max(v) | DP_TRAIN_MAX_PRE_EMPHASIS_REACHED;
+
+	DRM_DEBUG("using signal parameters: voltage %s pre_emph %s\n",
+		  voltage_names[(v & DP_TRAIN_VOLTAGE_SWING_MASK) >> DP_TRAIN_VOLTAGE_SWING_SHIFT],
+		  pre_emph_names[(p & DP_TRAIN_PRE_EMPHASIS_MASK) >> DP_TRAIN_PRE_EMPHASIS_SHIFT]);
+
+	for (lane = 0; lane < 4; lane++)
+		train_set[lane] = v | p;
+}
+
+union aux_channel_transaction {
+	PROCESS_AUX_CHANNEL_TRANSACTION_PS_ALLOCATION v1;
+	PROCESS_AUX_CHANNEL_TRANSACTION_PARAMETERS_V2 v2;
+};
+
+/* radeon aux chan functions */
+bool radeon_process_aux_ch(struct radeon_i2c_chan *chan, u8 *req_bytes,
+			   int num_bytes, u8 *read_byte,
+			   u8 read_buf_len, u8 delay)
+{
+	struct drm_device *dev = chan->dev;
+	struct radeon_device *rdev = dev->dev_private;
+	union aux_channel_transaction args;
+	int index = GetIndexIntoMasterTable(COMMAND, ProcessAuxChannelTransaction);
+	unsigned char *base;
+	int retry_count = 0;
+
+	memset(&args, 0, sizeof(args));
+
+	base = (unsigned char *)rdev->mode_info.atom_context->scratch;
+
+retry:
+	memcpy(base, req_bytes, num_bytes);
+
+	args.v1.lpAuxRequest = 0;
+	args.v1.lpDataOut = 16;
+	args.v1.ucDataOutLen = 0;
+	args.v1.ucChannelID = chan->rec.i2c_id;
+	args.v1.ucDelay = delay / 10;
+	if (ASIC_IS_DCE4(rdev))
+		args.v2.ucHPD_ID = chan->rec.hpd_id;
+
+	atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args);
+
+	if (args.v1.ucReplyStatus && !args.v1.ucDataOutLen) {
+		if (args.v1.ucReplyStatus == 0x20 && retry_count++ < 10)
+			goto retry;
+		DRM_DEBUG("failed to get auxch %02x%02x %02x %02x 0x%02x %02x after %d retries\n",
+			  req_bytes[1], req_bytes[0], req_bytes[2], req_bytes[3],
+			  chan->rec.i2c_id, args.v1.ucReplyStatus, retry_count);
+		return false;
+	}
+
+	if (args.v1.ucDataOutLen && read_byte && read_buf_len) {
+		if (read_buf_len < args.v1.ucDataOutLen) {
+			DRM_ERROR("Buffer to small for return answer %d %d\n",
+				  read_buf_len, args.v1.ucDataOutLen);
+			return false;
+		}
+		{
+			int len = min(read_buf_len, args.v1.ucDataOutLen);
+			memcpy(read_byte, base + 16, len);
+		}
+	}
+	return true;
+}
+
+bool radeon_dp_aux_native_write(struct radeon_connector *radeon_connector, uint16_t address,
+				uint8_t send_bytes, uint8_t *send)
+{
+	struct radeon_connector_atom_dig *dig_connector = radeon_connector->con_priv;
+	u8 msg[20];
+	u8 msg_len, dp_msg_len;
+	bool ret;
+
+	dp_msg_len = 4;
+	msg[0] = address;
+	msg[1] = address >> 8;
+	msg[2] = AUX_NATIVE_WRITE << 4;
+	dp_msg_len += send_bytes;
+	msg[3] = (dp_msg_len << 4) | (send_bytes - 1);
+
+	if (send_bytes > 16)
+		return false;
+
+	memcpy(&msg[4], send, send_bytes);
+	msg_len = 4 + send_bytes;
+	ret = radeon_process_aux_ch(dig_connector->dp_i2c_bus, msg, msg_len, NULL, 0, 0);
+	return ret;
+}
+
+bool radeon_dp_aux_native_read(struct radeon_connector *radeon_connector, uint16_t address,
+			       uint8_t delay, uint8_t expected_bytes,
+			       uint8_t *read_p)
+{
+	struct radeon_connector_atom_dig *dig_connector = radeon_connector->con_priv;
+	u8 msg[20];
+	u8 msg_len, dp_msg_len;
+	bool ret = false;
+	msg_len = 4;
+	dp_msg_len = 4;
+	msg[0] = address;
+	msg[1] = address >> 8;
+	msg[2] = AUX_NATIVE_READ << 4;
+	msg[3] = (dp_msg_len) << 4;
+	msg[3] |= expected_bytes - 1;
+
+	ret = radeon_process_aux_ch(dig_connector->dp_i2c_bus, msg, msg_len, read_p, expected_bytes, delay);
+	return ret;
+}
+
+/* radeon dp functions */
+static u8 radeon_dp_encoder_service(struct radeon_device *rdev, int action, int dp_clock,
+				    uint8_t ucconfig, uint8_t lane_num)
+{
+	DP_ENCODER_SERVICE_PARAMETERS args;
+	int index = GetIndexIntoMasterTable(COMMAND, DPEncoderService);
+
+	memset(&args, 0, sizeof(args));
+	args.ucLinkClock = dp_clock / 10;
+	args.ucConfig = ucconfig;
+	args.ucAction = action;
+	args.ucLaneNum = lane_num;
+	args.ucStatus = 0;
+
+	atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args);
+	return args.ucStatus;
+}
+
+u8 radeon_dp_getsinktype(struct radeon_connector *radeon_connector)
+{
+	struct radeon_connector_atom_dig *dig_connector = radeon_connector->con_priv;
+	struct drm_device *dev = radeon_connector->base.dev;
+	struct radeon_device *rdev = dev->dev_private;
+
+	return radeon_dp_encoder_service(rdev, ATOM_DP_ACTION_GET_SINK_TYPE, 0,
+					 dig_connector->dp_i2c_bus->rec.i2c_id, 0);
+}
+
+bool radeon_dp_getdpcd(struct radeon_connector *radeon_connector)
+{
+	struct radeon_connector_atom_dig *dig_connector = radeon_connector->con_priv;
+	u8 msg[25];
+	int ret;
+
+	ret = radeon_dp_aux_native_read(radeon_connector, DP_DPCD_REV, 0, 8, msg);
+	if (ret) {
+		memcpy(dig_connector->dpcd, msg, 8);
+		{
+			int i;
+			DRM_DEBUG("DPCD: ");
+			for (i = 0; i < 8; i++)
+				DRM_DEBUG("%02x ", msg[i]);
+			DRM_DEBUG("\n");
+		}
+		return true;
+	}
+	dig_connector->dpcd[0] = 0;
+	return false;
+}
+
+void radeon_dp_set_link_config(struct drm_connector *connector,
+			       struct drm_display_mode *mode)
+{
+	struct radeon_connector *radeon_connector;
+	struct radeon_connector_atom_dig *dig_connector;
+
+	if ((connector->connector_type != DRM_MODE_CONNECTOR_DisplayPort) &&
+	    (connector->connector_type != DRM_MODE_CONNECTOR_eDP))
+		return;
+
+	radeon_connector = to_radeon_connector(connector);
+	if (!radeon_connector->con_priv)
+		return;
+	dig_connector = radeon_connector->con_priv;
+
+	dig_connector->dp_clock =
+		dp_link_clock_for_mode_clock(dig_connector->dpcd, mode->clock);
+	dig_connector->dp_lane_count =
+		dp_lanes_for_mode_clock(dig_connector->dpcd, mode->clock);
+}
+
+int radeon_dp_mode_valid_helper(struct radeon_connector *radeon_connector,
+				struct drm_display_mode *mode)
+{
+	struct radeon_connector_atom_dig *dig_connector = radeon_connector->con_priv;
+
+	return dp_mode_valid(dig_connector->dpcd, mode->clock);
+}
+
+static bool atom_dp_get_link_status(struct radeon_connector *radeon_connector,
+				    u8 link_status[DP_LINK_STATUS_SIZE])
+{
+	int ret;
+	ret = radeon_dp_aux_native_read(radeon_connector, DP_LANE0_1_STATUS, 100,
+					DP_LINK_STATUS_SIZE, link_status);
+	if (!ret) {
+		DRM_ERROR("displayport link status failed\n");
+		return false;
+	}
+
+	DRM_DEBUG("link status %02x %02x %02x %02x %02x %02x\n",
+		  link_status[0], link_status[1], link_status[2],
+		  link_status[3], link_status[4], link_status[5]);
+	return true;
+}
+
+bool radeon_dp_needs_link_train(struct radeon_connector *radeon_connector)
+{
+	struct radeon_connector_atom_dig *dig_connector = radeon_connector->con_priv;
+	u8 link_status[DP_LINK_STATUS_SIZE];
+
+	if (!atom_dp_get_link_status(radeon_connector, link_status))
+		return false;
+	if (dp_channel_eq_ok(link_status, dig_connector->dp_lane_count))
+		return false;
+	return true;
+}
+
+static void dp_set_power(struct radeon_connector *radeon_connector, u8 power_state)
+{
+	struct radeon_connector_atom_dig *dig_connector = radeon_connector->con_priv;
+
+	if (dig_connector->dpcd[0] >= 0x11) {
+		radeon_dp_aux_native_write(radeon_connector, DP_SET_POWER, 1,
+					   &power_state);
+	}
+}
+
+static void dp_set_downspread(struct radeon_connector *radeon_connector, u8 downspread)
+{
+	radeon_dp_aux_native_write(radeon_connector, DP_DOWNSPREAD_CTRL, 1,
+				   &downspread);
+}
+
+static void dp_set_link_bw_lanes(struct radeon_connector *radeon_connector,
+				 u8 link_configuration[DP_LINK_CONFIGURATION_SIZE])
+{
+	radeon_dp_aux_native_write(radeon_connector, DP_LINK_BW_SET, 2,
+				   link_configuration);
+}
+
+static void dp_update_dpvs_emph(struct radeon_connector *radeon_connector,
+				struct drm_encoder *encoder,
+				u8 train_set[4])
+{
+	struct radeon_connector_atom_dig *dig_connector = radeon_connector->con_priv;
+	int i;
+
+	for (i = 0; i < dig_connector->dp_lane_count; i++)
+		atombios_dig_transmitter_setup(encoder,
+					       ATOM_TRANSMITTER_ACTION_SETUP_VSEMPH,
+					       i, train_set[i]);
+
+	radeon_dp_aux_native_write(radeon_connector, DP_TRAINING_LANE0_SET,
+				   dig_connector->dp_lane_count, train_set);
+}
+
+static void dp_set_training(struct radeon_connector *radeon_connector,
+			    u8 training)
+{
+	radeon_dp_aux_native_write(radeon_connector, DP_TRAINING_PATTERN_SET,
+				   1, &training);
+}
+
+void dp_link_train(struct drm_encoder *encoder,
+		   struct drm_connector *connector)
+{
+	struct drm_device *dev = encoder->dev;
+	struct radeon_device *rdev = dev->dev_private;
+	struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
+	struct radeon_encoder_atom_dig *dig;
+	struct radeon_connector *radeon_connector;
+	struct radeon_connector_atom_dig *dig_connector;
+	int enc_id = 0;
+	bool clock_recovery, channel_eq;
+	u8 link_status[DP_LINK_STATUS_SIZE];
+	u8 link_configuration[DP_LINK_CONFIGURATION_SIZE];
+	u8 tries, voltage;
+	u8 train_set[4];
+	int i;
+
+	if ((connector->connector_type != DRM_MODE_CONNECTOR_DisplayPort) &&
+	    (connector->connector_type != DRM_MODE_CONNECTOR_eDP))
+		return;
+
+	if (!radeon_encoder->enc_priv)
+		return;
+	dig = radeon_encoder->enc_priv;
+
+	radeon_connector = to_radeon_connector(connector);
+	if (!radeon_connector->con_priv)
+		return;
+	dig_connector = radeon_connector->con_priv;
+
+	if (dig->dig_encoder)
+		enc_id |= ATOM_DP_CONFIG_DIG2_ENCODER;
+	else
+		enc_id |= ATOM_DP_CONFIG_DIG1_ENCODER;
+	if (dig_connector->linkb)
+		enc_id |= ATOM_DP_CONFIG_LINK_B;
+	else
+		enc_id |= ATOM_DP_CONFIG_LINK_A;
+
+	memset(link_configuration, 0, DP_LINK_CONFIGURATION_SIZE);
+	if (dig_connector->dp_clock == 270000)
+		link_configuration[0] = DP_LINK_BW_2_7;
+	else
+		link_configuration[0] = DP_LINK_BW_1_62;
+	link_configuration[1] = dig_connector->dp_lane_count;
+	if (dig_connector->dpcd[0] >= 0x11)
+		link_configuration[1] |= DP_LANE_COUNT_ENHANCED_FRAME_EN;
+
+	/* power up the sink */
+	dp_set_power(radeon_connector, DP_SET_POWER_D0);
+	/* disable the training pattern on the sink */
+	dp_set_training(radeon_connector, DP_TRAINING_PATTERN_DISABLE);
+	/* set link bw and lanes on the sink */
+	dp_set_link_bw_lanes(radeon_connector, link_configuration);
+	/* disable downspread on the sink */
+	dp_set_downspread(radeon_connector, 0);
+	if (ASIC_IS_DCE4(rdev)) {
+		/* start training on the source */
+		atombios_dig_encoder_setup(encoder, ATOM_ENCODER_CMD_DP_LINK_TRAINING_START);
+		/* set training pattern 1 on the source */
+		atombios_dig_encoder_setup(encoder, ATOM_ENCODER_CMD_DP_LINK_TRAINING_PATTERN1);
+	} else {
+		/* start training on the source */
+		radeon_dp_encoder_service(rdev, ATOM_DP_ACTION_TRAINING_START,
+					  dig_connector->dp_clock, enc_id, 0);
+		/* set training pattern 1 on the source */
+		radeon_dp_encoder_service(rdev, ATOM_DP_ACTION_TRAINING_PATTERN_SEL,
+					  dig_connector->dp_clock, enc_id, 0);
+	}
+
+	/* set initial vs/emph */
+	memset(train_set, 0, 4);
+	udelay(400);
+	/* set training pattern 1 on the sink */
+	dp_set_training(radeon_connector, DP_TRAINING_PATTERN_1);
+
+	dp_update_dpvs_emph(radeon_connector, encoder, train_set);
+
+	/* clock recovery loop */
+	clock_recovery = false;
+	tries = 0;
+	voltage = 0xff;
+	for (;;) {
+		udelay(100);
+		if (!atom_dp_get_link_status(radeon_connector, link_status))
+			break;
+
+		if (dp_clock_recovery_ok(link_status, dig_connector->dp_lane_count)) {
+			clock_recovery = true;
+			break;
+		}
+
+		for (i = 0; i < dig_connector->dp_lane_count; i++) {
+			if ((train_set[i] & DP_TRAIN_MAX_SWING_REACHED) == 0)
+				break;
+		}
+		if (i == dig_connector->dp_lane_count) {
+			DRM_ERROR("clock recovery reached max voltage\n");
+			break;
+		}
+
+		if ((train_set[0] & DP_TRAIN_VOLTAGE_SWING_MASK) == voltage) {
+			++tries;
+			if (tries == 5) {
+				DRM_ERROR("clock recovery tried 5 times\n");
+				break;
+			}
+		} else
+			tries = 0;
+
+		voltage = train_set[0] & DP_TRAIN_VOLTAGE_SWING_MASK;
+
+		/* Compute new train_set as requested by sink */
+		dp_get_adjust_train(link_status, dig_connector->dp_lane_count, train_set);
+		dp_update_dpvs_emph(radeon_connector, encoder, train_set);
+	}
+	if (!clock_recovery)
+		DRM_ERROR("clock recovery failed\n");
+	else
+		DRM_DEBUG("clock recovery at voltage %d pre-emphasis %d\n",
+			  train_set[0] & DP_TRAIN_VOLTAGE_SWING_MASK,
+			  (train_set[0] & DP_TRAIN_PRE_EMPHASIS_MASK) >>
+			  DP_TRAIN_PRE_EMPHASIS_SHIFT);
+
+
+	/* set training pattern 2 on the sink */
+	dp_set_training(radeon_connector, DP_TRAINING_PATTERN_2);
+	/* set training pattern 2 on the source */
+	if (ASIC_IS_DCE4(rdev))
+		atombios_dig_encoder_setup(encoder, ATOM_ENCODER_CMD_DP_LINK_TRAINING_PATTERN2);
+	else
+		radeon_dp_encoder_service(rdev, ATOM_DP_ACTION_TRAINING_PATTERN_SEL,
+					  dig_connector->dp_clock, enc_id, 1);
+
+	/* channel equalization loop */
+	tries = 0;
+	channel_eq = false;
+	for (;;) {
+		udelay(400);
+		if (!atom_dp_get_link_status(radeon_connector, link_status))
+			break;
+
+		if (dp_channel_eq_ok(link_status, dig_connector->dp_lane_count)) {
+			channel_eq = true;
+			break;
+		}
+
+		/* Try 5 times */
+		if (tries > 5) {
+			DRM_ERROR("channel eq failed: 5 tries\n");
+			break;
+		}
+
+		/* Compute new train_set as requested by sink */
+		dp_get_adjust_train(link_status, dig_connector->dp_lane_count, train_set);
+		dp_update_dpvs_emph(radeon_connector, encoder, train_set);
+
+		tries++;
+	}
+
+	if (!channel_eq)
+		DRM_ERROR("channel eq failed\n");
+	else
+		DRM_DEBUG("channel eq at voltage %d pre-emphasis %d\n",
+			  train_set[0] & DP_TRAIN_VOLTAGE_SWING_MASK,
+			  (train_set[0] & DP_TRAIN_PRE_EMPHASIS_MASK)
+			  >> DP_TRAIN_PRE_EMPHASIS_SHIFT);
+
+	/* disable the training pattern on the sink */
+	if (ASIC_IS_DCE4(rdev))
+		atombios_dig_encoder_setup(encoder, ATOM_ENCODER_CMD_DP_LINK_TRAINING_COMPLETE);
+	else
+		radeon_dp_encoder_service(rdev, ATOM_DP_ACTION_TRAINING_COMPLETE,
+					  dig_connector->dp_clock, enc_id, 0);
+
+	radeon_dp_encoder_service(rdev, ATOM_DP_ACTION_TRAINING_COMPLETE,
+				  dig_connector->dp_clock, enc_id, 0);
+}
+
+int radeon_dp_i2c_aux_ch(struct i2c_adapter *adapter, int mode,
+			 uint8_t write_byte, uint8_t *read_byte)
+{
+	struct i2c_algo_dp_aux_data *algo_data = adapter->algo_data;
+	struct radeon_i2c_chan *auxch = (struct radeon_i2c_chan *)adapter;
+	int ret = 0;
+	uint16_t address = algo_data->address;
+	uint8_t msg[5];
+	uint8_t reply[2];
+	int msg_len, dp_msg_len;
+	int reply_bytes;
+
+	/* Set up the command byte */
+	if (mode & MODE_I2C_READ)
+		msg[2] = AUX_I2C_READ << 4;
+	else
+		msg[2] = AUX_I2C_WRITE << 4;
+
+	if (!(mode & MODE_I2C_STOP))
+		msg[2] |= AUX_I2C_MOT << 4;
+
+	msg[0] = address;
+	msg[1] = address >> 8;
+
+	reply_bytes = 1;
+
+	msg_len = 4;
+	dp_msg_len = 3;
+	switch (mode) {
+	case MODE_I2C_WRITE:
+		msg[4] = write_byte;
+		msg_len++;
+		dp_msg_len += 2;
+		break;
+	case MODE_I2C_READ:
+		dp_msg_len += 1;
+		break;
+	default:
+		break;
+	}
+
+	msg[3] = (dp_msg_len) << 4;
+	ret = radeon_process_aux_ch(auxch, msg, msg_len, reply, reply_bytes, 0);
+
+	if (ret) {
+		if (read_byte)
+			*read_byte = reply[0];
+		return reply_bytes;
+	}
+	return -EREMOTEIO;
+}
+
diff --git a/drivers/gpu/drm/radeon/avivod.h b/drivers/gpu/drm/radeon/avivod.h
index e2b92c445bab..3c391e7e9fd4 100644
--- a/drivers/gpu/drm/radeon/avivod.h
+++ b/drivers/gpu/drm/radeon/avivod.h
@@ -30,11 +30,13 @@
 
 #define	D1CRTC_CONTROL					0x6080
 #define		CRTC_EN						(1 << 0)
+#define	D1CRTC_STATUS					0x609c
 #define	D1CRTC_UPDATE_LOCK				0x60E8
 #define	D1GRPH_PRIMARY_SURFACE_ADDRESS			0x6110
 #define	D1GRPH_SECONDARY_SURFACE_ADDRESS		0x6118
 
 #define	D2CRTC_CONTROL					0x6880
+#define	D2CRTC_STATUS					0x689c
 #define	D2CRTC_UPDATE_LOCK				0x68E8
 #define	D2GRPH_PRIMARY_SURFACE_ADDRESS			0x6910
 #define	D2GRPH_SECONDARY_SURFACE_ADDRESS		0x6918
@@ -57,13 +59,4 @@
 #define	VGA_RENDER_CONTROL				0x0300
 #define		VGA_VSTATUS_CNTL_MASK				0x00030000
 
-/* AVIVO disable VGA rendering */
-static inline void radeon_avivo_vga_render_disable(struct radeon_device *rdev)
-{
-	u32 vga_render;
-	vga_render = RREG32(VGA_RENDER_CONTROL);
-	vga_render &= ~VGA_VSTATUS_CNTL_MASK;
-	WREG32(VGA_RENDER_CONTROL, vga_render);
-}
-
 #endif
diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c
new file mode 100644
index 000000000000..bd2e7aa85c1d
--- /dev/null
+++ b/drivers/gpu/drm/radeon/evergreen.c
@@ -0,0 +1,767 @@
+/*
+ * Copyright 2010 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Alex Deucher
+ */
+#include <linux/firmware.h>
+#include <linux/platform_device.h>
+#include "drmP.h"
+#include "radeon.h"
+#include "radeon_drm.h"
+#include "rv770d.h"
+#include "atom.h"
+#include "avivod.h"
+#include "evergreen_reg.h"
+
+static void evergreen_gpu_init(struct radeon_device *rdev);
+void evergreen_fini(struct radeon_device *rdev);
+
+bool evergreen_hpd_sense(struct radeon_device *rdev, enum radeon_hpd_id hpd)
+{
+	bool connected = false;
+	/* XXX */
+	return connected;
+}
+
+void evergreen_hpd_set_polarity(struct radeon_device *rdev,
+				enum radeon_hpd_id hpd)
+{
+	/* XXX */
+}
+
+void evergreen_hpd_init(struct radeon_device *rdev)
+{
+	/* XXX */
+}
+
+
+void evergreen_bandwidth_update(struct radeon_device *rdev)
+{
+	/* XXX */
+}
+
+void evergreen_hpd_fini(struct radeon_device *rdev)
+{
+	/* XXX */
+}
+
+static int evergreen_mc_wait_for_idle(struct radeon_device *rdev)
+{
+	unsigned i;
+	u32 tmp;
+
+	for (i = 0; i < rdev->usec_timeout; i++) {
+		/* read MC_STATUS */
+		tmp = RREG32(SRBM_STATUS) & 0x1F00;
+		if (!tmp)
+			return 0;
+		udelay(1);
+	}
+	return -1;
+}
+
+/*
+ * GART
+ */
+int evergreen_pcie_gart_enable(struct radeon_device *rdev)
+{
+	u32 tmp;
+	int r, i;
+
+	if (rdev->gart.table.vram.robj == NULL) {
+		dev_err(rdev->dev, "No VRAM object for PCIE GART.\n");
+		return -EINVAL;
+	}
+	r = radeon_gart_table_vram_pin(rdev);
+	if (r)
+		return r;
+	radeon_gart_restore(rdev);
+	/* Setup L2 cache */
+	WREG32(VM_L2_CNTL, ENABLE_L2_CACHE | ENABLE_L2_FRAGMENT_PROCESSING |
+				ENABLE_L2_PTE_CACHE_LRU_UPDATE_BY_WRITE |
+				EFFECTIVE_L2_QUEUE_SIZE(7));
+	WREG32(VM_L2_CNTL2, 0);
+	WREG32(VM_L2_CNTL3, BANK_SELECT(0) | CACHE_UPDATE_MODE(2));
+	/* Setup TLB control */
+	tmp = ENABLE_L1_TLB | ENABLE_L1_FRAGMENT_PROCESSING |
+		SYSTEM_ACCESS_MODE_NOT_IN_SYS |
+		SYSTEM_APERTURE_UNMAPPED_ACCESS_PASS_THRU |
+		EFFECTIVE_L1_TLB_SIZE(5) | EFFECTIVE_L1_QUEUE_SIZE(5);
+	WREG32(MC_VM_MD_L1_TLB0_CNTL, tmp);
+	WREG32(MC_VM_MD_L1_TLB1_CNTL, tmp);
+	WREG32(MC_VM_MD_L1_TLB2_CNTL, tmp);
+	WREG32(MC_VM_MB_L1_TLB0_CNTL, tmp);
+	WREG32(MC_VM_MB_L1_TLB1_CNTL, tmp);
+	WREG32(MC_VM_MB_L1_TLB2_CNTL, tmp);
+	WREG32(MC_VM_MB_L1_TLB3_CNTL, tmp);
+	WREG32(VM_CONTEXT0_PAGE_TABLE_START_ADDR, rdev->mc.gtt_start >> 12);
+	WREG32(VM_CONTEXT0_PAGE_TABLE_END_ADDR, rdev->mc.gtt_end >> 12);
+	WREG32(VM_CONTEXT0_PAGE_TABLE_BASE_ADDR, rdev->gart.table_addr >> 12);
+	WREG32(VM_CONTEXT0_CNTL, ENABLE_CONTEXT | PAGE_TABLE_DEPTH(0) |
+				RANGE_PROTECTION_FAULT_ENABLE_DEFAULT);
+	WREG32(VM_CONTEXT0_PROTECTION_FAULT_DEFAULT_ADDR,
+			(u32)(rdev->dummy_page.addr >> 12));
+	for (i = 1; i < 7; i++)
+		WREG32(VM_CONTEXT0_CNTL + (i * 4), 0);
+
+	r600_pcie_gart_tlb_flush(rdev);
+	rdev->gart.ready = true;
+	return 0;
+}
+
+void evergreen_pcie_gart_disable(struct radeon_device *rdev)
+{
+	u32 tmp;
+	int i, r;
+
+	/* Disable all tables */
+	for (i = 0; i < 7; i++)
+		WREG32(VM_CONTEXT0_CNTL + (i * 4), 0);
+
+	/* Setup L2 cache */
+	WREG32(VM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING |
+				EFFECTIVE_L2_QUEUE_SIZE(7));
+	WREG32(VM_L2_CNTL2, 0);
+	WREG32(VM_L2_CNTL3, BANK_SELECT(0) | CACHE_UPDATE_MODE(2));
+	/* Setup TLB control */
+	tmp = EFFECTIVE_L1_TLB_SIZE(5) | EFFECTIVE_L1_QUEUE_SIZE(5);
+	WREG32(MC_VM_MD_L1_TLB0_CNTL, tmp);
+	WREG32(MC_VM_MD_L1_TLB1_CNTL, tmp);
+	WREG32(MC_VM_MD_L1_TLB2_CNTL, tmp);
+	WREG32(MC_VM_MB_L1_TLB0_CNTL, tmp);
+	WREG32(MC_VM_MB_L1_TLB1_CNTL, tmp);
+	WREG32(MC_VM_MB_L1_TLB2_CNTL, tmp);
+	WREG32(MC_VM_MB_L1_TLB3_CNTL, tmp);
+	if (rdev->gart.table.vram.robj) {
+		r = radeon_bo_reserve(rdev->gart.table.vram.robj, false);
+		if (likely(r == 0)) {
+			radeon_bo_kunmap(rdev->gart.table.vram.robj);
+			radeon_bo_unpin(rdev->gart.table.vram.robj);
+			radeon_bo_unreserve(rdev->gart.table.vram.robj);
+		}
+	}
+}
+
+void evergreen_pcie_gart_fini(struct radeon_device *rdev)
+{
+	evergreen_pcie_gart_disable(rdev);
+	radeon_gart_table_vram_free(rdev);
+	radeon_gart_fini(rdev);
+}
+
+
+void evergreen_agp_enable(struct radeon_device *rdev)
+{
+	u32 tmp;
+	int i;
+
+	/* Setup L2 cache */
+	WREG32(VM_L2_CNTL, ENABLE_L2_CACHE | ENABLE_L2_FRAGMENT_PROCESSING |
+				ENABLE_L2_PTE_CACHE_LRU_UPDATE_BY_WRITE |
+				EFFECTIVE_L2_QUEUE_SIZE(7));
+	WREG32(VM_L2_CNTL2, 0);
+	WREG32(VM_L2_CNTL3, BANK_SELECT(0) | CACHE_UPDATE_MODE(2));
+	/* Setup TLB control */
+	tmp = ENABLE_L1_TLB | ENABLE_L1_FRAGMENT_PROCESSING |
+		SYSTEM_ACCESS_MODE_NOT_IN_SYS |
+		SYSTEM_APERTURE_UNMAPPED_ACCESS_PASS_THRU |
+		EFFECTIVE_L1_TLB_SIZE(5) | EFFECTIVE_L1_QUEUE_SIZE(5);
+	WREG32(MC_VM_MD_L1_TLB0_CNTL, tmp);
+	WREG32(MC_VM_MD_L1_TLB1_CNTL, tmp);
+	WREG32(MC_VM_MD_L1_TLB2_CNTL, tmp);
+	WREG32(MC_VM_MB_L1_TLB0_CNTL, tmp);
+	WREG32(MC_VM_MB_L1_TLB1_CNTL, tmp);
+	WREG32(MC_VM_MB_L1_TLB2_CNTL, tmp);
+	WREG32(MC_VM_MB_L1_TLB3_CNTL, tmp);
+	for (i = 0; i < 7; i++)
+		WREG32(VM_CONTEXT0_CNTL + (i * 4), 0);
+}
+
+static void evergreen_mc_stop(struct radeon_device *rdev, struct evergreen_mc_save *save)
+{
+	save->vga_control[0] = RREG32(D1VGA_CONTROL);
+	save->vga_control[1] = RREG32(D2VGA_CONTROL);
+	save->vga_control[2] = RREG32(EVERGREEN_D3VGA_CONTROL);
+	save->vga_control[3] = RREG32(EVERGREEN_D4VGA_CONTROL);
+	save->vga_control[4] = RREG32(EVERGREEN_D5VGA_CONTROL);
+	save->vga_control[5] = RREG32(EVERGREEN_D6VGA_CONTROL);
+	save->vga_render_control = RREG32(VGA_RENDER_CONTROL);
+	save->vga_hdp_control = RREG32(VGA_HDP_CONTROL);
+	save->crtc_control[0] = RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC0_REGISTER_OFFSET);
+	save->crtc_control[1] = RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC1_REGISTER_OFFSET);
+	save->crtc_control[2] = RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC2_REGISTER_OFFSET);
+	save->crtc_control[3] = RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC3_REGISTER_OFFSET);
+	save->crtc_control[4] = RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC4_REGISTER_OFFSET);
+	save->crtc_control[5] = RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC5_REGISTER_OFFSET);
+
+	/* Stop all video */
+	WREG32(VGA_RENDER_CONTROL, 0);
+	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC0_REGISTER_OFFSET, 1);
+	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC1_REGISTER_OFFSET, 1);
+	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC2_REGISTER_OFFSET, 1);
+	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC3_REGISTER_OFFSET, 1);
+	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC4_REGISTER_OFFSET, 1);
+	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC5_REGISTER_OFFSET, 1);
+	WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC0_REGISTER_OFFSET, 0);
+	WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC1_REGISTER_OFFSET, 0);
+	WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC2_REGISTER_OFFSET, 0);
+	WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC3_REGISTER_OFFSET, 0);
+	WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC4_REGISTER_OFFSET, 0);
+	WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC5_REGISTER_OFFSET, 0);
+	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC0_REGISTER_OFFSET, 0);
+	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC1_REGISTER_OFFSET, 0);
+	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC2_REGISTER_OFFSET, 0);
+	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC3_REGISTER_OFFSET, 0);
+	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC4_REGISTER_OFFSET, 0);
+	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC5_REGISTER_OFFSET, 0);
+
+	WREG32(D1VGA_CONTROL, 0);
+	WREG32(D2VGA_CONTROL, 0);
+	WREG32(EVERGREEN_D3VGA_CONTROL, 0);
+	WREG32(EVERGREEN_D4VGA_CONTROL, 0);
+	WREG32(EVERGREEN_D5VGA_CONTROL, 0);
+	WREG32(EVERGREEN_D6VGA_CONTROL, 0);
+}
+
+static void evergreen_mc_resume(struct radeon_device *rdev, struct evergreen_mc_save *save)
+{
+	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC0_REGISTER_OFFSET,
+	       upper_32_bits(rdev->mc.vram_start));
+	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC0_REGISTER_OFFSET,
+	       upper_32_bits(rdev->mc.vram_start));
+	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS + EVERGREEN_CRTC0_REGISTER_OFFSET,
+	       (u32)rdev->mc.vram_start);
+	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS + EVERGREEN_CRTC0_REGISTER_OFFSET,
+	       (u32)rdev->mc.vram_start);
+
+	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC1_REGISTER_OFFSET,
+	       upper_32_bits(rdev->mc.vram_start));
+	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC1_REGISTER_OFFSET,
+	       upper_32_bits(rdev->mc.vram_start));
+	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS + EVERGREEN_CRTC1_REGISTER_OFFSET,
+	       (u32)rdev->mc.vram_start);
+	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS + EVERGREEN_CRTC1_REGISTER_OFFSET,
+	       (u32)rdev->mc.vram_start);
+
+	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC2_REGISTER_OFFSET,
+	       upper_32_bits(rdev->mc.vram_start));
+	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC2_REGISTER_OFFSET,
+	       upper_32_bits(rdev->mc.vram_start));
+	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS + EVERGREEN_CRTC2_REGISTER_OFFSET,
+	       (u32)rdev->mc.vram_start);
+	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS + EVERGREEN_CRTC2_REGISTER_OFFSET,
+	       (u32)rdev->mc.vram_start);
+
+	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC3_REGISTER_OFFSET,
+	       upper_32_bits(rdev->mc.vram_start));
+	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC3_REGISTER_OFFSET,
+	       upper_32_bits(rdev->mc.vram_start));
+	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS + EVERGREEN_CRTC3_REGISTER_OFFSET,
+	       (u32)rdev->mc.vram_start);
+	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS + EVERGREEN_CRTC3_REGISTER_OFFSET,
+	       (u32)rdev->mc.vram_start);
+
+	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC4_REGISTER_OFFSET,
+	       upper_32_bits(rdev->mc.vram_start));
+	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC4_REGISTER_OFFSET,
+	       upper_32_bits(rdev->mc.vram_start));
+	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS + EVERGREEN_CRTC4_REGISTER_OFFSET,
+	       (u32)rdev->mc.vram_start);
+	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS + EVERGREEN_CRTC4_REGISTER_OFFSET,
+	       (u32)rdev->mc.vram_start);
+
+	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC5_REGISTER_OFFSET,
+	       upper_32_bits(rdev->mc.vram_start));
+	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC5_REGISTER_OFFSET,
+	       upper_32_bits(rdev->mc.vram_start));
+	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS + EVERGREEN_CRTC5_REGISTER_OFFSET,
+	       (u32)rdev->mc.vram_start);
+	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS + EVERGREEN_CRTC5_REGISTER_OFFSET,
+	       (u32)rdev->mc.vram_start);
+
+	WREG32(EVERGREEN_VGA_MEMORY_BASE_ADDRESS_HIGH, upper_32_bits(rdev->mc.vram_start));
+	WREG32(EVERGREEN_VGA_MEMORY_BASE_ADDRESS, (u32)rdev->mc.vram_start);
+	/* Unlock host access */
+	WREG32(VGA_HDP_CONTROL, save->vga_hdp_control);
+	mdelay(1);
+	/* Restore video state */
+	WREG32(D1VGA_CONTROL, save->vga_control[0]);
+	WREG32(D2VGA_CONTROL, save->vga_control[1]);
+	WREG32(EVERGREEN_D3VGA_CONTROL, save->vga_control[2]);
+	WREG32(EVERGREEN_D4VGA_CONTROL, save->vga_control[3]);
+	WREG32(EVERGREEN_D5VGA_CONTROL, save->vga_control[4]);
+	WREG32(EVERGREEN_D6VGA_CONTROL, save->vga_control[5]);
+	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC0_REGISTER_OFFSET, 1);
+	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC1_REGISTER_OFFSET, 1);
+	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC2_REGISTER_OFFSET, 1);
+	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC3_REGISTER_OFFSET, 1);
+	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC4_REGISTER_OFFSET, 1);
+	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC5_REGISTER_OFFSET, 1);
+	WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC0_REGISTER_OFFSET, save->crtc_control[0]);
+	WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC1_REGISTER_OFFSET, save->crtc_control[1]);
+	WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC2_REGISTER_OFFSET, save->crtc_control[2]);
+	WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC3_REGISTER_OFFSET, save->crtc_control[3]);
+	WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC4_REGISTER_OFFSET, save->crtc_control[4]);
+	WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC5_REGISTER_OFFSET, save->crtc_control[5]);
+	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC0_REGISTER_OFFSET, 0);
+	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC1_REGISTER_OFFSET, 0);
+	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC2_REGISTER_OFFSET, 0);
+	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC3_REGISTER_OFFSET, 0);
+	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC4_REGISTER_OFFSET, 0);
+	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC5_REGISTER_OFFSET, 0);
+	WREG32(VGA_RENDER_CONTROL, save->vga_render_control);
+}
+
+static void evergreen_mc_program(struct radeon_device *rdev)
+{
+	struct evergreen_mc_save save;
+	u32 tmp;
+	int i, j;
+
+	/* Initialize HDP */
+	for (i = 0, j = 0; i < 32; i++, j += 0x18) {
+		WREG32((0x2c14 + j), 0x00000000);
+		WREG32((0x2c18 + j), 0x00000000);
+		WREG32((0x2c1c + j), 0x00000000);
+		WREG32((0x2c20 + j), 0x00000000);
+		WREG32((0x2c24 + j), 0x00000000);
+	}
+	WREG32(HDP_REG_COHERENCY_FLUSH_CNTL, 0);
+
+	evergreen_mc_stop(rdev, &save);
+	if (evergreen_mc_wait_for_idle(rdev)) {
+		dev_warn(rdev->dev, "Wait for MC idle timedout !\n");
+	}
+	/* Lockout access through VGA aperture*/
+	WREG32(VGA_HDP_CONTROL, VGA_MEMORY_DISABLE);
+	/* Update configuration */
+	if (rdev->flags & RADEON_IS_AGP) {
+		if (rdev->mc.vram_start < rdev->mc.gtt_start) {
+			/* VRAM before AGP */
+			WREG32(MC_VM_SYSTEM_APERTURE_LOW_ADDR,
+				rdev->mc.vram_start >> 12);
+			WREG32(MC_VM_SYSTEM_APERTURE_HIGH_ADDR,
+				rdev->mc.gtt_end >> 12);
+		} else {
+			/* VRAM after AGP */
+			WREG32(MC_VM_SYSTEM_APERTURE_LOW_ADDR,
+				rdev->mc.gtt_start >> 12);
+			WREG32(MC_VM_SYSTEM_APERTURE_HIGH_ADDR,
+				rdev->mc.vram_end >> 12);
+		}
+	} else {
+		WREG32(MC_VM_SYSTEM_APERTURE_LOW_ADDR,
+			rdev->mc.vram_start >> 12);
+		WREG32(MC_VM_SYSTEM_APERTURE_HIGH_ADDR,
+			rdev->mc.vram_end >> 12);
+	}
+	WREG32(MC_VM_SYSTEM_APERTURE_DEFAULT_ADDR, 0);
+	tmp = ((rdev->mc.vram_end >> 24) & 0xFFFF) << 16;
+	tmp |= ((rdev->mc.vram_start >> 24) & 0xFFFF);
+	WREG32(MC_VM_FB_LOCATION, tmp);
+	WREG32(HDP_NONSURFACE_BASE, (rdev->mc.vram_start >> 8));
+	WREG32(HDP_NONSURFACE_INFO, (2 << 7));
+	WREG32(HDP_NONSURFACE_SIZE, (rdev->mc.mc_vram_size - 1) | 0x3FF);
+	if (rdev->flags & RADEON_IS_AGP) {
+		WREG32(MC_VM_AGP_TOP, rdev->mc.gtt_end >> 16);
+		WREG32(MC_VM_AGP_BOT, rdev->mc.gtt_start >> 16);
+		WREG32(MC_VM_AGP_BASE, rdev->mc.agp_base >> 22);
+	} else {
+		WREG32(MC_VM_AGP_BASE, 0);
+		WREG32(MC_VM_AGP_TOP, 0x0FFFFFFF);
+		WREG32(MC_VM_AGP_BOT, 0x0FFFFFFF);
+	}
+	if (evergreen_mc_wait_for_idle(rdev)) {
+		dev_warn(rdev->dev, "Wait for MC idle timedout !\n");
+	}
+	evergreen_mc_resume(rdev, &save);
+	/* we need to own VRAM, so turn off the VGA renderer here
+	 * to stop it overwriting our objects */
+	rv515_vga_render_disable(rdev);
+}
+
+#if 0
+/*
+ * CP.
+ */
+static void evergreen_cp_stop(struct radeon_device *rdev)
+{
+	/* XXX */
+}
+
+
+static int evergreen_cp_load_microcode(struct radeon_device *rdev)
+{
+	/* XXX */
+
+	return 0;
+}
+
+
+/*
+ * Core functions
+ */
+static u32 evergreen_get_tile_pipe_to_backend_map(u32 num_tile_pipes,
+						  u32 num_backends,
+						  u32 backend_disable_mask)
+{
+	u32 backend_map = 0;
+
+	return backend_map;
+}
+#endif
+
+static void evergreen_gpu_init(struct radeon_device *rdev)
+{
+	/* XXX */
+}
+
+int evergreen_mc_init(struct radeon_device *rdev)
+{
+	fixed20_12 a;
+	u32 tmp;
+	int chansize, numchan;
+
+	/* Get VRAM informations */
+	rdev->mc.vram_is_ddr = true;
+	tmp = RREG32(MC_ARB_RAMCFG);
+	if (tmp & CHANSIZE_OVERRIDE) {
+		chansize = 16;
+	} else if (tmp & CHANSIZE_MASK) {
+		chansize = 64;
+	} else {
+		chansize = 32;
+	}
+	tmp = RREG32(MC_SHARED_CHMAP);
+	switch ((tmp & NOOFCHAN_MASK) >> NOOFCHAN_SHIFT) {
+	case 0:
+	default:
+		numchan = 1;
+		break;
+	case 1:
+		numchan = 2;
+		break;
+	case 2:
+		numchan = 4;
+		break;
+	case 3:
+		numchan = 8;
+		break;
+	}
+	rdev->mc.vram_width = numchan * chansize;
+	/* Could aper size report 0 ? */
+	rdev->mc.aper_base = drm_get_resource_start(rdev->ddev, 0);
+	rdev->mc.aper_size = drm_get_resource_len(rdev->ddev, 0);
+	/* Setup GPU memory space */
+	/* size in MB on evergreen */
+	rdev->mc.mc_vram_size = RREG32(CONFIG_MEMSIZE) * 1024 * 1024;
+	rdev->mc.real_vram_size = RREG32(CONFIG_MEMSIZE) * 1024 * 1024;
+	rdev->mc.visible_vram_size = rdev->mc.aper_size;
+	/* FIXME remove this once we support unmappable VRAM */
+	if (rdev->mc.mc_vram_size > rdev->mc.aper_size) {
+		rdev->mc.mc_vram_size = rdev->mc.aper_size;
+		rdev->mc.real_vram_size = rdev->mc.aper_size;
+	}
+	r600_vram_gtt_location(rdev, &rdev->mc);
+	/* FIXME: we should enforce default clock in case GPU is not in
+	 * default setup
+	 */
+	a.full = rfixed_const(100);
+	rdev->pm.sclk.full = rfixed_const(rdev->clock.default_sclk);
+	rdev->pm.sclk.full = rfixed_div(rdev->pm.sclk, a);
+	return 0;
+}
+
+int evergreen_gpu_reset(struct radeon_device *rdev)
+{
+	/* FIXME: implement for evergreen */
+	return 0;
+}
+
+static int evergreen_startup(struct radeon_device *rdev)
+{
+#if 0
+	int r;
+
+	if (!rdev->me_fw || !rdev->pfp_fw || !rdev->rlc_fw) {
+		r = r600_init_microcode(rdev);
+		if (r) {
+			DRM_ERROR("Failed to load firmware!\n");
+			return r;
+		}
+	}
+#endif
+	evergreen_mc_program(rdev);
+#if 0
+	if (rdev->flags & RADEON_IS_AGP) {
+		evergreem_agp_enable(rdev);
+	} else {
+		r = evergreen_pcie_gart_enable(rdev);
+		if (r)
+			return r;
+	}
+#endif
+	evergreen_gpu_init(rdev);
+#if 0
+	if (!rdev->r600_blit.shader_obj) {
+		r = r600_blit_init(rdev);
+		if (r) {
+			DRM_ERROR("radeon: failed blitter (%d).\n", r);
+			return r;
+		}
+	}
+
+	r = radeon_bo_reserve(rdev->r600_blit.shader_obj, false);
+	if (unlikely(r != 0))
+		return r;
+	r = radeon_bo_pin(rdev->r600_blit.shader_obj, RADEON_GEM_DOMAIN_VRAM,
+			&rdev->r600_blit.shader_gpu_addr);
+	radeon_bo_unreserve(rdev->r600_blit.shader_obj);
+	if (r) {
+		DRM_ERROR("failed to pin blit object %d\n", r);
+		return r;
+	}
+
+	/* Enable IRQ */
+	r = r600_irq_init(rdev);
+	if (r) {
+		DRM_ERROR("radeon: IH init failed (%d).\n", r);
+		radeon_irq_kms_fini(rdev);
+		return r;
+	}
+	r600_irq_set(rdev);
+
+	r = radeon_ring_init(rdev, rdev->cp.ring_size);
+	if (r)
+		return r;
+	r = evergreen_cp_load_microcode(rdev);
+	if (r)
+		return r;
+	r = r600_cp_resume(rdev);
+	if (r)
+		return r;
+	/* write back buffer are not vital so don't worry about failure */
+	r600_wb_enable(rdev);
+#endif
+	return 0;
+}
+
+int evergreen_resume(struct radeon_device *rdev)
+{
+	int r;
+
+	/* Do not reset GPU before posting, on rv770 hw unlike on r500 hw,
+	 * posting will perform necessary task to bring back GPU into good
+	 * shape.
+	 */
+	/* post card */
+	atom_asic_init(rdev->mode_info.atom_context);
+	/* Initialize clocks */
+	r = radeon_clocks_init(rdev);
+	if (r) {
+		return r;
+	}
+
+	r = evergreen_startup(rdev);
+	if (r) {
+		DRM_ERROR("r600 startup failed on resume\n");
+		return r;
+	}
+#if 0
+	r = r600_ib_test(rdev);
+	if (r) {
+		DRM_ERROR("radeon: failled testing IB (%d).\n", r);
+		return r;
+	}
+#endif
+	return r;
+
+}
+
+int evergreen_suspend(struct radeon_device *rdev)
+{
+#if 0
+	int r;
+
+	/* FIXME: we should wait for ring to be empty */
+	r700_cp_stop(rdev);
+	rdev->cp.ready = false;
+	r600_wb_disable(rdev);
+	evergreen_pcie_gart_disable(rdev);
+	/* unpin shaders bo */
+	r = radeon_bo_reserve(rdev->r600_blit.shader_obj, false);
+	if (likely(r == 0)) {
+		radeon_bo_unpin(rdev->r600_blit.shader_obj);
+		radeon_bo_unreserve(rdev->r600_blit.shader_obj);
+	}
+#endif
+	return 0;
+}
+
+static bool evergreen_card_posted(struct radeon_device *rdev)
+{
+	u32 reg;
+
+	/* first check CRTCs */
+	reg = RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC0_REGISTER_OFFSET) |
+		RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC1_REGISTER_OFFSET) |
+		RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC2_REGISTER_OFFSET) |
+		RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC3_REGISTER_OFFSET) |
+		RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC4_REGISTER_OFFSET) |
+		RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC5_REGISTER_OFFSET);
+	if (reg & EVERGREEN_CRTC_MASTER_EN)
+		return true;
+
+	/* then check MEM_SIZE, in case the crtcs are off */
+	if (RREG32(CONFIG_MEMSIZE))
+		return true;
+
+	return false;
+}
+
+/* Plan is to move initialization in that function and use
+ * helper function so that radeon_device_init pretty much
+ * do nothing more than calling asic specific function. This
+ * should also allow to remove a bunch of callback function
+ * like vram_info.
+ */
+int evergreen_init(struct radeon_device *rdev)
+{
+	int r;
+
+	r = radeon_dummy_page_init(rdev);
+	if (r)
+		return r;
+	/* This don't do much */
+	r = radeon_gem_init(rdev);
+	if (r)
+		return r;
+	/* Read BIOS */
+	if (!radeon_get_bios(rdev)) {
+		if (ASIC_IS_AVIVO(rdev))
+			return -EINVAL;
+	}
+	/* Must be an ATOMBIOS */
+	if (!rdev->is_atom_bios) {
+		dev_err(rdev->dev, "Expecting atombios for R600 GPU\n");
+		return -EINVAL;
+	}
+	r = radeon_atombios_init(rdev);
+	if (r)
+		return r;
+	/* Post card if necessary */
+	if (!evergreen_card_posted(rdev)) {
+		if (!rdev->bios) {
+			dev_err(rdev->dev, "Card not posted and no BIOS - ignoring\n");
+			return -EINVAL;
+		}
+		DRM_INFO("GPU not posted. posting now...\n");
+		atom_asic_init(rdev->mode_info.atom_context);
+	}
+	/* Initialize scratch registers */
+	r600_scratch_init(rdev);
+	/* Initialize surface registers */
+	radeon_surface_init(rdev);
+	/* Initialize clocks */
+	radeon_get_clock_info(rdev->ddev);
+	r = radeon_clocks_init(rdev);
+	if (r)
+		return r;
+	/* Initialize power management */
+	radeon_pm_init(rdev);
+	/* Fence driver */
+	r = radeon_fence_driver_init(rdev);
+	if (r)
+		return r;
+	/* initialize AGP */
+	if (rdev->flags & RADEON_IS_AGP) {
+		r = radeon_agp_init(rdev);
+		if (r)
+			radeon_agp_disable(rdev);
+	}
+	/* initialize memory controller */
+	r = evergreen_mc_init(rdev);
+	if (r)
+		return r;
+	/* Memory manager */
+	r = radeon_bo_init(rdev);
+	if (r)
+		return r;
+#if 0
+	r = radeon_irq_kms_init(rdev);
+	if (r)
+		return r;
+
+	rdev->cp.ring_obj = NULL;
+	r600_ring_init(rdev, 1024 * 1024);
+
+	rdev->ih.ring_obj = NULL;
+	r600_ih_ring_init(rdev, 64 * 1024);
+
+	r = r600_pcie_gart_init(rdev);
+	if (r)
+		return r;
+#endif
+	rdev->accel_working = false;
+	r = evergreen_startup(rdev);
+	if (r) {
+		evergreen_suspend(rdev);
+		/*r600_wb_fini(rdev);*/
+		/*radeon_ring_fini(rdev);*/
+		/*evergreen_pcie_gart_fini(rdev);*/
+		rdev->accel_working = false;
+	}
+	if (rdev->accel_working) {
+		r = radeon_ib_pool_init(rdev);
+		if (r) {
+			DRM_ERROR("radeon: failed initializing IB pool (%d).\n", r);
+			rdev->accel_working = false;
+		}
+		r = r600_ib_test(rdev);
+		if (r) {
+			DRM_ERROR("radeon: failed testing IB (%d).\n", r);
+			rdev->accel_working = false;
+		}
+	}
+	return 0;
+}
+
+void evergreen_fini(struct radeon_device *rdev)
+{
+	evergreen_suspend(rdev);
+#if 0
+	r600_blit_fini(rdev);
+	r600_irq_fini(rdev);
+	radeon_irq_kms_fini(rdev);
+	radeon_ring_fini(rdev);
+	r600_wb_fini(rdev);
+	evergreen_pcie_gart_fini(rdev);
+#endif
+	radeon_gem_fini(rdev);
+	radeon_fence_driver_fini(rdev);
+	radeon_clocks_fini(rdev);
+	radeon_agp_fini(rdev);
+	radeon_bo_fini(rdev);
+	radeon_atombios_fini(rdev);
+	kfree(rdev->bios);
+	rdev->bios = NULL;
+	radeon_dummy_page_fini(rdev);
+}
diff --git a/drivers/gpu/drm/radeon/evergreen_reg.h b/drivers/gpu/drm/radeon/evergreen_reg.h
new file mode 100644
index 000000000000..f7c7c9643433
--- /dev/null
+++ b/drivers/gpu/drm/radeon/evergreen_reg.h
@@ -0,0 +1,176 @@
+/*
+ * Copyright 2010 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Alex Deucher
+ */
+#ifndef __EVERGREEN_REG_H__
+#define __EVERGREEN_REG_H__
+
+/* evergreen */
+#define EVERGREEN_VGA_MEMORY_BASE_ADDRESS               0x310
+#define EVERGREEN_VGA_MEMORY_BASE_ADDRESS_HIGH          0x324
+#define EVERGREEN_D3VGA_CONTROL                         0x3e0
+#define EVERGREEN_D4VGA_CONTROL                         0x3e4
+#define EVERGREEN_D5VGA_CONTROL                         0x3e8
+#define EVERGREEN_D6VGA_CONTROL                         0x3ec
+
+#define EVERGREEN_P1PLL_SS_CNTL                         0x414
+#define EVERGREEN_P2PLL_SS_CNTL                         0x454
+#       define EVERGREEN_PxPLL_SS_EN                    (1 << 12)
+/* GRPH blocks at 0x6800, 0x7400, 0x10000, 0x10c00, 0x11800, 0x12400 */
+#define EVERGREEN_GRPH_ENABLE                           0x6800
+#define EVERGREEN_GRPH_CONTROL                          0x6804
+#       define EVERGREEN_GRPH_DEPTH(x)                  (((x) & 0x3) << 0)
+#       define EVERGREEN_GRPH_DEPTH_8BPP                0
+#       define EVERGREEN_GRPH_DEPTH_16BPP               1
+#       define EVERGREEN_GRPH_DEPTH_32BPP               2
+#       define EVERGREEN_GRPH_FORMAT(x)                 (((x) & 0x7) << 8)
+/* 8 BPP */
+#       define EVERGREEN_GRPH_FORMAT_INDEXED            0
+/* 16 BPP */
+#       define EVERGREEN_GRPH_FORMAT_ARGB1555           0
+#       define EVERGREEN_GRPH_FORMAT_ARGB565            1
+#       define EVERGREEN_GRPH_FORMAT_ARGB4444           2
+#       define EVERGREEN_GRPH_FORMAT_AI88               3
+#       define EVERGREEN_GRPH_FORMAT_MONO16             4
+#       define EVERGREEN_GRPH_FORMAT_BGRA5551           5
+/* 32 BPP */
+#       define EVERGREEN_GRPH_FORMAT_ARGB8888           0
+#       define EVERGREEN_GRPH_FORMAT_ARGB2101010        1
+#       define EVERGREEN_GRPH_FORMAT_32BPP_DIG          2
+#       define EVERGREEN_GRPH_FORMAT_8B_ARGB2101010     3
+#       define EVERGREEN_GRPH_FORMAT_BGRA1010102        4
+#       define EVERGREEN_GRPH_FORMAT_8B_BGRA1010102     5
+#       define EVERGREEN_GRPH_FORMAT_RGB111110          6
+#       define EVERGREEN_GRPH_FORMAT_BGR101111          7
+#define EVERGREEN_GRPH_SWAP_CONTROL                     0x680c
+#       define EVERGREEN_GRPH_ENDIAN_SWAP(x)            (((x) & 0x3) << 0)
+#       define EVERGREEN_GRPH_ENDIAN_NONE               0
+#       define EVERGREEN_GRPH_ENDIAN_8IN16              1
+#       define EVERGREEN_GRPH_ENDIAN_8IN32              2
+#       define EVERGREEN_GRPH_ENDIAN_8IN64              3
+#       define EVERGREEN_GRPH_RED_CROSSBAR(x)           (((x) & 0x3) << 4)
+#       define EVERGREEN_GRPH_RED_SEL_R                 0
+#       define EVERGREEN_GRPH_RED_SEL_G                 1
+#       define EVERGREEN_GRPH_RED_SEL_B                 2
+#       define EVERGREEN_GRPH_RED_SEL_A                 3
+#       define EVERGREEN_GRPH_GREEN_CROSSBAR(x)         (((x) & 0x3) << 6)
+#       define EVERGREEN_GRPH_GREEN_SEL_G               0
+#       define EVERGREEN_GRPH_GREEN_SEL_B               1
+#       define EVERGREEN_GRPH_GREEN_SEL_A               2
+#       define EVERGREEN_GRPH_GREEN_SEL_R               3
+#       define EVERGREEN_GRPH_BLUE_CROSSBAR(x)          (((x) & 0x3) << 8)
+#       define EVERGREEN_GRPH_BLUE_SEL_B                0
+#       define EVERGREEN_GRPH_BLUE_SEL_A                1
+#       define EVERGREEN_GRPH_BLUE_SEL_R                2
+#       define EVERGREEN_GRPH_BLUE_SEL_G                3
+#       define EVERGREEN_GRPH_ALPHA_CROSSBAR(x)         (((x) & 0x3) << 10)
+#       define EVERGREEN_GRPH_ALPHA_SEL_A               0
+#       define EVERGREEN_GRPH_ALPHA_SEL_R               1
+#       define EVERGREEN_GRPH_ALPHA_SEL_G               2
+#       define EVERGREEN_GRPH_ALPHA_SEL_B               3
+#define EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS          0x6810
+#define EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS        0x6814
+#       define EVERGREEN_GRPH_DFQ_ENABLE                (1 << 0)
+#       define EVERGREEN_GRPH_SURFACE_ADDRESS_MASK      0xffffff00
+#define EVERGREEN_GRPH_PITCH                            0x6818
+#define EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS_HIGH     0x681c
+#define EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS_HIGH   0x6820
+#define EVERGREEN_GRPH_SURFACE_OFFSET_X                 0x6824
+#define EVERGREEN_GRPH_SURFACE_OFFSET_Y                 0x6828
+#define EVERGREEN_GRPH_X_START                          0x682c
+#define EVERGREEN_GRPH_Y_START                          0x6830
+#define EVERGREEN_GRPH_X_END                            0x6834
+#define EVERGREEN_GRPH_Y_END                            0x6838
+
+/* CUR blocks at 0x6998, 0x7598, 0x10198, 0x10d98, 0x11998, 0x12598 */
+#define EVERGREEN_CUR_CONTROL                           0x6998
+#       define EVERGREEN_CURSOR_EN                      (1 << 0)
+#       define EVERGREEN_CURSOR_MODE(x)                 (((x) & 0x3) << 8)
+#       define EVERGREEN_CURSOR_MONO                    0
+#       define EVERGREEN_CURSOR_24_1                    1
+#       define EVERGREEN_CURSOR_24_8_PRE_MULT           2
+#       define EVERGREEN_CURSOR_24_8_UNPRE_MULT         3
+#       define EVERGREEN_CURSOR_2X_MAGNIFY              (1 << 16)
+#       define EVERGREEN_CURSOR_FORCE_MC_ON             (1 << 20)
+#       define EVERGREEN_CURSOR_URGENT_CONTROL(x)       (((x) & 0x7) << 24)
+#       define EVERGREEN_CURSOR_URGENT_ALWAYS           0
+#       define EVERGREEN_CURSOR_URGENT_1_8              1
+#       define EVERGREEN_CURSOR_URGENT_1_4              2
+#       define EVERGREEN_CURSOR_URGENT_3_8              3
+#       define EVERGREEN_CURSOR_URGENT_1_2              4
+#define EVERGREEN_CUR_SURFACE_ADDRESS                   0x699c
+#       define EVERGREEN_CUR_SURFACE_ADDRESS_MASK       0xfffff000
+#define EVERGREEN_CUR_SIZE                              0x69a0
+#define EVERGREEN_CUR_SURFACE_ADDRESS_HIGH              0x69a4
+#define EVERGREEN_CUR_POSITION                          0x69a8
+#define EVERGREEN_CUR_HOT_SPOT                          0x69ac
+#define EVERGREEN_CUR_COLOR1                            0x69b0
+#define EVERGREEN_CUR_COLOR2                            0x69b4
+#define EVERGREEN_CUR_UPDATE                            0x69b8
+#       define EVERGREEN_CURSOR_UPDATE_PENDING          (1 << 0)
+#       define EVERGREEN_CURSOR_UPDATE_TAKEN            (1 << 1)
+#       define EVERGREEN_CURSOR_UPDATE_LOCK             (1 << 16)
+#       define EVERGREEN_CURSOR_DISABLE_MULTIPLE_UPDATE (1 << 24)
+
+/* LUT blocks at 0x69e0, 0x75e0, 0x101e0, 0x10de0, 0x119e0, 0x125e0 */
+#define EVERGREEN_DC_LUT_RW_MODE                        0x69e0
+#define EVERGREEN_DC_LUT_RW_INDEX                       0x69e4
+#define EVERGREEN_DC_LUT_SEQ_COLOR                      0x69e8
+#define EVERGREEN_DC_LUT_PWL_DATA                       0x69ec
+#define EVERGREEN_DC_LUT_30_COLOR                       0x69f0
+#define EVERGREEN_DC_LUT_VGA_ACCESS_ENABLE              0x69f4
+#define EVERGREEN_DC_LUT_WRITE_EN_MASK                  0x69f8
+#define EVERGREEN_DC_LUT_AUTOFILL                       0x69fc
+#define EVERGREEN_DC_LUT_CONTROL                        0x6a00
+#define EVERGREEN_DC_LUT_BLACK_OFFSET_BLUE              0x6a04
+#define EVERGREEN_DC_LUT_BLACK_OFFSET_GREEN             0x6a08
+#define EVERGREEN_DC_LUT_BLACK_OFFSET_RED               0x6a0c
+#define EVERGREEN_DC_LUT_WHITE_OFFSET_BLUE              0x6a10
+#define EVERGREEN_DC_LUT_WHITE_OFFSET_GREEN             0x6a14
+#define EVERGREEN_DC_LUT_WHITE_OFFSET_RED               0x6a18
+
+#define EVERGREEN_DATA_FORMAT                           0x6b00
+#       define EVERGREEN_INTERLEAVE_EN                  (1 << 0)
+#define EVERGREEN_DESKTOP_HEIGHT                        0x6b04
+
+#define EVERGREEN_VIEWPORT_START                        0x6d70
+#define EVERGREEN_VIEWPORT_SIZE                         0x6d74
+
+/* display controller offsets used for crtc/cur/lut/grph/viewport/etc. */
+#define EVERGREEN_CRTC0_REGISTER_OFFSET                 (0x6df0 - 0x6df0)
+#define EVERGREEN_CRTC1_REGISTER_OFFSET                 (0x79f0 - 0x6df0)
+#define EVERGREEN_CRTC2_REGISTER_OFFSET                 (0x105f0 - 0x6df0)
+#define EVERGREEN_CRTC3_REGISTER_OFFSET                 (0x111f0 - 0x6df0)
+#define EVERGREEN_CRTC4_REGISTER_OFFSET                 (0x11df0 - 0x6df0)
+#define EVERGREEN_CRTC5_REGISTER_OFFSET                 (0x129f0 - 0x6df0)
+
+/* CRTC blocks at 0x6df0, 0x79f0, 0x105f0, 0x111f0, 0x11df0, 0x129f0 */
+#define EVERGREEN_CRTC_CONTROL                          0x6e70
+#       define EVERGREEN_CRTC_MASTER_EN                 (1 << 0)
+#define EVERGREEN_CRTC_UPDATE_LOCK                      0x6ed4
+
+#define EVERGREEN_DC_GPIO_HPD_MASK                      0x64b0
+#define EVERGREEN_DC_GPIO_HPD_A                         0x64b4
+#define EVERGREEN_DC_GPIO_HPD_EN                        0x64b8
+#define EVERGREEN_DC_GPIO_HPD_Y                         0x64bc
+
+#endif
diff --git a/drivers/gpu/drm/radeon/mkregtable.c b/drivers/gpu/drm/radeon/mkregtable.c
index fb211e585dea..607241c6a8a9 100644
--- a/drivers/gpu/drm/radeon/mkregtable.c
+++ b/drivers/gpu/drm/radeon/mkregtable.c
@@ -561,7 +561,7 @@ struct table {
 	char *gpu_prefix;
 };
 
-struct offset *offset_new(unsigned o)
+static struct offset *offset_new(unsigned o)
 {
 	struct offset *offset;
 
@@ -573,12 +573,12 @@ struct offset *offset_new(unsigned o)
 	return offset;
 }
 
-void table_offset_add(struct table *t, struct offset *offset)
+static void table_offset_add(struct table *t, struct offset *offset)
 {
 	list_add_tail(&offset->list, &t->offsets);
 }
 
-void table_init(struct table *t)
+static void table_init(struct table *t)
 {
 	INIT_LIST_HEAD(&t->offsets);
 	t->offset_max = 0;
@@ -586,7 +586,7 @@ void table_init(struct table *t)
 	t->table = NULL;
 }
 
-void table_print(struct table *t)
+static void table_print(struct table *t)
 {
 	unsigned nlloop, i, j, n, c, id;
 
@@ -611,7 +611,7 @@ void table_print(struct table *t)
 	printf("};\n");
 }
 
-int table_build(struct table *t)
+static int table_build(struct table *t)
 {
 	struct offset *offset;
 	unsigned i, m;
@@ -631,7 +631,7 @@ int table_build(struct table *t)
 }
 
 static char gpu_name[10];
-int parser_auth(struct table *t, const char *filename)
+static int parser_auth(struct table *t, const char *filename)
 {
 	FILE *file;
 	regex_t mask_rex;
@@ -661,8 +661,10 @@ int parser_auth(struct table *t, const char *filename)
 	fseek(file, 0, SEEK_SET);
 
 	/* get header */
-	if (fgets(buf, 1024, file) == NULL)
+	if (fgets(buf, 1024, file) == NULL) {
+		fclose(file);
 		return -1;
+	}
 
 	/* first line will contain the last register
 	 * and gpu name */
diff --git a/drivers/gpu/drm/radeon/r100.c b/drivers/gpu/drm/radeon/r100.c
index be51c5f7d0f6..91eb762eb3f9 100644
--- a/drivers/gpu/drm/radeon/r100.c
+++ b/drivers/gpu/drm/radeon/r100.c
@@ -32,6 +32,9 @@
 #include "radeon_reg.h"
 #include "radeon.h"
 #include "r100d.h"
+#include "rs100d.h"
+#include "rv200d.h"
+#include "rv250d.h"
 
 #include <linux/firmware.h>
 #include <linux/platform_device.h>
@@ -60,18 +63,97 @@ MODULE_FIRMWARE(FIRMWARE_R520);
 
 /* This files gather functions specifics to:
  * r100,rv100,rs100,rv200,rs200,r200,rv250,rs300,rv280
- *
- * Some of these functions might be used by newer ASICs.
  */
-int r200_init(struct radeon_device *rdev);
-void r100_hdp_reset(struct radeon_device *rdev);
-void r100_gpu_init(struct radeon_device *rdev);
-int r100_gui_wait_for_idle(struct radeon_device *rdev);
-int r100_mc_wait_for_idle(struct radeon_device *rdev);
-void r100_gpu_wait_for_vsync(struct radeon_device *rdev);
-void r100_gpu_wait_for_vsync2(struct radeon_device *rdev);
-int r100_debugfs_mc_info_init(struct radeon_device *rdev);
 
+/* hpd for digital panel detect/disconnect */
+bool r100_hpd_sense(struct radeon_device *rdev, enum radeon_hpd_id hpd)
+{
+	bool connected = false;
+
+	switch (hpd) {
+	case RADEON_HPD_1:
+		if (RREG32(RADEON_FP_GEN_CNTL) & RADEON_FP_DETECT_SENSE)
+			connected = true;
+		break;
+	case RADEON_HPD_2:
+		if (RREG32(RADEON_FP2_GEN_CNTL) & RADEON_FP2_DETECT_SENSE)
+			connected = true;
+		break;
+	default:
+		break;
+	}
+	return connected;
+}
+
+void r100_hpd_set_polarity(struct radeon_device *rdev,
+			   enum radeon_hpd_id hpd)
+{
+	u32 tmp;
+	bool connected = r100_hpd_sense(rdev, hpd);
+
+	switch (hpd) {
+	case RADEON_HPD_1:
+		tmp = RREG32(RADEON_FP_GEN_CNTL);
+		if (connected)
+			tmp &= ~RADEON_FP_DETECT_INT_POL;
+		else
+			tmp |= RADEON_FP_DETECT_INT_POL;
+		WREG32(RADEON_FP_GEN_CNTL, tmp);
+		break;
+	case RADEON_HPD_2:
+		tmp = RREG32(RADEON_FP2_GEN_CNTL);
+		if (connected)
+			tmp &= ~RADEON_FP2_DETECT_INT_POL;
+		else
+			tmp |= RADEON_FP2_DETECT_INT_POL;
+		WREG32(RADEON_FP2_GEN_CNTL, tmp);
+		break;
+	default:
+		break;
+	}
+}
+
+void r100_hpd_init(struct radeon_device *rdev)
+{
+	struct drm_device *dev = rdev->ddev;
+	struct drm_connector *connector;
+
+	list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
+		struct radeon_connector *radeon_connector = to_radeon_connector(connector);
+		switch (radeon_connector->hpd.hpd) {
+		case RADEON_HPD_1:
+			rdev->irq.hpd[0] = true;
+			break;
+		case RADEON_HPD_2:
+			rdev->irq.hpd[1] = true;
+			break;
+		default:
+			break;
+		}
+	}
+	if (rdev->irq.installed)
+		r100_irq_set(rdev);
+}
+
+void r100_hpd_fini(struct radeon_device *rdev)
+{
+	struct drm_device *dev = rdev->ddev;
+	struct drm_connector *connector;
+
+	list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
+		struct radeon_connector *radeon_connector = to_radeon_connector(connector);
+		switch (radeon_connector->hpd.hpd) {
+		case RADEON_HPD_1:
+			rdev->irq.hpd[0] = false;
+			break;
+		case RADEON_HPD_2:
+			rdev->irq.hpd[1] = false;
+			break;
+		default:
+			break;
+		}
+	}
+}
 
 /*
  * PCI GART
@@ -102,20 +184,26 @@ int r100_pci_gart_init(struct radeon_device *rdev)
 	return radeon_gart_table_ram_alloc(rdev);
 }
 
+/* required on r1xx, r2xx, r300, r(v)350, r420/r481, rs400/rs480 */
+void r100_enable_bm(struct radeon_device *rdev)
+{
+	uint32_t tmp;
+	/* Enable bus mastering */
+	tmp = RREG32(RADEON_BUS_CNTL) & ~RADEON_BUS_MASTER_DIS;
+	WREG32(RADEON_BUS_CNTL, tmp);
+}
+
 int r100_pci_gart_enable(struct radeon_device *rdev)
 {
 	uint32_t tmp;
 
+	radeon_gart_restore(rdev);
 	/* discard memory request outside of configured range */
 	tmp = RREG32(RADEON_AIC_CNTL) | RADEON_DIS_OUT_OF_PCI_GART_ACCESS;
 	WREG32(RADEON_AIC_CNTL, tmp);
 	/* set address range for PCI address translate */
-	WREG32(RADEON_AIC_LO_ADDR, rdev->mc.gtt_location);
-	tmp = rdev->mc.gtt_location + rdev->mc.gtt_size - 1;
-	WREG32(RADEON_AIC_HI_ADDR, tmp);
-	/* Enable bus mastering */
-	tmp = RREG32(RADEON_BUS_CNTL) & ~RADEON_BUS_MASTER_DIS;
-	WREG32(RADEON_BUS_CNTL, tmp);
+	WREG32(RADEON_AIC_LO_ADDR, rdev->mc.gtt_start);
+	WREG32(RADEON_AIC_HI_ADDR, rdev->mc.gtt_end);
 	/* set PCI GART page-table base address */
 	WREG32(RADEON_AIC_PT_BASE, rdev->gart.table_addr);
 	tmp = RREG32(RADEON_AIC_CNTL) | RADEON_PCIGART_TRANSLATE_EN;
@@ -152,140 +240,15 @@ void r100_pci_gart_fini(struct radeon_device *rdev)
 	radeon_gart_fini(rdev);
 }
 
-
-/*
- * MC
- */
-void r100_mc_disable_clients(struct radeon_device *rdev)
-{
-	uint32_t ov0_scale_cntl, crtc_ext_cntl, crtc_gen_cntl, crtc2_gen_cntl;
-
-	/* FIXME: is this function correct for rs100,rs200,rs300 ? */
-	if (r100_gui_wait_for_idle(rdev)) {
-		printk(KERN_WARNING "Failed to wait GUI idle while "
-		       "programming pipes. Bad things might happen.\n");
-	}
-
-	/* stop display and memory access */
-	ov0_scale_cntl = RREG32(RADEON_OV0_SCALE_CNTL);
-	WREG32(RADEON_OV0_SCALE_CNTL, ov0_scale_cntl & ~RADEON_SCALER_ENABLE);
-	crtc_ext_cntl = RREG32(RADEON_CRTC_EXT_CNTL);
-	WREG32(RADEON_CRTC_EXT_CNTL, crtc_ext_cntl | RADEON_CRTC_DISPLAY_DIS);
-	crtc_gen_cntl = RREG32(RADEON_CRTC_GEN_CNTL);
-
-	r100_gpu_wait_for_vsync(rdev);
-
-	WREG32(RADEON_CRTC_GEN_CNTL,
-	       (crtc_gen_cntl & ~(RADEON_CRTC_CUR_EN | RADEON_CRTC_ICON_EN)) |
-	       RADEON_CRTC_DISP_REQ_EN_B | RADEON_CRTC_EXT_DISP_EN);
-
-	if (!(rdev->flags & RADEON_SINGLE_CRTC)) {
-		crtc2_gen_cntl = RREG32(RADEON_CRTC2_GEN_CNTL);
-
-		r100_gpu_wait_for_vsync2(rdev);
-		WREG32(RADEON_CRTC2_GEN_CNTL,
-		       (crtc2_gen_cntl &
-		        ~(RADEON_CRTC2_CUR_EN | RADEON_CRTC2_ICON_EN)) |
-		       RADEON_CRTC2_DISP_REQ_EN_B);
-	}
-
-	udelay(500);
-}
-
-void r100_mc_setup(struct radeon_device *rdev)
-{
-	uint32_t tmp;
-	int r;
-
-	r = r100_debugfs_mc_info_init(rdev);
-	if (r) {
-		DRM_ERROR("Failed to register debugfs file for R100 MC !\n");
-	}
-	/* Write VRAM size in case we are limiting it */
-	WREG32(RADEON_CONFIG_MEMSIZE, rdev->mc.real_vram_size);
-	/* Novell bug 204882 for RN50/M6/M7 with 8/16/32MB VRAM,
-	 * if the aperture is 64MB but we have 32MB VRAM
-	 * we report only 32MB VRAM but we have to set MC_FB_LOCATION
-	 * to 64MB, otherwise the gpu accidentially dies */
-	tmp = rdev->mc.vram_location + rdev->mc.mc_vram_size - 1;
-	tmp = REG_SET(RADEON_MC_FB_TOP, tmp >> 16);
-	tmp |= REG_SET(RADEON_MC_FB_START, rdev->mc.vram_location >> 16);
-	WREG32(RADEON_MC_FB_LOCATION, tmp);
-
-	/* Enable bus mastering */
-	tmp = RREG32(RADEON_BUS_CNTL) & ~RADEON_BUS_MASTER_DIS;
-	WREG32(RADEON_BUS_CNTL, tmp);
-
-	if (rdev->flags & RADEON_IS_AGP) {
-		tmp = rdev->mc.gtt_location + rdev->mc.gtt_size - 1;
-		tmp = REG_SET(RADEON_MC_AGP_TOP, tmp >> 16);
-		tmp |= REG_SET(RADEON_MC_AGP_START, rdev->mc.gtt_location >> 16);
-		WREG32(RADEON_MC_AGP_LOCATION, tmp);
-		WREG32(RADEON_AGP_BASE, rdev->mc.agp_base);
-	} else {
-		WREG32(RADEON_MC_AGP_LOCATION, 0x0FFFFFFF);
-		WREG32(RADEON_AGP_BASE, 0);
-	}
-
-	tmp = RREG32(RADEON_HOST_PATH_CNTL) & RADEON_HDP_APER_CNTL;
-	tmp |= (7 << 28);
-	WREG32(RADEON_HOST_PATH_CNTL, tmp | RADEON_HDP_SOFT_RESET | RADEON_HDP_READ_BUFFER_INVALIDATE);
-	(void)RREG32(RADEON_HOST_PATH_CNTL);
-	WREG32(RADEON_HOST_PATH_CNTL, tmp);
-	(void)RREG32(RADEON_HOST_PATH_CNTL);
-}
-
-int r100_mc_init(struct radeon_device *rdev)
-{
-	int r;
-
-	if (r100_debugfs_rbbm_init(rdev)) {
-		DRM_ERROR("Failed to register debugfs file for RBBM !\n");
-	}
-
-	r100_gpu_init(rdev);
-	/* Disable gart which also disable out of gart access */
-	r100_pci_gart_disable(rdev);
-
-	/* Setup GPU memory space */
-	rdev->mc.gtt_location = 0xFFFFFFFFUL;
-	if (rdev->flags & RADEON_IS_AGP) {
-		r = radeon_agp_init(rdev);
-		if (r) {
-			printk(KERN_WARNING "[drm] Disabling AGP\n");
-			rdev->flags &= ~RADEON_IS_AGP;
-			rdev->mc.gtt_size = radeon_gart_size * 1024 * 1024;
-		} else {
-			rdev->mc.gtt_location = rdev->mc.agp_base;
-		}
-	}
-	r = radeon_mc_setup(rdev);
-	if (r) {
-		return r;
-	}
-
-	r100_mc_disable_clients(rdev);
-	if (r100_mc_wait_for_idle(rdev)) {
-		printk(KERN_WARNING "Failed to wait MC idle while "
-		       "programming pipes. Bad things might happen.\n");
-	}
-
-	r100_mc_setup(rdev);
-	return 0;
-}
-
-void r100_mc_fini(struct radeon_device *rdev)
-{
-}
-
-
-/*
- * Interrupts
- */
 int r100_irq_set(struct radeon_device *rdev)
 {
 	uint32_t tmp = 0;
 
+	if (!rdev->irq.installed) {
+		WARN(1, "Can't enable IRQ/MSI because no handler is installed.\n");
+		WREG32(R_000040_GEN_INT_CNTL, 0);
+		return -EINVAL;
+	}
 	if (rdev->irq.sw_int) {
 		tmp |= RADEON_SW_INT_ENABLE;
 	}
@@ -295,6 +258,12 @@ int r100_irq_set(struct radeon_device *rdev)
 	if (rdev->irq.crtc_vblank_int[1]) {
 		tmp |= RADEON_CRTC2_VBLANK_MASK;
 	}
+	if (rdev->irq.hpd[0]) {
+		tmp |= RADEON_FP_DETECT_MASK;
+	}
+	if (rdev->irq.hpd[1]) {
+		tmp |= RADEON_FP2_DETECT_MASK;
+	}
 	WREG32(RADEON_GEN_INT_CNTL, tmp);
 	return 0;
 }
@@ -313,8 +282,9 @@ void r100_irq_disable(struct radeon_device *rdev)
 static inline uint32_t r100_irq_ack(struct radeon_device *rdev)
 {
 	uint32_t irqs = RREG32(RADEON_GEN_INT_STATUS);
-	uint32_t irq_mask = RADEON_SW_INT_TEST | RADEON_CRTC_VBLANK_STAT |
-		RADEON_CRTC2_VBLANK_STAT;
+	uint32_t irq_mask = RADEON_SW_INT_TEST |
+		RADEON_CRTC_VBLANK_STAT | RADEON_CRTC2_VBLANK_STAT |
+		RADEON_FP_DETECT_STAT | RADEON_FP2_DETECT_STAT;
 
 	if (irqs) {
 		WREG32(RADEON_GEN_INT_STATUS, irqs);
@@ -324,7 +294,8 @@ static inline uint32_t r100_irq_ack(struct radeon_device *rdev)
 
 int r100_irq_process(struct radeon_device *rdev)
 {
-	uint32_t status;
+	uint32_t status, msi_rearm;
+	bool queue_hotplug = false;
 
 	status = r100_irq_ack(rdev);
 	if (!status) {
@@ -341,12 +312,39 @@ int r100_irq_process(struct radeon_device *rdev)
 		/* Vertical blank interrupts */
 		if (status & RADEON_CRTC_VBLANK_STAT) {
 			drm_handle_vblank(rdev->ddev, 0);
+			wake_up(&rdev->irq.vblank_queue);
 		}
 		if (status & RADEON_CRTC2_VBLANK_STAT) {
 			drm_handle_vblank(rdev->ddev, 1);
+			wake_up(&rdev->irq.vblank_queue);
+		}
+		if (status & RADEON_FP_DETECT_STAT) {
+			queue_hotplug = true;
+			DRM_DEBUG("HPD1\n");
+		}
+		if (status & RADEON_FP2_DETECT_STAT) {
+			queue_hotplug = true;
+			DRM_DEBUG("HPD2\n");
 		}
 		status = r100_irq_ack(rdev);
 	}
+	if (queue_hotplug)
+		queue_work(rdev->wq, &rdev->hotplug_work);
+	if (rdev->msi_enabled) {
+		switch (rdev->family) {
+		case CHIP_RS400:
+		case CHIP_RS480:
+			msi_rearm = RREG32(RADEON_AIC_CNTL) & ~RS400_MSI_REARM;
+			WREG32(RADEON_AIC_CNTL, msi_rearm);
+			WREG32(RADEON_AIC_CNTL, msi_rearm | RS400_MSI_REARM);
+			break;
+		default:
+			msi_rearm = RREG32(RADEON_MSI_REARM_EN) & ~RV370_MSI_REARM_EN;
+			WREG32(RADEON_MSI_REARM_EN, msi_rearm);
+			WREG32(RADEON_MSI_REARM_EN, msi_rearm | RV370_MSI_REARM_EN);
+			break;
+		}
+	}
 	return IRQ_HANDLED;
 }
 
@@ -358,18 +356,25 @@ u32 r100_get_vblank_counter(struct radeon_device *rdev, int crtc)
 		return RREG32(RADEON_CRTC2_CRNT_FRAME);
 }
 
-
-/*
- * Fence emission
- */
+/* Who ever call radeon_fence_emit should call ring_lock and ask
+ * for enough space (today caller are ib schedule and buffer move) */
 void r100_fence_ring_emit(struct radeon_device *rdev,
 			  struct radeon_fence *fence)
 {
-	/* Who ever call radeon_fence_emit should call ring_lock and ask
-	 * for enough space (today caller are ib schedule and buffer move) */
+	/* We have to make sure that caches are flushed before
+	 * CPU might read something from VRAM. */
+	radeon_ring_write(rdev, PACKET0(RADEON_RB3D_DSTCACHE_CTLSTAT, 0));
+	radeon_ring_write(rdev, RADEON_RB3D_DC_FLUSH_ALL);
+	radeon_ring_write(rdev, PACKET0(RADEON_RB3D_ZCACHE_CTLSTAT, 0));
+	radeon_ring_write(rdev, RADEON_RB3D_ZC_FLUSH_ALL);
 	/* Wait until IDLE & CLEAN */
-	radeon_ring_write(rdev, PACKET0(0x1720, 0));
-	radeon_ring_write(rdev, (1 << 16) | (1 << 17));
+	radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0));
+	radeon_ring_write(rdev, RADEON_WAIT_2D_IDLECLEAN | RADEON_WAIT_3D_IDLECLEAN);
+	radeon_ring_write(rdev, PACKET0(RADEON_HOST_PATH_CNTL, 0));
+	radeon_ring_write(rdev, rdev->config.r100.hdp_cntl |
+				RADEON_HDP_READ_BUFFER_INVALIDATE);
+	radeon_ring_write(rdev, PACKET0(RADEON_HOST_PATH_CNTL, 0));
+	radeon_ring_write(rdev, rdev->config.r100.hdp_cntl);
 	/* Emit fence sequence & fire IRQ */
 	radeon_ring_write(rdev, PACKET0(rdev->fence_drv.scratch_reg, 0));
 	radeon_ring_write(rdev, fence->seq);
@@ -377,33 +382,32 @@ void r100_fence_ring_emit(struct radeon_device *rdev,
 	radeon_ring_write(rdev, RADEON_SW_INT_FIRE);
 }
 
-
-/*
- * Writeback
- */
 int r100_wb_init(struct radeon_device *rdev)
 {
 	int r;
 
 	if (rdev->wb.wb_obj == NULL) {
-		r = radeon_object_create(rdev, NULL, 4096,
-					 true,
-					 RADEON_GEM_DOMAIN_GTT,
-					 false, &rdev->wb.wb_obj);
+		r = radeon_bo_create(rdev, NULL, RADEON_GPU_PAGE_SIZE, true,
+					RADEON_GEM_DOMAIN_GTT,
+					&rdev->wb.wb_obj);
 		if (r) {
-			DRM_ERROR("radeon: failed to create WB buffer (%d).\n", r);
+			dev_err(rdev->dev, "(%d) create WB buffer failed\n", r);
 			return r;
 		}
-		r = radeon_object_pin(rdev->wb.wb_obj,
-				      RADEON_GEM_DOMAIN_GTT,
-				      &rdev->wb.gpu_addr);
+		r = radeon_bo_reserve(rdev->wb.wb_obj, false);
+		if (unlikely(r != 0))
+			return r;
+		r = radeon_bo_pin(rdev->wb.wb_obj, RADEON_GEM_DOMAIN_GTT,
+					&rdev->wb.gpu_addr);
 		if (r) {
-			DRM_ERROR("radeon: failed to pin WB buffer (%d).\n", r);
+			dev_err(rdev->dev, "(%d) pin WB buffer failed\n", r);
+			radeon_bo_unreserve(rdev->wb.wb_obj);
 			return r;
 		}
-		r = radeon_object_kmap(rdev->wb.wb_obj, (void **)&rdev->wb.wb);
+		r = radeon_bo_kmap(rdev->wb.wb_obj, (void **)&rdev->wb.wb);
+		radeon_bo_unreserve(rdev->wb.wb_obj);
 		if (r) {
-			DRM_ERROR("radeon: failed to map WB buffer (%d).\n", r);
+			dev_err(rdev->dev, "(%d) map WB buffer failed\n", r);
 			return r;
 		}
 	}
@@ -421,11 +425,19 @@ void r100_wb_disable(struct radeon_device *rdev)
 
 void r100_wb_fini(struct radeon_device *rdev)
 {
+	int r;
+
 	r100_wb_disable(rdev);
 	if (rdev->wb.wb_obj) {
-		radeon_object_kunmap(rdev->wb.wb_obj);
-		radeon_object_unpin(rdev->wb.wb_obj);
-		radeon_object_unref(&rdev->wb.wb_obj);
+		r = radeon_bo_reserve(rdev->wb.wb_obj, false);
+		if (unlikely(r != 0)) {
+			dev_err(rdev->dev, "(%d) can't finish WB\n", r);
+			return;
+		}
+		radeon_bo_kunmap(rdev->wb.wb_obj);
+		radeon_bo_unpin(rdev->wb.wb_obj);
+		radeon_bo_unreserve(rdev->wb.wb_obj);
+		radeon_bo_unref(&rdev->wb.wb_obj);
 		rdev->wb.wb = NULL;
 		rdev->wb.wb_obj = NULL;
 	}
@@ -504,10 +516,6 @@ int r100_copy_blit(struct radeon_device *rdev,
 	return r;
 }
 
-
-/*
- * CP
- */
 static int r100_cp_wait_for_idle(struct radeon_device *rdev)
 {
 	unsigned i;
@@ -612,6 +620,7 @@ static int r100_cp_init_microcode(struct radeon_device *rdev)
 	}
 	return err;
 }
+
 static void r100_cp_load_microcode(struct radeon_device *rdev)
 {
 	const __be32 *fw_data;
@@ -712,19 +721,19 @@ int r100_cp_init(struct radeon_device *rdev, unsigned ring_size)
 	indirect1_start = 16;
 	/* cp setup */
 	WREG32(0x718, pre_write_timer | (pre_write_limit << 28));
-	WREG32(RADEON_CP_RB_CNTL,
-#ifdef __BIG_ENDIAN
-	       RADEON_BUF_SWAP_32BIT |
-#endif
-	       REG_SET(RADEON_RB_BUFSZ, rb_bufsz) |
+	tmp = (REG_SET(RADEON_RB_BUFSZ, rb_bufsz) |
 	       REG_SET(RADEON_RB_BLKSZ, rb_blksz) |
 	       REG_SET(RADEON_MAX_FETCH, max_fetch) |
 	       RADEON_RB_NO_UPDATE);
+#ifdef __BIG_ENDIAN
+	tmp |= RADEON_BUF_SWAP_32BIT;
+#endif
+	WREG32(RADEON_CP_RB_CNTL, tmp);
+
 	/* Set ring address */
 	DRM_INFO("radeon: ring at 0x%016lX\n", (unsigned long)rdev->cp.gpu_addr);
 	WREG32(RADEON_CP_RB_BASE, rdev->cp.gpu_addr);
 	/* Force read & write ptr to 0 */
-	tmp = RREG32(RADEON_CP_RB_CNTL);
 	WREG32(RADEON_CP_RB_CNTL, tmp | RADEON_RB_RPTR_WR_ENA);
 	WREG32(RADEON_CP_RB_RPTR_WR, 0);
 	WREG32(RADEON_CP_RB_WPTR, 0);
@@ -863,13 +872,11 @@ int r100_cs_parse_packet0(struct radeon_cs_parser *p,
 void r100_cs_dump_packet(struct radeon_cs_parser *p,
 			 struct radeon_cs_packet *pkt)
 {
-	struct radeon_cs_chunk *ib_chunk;
 	volatile uint32_t *ib;
 	unsigned i;
 	unsigned idx;
 
 	ib = p->ib->ptr;
-	ib_chunk = &p->chunks[p->chunk_ib_idx];
 	idx = pkt->idx;
 	for (i = 0; i <= (pkt->count + 1); i++, idx++) {
 		DRM_INFO("ib[%d]=0x%08X\n", idx, ib[idx]);
@@ -896,7 +903,7 @@ int r100_cs_packet_parse(struct radeon_cs_parser *p,
 			  idx, ib_chunk->length_dw);
 		return -EINVAL;
 	}
-	header = ib_chunk->kdata[idx];
+	header = radeon_get_ib_value(p, idx);
 	pkt->idx = idx;
 	pkt->type = CP_PACKET_GET_TYPE(header);
 	pkt->count = CP_PACKET_GET_COUNT(header);
@@ -939,7 +946,6 @@ int r100_cs_packet_parse(struct radeon_cs_parser *p,
  */
 int r100_cs_packet_parse_vline(struct radeon_cs_parser *p)
 {
-	struct radeon_cs_chunk *ib_chunk;
 	struct drm_mode_object *obj;
 	struct drm_crtc *crtc;
 	struct radeon_crtc *radeon_crtc;
@@ -947,8 +953,9 @@ int r100_cs_packet_parse_vline(struct radeon_cs_parser *p)
 	int crtc_id;
 	int r;
 	uint32_t header, h_idx, reg;
+	volatile uint32_t *ib;
 
-	ib_chunk = &p->chunks[p->chunk_ib_idx];
+	ib = p->ib->ptr;
 
 	/* parse the wait until */
 	r = r100_cs_packet_parse(p, &waitreloc, p->idx);
@@ -963,24 +970,24 @@ int r100_cs_packet_parse_vline(struct radeon_cs_parser *p)
 		return r;
 	}
 
-	if (ib_chunk->kdata[waitreloc.idx + 1] != RADEON_WAIT_CRTC_VLINE) {
+	if (radeon_get_ib_value(p, waitreloc.idx + 1) != RADEON_WAIT_CRTC_VLINE) {
 		DRM_ERROR("vline wait had illegal wait until\n");
 		r = -EINVAL;
 		return r;
 	}
 
 	/* jump over the NOP */
-	r = r100_cs_packet_parse(p, &p3reloc, p->idx);
+	r = r100_cs_packet_parse(p, &p3reloc, p->idx + waitreloc.count + 2);
 	if (r)
 		return r;
 
 	h_idx = p->idx - 2;
-	p->idx += waitreloc.count;
-	p->idx += p3reloc.count;
+	p->idx += waitreloc.count + 2;
+	p->idx += p3reloc.count + 2;
 
-	header = ib_chunk->kdata[h_idx];
-	crtc_id = ib_chunk->kdata[h_idx + 5];
-	reg = ib_chunk->kdata[h_idx] >> 2;
+	header = radeon_get_ib_value(p, h_idx);
+	crtc_id = radeon_get_ib_value(p, h_idx + 5);
+	reg = CP_PACKET0_GET_REG(header);
 	mutex_lock(&p->rdev->ddev->mode_config.mutex);
 	obj = drm_mode_object_find(p->rdev->ddev, crtc_id, DRM_MODE_OBJECT_CRTC);
 	if (!obj) {
@@ -994,16 +1001,16 @@ int r100_cs_packet_parse_vline(struct radeon_cs_parser *p)
 
 	if (!crtc->enabled) {
 		/* if the CRTC isn't enabled - we need to nop out the wait until */
-		ib_chunk->kdata[h_idx + 2] = PACKET2(0);
-		ib_chunk->kdata[h_idx + 3] = PACKET2(0);
+		ib[h_idx + 2] = PACKET2(0);
+		ib[h_idx + 3] = PACKET2(0);
 	} else if (crtc_id == 1) {
 		switch (reg) {
 		case AVIVO_D1MODE_VLINE_START_END:
-			header &= R300_CP_PACKET0_REG_MASK;
+			header &= ~R300_CP_PACKET0_REG_MASK;
 			header |= AVIVO_D2MODE_VLINE_START_END >> 2;
 			break;
 		case RADEON_CRTC_GUI_TRIG_VLINE:
-			header &= R300_CP_PACKET0_REG_MASK;
+			header &= ~R300_CP_PACKET0_REG_MASK;
 			header |= RADEON_CRTC2_GUI_TRIG_VLINE >> 2;
 			break;
 		default:
@@ -1011,8 +1018,8 @@ int r100_cs_packet_parse_vline(struct radeon_cs_parser *p)
 			r = -EINVAL;
 			goto out;
 		}
-		ib_chunk->kdata[h_idx] = header;
-		ib_chunk->kdata[h_idx + 3] |= RADEON_ENG_DISPLAY_SELECT_CRTC1;
+		ib[h_idx] = header;
+		ib[h_idx + 3] |= RADEON_ENG_DISPLAY_SELECT_CRTC1;
 	}
 out:
 	mutex_unlock(&p->rdev->ddev->mode_config.mutex);
@@ -1033,7 +1040,6 @@ out:
 int r100_cs_packet_next_reloc(struct radeon_cs_parser *p,
 			      struct radeon_cs_reloc **cs_reloc)
 {
-	struct radeon_cs_chunk *ib_chunk;
 	struct radeon_cs_chunk *relocs_chunk;
 	struct radeon_cs_packet p3reloc;
 	unsigned idx;
@@ -1044,7 +1050,6 @@ int r100_cs_packet_next_reloc(struct radeon_cs_parser *p,
 		return -EINVAL;
 	}
 	*cs_reloc = NULL;
-	ib_chunk = &p->chunks[p->chunk_ib_idx];
 	relocs_chunk = &p->chunks[p->chunk_relocs_idx];
 	r = r100_cs_packet_parse(p, &p3reloc, p->idx);
 	if (r) {
@@ -1057,7 +1062,7 @@ int r100_cs_packet_next_reloc(struct radeon_cs_parser *p,
 		r100_cs_dump_packet(p, &p3reloc);
 		return -EINVAL;
 	}
-	idx = ib_chunk->kdata[p3reloc.idx + 1];
+	idx = radeon_get_ib_value(p, p3reloc.idx + 1);
 	if (idx >= relocs_chunk->length_dw) {
 		DRM_ERROR("Relocs at %d after relocations chunk end %d !\n",
 			  idx, relocs_chunk->length_dw);
@@ -1126,7 +1131,6 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
 			      struct radeon_cs_packet *pkt,
 			      unsigned idx, unsigned reg)
 {
-	struct radeon_cs_chunk *ib_chunk;
 	struct radeon_cs_reloc *reloc;
 	struct r100_cs_track *track;
 	volatile uint32_t *ib;
@@ -1134,11 +1138,13 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
 	int r;
 	int i, face;
 	u32 tile_flags = 0;
+	u32 idx_value;
 
 	ib = p->ib->ptr;
-	ib_chunk = &p->chunks[p->chunk_ib_idx];
 	track = (struct r100_cs_track *)p->track;
 
+	idx_value = radeon_get_ib_value(p, idx);
+
 	switch (reg) {
 	case RADEON_CRTC_GUI_TRIG_VLINE:
 		r = r100_cs_packet_parse_vline(p);
@@ -1166,8 +1172,8 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
 			return r;
 		}
 		track->zb.robj = reloc->robj;
-		track->zb.offset = ib_chunk->kdata[idx];
-		ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset);
+		track->zb.offset = idx_value;
+		ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
 		break;
 	case RADEON_RB3D_COLOROFFSET:
 		r = r100_cs_packet_next_reloc(p, &reloc);
@@ -1178,8 +1184,8 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
 			return r;
 		}
 		track->cb[0].robj = reloc->robj;
-		track->cb[0].offset = ib_chunk->kdata[idx];
-		ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset);
+		track->cb[0].offset = idx_value;
+		ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
 		break;
 	case RADEON_PP_TXOFFSET_0:
 	case RADEON_PP_TXOFFSET_1:
@@ -1192,7 +1198,7 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
 			r100_cs_dump_packet(p, pkt);
 			return r;
 		}
-		ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset);
+		ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
 		track->textures[i].robj = reloc->robj;
 		break;
 	case RADEON_PP_CUBIC_OFFSET_T0_0:
@@ -1208,8 +1214,8 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
 			r100_cs_dump_packet(p, pkt);
 			return r;
 		}
-		track->textures[0].cube_info[i].offset = ib_chunk->kdata[idx];
-		ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset);
+		track->textures[0].cube_info[i].offset = idx_value;
+		ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
 		track->textures[0].cube_info[i].robj = reloc->robj;
 		break;
 	case RADEON_PP_CUBIC_OFFSET_T1_0:
@@ -1225,8 +1231,8 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
 			r100_cs_dump_packet(p, pkt);
 			return r;
 		}
-		track->textures[1].cube_info[i].offset = ib_chunk->kdata[idx];
-		ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset);
+		track->textures[1].cube_info[i].offset = idx_value;
+		ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
 		track->textures[1].cube_info[i].robj = reloc->robj;
 		break;
 	case RADEON_PP_CUBIC_OFFSET_T2_0:
@@ -1242,12 +1248,12 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
 			r100_cs_dump_packet(p, pkt);
 			return r;
 		}
-		track->textures[2].cube_info[i].offset = ib_chunk->kdata[idx];
-		ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset);
+		track->textures[2].cube_info[i].offset = idx_value;
+		ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
 		track->textures[2].cube_info[i].robj = reloc->robj;
 		break;
 	case RADEON_RE_WIDTH_HEIGHT:
-		track->maxy = ((ib_chunk->kdata[idx] >> 16) & 0x7FF);
+		track->maxy = ((idx_value >> 16) & 0x7FF);
 		break;
 	case RADEON_RB3D_COLORPITCH:
 		r = r100_cs_packet_next_reloc(p, &reloc);
@@ -1263,17 +1269,17 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
 		if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO)
 			tile_flags |= RADEON_COLOR_MICROTILE_ENABLE;
 
-		tmp = ib_chunk->kdata[idx] & ~(0x7 << 16);
+		tmp = idx_value & ~(0x7 << 16);
 		tmp |= tile_flags;
 		ib[idx] = tmp;
 
-		track->cb[0].pitch = ib_chunk->kdata[idx] & RADEON_COLORPITCH_MASK;
+		track->cb[0].pitch = idx_value & RADEON_COLORPITCH_MASK;
 		break;
 	case RADEON_RB3D_DEPTHPITCH:
-		track->zb.pitch = ib_chunk->kdata[idx] & RADEON_DEPTHPITCH_MASK;
+		track->zb.pitch = idx_value & RADEON_DEPTHPITCH_MASK;
 		break;
 	case RADEON_RB3D_CNTL:
-		switch ((ib_chunk->kdata[idx] >> RADEON_RB3D_COLOR_FORMAT_SHIFT) & 0x1f) {
+		switch ((idx_value >> RADEON_RB3D_COLOR_FORMAT_SHIFT) & 0x1f) {
 		case 7:
 		case 8:
 		case 9:
@@ -1291,13 +1297,13 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
 			break;
 		default:
 			DRM_ERROR("Invalid color buffer format (%d) !\n",
-				  ((ib_chunk->kdata[idx] >> RADEON_RB3D_COLOR_FORMAT_SHIFT) & 0x1f));
+				  ((idx_value >> RADEON_RB3D_COLOR_FORMAT_SHIFT) & 0x1f));
 			return -EINVAL;
 		}
-		track->z_enabled = !!(ib_chunk->kdata[idx] & RADEON_Z_ENABLE);
+		track->z_enabled = !!(idx_value & RADEON_Z_ENABLE);
 		break;
 	case RADEON_RB3D_ZSTENCILCNTL:
-		switch (ib_chunk->kdata[idx] & 0xf) {
+		switch (idx_value & 0xf) {
 		case 0:
 			track->zb.cpp = 2;
 			break;
@@ -1321,44 +1327,44 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
 			r100_cs_dump_packet(p, pkt);
 			return r;
 		}
-		ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset);
+		ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
 		break;
 	case RADEON_PP_CNTL:
 		{
-			uint32_t temp = ib_chunk->kdata[idx] >> 4;
+			uint32_t temp = idx_value >> 4;
 			for (i = 0; i < track->num_texture; i++)
 				track->textures[i].enabled = !!(temp & (1 << i));
 		}
 		break;
 	case RADEON_SE_VF_CNTL:
-		track->vap_vf_cntl = ib_chunk->kdata[idx];
+		track->vap_vf_cntl = idx_value;
 		break;
 	case RADEON_SE_VTX_FMT:
-		track->vtx_size = r100_get_vtx_size(ib_chunk->kdata[idx]);
+		track->vtx_size = r100_get_vtx_size(idx_value);
 		break;
 	case RADEON_PP_TEX_SIZE_0:
 	case RADEON_PP_TEX_SIZE_1:
 	case RADEON_PP_TEX_SIZE_2:
 		i = (reg - RADEON_PP_TEX_SIZE_0) / 8;
-		track->textures[i].width = (ib_chunk->kdata[idx] & RADEON_TEX_USIZE_MASK) + 1;
-		track->textures[i].height = ((ib_chunk->kdata[idx] & RADEON_TEX_VSIZE_MASK) >> RADEON_TEX_VSIZE_SHIFT) + 1;
+		track->textures[i].width = (idx_value & RADEON_TEX_USIZE_MASK) + 1;
+		track->textures[i].height = ((idx_value & RADEON_TEX_VSIZE_MASK) >> RADEON_TEX_VSIZE_SHIFT) + 1;
 		break;
 	case RADEON_PP_TEX_PITCH_0:
 	case RADEON_PP_TEX_PITCH_1:
 	case RADEON_PP_TEX_PITCH_2:
 		i = (reg - RADEON_PP_TEX_PITCH_0) / 8;
-		track->textures[i].pitch = ib_chunk->kdata[idx] + 32;
+		track->textures[i].pitch = idx_value + 32;
 		break;
 	case RADEON_PP_TXFILTER_0:
 	case RADEON_PP_TXFILTER_1:
 	case RADEON_PP_TXFILTER_2:
 		i = (reg - RADEON_PP_TXFILTER_0) / 24;
-		track->textures[i].num_levels = ((ib_chunk->kdata[idx] & RADEON_MAX_MIP_LEVEL_MASK)
+		track->textures[i].num_levels = ((idx_value & RADEON_MAX_MIP_LEVEL_MASK)
 						 >> RADEON_MAX_MIP_LEVEL_SHIFT);
-		tmp = (ib_chunk->kdata[idx] >> 23) & 0x7;
+		tmp = (idx_value >> 23) & 0x7;
 		if (tmp == 2 || tmp == 6)
 			track->textures[i].roundup_w = false;
-		tmp = (ib_chunk->kdata[idx] >> 27) & 0x7;
+		tmp = (idx_value >> 27) & 0x7;
 		if (tmp == 2 || tmp == 6)
 			track->textures[i].roundup_h = false;
 		break;
@@ -1366,16 +1372,16 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
 	case RADEON_PP_TXFORMAT_1:
 	case RADEON_PP_TXFORMAT_2:
 		i = (reg - RADEON_PP_TXFORMAT_0) / 24;
-		if (ib_chunk->kdata[idx] & RADEON_TXFORMAT_NON_POWER2) {
+		if (idx_value & RADEON_TXFORMAT_NON_POWER2) {
 			track->textures[i].use_pitch = 1;
 		} else {
 			track->textures[i].use_pitch = 0;
-			track->textures[i].width = 1 << ((ib_chunk->kdata[idx] >> RADEON_TXFORMAT_WIDTH_SHIFT) & RADEON_TXFORMAT_WIDTH_MASK);
-			track->textures[i].height = 1 << ((ib_chunk->kdata[idx] >> RADEON_TXFORMAT_HEIGHT_SHIFT) & RADEON_TXFORMAT_HEIGHT_MASK);
+			track->textures[i].width = 1 << ((idx_value >> RADEON_TXFORMAT_WIDTH_SHIFT) & RADEON_TXFORMAT_WIDTH_MASK);
+			track->textures[i].height = 1 << ((idx_value >> RADEON_TXFORMAT_HEIGHT_SHIFT) & RADEON_TXFORMAT_HEIGHT_MASK);
 		}
-		if (ib_chunk->kdata[idx] & RADEON_TXFORMAT_CUBIC_MAP_ENABLE)
+		if (idx_value & RADEON_TXFORMAT_CUBIC_MAP_ENABLE)
 			track->textures[i].tex_coord_type = 2;
-		switch ((ib_chunk->kdata[idx] & RADEON_TXFORMAT_FORMAT_MASK)) {
+		switch ((idx_value & RADEON_TXFORMAT_FORMAT_MASK)) {
 		case RADEON_TXFORMAT_I8:
 		case RADEON_TXFORMAT_RGB332:
 		case RADEON_TXFORMAT_Y8:
@@ -1387,7 +1393,6 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
 		case RADEON_TXFORMAT_ARGB4444:
 		case RADEON_TXFORMAT_VYUY422:
 		case RADEON_TXFORMAT_YVYU422:
-		case RADEON_TXFORMAT_DXT1:
 		case RADEON_TXFORMAT_SHADOW16:
 		case RADEON_TXFORMAT_LDUDV655:
 		case RADEON_TXFORMAT_DUDV88:
@@ -1395,20 +1400,27 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
 			break;
 		case RADEON_TXFORMAT_ARGB8888:
 		case RADEON_TXFORMAT_RGBA8888:
-		case RADEON_TXFORMAT_DXT23:
-		case RADEON_TXFORMAT_DXT45:
 		case RADEON_TXFORMAT_SHADOW32:
 		case RADEON_TXFORMAT_LDUDUV8888:
 			track->textures[i].cpp = 4;
 			break;
+		case RADEON_TXFORMAT_DXT1:
+			track->textures[i].cpp = 1;
+			track->textures[i].compress_format = R100_TRACK_COMP_DXT1;
+			break;
+		case RADEON_TXFORMAT_DXT23:
+		case RADEON_TXFORMAT_DXT45:
+			track->textures[i].cpp = 1;
+			track->textures[i].compress_format = R100_TRACK_COMP_DXT35;
+			break;
 		}
-		track->textures[i].cube_info[4].width = 1 << ((ib_chunk->kdata[idx] >> 16) & 0xf);
-		track->textures[i].cube_info[4].height = 1 << ((ib_chunk->kdata[idx] >> 20) & 0xf);
+		track->textures[i].cube_info[4].width = 1 << ((idx_value >> 16) & 0xf);
+		track->textures[i].cube_info[4].height = 1 << ((idx_value >> 20) & 0xf);
 		break;
 	case RADEON_PP_CUBIC_FACES_0:
 	case RADEON_PP_CUBIC_FACES_1:
 	case RADEON_PP_CUBIC_FACES_2:
-		tmp = ib_chunk->kdata[idx];
+		tmp = idx_value;
 		i = (reg - RADEON_PP_CUBIC_FACES_0) / 4;
 		for (face = 0; face < 4; face++) {
 			track->textures[i].cube_info[face].width = 1 << ((tmp >> (face * 8)) & 0xf);
@@ -1425,18 +1437,17 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
 
 int r100_cs_track_check_pkt3_indx_buffer(struct radeon_cs_parser *p,
 					 struct radeon_cs_packet *pkt,
-					 struct radeon_object *robj)
+					 struct radeon_bo *robj)
 {
-	struct radeon_cs_chunk *ib_chunk;
 	unsigned idx;
-
-	ib_chunk = &p->chunks[p->chunk_ib_idx];
+	u32 value;
 	idx = pkt->idx + 1;
-	if ((ib_chunk->kdata[idx+2] + 1) > radeon_object_size(robj)) {
+	value = radeon_get_ib_value(p, idx + 2);
+	if ((value + 1) > radeon_bo_size(robj)) {
 		DRM_ERROR("[drm] Buffer too small for PACKET3 INDX_BUFFER "
 			  "(need %u have %lu) !\n",
-			  ib_chunk->kdata[idx+2] + 1,
-			  radeon_object_size(robj));
+			  value + 1,
+			  radeon_bo_size(robj));
 		return -EINVAL;
 	}
 	return 0;
@@ -1445,59 +1456,20 @@ int r100_cs_track_check_pkt3_indx_buffer(struct radeon_cs_parser *p,
 static int r100_packet3_check(struct radeon_cs_parser *p,
 			      struct radeon_cs_packet *pkt)
 {
-	struct radeon_cs_chunk *ib_chunk;
 	struct radeon_cs_reloc *reloc;
 	struct r100_cs_track *track;
 	unsigned idx;
-	unsigned i, c;
 	volatile uint32_t *ib;
 	int r;
 
 	ib = p->ib->ptr;
-	ib_chunk = &p->chunks[p->chunk_ib_idx];
 	idx = pkt->idx + 1;
 	track = (struct r100_cs_track *)p->track;
 	switch (pkt->opcode) {
 	case PACKET3_3D_LOAD_VBPNTR:
-		c = ib_chunk->kdata[idx++];
-		track->num_arrays = c;
-		for (i = 0; i < (c - 1); i += 2, idx += 3) {
-			r = r100_cs_packet_next_reloc(p, &reloc);
-			if (r) {
-				DRM_ERROR("No reloc for packet3 %d\n",
-					  pkt->opcode);
-				r100_cs_dump_packet(p, pkt);
-				return r;
-			}
-			ib[idx+1] = ib_chunk->kdata[idx+1] + ((u32)reloc->lobj.gpu_offset);
-			track->arrays[i + 0].robj = reloc->robj;
-			track->arrays[i + 0].esize = ib_chunk->kdata[idx] >> 8;
-			track->arrays[i + 0].esize &= 0x7F;
-			r = r100_cs_packet_next_reloc(p, &reloc);
-			if (r) {
-				DRM_ERROR("No reloc for packet3 %d\n",
-					  pkt->opcode);
-				r100_cs_dump_packet(p, pkt);
-				return r;
-			}
-			ib[idx+2] = ib_chunk->kdata[idx+2] + ((u32)reloc->lobj.gpu_offset);
-			track->arrays[i + 1].robj = reloc->robj;
-			track->arrays[i + 1].esize = ib_chunk->kdata[idx] >> 24;
-			track->arrays[i + 1].esize &= 0x7F;
-		}
-		if (c & 1) {
-			r = r100_cs_packet_next_reloc(p, &reloc);
-			if (r) {
-				DRM_ERROR("No reloc for packet3 %d\n",
-					  pkt->opcode);
-				r100_cs_dump_packet(p, pkt);
-				return r;
-			}
-			ib[idx+1] = ib_chunk->kdata[idx+1] + ((u32)reloc->lobj.gpu_offset);
-			track->arrays[i + 0].robj = reloc->robj;
-			track->arrays[i + 0].esize = ib_chunk->kdata[idx] >> 8;
-			track->arrays[i + 0].esize &= 0x7F;
-		}
+		r = r100_packet3_load_vbpntr(p, pkt, idx);
+		if (r)
+			return r;
 		break;
 	case PACKET3_INDX_BUFFER:
 		r = r100_cs_packet_next_reloc(p, &reloc);
@@ -1506,7 +1478,7 @@ static int r100_packet3_check(struct radeon_cs_parser *p,
 			r100_cs_dump_packet(p, pkt);
 			return r;
 		}
-		ib[idx+1] = ib_chunk->kdata[idx+1] + ((u32)reloc->lobj.gpu_offset);
+		ib[idx+1] = radeon_get_ib_value(p, idx+1) + ((u32)reloc->lobj.gpu_offset);
 		r = r100_cs_track_check_pkt3_indx_buffer(p, pkt, reloc->robj);
 		if (r) {
 			return r;
@@ -1520,27 +1492,28 @@ static int r100_packet3_check(struct radeon_cs_parser *p,
 			r100_cs_dump_packet(p, pkt);
 			return r;
 		}
-		ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset);
+		ib[idx] = radeon_get_ib_value(p, idx) + ((u32)reloc->lobj.gpu_offset);
 		track->num_arrays = 1;
-		track->vtx_size = r100_get_vtx_size(ib_chunk->kdata[idx+2]);
+		track->vtx_size = r100_get_vtx_size(radeon_get_ib_value(p, idx + 2));
 
 		track->arrays[0].robj = reloc->robj;
 		track->arrays[0].esize = track->vtx_size;
 
-		track->max_indx = ib_chunk->kdata[idx+1];
+		track->max_indx = radeon_get_ib_value(p, idx+1);
 
-		track->vap_vf_cntl = ib_chunk->kdata[idx+3];
+		track->vap_vf_cntl = radeon_get_ib_value(p, idx+3);
 		track->immd_dwords = pkt->count - 1;
 		r = r100_cs_track_check(p->rdev, track);
 		if (r)
 			return r;
 		break;
 	case PACKET3_3D_DRAW_IMMD:
-		if (((ib_chunk->kdata[idx+1] >> 4) & 0x3) != 3) {
+		if (((radeon_get_ib_value(p, idx + 1) >> 4) & 0x3) != 3) {
 			DRM_ERROR("PRIM_WALK must be 3 for IMMD draw\n");
 			return -EINVAL;
 		}
-		track->vap_vf_cntl = ib_chunk->kdata[idx+1];
+		track->vtx_size = r100_get_vtx_size(radeon_get_ib_value(p, idx + 0));
+		track->vap_vf_cntl = radeon_get_ib_value(p, idx + 1);
 		track->immd_dwords = pkt->count - 1;
 		r = r100_cs_track_check(p->rdev, track);
 		if (r)
@@ -1548,11 +1521,11 @@ static int r100_packet3_check(struct radeon_cs_parser *p,
 		break;
 		/* triggers drawing using in-packet vertex data */
 	case PACKET3_3D_DRAW_IMMD_2:
-		if (((ib_chunk->kdata[idx] >> 4) & 0x3) != 3) {
+		if (((radeon_get_ib_value(p, idx) >> 4) & 0x3) != 3) {
 			DRM_ERROR("PRIM_WALK must be 3 for IMMD draw\n");
 			return -EINVAL;
 		}
-		track->vap_vf_cntl = ib_chunk->kdata[idx];
+		track->vap_vf_cntl = radeon_get_ib_value(p, idx);
 		track->immd_dwords = pkt->count;
 		r = r100_cs_track_check(p->rdev, track);
 		if (r)
@@ -1560,28 +1533,28 @@ static int r100_packet3_check(struct radeon_cs_parser *p,
 		break;
 		/* triggers drawing using in-packet vertex data */
 	case PACKET3_3D_DRAW_VBUF_2:
-		track->vap_vf_cntl = ib_chunk->kdata[idx];
+		track->vap_vf_cntl = radeon_get_ib_value(p, idx);
 		r = r100_cs_track_check(p->rdev, track);
 		if (r)
 			return r;
 		break;
 		/* triggers drawing of vertex buffers setup elsewhere */
 	case PACKET3_3D_DRAW_INDX_2:
-		track->vap_vf_cntl = ib_chunk->kdata[idx];
+		track->vap_vf_cntl = radeon_get_ib_value(p, idx);
 		r = r100_cs_track_check(p->rdev, track);
 		if (r)
 			return r;
 		break;
 		/* triggers drawing using indices to vertex buffer */
 	case PACKET3_3D_DRAW_VBUF:
-		track->vap_vf_cntl = ib_chunk->kdata[idx + 1];
+		track->vap_vf_cntl = radeon_get_ib_value(p, idx + 1);
 		r = r100_cs_track_check(p->rdev, track);
 		if (r)
 			return r;
 		break;
 		/* triggers drawing of vertex buffers setup elsewhere */
 	case PACKET3_3D_DRAW_INDX:
-		track->vap_vf_cntl = ib_chunk->kdata[idx + 1];
+		track->vap_vf_cntl = radeon_get_ib_value(p, idx + 1);
 		r = r100_cs_track_check(p->rdev, track);
 		if (r)
 			return r;
@@ -1730,7 +1703,7 @@ int r100_gui_wait_for_idle(struct radeon_device *rdev)
 	}
 	for (i = 0; i < rdev->usec_timeout; i++) {
 		tmp = RREG32(RADEON_RBBM_STATUS);
-		if (!(tmp & (1 << 31))) {
+		if (!(tmp & RADEON_RBBM_ACTIVE)) {
 			return 0;
 		}
 		DRM_UDELAY(1);
@@ -1745,8 +1718,8 @@ int r100_mc_wait_for_idle(struct radeon_device *rdev)
 
 	for (i = 0; i < rdev->usec_timeout; i++) {
 		/* read MC_STATUS */
-		tmp = RREG32(0x0150);
-		if (tmp & (1 << 2)) {
+		tmp = RREG32(RADEON_MC_STATUS);
+		if (tmp & RADEON_MC_IDLE) {
 			return 0;
 		}
 		DRM_UDELAY(1);
@@ -1819,7 +1792,7 @@ int r100_gpu_reset(struct radeon_device *rdev)
 	}
 	/* Check if GPU is idle */
 	status = RREG32(RADEON_RBBM_STATUS);
-	if (status & (1 << 31)) {
+	if (status & RADEON_RBBM_ACTIVE) {
 		DRM_ERROR("Failed to reset GPU (RBBM_STATUS=0x%08X)\n", status);
 		return -1;
 	}
@@ -1827,6 +1800,82 @@ int r100_gpu_reset(struct radeon_device *rdev)
 	return 0;
 }
 
+void r100_set_common_regs(struct radeon_device *rdev)
+{
+	struct drm_device *dev = rdev->ddev;
+	bool force_dac2 = false;
+
+	/* set these so they don't interfere with anything */
+	WREG32(RADEON_OV0_SCALE_CNTL, 0);
+	WREG32(RADEON_SUBPIC_CNTL, 0);
+	WREG32(RADEON_VIPH_CONTROL, 0);
+	WREG32(RADEON_I2C_CNTL_1, 0);
+	WREG32(RADEON_DVI_I2C_CNTL_1, 0);
+	WREG32(RADEON_CAP0_TRIG_CNTL, 0);
+	WREG32(RADEON_CAP1_TRIG_CNTL, 0);
+
+	/* always set up dac2 on rn50 and some rv100 as lots
+	 * of servers seem to wire it up to a VGA port but
+	 * don't report it in the bios connector
+	 * table.
+	 */
+	switch (dev->pdev->device) {
+		/* RN50 */
+	case 0x515e:
+	case 0x5969:
+		force_dac2 = true;
+		break;
+		/* RV100*/
+	case 0x5159:
+	case 0x515a:
+		/* DELL triple head servers */
+		if ((dev->pdev->subsystem_vendor == 0x1028 /* DELL */) &&
+		    ((dev->pdev->subsystem_device == 0x016c) ||
+		     (dev->pdev->subsystem_device == 0x016d) ||
+		     (dev->pdev->subsystem_device == 0x016e) ||
+		     (dev->pdev->subsystem_device == 0x016f) ||
+		     (dev->pdev->subsystem_device == 0x0170) ||
+		     (dev->pdev->subsystem_device == 0x017d) ||
+		     (dev->pdev->subsystem_device == 0x017e) ||
+		     (dev->pdev->subsystem_device == 0x0183) ||
+		     (dev->pdev->subsystem_device == 0x018a) ||
+		     (dev->pdev->subsystem_device == 0x019a)))
+			force_dac2 = true;
+		break;
+	}
+
+	if (force_dac2) {
+		u32 disp_hw_debug = RREG32(RADEON_DISP_HW_DEBUG);
+		u32 tv_dac_cntl = RREG32(RADEON_TV_DAC_CNTL);
+		u32 dac2_cntl = RREG32(RADEON_DAC_CNTL2);
+
+		/* For CRT on DAC2, don't turn it on if BIOS didn't
+		   enable it, even it's detected.
+		*/
+
+		/* force it to crtc0 */
+		dac2_cntl &= ~RADEON_DAC2_DAC_CLK_SEL;
+		dac2_cntl |= RADEON_DAC2_DAC2_CLK_SEL;
+		disp_hw_debug |= RADEON_CRT2_DISP1_SEL;
+
+		/* set up the TV DAC */
+		tv_dac_cntl &= ~(RADEON_TV_DAC_PEDESTAL |
+				 RADEON_TV_DAC_STD_MASK |
+				 RADEON_TV_DAC_RDACPD |
+				 RADEON_TV_DAC_GDACPD |
+				 RADEON_TV_DAC_BDACPD |
+				 RADEON_TV_DAC_BGADJ_MASK |
+				 RADEON_TV_DAC_DACADJ_MASK);
+		tv_dac_cntl |= (RADEON_TV_DAC_NBLANK |
+				RADEON_TV_DAC_NHOLD |
+				RADEON_TV_DAC_STD_PS2 |
+				(0x58 << 16));
+
+		WREG32(RADEON_TV_DAC_CNTL, tv_dac_cntl);
+		WREG32(RADEON_DISP_HW_DEBUG, disp_hw_debug);
+		WREG32(RADEON_DAC_CNTL2, dac2_cntl);
+	}
+}
 
 /*
  * VRAM info
@@ -1907,17 +1956,20 @@ static u32 r100_get_accessible_vram(struct radeon_device *rdev)
 void r100_vram_init_sizes(struct radeon_device *rdev)
 {
 	u64 config_aper_size;
-	u32 accessible;
 
+	/* work out accessible VRAM */
+	rdev->mc.aper_base = drm_get_resource_start(rdev->ddev, 0);
+	rdev->mc.aper_size = drm_get_resource_len(rdev->ddev, 0);
+	rdev->mc.visible_vram_size = r100_get_accessible_vram(rdev);
+	/* FIXME we don't use the second aperture yet when we could use it */
+	if (rdev->mc.visible_vram_size > rdev->mc.aper_size)
+		rdev->mc.visible_vram_size = rdev->mc.aper_size;
 	config_aper_size = RREG32(RADEON_CONFIG_APER_SIZE);
-
 	if (rdev->flags & RADEON_IS_IGP) {
 		uint32_t tom;
 		/* read NB_TOM to get the amount of ram stolen for the GPU */
 		tom = RREG32(RADEON_NB_TOM);
 		rdev->mc.real_vram_size = (((tom >> 16) - (tom & 0xffff) + 1) << 16);
-		/* for IGPs we need to keep VRAM where it was put by the BIOS */
-		rdev->mc.vram_location = (tom & 0xffff) << 16;
 		WREG32(RADEON_CONFIG_MEMSIZE, rdev->mc.real_vram_size);
 		rdev->mc.mc_vram_size = rdev->mc.real_vram_size;
 	} else {
@@ -1929,30 +1981,19 @@ void r100_vram_init_sizes(struct radeon_device *rdev)
 			rdev->mc.real_vram_size = 8192 * 1024;
 			WREG32(RADEON_CONFIG_MEMSIZE, rdev->mc.real_vram_size);
 		}
-		/* let driver place VRAM */
-		rdev->mc.vram_location = 0xFFFFFFFFUL;
-		 /* Fix for RN50, M6, M7 with 8/16/32(??) MBs of VRAM - 
-		  * Novell bug 204882 + along with lots of ubuntu ones */
+		/* Fix for RN50, M6, M7 with 8/16/32(??) MBs of VRAM - 
+		 * Novell bug 204882 + along with lots of ubuntu ones
+		 */
 		if (config_aper_size > rdev->mc.real_vram_size)
 			rdev->mc.mc_vram_size = config_aper_size;
 		else
 			rdev->mc.mc_vram_size = rdev->mc.real_vram_size;
 	}
-
-	/* work out accessible VRAM */
-	accessible = r100_get_accessible_vram(rdev);
-
-	rdev->mc.aper_base = drm_get_resource_start(rdev->ddev, 0);
-	rdev->mc.aper_size = drm_get_resource_len(rdev->ddev, 0);
-
-	if (accessible > rdev->mc.aper_size)
-		accessible = rdev->mc.aper_size;
-
-	if (rdev->mc.mc_vram_size > rdev->mc.aper_size)
+	/* FIXME remove this once we support unmappable VRAM */
+	if (rdev->mc.mc_vram_size > rdev->mc.aper_size) {
 		rdev->mc.mc_vram_size = rdev->mc.aper_size;
-
-	if (rdev->mc.real_vram_size > rdev->mc.aper_size)
 		rdev->mc.real_vram_size = rdev->mc.aper_size;
+	}
 }
 
 void r100_vga_set_state(struct radeon_device *rdev, bool state)
@@ -1969,11 +2010,18 @@ void r100_vga_set_state(struct radeon_device *rdev, bool state)
 	WREG32(RADEON_CONFIG_CNTL, temp);
 }
 
-void r100_vram_info(struct radeon_device *rdev)
+void r100_mc_init(struct radeon_device *rdev)
 {
-	r100_vram_get_type(rdev);
+	u64 base;
 
+	r100_vram_get_type(rdev);
 	r100_vram_init_sizes(rdev);
+	base = rdev->mc.aper_base;
+	if (rdev->flags & RADEON_IS_IGP)
+		base = (RREG32(RADEON_NB_TOM) & 0xffff) << 16;
+	radeon_vram_location(rdev, &rdev->mc, base);
+	if (!(rdev->flags & RADEON_IS_AGP))
+		radeon_gtt_location(rdev, &rdev->mc);
 }
 
 
@@ -2033,7 +2081,7 @@ void r100_pll_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v)
 	r100_pll_errata_after_data(rdev);
 }
 
-int r100_init(struct radeon_device *rdev)
+void r100_set_safe_registers(struct radeon_device *rdev)
 {
 	if (ASIC_IS_RN50(rdev)) {
 		rdev->config.r100.reg_safe_bm = rn50_reg_safe_bm;
@@ -2042,9 +2090,8 @@ int r100_init(struct radeon_device *rdev)
 		rdev->config.r100.reg_safe_bm = r100_reg_safe_bm;
 		rdev->config.r100.reg_safe_bm_size = ARRAY_SIZE(r100_reg_safe_bm);
 	} else {
-		return r200_init(rdev);
+		r200_set_safe_registers(rdev);
 	}
-	return 0;
 }
 
 /*
@@ -2342,9 +2389,11 @@ void r100_bandwidth_update(struct radeon_device *rdev)
 		mode1 = &rdev->mode_info.crtcs[0]->base.mode;
 		pixel_bytes1 = rdev->mode_info.crtcs[0]->base.fb->bits_per_pixel / 8;
 	}
-	if (rdev->mode_info.crtcs[1]->base.enabled) {
-		mode2 = &rdev->mode_info.crtcs[1]->base.mode;
-		pixel_bytes2 = rdev->mode_info.crtcs[1]->base.fb->bits_per_pixel / 8;
+	if (!(rdev->flags & RADEON_SINGLE_CRTC)) {
+		if (rdev->mode_info.crtcs[1]->base.enabled) {
+			mode2 = &rdev->mode_info.crtcs[1]->base.mode;
+			pixel_bytes2 = rdev->mode_info.crtcs[1]->base.fb->bits_per_pixel / 8;
+		}
 	}
 
 	min_mem_eff.full = rfixed_const_8(0);
@@ -2555,7 +2604,7 @@ void r100_bandwidth_update(struct radeon_device *rdev)
 	/*
 	  Find the total latency for the display data.
 	*/
-	disp_latency_overhead.full = rfixed_const(80);
+	disp_latency_overhead.full = rfixed_const(8);
 	disp_latency_overhead.full = rfixed_div(disp_latency_overhead, sclk_ff);
 	mc_latency_mclk.full += disp_latency_overhead.full + cur_latency_mclk.full;
 	mc_latency_sclk.full += disp_latency_overhead.full + cur_latency_sclk.full;
@@ -2753,21 +2802,25 @@ void r100_bandwidth_update(struct radeon_device *rdev)
 static inline void r100_cs_track_texture_print(struct r100_cs_track_texture *t)
 {
 	DRM_ERROR("pitch                      %d\n", t->pitch);
+	DRM_ERROR("use_pitch                  %d\n", t->use_pitch);
 	DRM_ERROR("width                      %d\n", t->width);
+	DRM_ERROR("width_11                   %d\n", t->width_11);
 	DRM_ERROR("height                     %d\n", t->height);
+	DRM_ERROR("height_11                  %d\n", t->height_11);
 	DRM_ERROR("num levels                 %d\n", t->num_levels);
 	DRM_ERROR("depth                      %d\n", t->txdepth);
 	DRM_ERROR("bpp                        %d\n", t->cpp);
 	DRM_ERROR("coordinate type            %d\n", t->tex_coord_type);
 	DRM_ERROR("width round to power of 2  %d\n", t->roundup_w);
 	DRM_ERROR("height round to power of 2 %d\n", t->roundup_h);
+	DRM_ERROR("compress format            %d\n", t->compress_format);
 }
 
 static int r100_cs_track_cube(struct radeon_device *rdev,
 			      struct r100_cs_track *track, unsigned idx)
 {
 	unsigned face, w, h;
-	struct radeon_object *cube_robj;
+	struct radeon_bo *cube_robj;
 	unsigned long size;
 
 	for (face = 0; face < 5; face++) {
@@ -2780,9 +2833,9 @@ static int r100_cs_track_cube(struct radeon_device *rdev,
 
 		size += track->textures[idx].cube_info[face].offset;
 
-		if (size > radeon_object_size(cube_robj)) {
+		if (size > radeon_bo_size(cube_robj)) {
 			DRM_ERROR("Cube texture offset greater than object size %lu %lu\n",
-				  size, radeon_object_size(cube_robj));
+				  size, radeon_bo_size(cube_robj));
 			r100_cs_track_texture_print(&track->textures[idx]);
 			return -1;
 		}
@@ -2790,10 +2843,40 @@ static int r100_cs_track_cube(struct radeon_device *rdev,
 	return 0;
 }
 
+static int r100_track_compress_size(int compress_format, int w, int h)
+{
+	int block_width, block_height, block_bytes;
+	int wblocks, hblocks;
+	int min_wblocks;
+	int sz;
+
+	block_width = 4;
+	block_height = 4;
+
+	switch (compress_format) {
+	case R100_TRACK_COMP_DXT1:
+		block_bytes = 8;
+		min_wblocks = 4;
+		break;
+	default:
+	case R100_TRACK_COMP_DXT35:
+		block_bytes = 16;
+		min_wblocks = 2;
+		break;
+	}
+
+	hblocks = (h + block_height - 1) / block_height;
+	wblocks = (w + block_width - 1) / block_width;
+	if (wblocks < min_wblocks)
+		wblocks = min_wblocks;
+	sz = wblocks * hblocks * block_bytes;
+	return sz;
+}
+
 static int r100_cs_track_texture_check(struct radeon_device *rdev,
 				       struct r100_cs_track *track)
 {
-	struct radeon_object *robj;
+	struct radeon_bo *robj;
 	unsigned long size;
 	unsigned u, i, w, h;
 	int ret;
@@ -2814,20 +2897,28 @@ static int r100_cs_track_texture_check(struct radeon_device *rdev,
 				else
 					w = track->textures[u].pitch / (1 << i);
 			} else {
-				w = track->textures[u].width / (1 << i);
+				w = track->textures[u].width;
 				if (rdev->family >= CHIP_RV515)
 					w |= track->textures[u].width_11;
+				w = w / (1 << i);
 				if (track->textures[u].roundup_w)
 					w = roundup_pow_of_two(w);
 			}
-			h = track->textures[u].height / (1 << i);
+			h = track->textures[u].height;
 			if (rdev->family >= CHIP_RV515)
 				h |= track->textures[u].height_11;
+			h = h / (1 << i);
 			if (track->textures[u].roundup_h)
 				h = roundup_pow_of_two(h);
-			size += w * h;
+			if (track->textures[u].compress_format) {
+
+				size += r100_track_compress_size(track->textures[u].compress_format, w, h);
+				/* compressed textures are block based */
+			} else
+				size += w * h;
 		}
 		size *= track->textures[u].cpp;
+
 		switch (track->textures[u].tex_coord_type) {
 		case 0:
 			break;
@@ -2847,9 +2938,9 @@ static int r100_cs_track_texture_check(struct radeon_device *rdev,
 				  "%u\n", track->textures[u].tex_coord_type, u);
 			return -EINVAL;
 		}
-		if (size > radeon_object_size(robj)) {
+		if (size > radeon_bo_size(robj)) {
 			DRM_ERROR("Texture of unit %u needs %lu bytes but is "
-				  "%lu\n", u, size, radeon_object_size(robj));
+				  "%lu\n", u, size, radeon_bo_size(robj));
 			r100_cs_track_texture_print(&track->textures[u]);
 			return -EINVAL;
 		}
@@ -2866,15 +2957,19 @@ int r100_cs_track_check(struct radeon_device *rdev, struct r100_cs_track *track)
 
 	for (i = 0; i < track->num_cb; i++) {
 		if (track->cb[i].robj == NULL) {
+			if (!(track->fastfill || track->color_channel_mask ||
+			      track->blend_read_enable)) {
+				continue;
+			}
 			DRM_ERROR("[drm] No buffer for color buffer %d !\n", i);
 			return -EINVAL;
 		}
 		size = track->cb[i].pitch * track->cb[i].cpp * track->maxy;
 		size += track->cb[i].offset;
-		if (size > radeon_object_size(track->cb[i].robj)) {
+		if (size > radeon_bo_size(track->cb[i].robj)) {
 			DRM_ERROR("[drm] Buffer too small for color buffer %d "
 				  "(need %lu have %lu) !\n", i, size,
-				  radeon_object_size(track->cb[i].robj));
+				  radeon_bo_size(track->cb[i].robj));
 			DRM_ERROR("[drm] color buffer %d (%u %u %u %u)\n",
 				  i, track->cb[i].pitch, track->cb[i].cpp,
 				  track->cb[i].offset, track->maxy);
@@ -2888,10 +2983,10 @@ int r100_cs_track_check(struct radeon_device *rdev, struct r100_cs_track *track)
 		}
 		size = track->zb.pitch * track->zb.cpp * track->maxy;
 		size += track->zb.offset;
-		if (size > radeon_object_size(track->zb.robj)) {
+		if (size > radeon_bo_size(track->zb.robj)) {
 			DRM_ERROR("[drm] Buffer too small for z buffer "
 				  "(need %lu have %lu) !\n", size,
-				  radeon_object_size(track->zb.robj));
+				  radeon_bo_size(track->zb.robj));
 			DRM_ERROR("[drm] zbuffer (%u %u %u %u)\n",
 				  track->zb.pitch, track->zb.cpp,
 				  track->zb.offset, track->maxy);
@@ -2909,11 +3004,12 @@ int r100_cs_track_check(struct radeon_device *rdev, struct r100_cs_track *track)
 					  "bound\n", prim_walk, i);
 				return -EINVAL;
 			}
-			if (size > radeon_object_size(track->arrays[i].robj)) {
-				DRM_ERROR("(PW %u) Vertex array %u need %lu dwords "
-					   "have %lu dwords\n", prim_walk, i,
-					   size >> 2,
-					   radeon_object_size(track->arrays[i].robj) >> 2);
+			if (size > radeon_bo_size(track->arrays[i].robj)) {
+				dev_err(rdev->dev, "(PW %u) Vertex array %u "
+					"need %lu dwords have %lu dwords\n",
+					prim_walk, i, size >> 2,
+					radeon_bo_size(track->arrays[i].robj)
+					>> 2);
 				DRM_ERROR("Max indices %u\n", track->max_indx);
 				return -EINVAL;
 			}
@@ -2927,10 +3023,12 @@ int r100_cs_track_check(struct radeon_device *rdev, struct r100_cs_track *track)
 					  "bound\n", prim_walk, i);
 				return -EINVAL;
 			}
-			if (size > radeon_object_size(track->arrays[i].robj)) {
-				DRM_ERROR("(PW %u) Vertex array %u need %lu dwords "
-					   "have %lu dwords\n", prim_walk, i, size >> 2,
-					   radeon_object_size(track->arrays[i].robj) >> 2);
+			if (size > radeon_bo_size(track->arrays[i].robj)) {
+				dev_err(rdev->dev, "(PW %u) Vertex array %u "
+					"need %lu dwords have %lu dwords\n",
+					prim_walk, i, size >> 2,
+					radeon_bo_size(track->arrays[i].robj)
+					>> 2);
 				return -EINVAL;
 			}
 		}
@@ -2992,6 +3090,7 @@ void r100_cs_track_clear(struct radeon_device *rdev, struct r100_cs_track *track
 		track->arrays[i].esize = 0x7F;
 	}
 	for (i = 0; i < track->num_texture; i++) {
+		track->textures[i].compress_format = R100_TRACK_COMP_NONE;
 		track->textures[i].pitch = 16536;
 		track->textures[i].width = 16536;
 		track->textures[i].height = 16536;
@@ -3157,7 +3256,7 @@ void r100_mc_stop(struct radeon_device *rdev, struct r100_mc_save *save)
 	WREG32(R_000740_CP_CSQ_CNTL, 0);
 
 	/* Save few CRTC registers */
-	save->GENMO_WT = RREG32(R_0003C0_GENMO_WT);
+	save->GENMO_WT = RREG8(R_0003C2_GENMO_WT);
 	save->CRTC_EXT_CNTL = RREG32(R_000054_CRTC_EXT_CNTL);
 	save->CRTC_GEN_CNTL = RREG32(R_000050_CRTC_GEN_CNTL);
 	save->CUR_OFFSET = RREG32(R_000260_CUR_OFFSET);
@@ -3167,7 +3266,7 @@ void r100_mc_stop(struct radeon_device *rdev, struct r100_mc_save *save)
 	}
 
 	/* Disable VGA aperture access */
-	WREG32(R_0003C0_GENMO_WT, C_0003C0_VGA_RAM_EN & save->GENMO_WT);
+	WREG8(R_0003C2_GENMO_WT, C_0003C2_VGA_RAM_EN & save->GENMO_WT);
 	/* Disable cursor, overlay, crtc */
 	WREG32(R_000260_CUR_OFFSET, save->CUR_OFFSET | S_000260_CUR_LOCK(1));
 	WREG32(R_000054_CRTC_EXT_CNTL, save->CRTC_EXT_CNTL |
@@ -3193,16 +3292,249 @@ void r100_mc_stop(struct radeon_device *rdev, struct r100_mc_save *save)
 void r100_mc_resume(struct radeon_device *rdev, struct r100_mc_save *save)
 {
 	/* Update base address for crtc */
-	WREG32(R_00023C_DISPLAY_BASE_ADDR, rdev->mc.vram_location);
+	WREG32(R_00023C_DISPLAY_BASE_ADDR, rdev->mc.vram_start);
 	if (!(rdev->flags & RADEON_SINGLE_CRTC)) {
-		WREG32(R_00033C_CRTC2_DISPLAY_BASE_ADDR,
-				rdev->mc.vram_location);
+		WREG32(R_00033C_CRTC2_DISPLAY_BASE_ADDR, rdev->mc.vram_start);
 	}
 	/* Restore CRTC registers */
-	WREG32(R_0003C0_GENMO_WT, save->GENMO_WT);
+	WREG8(R_0003C2_GENMO_WT, save->GENMO_WT);
 	WREG32(R_000054_CRTC_EXT_CNTL, save->CRTC_EXT_CNTL);
 	WREG32(R_000050_CRTC_GEN_CNTL, save->CRTC_GEN_CNTL);
 	if (!(rdev->flags & RADEON_SINGLE_CRTC)) {
 		WREG32(R_0003F8_CRTC2_GEN_CNTL, save->CRTC2_GEN_CNTL);
 	}
 }
+
+void r100_vga_render_disable(struct radeon_device *rdev)
+{
+	u32 tmp;
+
+	tmp = RREG8(R_0003C2_GENMO_WT);
+	WREG8(R_0003C2_GENMO_WT, C_0003C2_VGA_RAM_EN & tmp);
+}
+
+static void r100_debugfs(struct radeon_device *rdev)
+{
+	int r;
+
+	r = r100_debugfs_mc_info_init(rdev);
+	if (r)
+		dev_warn(rdev->dev, "Failed to create r100_mc debugfs file.\n");
+}
+
+static void r100_mc_program(struct radeon_device *rdev)
+{
+	struct r100_mc_save save;
+
+	/* Stops all mc clients */
+	r100_mc_stop(rdev, &save);
+	if (rdev->flags & RADEON_IS_AGP) {
+		WREG32(R_00014C_MC_AGP_LOCATION,
+			S_00014C_MC_AGP_START(rdev->mc.gtt_start >> 16) |
+			S_00014C_MC_AGP_TOP(rdev->mc.gtt_end >> 16));
+		WREG32(R_000170_AGP_BASE, lower_32_bits(rdev->mc.agp_base));
+		if (rdev->family > CHIP_RV200)
+			WREG32(R_00015C_AGP_BASE_2,
+				upper_32_bits(rdev->mc.agp_base) & 0xff);
+	} else {
+		WREG32(R_00014C_MC_AGP_LOCATION, 0x0FFFFFFF);
+		WREG32(R_000170_AGP_BASE, 0);
+		if (rdev->family > CHIP_RV200)
+			WREG32(R_00015C_AGP_BASE_2, 0);
+	}
+	/* Wait for mc idle */
+	if (r100_mc_wait_for_idle(rdev))
+		dev_warn(rdev->dev, "Wait for MC idle timeout.\n");
+	/* Program MC, should be a 32bits limited address space */
+	WREG32(R_000148_MC_FB_LOCATION,
+		S_000148_MC_FB_START(rdev->mc.vram_start >> 16) |
+		S_000148_MC_FB_TOP(rdev->mc.vram_end >> 16));
+	r100_mc_resume(rdev, &save);
+}
+
+void r100_clock_startup(struct radeon_device *rdev)
+{
+	u32 tmp;
+
+	if (radeon_dynclks != -1 && radeon_dynclks)
+		radeon_legacy_set_clock_gating(rdev, 1);
+	/* We need to force on some of the block */
+	tmp = RREG32_PLL(R_00000D_SCLK_CNTL);
+	tmp |= S_00000D_FORCE_CP(1) | S_00000D_FORCE_VIP(1);
+	if ((rdev->family == CHIP_RV250) || (rdev->family == CHIP_RV280))
+		tmp |= S_00000D_FORCE_DISP1(1) | S_00000D_FORCE_DISP2(1);
+	WREG32_PLL(R_00000D_SCLK_CNTL, tmp);
+}
+
+static int r100_startup(struct radeon_device *rdev)
+{
+	int r;
+
+	/* set common regs */
+	r100_set_common_regs(rdev);
+	/* program mc */
+	r100_mc_program(rdev);
+	/* Resume clock */
+	r100_clock_startup(rdev);
+	/* Initialize GPU configuration (# pipes, ...) */
+	r100_gpu_init(rdev);
+	/* Initialize GART (initialize after TTM so we can allocate
+	 * memory through TTM but finalize after TTM) */
+	r100_enable_bm(rdev);
+	if (rdev->flags & RADEON_IS_PCI) {
+		r = r100_pci_gart_enable(rdev);
+		if (r)
+			return r;
+	}
+	/* Enable IRQ */
+	r100_irq_set(rdev);
+	rdev->config.r100.hdp_cntl = RREG32(RADEON_HOST_PATH_CNTL);
+	/* 1M ring buffer */
+	r = r100_cp_init(rdev, 1024 * 1024);
+	if (r) {
+		dev_err(rdev->dev, "failled initializing CP (%d).\n", r);
+		return r;
+	}
+	r = r100_wb_init(rdev);
+	if (r)
+		dev_err(rdev->dev, "failled initializing WB (%d).\n", r);
+	r = r100_ib_init(rdev);
+	if (r) {
+		dev_err(rdev->dev, "failled initializing IB (%d).\n", r);
+		return r;
+	}
+	return 0;
+}
+
+int r100_resume(struct radeon_device *rdev)
+{
+	/* Make sur GART are not working */
+	if (rdev->flags & RADEON_IS_PCI)
+		r100_pci_gart_disable(rdev);
+	/* Resume clock before doing reset */
+	r100_clock_startup(rdev);
+	/* Reset gpu before posting otherwise ATOM will enter infinite loop */
+	if (radeon_gpu_reset(rdev)) {
+		dev_warn(rdev->dev, "GPU reset failed ! (0xE40=0x%08X, 0x7C0=0x%08X)\n",
+			RREG32(R_000E40_RBBM_STATUS),
+			RREG32(R_0007C0_CP_STAT));
+	}
+	/* post */
+	radeon_combios_asic_init(rdev->ddev);
+	/* Resume clock after posting */
+	r100_clock_startup(rdev);
+	/* Initialize surface registers */
+	radeon_surface_init(rdev);
+	return r100_startup(rdev);
+}
+
+int r100_suspend(struct radeon_device *rdev)
+{
+	r100_cp_disable(rdev);
+	r100_wb_disable(rdev);
+	r100_irq_disable(rdev);
+	if (rdev->flags & RADEON_IS_PCI)
+		r100_pci_gart_disable(rdev);
+	return 0;
+}
+
+void r100_fini(struct radeon_device *rdev)
+{
+	r100_cp_fini(rdev);
+	r100_wb_fini(rdev);
+	r100_ib_fini(rdev);
+	radeon_gem_fini(rdev);
+	if (rdev->flags & RADEON_IS_PCI)
+		r100_pci_gart_fini(rdev);
+	radeon_agp_fini(rdev);
+	radeon_irq_kms_fini(rdev);
+	radeon_fence_driver_fini(rdev);
+	radeon_bo_fini(rdev);
+	radeon_atombios_fini(rdev);
+	kfree(rdev->bios);
+	rdev->bios = NULL;
+}
+
+int r100_init(struct radeon_device *rdev)
+{
+	int r;
+
+	/* Register debugfs file specific to this group of asics */
+	r100_debugfs(rdev);
+	/* Disable VGA */
+	r100_vga_render_disable(rdev);
+	/* Initialize scratch registers */
+	radeon_scratch_init(rdev);
+	/* Initialize surface registers */
+	radeon_surface_init(rdev);
+	/* TODO: disable VGA need to use VGA request */
+	/* BIOS*/
+	if (!radeon_get_bios(rdev)) {
+		if (ASIC_IS_AVIVO(rdev))
+			return -EINVAL;
+	}
+	if (rdev->is_atom_bios) {
+		dev_err(rdev->dev, "Expecting combios for RS400/RS480 GPU\n");
+		return -EINVAL;
+	} else {
+		r = radeon_combios_init(rdev);
+		if (r)
+			return r;
+	}
+	/* Reset gpu before posting otherwise ATOM will enter infinite loop */
+	if (radeon_gpu_reset(rdev)) {
+		dev_warn(rdev->dev,
+			"GPU reset failed ! (0xE40=0x%08X, 0x7C0=0x%08X)\n",
+			RREG32(R_000E40_RBBM_STATUS),
+			RREG32(R_0007C0_CP_STAT));
+	}
+	/* check if cards are posted or not */
+	if (radeon_boot_test_post_card(rdev) == false)
+		return -EINVAL;
+	/* Set asic errata */
+	r100_errata(rdev);
+	/* Initialize clocks */
+	radeon_get_clock_info(rdev->ddev);
+	/* Initialize power management */
+	radeon_pm_init(rdev);
+	/* initialize AGP */
+	if (rdev->flags & RADEON_IS_AGP) {
+		r = radeon_agp_init(rdev);
+		if (r) {
+			radeon_agp_disable(rdev);
+		}
+	}
+	/* initialize VRAM */
+	r100_mc_init(rdev);
+	/* Fence driver */
+	r = radeon_fence_driver_init(rdev);
+	if (r)
+		return r;
+	r = radeon_irq_kms_init(rdev);
+	if (r)
+		return r;
+	/* Memory manager */
+	r = radeon_bo_init(rdev);
+	if (r)
+		return r;
+	if (rdev->flags & RADEON_IS_PCI) {
+		r = r100_pci_gart_init(rdev);
+		if (r)
+			return r;
+	}
+	r100_set_safe_registers(rdev);
+	rdev->accel_working = true;
+	r = r100_startup(rdev);
+	if (r) {
+		/* Somethings want wront with the accel init stop accel */
+		dev_err(rdev->dev, "Disabling GPU acceleration\n");
+		r100_cp_fini(rdev);
+		r100_wb_fini(rdev);
+		r100_ib_fini(rdev);
+		radeon_irq_kms_fini(rdev);
+		if (rdev->flags & RADEON_IS_PCI)
+			r100_pci_gart_fini(rdev);
+		rdev->accel_working = false;
+	}
+	return 0;
+}
diff --git a/drivers/gpu/drm/radeon/r100_track.h b/drivers/gpu/drm/radeon/r100_track.h
index 70a82eda394a..b27a6999d219 100644
--- a/drivers/gpu/drm/radeon/r100_track.h
+++ b/drivers/gpu/drm/radeon/r100_track.h
@@ -10,26 +10,30 @@
  * CS functions
  */
 struct r100_cs_track_cb {
-	struct radeon_object	*robj;
+	struct radeon_bo	*robj;
 	unsigned		pitch;
 	unsigned		cpp;
 	unsigned		offset;
 };
 
 struct r100_cs_track_array {
-	struct radeon_object	*robj;
+	struct radeon_bo	*robj;
 	unsigned		esize;
 };
 
 struct r100_cs_cube_info {
-	struct radeon_object	*robj;
-	unsigned                offset;
+	struct radeon_bo	*robj;
+	unsigned		offset;
 	unsigned		width;
 	unsigned		height;
 };
 
+#define R100_TRACK_COMP_NONE   0
+#define R100_TRACK_COMP_DXT1   1
+#define R100_TRACK_COMP_DXT35  2
+
 struct r100_cs_track_texture {
-	struct radeon_object	*robj;
+	struct radeon_bo	*robj;
 	struct r100_cs_cube_info cube_info[5]; /* info for 5 non-primary faces */
 	unsigned		pitch;
 	unsigned		width;
@@ -44,6 +48,7 @@ struct r100_cs_track_texture {
 	bool			enabled;
 	bool			roundup_w;
 	bool			roundup_h;
+	unsigned                compress_format;
 };
 
 struct r100_cs_track_limits {
@@ -62,13 +67,15 @@ struct r100_cs_track {
 	unsigned			immd_dwords;
 	unsigned			num_arrays;
 	unsigned			max_indx;
+	unsigned			color_channel_mask;
 	struct r100_cs_track_array	arrays[11];
 	struct r100_cs_track_cb 	cb[R300_MAX_CB];
 	struct r100_cs_track_cb 	zb;
 	struct r100_cs_track_texture	textures[R300_TRACK_MAX_TEXTURE];
 	bool				z_enabled;
 	bool                            separate_cube;
-
+	bool				fastfill;
+	bool				blend_read_enable;
 };
 
 int r100_cs_track_check(struct radeon_device *rdev, struct r100_cs_track *track);
@@ -84,6 +91,8 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 		       struct radeon_cs_packet *pkt,
 		       unsigned idx, unsigned reg);
 
+
+
 static inline int r100_reloc_pitch_offset(struct radeon_cs_parser *p,
 					  struct radeon_cs_packet *pkt,
 					  unsigned idx,
@@ -93,9 +102,7 @@ static inline int r100_reloc_pitch_offset(struct radeon_cs_parser *p,
 	u32 tile_flags = 0;
 	u32 tmp;
 	struct radeon_cs_reloc *reloc;
-	struct radeon_cs_chunk *ib_chunk;
-
-	ib_chunk = &p->chunks[p->chunk_ib_idx];
+	u32 value;
 
 	r = r100_cs_packet_next_reloc(p, &reloc);
 	if (r) {
@@ -104,7 +111,8 @@ static inline int r100_reloc_pitch_offset(struct radeon_cs_parser *p,
 		r100_cs_dump_packet(p, pkt);
 		return r;
 	}
-	tmp = ib_chunk->kdata[idx] & 0x003fffff;
+	value = radeon_get_ib_value(p, idx);
+	tmp = value & 0x003fffff;
 	tmp += (((u32)reloc->lobj.gpu_offset) >> 10);
 
 	if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO)
@@ -119,6 +127,64 @@ static inline int r100_reloc_pitch_offset(struct radeon_cs_parser *p,
 	}
 
 	tmp |= tile_flags;
-	p->ib->ptr[idx] = (ib_chunk->kdata[idx] & 0x3fc00000) | tmp;
+	p->ib->ptr[idx] = (value & 0x3fc00000) | tmp;
 	return 0;
 }
+
+static inline int r100_packet3_load_vbpntr(struct radeon_cs_parser *p,
+					   struct radeon_cs_packet *pkt,
+					   int idx)
+{
+	unsigned c, i;
+	struct radeon_cs_reloc *reloc;
+	struct r100_cs_track *track;
+	int r = 0;
+	volatile uint32_t *ib;
+	u32 idx_value;
+
+	ib = p->ib->ptr;
+	track = (struct r100_cs_track *)p->track;
+	c = radeon_get_ib_value(p, idx++) & 0x1F;
+	track->num_arrays = c;
+	for (i = 0; i < (c - 1); i+=2, idx+=3) {
+		r = r100_cs_packet_next_reloc(p, &reloc);
+		if (r) {
+			DRM_ERROR("No reloc for packet3 %d\n",
+				  pkt->opcode);
+			r100_cs_dump_packet(p, pkt);
+			return r;
+		}
+		idx_value = radeon_get_ib_value(p, idx);
+		ib[idx+1] = radeon_get_ib_value(p, idx + 1) + ((u32)reloc->lobj.gpu_offset);
+
+		track->arrays[i + 0].esize = idx_value >> 8;
+		track->arrays[i + 0].robj = reloc->robj;
+		track->arrays[i + 0].esize &= 0x7F;
+		r = r100_cs_packet_next_reloc(p, &reloc);
+		if (r) {
+			DRM_ERROR("No reloc for packet3 %d\n",
+				  pkt->opcode);
+			r100_cs_dump_packet(p, pkt);
+			return r;
+		}
+		ib[idx+2] = radeon_get_ib_value(p, idx + 2) + ((u32)reloc->lobj.gpu_offset);
+		track->arrays[i + 1].robj = reloc->robj;
+		track->arrays[i + 1].esize = idx_value >> 24;
+		track->arrays[i + 1].esize &= 0x7F;
+	}
+	if (c & 1) {
+		r = r100_cs_packet_next_reloc(p, &reloc);
+		if (r) {
+			DRM_ERROR("No reloc for packet3 %d\n",
+					  pkt->opcode);
+			r100_cs_dump_packet(p, pkt);
+			return r;
+		}
+		idx_value = radeon_get_ib_value(p, idx);
+		ib[idx+1] = radeon_get_ib_value(p, idx + 1) + ((u32)reloc->lobj.gpu_offset);
+		track->arrays[i + 0].robj = reloc->robj;
+		track->arrays[i + 0].esize = idx_value >> 8;
+		track->arrays[i + 0].esize &= 0x7F;
+	}
+	return r;
+}
diff --git a/drivers/gpu/drm/radeon/r100d.h b/drivers/gpu/drm/radeon/r100d.h
index c4b257ec920e..df29a630c466 100644
--- a/drivers/gpu/drm/radeon/r100d.h
+++ b/drivers/gpu/drm/radeon/r100d.h
@@ -381,6 +381,24 @@
 #define   S_000054_VCRTC_IDX_MASTER(x)                 (((x) & 0x7F) << 24)
 #define   G_000054_VCRTC_IDX_MASTER(x)                 (((x) >> 24) & 0x7F)
 #define   C_000054_VCRTC_IDX_MASTER                    0x80FFFFFF
+#define R_000148_MC_FB_LOCATION                      0x000148
+#define   S_000148_MC_FB_START(x)                      (((x) & 0xFFFF) << 0)
+#define   G_000148_MC_FB_START(x)                      (((x) >> 0) & 0xFFFF)
+#define   C_000148_MC_FB_START                         0xFFFF0000
+#define   S_000148_MC_FB_TOP(x)                        (((x) & 0xFFFF) << 16)
+#define   G_000148_MC_FB_TOP(x)                        (((x) >> 16) & 0xFFFF)
+#define   C_000148_MC_FB_TOP                           0x0000FFFF
+#define R_00014C_MC_AGP_LOCATION                     0x00014C
+#define   S_00014C_MC_AGP_START(x)                     (((x) & 0xFFFF) << 0)
+#define   G_00014C_MC_AGP_START(x)                     (((x) >> 0) & 0xFFFF)
+#define   C_00014C_MC_AGP_START                        0xFFFF0000
+#define   S_00014C_MC_AGP_TOP(x)                       (((x) & 0xFFFF) << 16)
+#define   G_00014C_MC_AGP_TOP(x)                       (((x) >> 16) & 0xFFFF)
+#define   C_00014C_MC_AGP_TOP                          0x0000FFFF
+#define R_000170_AGP_BASE                            0x000170
+#define   S_000170_AGP_BASE_ADDR(x)                    (((x) & 0xFFFFFFFF) << 0)
+#define   G_000170_AGP_BASE_ADDR(x)                    (((x) >> 0) & 0xFFFFFFFF)
+#define   C_000170_AGP_BASE_ADDR                       0x00000000
 #define R_00023C_DISPLAY_BASE_ADDR                   0x00023C
 #define   S_00023C_DISPLAY_BASE_ADDR(x)                (((x) & 0xFFFFFFFF) << 0)
 #define   G_00023C_DISPLAY_BASE_ADDR(x)                (((x) >> 0) & 0xFFFFFFFF)
@@ -403,25 +421,25 @@
 #define   S_000360_CUR2_LOCK(x)                        (((x) & 0x1) << 31)
 #define   G_000360_CUR2_LOCK(x)                        (((x) >> 31) & 0x1)
 #define   C_000360_CUR2_LOCK                           0x7FFFFFFF
-#define R_0003C0_GENMO_WT                            0x0003C0
-#define   S_0003C0_GENMO_MONO_ADDRESS_B(x)             (((x) & 0x1) << 0)
-#define   G_0003C0_GENMO_MONO_ADDRESS_B(x)             (((x) >> 0) & 0x1)
-#define   C_0003C0_GENMO_MONO_ADDRESS_B                0xFFFFFFFE
-#define   S_0003C0_VGA_RAM_EN(x)                       (((x) & 0x1) << 1)
-#define   G_0003C0_VGA_RAM_EN(x)                       (((x) >> 1) & 0x1)
-#define   C_0003C0_VGA_RAM_EN                          0xFFFFFFFD
-#define   S_0003C0_VGA_CKSEL(x)                        (((x) & 0x3) << 2)
-#define   G_0003C0_VGA_CKSEL(x)                        (((x) >> 2) & 0x3)
-#define   C_0003C0_VGA_CKSEL                           0xFFFFFFF3
-#define   S_0003C0_ODD_EVEN_MD_PGSEL(x)                (((x) & 0x1) << 5)
-#define   G_0003C0_ODD_EVEN_MD_PGSEL(x)                (((x) >> 5) & 0x1)
-#define   C_0003C0_ODD_EVEN_MD_PGSEL                   0xFFFFFFDF
-#define   S_0003C0_VGA_HSYNC_POL(x)                    (((x) & 0x1) << 6)
-#define   G_0003C0_VGA_HSYNC_POL(x)                    (((x) >> 6) & 0x1)
-#define   C_0003C0_VGA_HSYNC_POL                       0xFFFFFFBF
-#define   S_0003C0_VGA_VSYNC_POL(x)                    (((x) & 0x1) << 7)
-#define   G_0003C0_VGA_VSYNC_POL(x)                    (((x) >> 7) & 0x1)
-#define   C_0003C0_VGA_VSYNC_POL                       0xFFFFFF7F
+#define R_0003C2_GENMO_WT                            0x0003C0
+#define   S_0003C2_GENMO_MONO_ADDRESS_B(x)             (((x) & 0x1) << 0)
+#define   G_0003C2_GENMO_MONO_ADDRESS_B(x)             (((x) >> 0) & 0x1)
+#define   C_0003C2_GENMO_MONO_ADDRESS_B                0xFE
+#define   S_0003C2_VGA_RAM_EN(x)                       (((x) & 0x1) << 1)
+#define   G_0003C2_VGA_RAM_EN(x)                       (((x) >> 1) & 0x1)
+#define   C_0003C2_VGA_RAM_EN                          0xFD
+#define   S_0003C2_VGA_CKSEL(x)                        (((x) & 0x3) << 2)
+#define   G_0003C2_VGA_CKSEL(x)                        (((x) >> 2) & 0x3)
+#define   C_0003C2_VGA_CKSEL                           0xF3
+#define   S_0003C2_ODD_EVEN_MD_PGSEL(x)                (((x) & 0x1) << 5)
+#define   G_0003C2_ODD_EVEN_MD_PGSEL(x)                (((x) >> 5) & 0x1)
+#define   C_0003C2_ODD_EVEN_MD_PGSEL                   0xDF
+#define   S_0003C2_VGA_HSYNC_POL(x)                    (((x) & 0x1) << 6)
+#define   G_0003C2_VGA_HSYNC_POL(x)                    (((x) >> 6) & 0x1)
+#define   C_0003C2_VGA_HSYNC_POL                       0xBF
+#define   S_0003C2_VGA_VSYNC_POL(x)                    (((x) & 0x1) << 7)
+#define   G_0003C2_VGA_VSYNC_POL(x)                    (((x) >> 7) & 0x1)
+#define   C_0003C2_VGA_VSYNC_POL                       0x7F
 #define R_0003F8_CRTC2_GEN_CNTL                      0x0003F8
 #define   S_0003F8_CRTC2_DBL_SCAN_EN(x)                (((x) & 0x1) << 0)
 #define   G_0003F8_CRTC2_DBL_SCAN_EN(x)                (((x) >> 0) & 0x1)
@@ -545,6 +563,46 @@
 #define   S_000774_SCRATCH_ADDR(x)                     (((x) & 0x7FFFFFF) << 5)
 #define   G_000774_SCRATCH_ADDR(x)                     (((x) >> 5) & 0x7FFFFFF)
 #define   C_000774_SCRATCH_ADDR                        0x0000001F
+#define R_0007C0_CP_STAT                             0x0007C0
+#define   S_0007C0_MRU_BUSY(x)                         (((x) & 0x1) << 0)
+#define   G_0007C0_MRU_BUSY(x)                         (((x) >> 0) & 0x1)
+#define   C_0007C0_MRU_BUSY                            0xFFFFFFFE
+#define   S_0007C0_MWU_BUSY(x)                         (((x) & 0x1) << 1)
+#define   G_0007C0_MWU_BUSY(x)                         (((x) >> 1) & 0x1)
+#define   C_0007C0_MWU_BUSY                            0xFFFFFFFD
+#define   S_0007C0_RSIU_BUSY(x)                        (((x) & 0x1) << 2)
+#define   G_0007C0_RSIU_BUSY(x)                        (((x) >> 2) & 0x1)
+#define   C_0007C0_RSIU_BUSY                           0xFFFFFFFB
+#define   S_0007C0_RCIU_BUSY(x)                        (((x) & 0x1) << 3)
+#define   G_0007C0_RCIU_BUSY(x)                        (((x) >> 3) & 0x1)
+#define   C_0007C0_RCIU_BUSY                           0xFFFFFFF7
+#define   S_0007C0_CSF_PRIMARY_BUSY(x)                 (((x) & 0x1) << 9)
+#define   G_0007C0_CSF_PRIMARY_BUSY(x)                 (((x) >> 9) & 0x1)
+#define   C_0007C0_CSF_PRIMARY_BUSY                    0xFFFFFDFF
+#define   S_0007C0_CSF_INDIRECT_BUSY(x)                (((x) & 0x1) << 10)
+#define   G_0007C0_CSF_INDIRECT_BUSY(x)                (((x) >> 10) & 0x1)
+#define   C_0007C0_CSF_INDIRECT_BUSY                   0xFFFFFBFF
+#define   S_0007C0_CSQ_PRIMARY_BUSY(x)                 (((x) & 0x1) << 11)
+#define   G_0007C0_CSQ_PRIMARY_BUSY(x)                 (((x) >> 11) & 0x1)
+#define   C_0007C0_CSQ_PRIMARY_BUSY                    0xFFFFF7FF
+#define   S_0007C0_CSQ_INDIRECT_BUSY(x)                (((x) & 0x1) << 12)
+#define   G_0007C0_CSQ_INDIRECT_BUSY(x)                (((x) >> 12) & 0x1)
+#define   C_0007C0_CSQ_INDIRECT_BUSY                   0xFFFFEFFF
+#define   S_0007C0_CSI_BUSY(x)                         (((x) & 0x1) << 13)
+#define   G_0007C0_CSI_BUSY(x)                         (((x) >> 13) & 0x1)
+#define   C_0007C0_CSI_BUSY                            0xFFFFDFFF
+#define   S_0007C0_GUIDMA_BUSY(x)                      (((x) & 0x1) << 28)
+#define   G_0007C0_GUIDMA_BUSY(x)                      (((x) >> 28) & 0x1)
+#define   C_0007C0_GUIDMA_BUSY                         0xEFFFFFFF
+#define   S_0007C0_VIDDMA_BUSY(x)                      (((x) & 0x1) << 29)
+#define   G_0007C0_VIDDMA_BUSY(x)                      (((x) >> 29) & 0x1)
+#define   C_0007C0_VIDDMA_BUSY                         0xDFFFFFFF
+#define   S_0007C0_CMDSTRM_BUSY(x)                     (((x) & 0x1) << 30)
+#define   G_0007C0_CMDSTRM_BUSY(x)                     (((x) >> 30) & 0x1)
+#define   C_0007C0_CMDSTRM_BUSY                        0xBFFFFFFF
+#define   S_0007C0_CP_BUSY(x)                          (((x) & 0x1) << 31)
+#define   G_0007C0_CP_BUSY(x)                          (((x) >> 31) & 0x1)
+#define   C_0007C0_CP_BUSY                             0x7FFFFFFF
 #define R_000E40_RBBM_STATUS                         0x000E40
 #define   S_000E40_CMDFIFO_AVAIL(x)                    (((x) & 0x7F) << 0)
 #define   G_000E40_CMDFIFO_AVAIL(x)                    (((x) >> 0) & 0x7F)
@@ -604,4 +662,53 @@
 #define   G_000E40_GUI_ACTIVE(x)                       (((x) >> 31) & 0x1)
 #define   C_000E40_GUI_ACTIVE                          0x7FFFFFFF
 
+
+#define R_00000D_SCLK_CNTL                           0x00000D
+#define   S_00000D_SCLK_SRC_SEL(x)                     (((x) & 0x7) << 0)
+#define   G_00000D_SCLK_SRC_SEL(x)                     (((x) >> 0) & 0x7)
+#define   C_00000D_SCLK_SRC_SEL                        0xFFFFFFF8
+#define   S_00000D_TCLK_SRC_SEL(x)                     (((x) & 0x7) << 8)
+#define   G_00000D_TCLK_SRC_SEL(x)                     (((x) >> 8) & 0x7)
+#define   C_00000D_TCLK_SRC_SEL                        0xFFFFF8FF
+#define   S_00000D_FORCE_CP(x)                         (((x) & 0x1) << 16)
+#define   G_00000D_FORCE_CP(x)                         (((x) >> 16) & 0x1)
+#define   C_00000D_FORCE_CP                            0xFFFEFFFF
+#define   S_00000D_FORCE_HDP(x)                        (((x) & 0x1) << 17)
+#define   G_00000D_FORCE_HDP(x)                        (((x) >> 17) & 0x1)
+#define   C_00000D_FORCE_HDP                           0xFFFDFFFF
+#define   S_00000D_FORCE_DISP(x)                       (((x) & 0x1) << 18)
+#define   G_00000D_FORCE_DISP(x)                       (((x) >> 18) & 0x1)
+#define   C_00000D_FORCE_DISP                          0xFFFBFFFF
+#define   S_00000D_FORCE_TOP(x)                        (((x) & 0x1) << 19)
+#define   G_00000D_FORCE_TOP(x)                        (((x) >> 19) & 0x1)
+#define   C_00000D_FORCE_TOP                           0xFFF7FFFF
+#define   S_00000D_FORCE_E2(x)                         (((x) & 0x1) << 20)
+#define   G_00000D_FORCE_E2(x)                         (((x) >> 20) & 0x1)
+#define   C_00000D_FORCE_E2                            0xFFEFFFFF
+#define   S_00000D_FORCE_SE(x)                         (((x) & 0x1) << 21)
+#define   G_00000D_FORCE_SE(x)                         (((x) >> 21) & 0x1)
+#define   C_00000D_FORCE_SE                            0xFFDFFFFF
+#define   S_00000D_FORCE_IDCT(x)                       (((x) & 0x1) << 22)
+#define   G_00000D_FORCE_IDCT(x)                       (((x) >> 22) & 0x1)
+#define   C_00000D_FORCE_IDCT                          0xFFBFFFFF
+#define   S_00000D_FORCE_VIP(x)                        (((x) & 0x1) << 23)
+#define   G_00000D_FORCE_VIP(x)                        (((x) >> 23) & 0x1)
+#define   C_00000D_FORCE_VIP                           0xFF7FFFFF
+#define   S_00000D_FORCE_RE(x)                         (((x) & 0x1) << 24)
+#define   G_00000D_FORCE_RE(x)                         (((x) >> 24) & 0x1)
+#define   C_00000D_FORCE_RE                            0xFEFFFFFF
+#define   S_00000D_FORCE_PB(x)                         (((x) & 0x1) << 25)
+#define   G_00000D_FORCE_PB(x)                         (((x) >> 25) & 0x1)
+#define   C_00000D_FORCE_PB                            0xFDFFFFFF
+#define   S_00000D_FORCE_TAM(x)                        (((x) & 0x1) << 26)
+#define   G_00000D_FORCE_TAM(x)                        (((x) >> 26) & 0x1)
+#define   C_00000D_FORCE_TAM                           0xFBFFFFFF
+#define   S_00000D_FORCE_TDM(x)                        (((x) & 0x1) << 27)
+#define   G_00000D_FORCE_TDM(x)                        (((x) >> 27) & 0x1)
+#define   C_00000D_FORCE_TDM                           0xF7FFFFFF
+#define   S_00000D_FORCE_RB(x)                         (((x) & 0x1) << 28)
+#define   G_00000D_FORCE_RB(x)                         (((x) >> 28) & 0x1)
+#define   C_00000D_FORCE_RB                            0xEFFFFFFF
+
+
 #endif
diff --git a/drivers/gpu/drm/radeon/r200.c b/drivers/gpu/drm/radeon/r200.c
index 568c74bfba3d..1146c9909c2c 100644
--- a/drivers/gpu/drm/radeon/r200.c
+++ b/drivers/gpu/drm/radeon/r200.c
@@ -31,6 +31,7 @@
 #include "radeon_reg.h"
 #include "radeon.h"
 
+#include "r100d.h"
 #include "r200_reg_safe.h"
 
 #include "r100_track.h"
@@ -79,6 +80,51 @@ static int r200_get_vtx_size_0(uint32_t vtx_fmt_0)
 	return vtx_size;
 }
 
+int r200_copy_dma(struct radeon_device *rdev,
+		  uint64_t src_offset,
+		  uint64_t dst_offset,
+		  unsigned num_pages,
+		  struct radeon_fence *fence)
+{
+	uint32_t size;
+	uint32_t cur_size;
+	int i, num_loops;
+	int r = 0;
+
+	/* radeon pitch is /64 */
+	size = num_pages << PAGE_SHIFT;
+	num_loops = DIV_ROUND_UP(size, 0x1FFFFF);
+	r = radeon_ring_lock(rdev, num_loops * 4 + 64);
+	if (r) {
+		DRM_ERROR("radeon: moving bo (%d).\n", r);
+		return r;
+	}
+	/* Must wait for 2D idle & clean before DMA or hangs might happen */
+	radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0));
+	radeon_ring_write(rdev, (1 << 16));
+	for (i = 0; i < num_loops; i++) {
+		cur_size = size;
+		if (cur_size > 0x1FFFFF) {
+			cur_size = 0x1FFFFF;
+		}
+		size -= cur_size;
+		radeon_ring_write(rdev, PACKET0(0x720, 2));
+		radeon_ring_write(rdev, src_offset);
+		radeon_ring_write(rdev, dst_offset);
+		radeon_ring_write(rdev, cur_size | (1 << 31) | (1 << 30));
+		src_offset += cur_size;
+		dst_offset += cur_size;
+	}
+	radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0));
+	radeon_ring_write(rdev, RADEON_WAIT_DMA_GUI_IDLE);
+	if (fence) {
+		r = radeon_fence_emit(rdev, fence);
+	}
+	radeon_ring_unlock_commit(rdev);
+	return r;
+}
+
+
 static int r200_get_vtx_size_1(uint32_t vtx_fmt_1)
 {
 	int vtx_size, i, tex_size;
@@ -96,7 +142,6 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 		       struct radeon_cs_packet *pkt,
 		       unsigned idx, unsigned reg)
 {
-	struct radeon_cs_chunk *ib_chunk;
 	struct radeon_cs_reloc *reloc;
 	struct r100_cs_track *track;
 	volatile uint32_t *ib;
@@ -105,11 +150,11 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 	int i;
 	int face;
 	u32 tile_flags = 0;
+	u32 idx_value;
 
 	ib = p->ib->ptr;
-	ib_chunk = &p->chunks[p->chunk_ib_idx];
 	track = (struct r100_cs_track *)p->track;
-
+	idx_value = radeon_get_ib_value(p, idx);
 	switch (reg) {
 	case RADEON_CRTC_GUI_TRIG_VLINE:
 		r = r100_cs_packet_parse_vline(p);
@@ -137,8 +182,8 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 			return r;
 		}
 		track->zb.robj = reloc->robj;
-		track->zb.offset = ib_chunk->kdata[idx];
-		ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset);
+		track->zb.offset = idx_value;
+		ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
 		break;
 	case RADEON_RB3D_COLOROFFSET:
 		r = r100_cs_packet_next_reloc(p, &reloc);
@@ -149,8 +194,8 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 			return r;
 		}
 		track->cb[0].robj = reloc->robj;
-		track->cb[0].offset = ib_chunk->kdata[idx];
-		ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset);
+		track->cb[0].offset = idx_value;
+		ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
 		break;
 	case R200_PP_TXOFFSET_0:
 	case R200_PP_TXOFFSET_1:
@@ -166,7 +211,7 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 			r100_cs_dump_packet(p, pkt);
 			return r;
 		}
-		ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset);
+		ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
 		track->textures[i].robj = reloc->robj;
 		break;
 	case R200_PP_CUBIC_OFFSET_F1_0:
@@ -208,12 +253,12 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 			r100_cs_dump_packet(p, pkt);
 			return r;
 		}
-		track->textures[i].cube_info[face - 1].offset = ib_chunk->kdata[idx];
-		ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset);
+		track->textures[i].cube_info[face - 1].offset = idx_value;
+		ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
 		track->textures[i].cube_info[face - 1].robj = reloc->robj;
 		break;
 	case RADEON_RE_WIDTH_HEIGHT:
-		track->maxy = ((ib_chunk->kdata[idx] >> 16) & 0x7FF);
+		track->maxy = ((idx_value >> 16) & 0x7FF);
 		break;
 	case RADEON_RB3D_COLORPITCH:
 		r = r100_cs_packet_next_reloc(p, &reloc);
@@ -229,17 +274,17 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 		if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO)
 			tile_flags |= RADEON_COLOR_MICROTILE_ENABLE;
 
-		tmp = ib_chunk->kdata[idx] & ~(0x7 << 16);
+		tmp = idx_value & ~(0x7 << 16);
 		tmp |= tile_flags;
 		ib[idx] = tmp;
 
-		track->cb[0].pitch = ib_chunk->kdata[idx] & RADEON_COLORPITCH_MASK;
+		track->cb[0].pitch = idx_value & RADEON_COLORPITCH_MASK;
 		break;
 	case RADEON_RB3D_DEPTHPITCH:
-		track->zb.pitch = ib_chunk->kdata[idx] & RADEON_DEPTHPITCH_MASK;
+		track->zb.pitch = idx_value & RADEON_DEPTHPITCH_MASK;
 		break;
 	case RADEON_RB3D_CNTL:
-		switch ((ib_chunk->kdata[idx] >> RADEON_RB3D_COLOR_FORMAT_SHIFT) & 0x1f) {
+		switch ((idx_value >> RADEON_RB3D_COLOR_FORMAT_SHIFT) & 0x1f) {
 		case 7:
 		case 8:
 		case 9:
@@ -257,18 +302,18 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 			break;
 		default:
 			DRM_ERROR("Invalid color buffer format (%d) !\n",
-				  ((ib_chunk->kdata[idx] >> RADEON_RB3D_COLOR_FORMAT_SHIFT) & 0x1f));
+				  ((idx_value >> RADEON_RB3D_COLOR_FORMAT_SHIFT) & 0x1f));
 			return -EINVAL;
 		}
-		if (ib_chunk->kdata[idx] & RADEON_DEPTHXY_OFFSET_ENABLE) {
+		if (idx_value & RADEON_DEPTHXY_OFFSET_ENABLE) {
 			DRM_ERROR("No support for depth xy offset in kms\n");
 			return -EINVAL;
 		}
 
-		track->z_enabled = !!(ib_chunk->kdata[idx] & RADEON_Z_ENABLE);
+		track->z_enabled = !!(idx_value & RADEON_Z_ENABLE);
 		break;
 	case RADEON_RB3D_ZSTENCILCNTL:
-		switch (ib_chunk->kdata[idx] & 0xf) {
+		switch (idx_value & 0xf) {
 		case 0:
 			track->zb.cpp = 2;
 			break;
@@ -292,27 +337,27 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 			r100_cs_dump_packet(p, pkt);
 			return r;
 		}
-		ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset);
+		ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
 		break;
 	case RADEON_PP_CNTL:
 		{
-			uint32_t temp = ib_chunk->kdata[idx] >> 4;
+			uint32_t temp = idx_value >> 4;
 			for (i = 0; i < track->num_texture; i++)
 				track->textures[i].enabled = !!(temp & (1 << i));
 		}
 		break;
 	case RADEON_SE_VF_CNTL:
-		track->vap_vf_cntl = ib_chunk->kdata[idx];
+		track->vap_vf_cntl = idx_value;
 		break;
 	case 0x210c:
 		/* VAP_VF_MAX_VTX_INDX */
-		track->max_indx = ib_chunk->kdata[idx] & 0x00FFFFFFUL;
+		track->max_indx = idx_value & 0x00FFFFFFUL;
 		break;
 	case R200_SE_VTX_FMT_0:
-		track->vtx_size = r200_get_vtx_size_0(ib_chunk->kdata[idx]);
+		track->vtx_size = r200_get_vtx_size_0(idx_value);
 		break;
 	case R200_SE_VTX_FMT_1:
-		track->vtx_size += r200_get_vtx_size_1(ib_chunk->kdata[idx]);
+		track->vtx_size += r200_get_vtx_size_1(idx_value);
 		break;
 	case R200_PP_TXSIZE_0:
 	case R200_PP_TXSIZE_1:
@@ -321,8 +366,8 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 	case R200_PP_TXSIZE_4:
 	case R200_PP_TXSIZE_5:
 		i = (reg - R200_PP_TXSIZE_0) / 32;
-		track->textures[i].width = (ib_chunk->kdata[idx] & RADEON_TEX_USIZE_MASK) + 1;
-		track->textures[i].height = ((ib_chunk->kdata[idx] & RADEON_TEX_VSIZE_MASK) >> RADEON_TEX_VSIZE_SHIFT) + 1;
+		track->textures[i].width = (idx_value & RADEON_TEX_USIZE_MASK) + 1;
+		track->textures[i].height = ((idx_value & RADEON_TEX_VSIZE_MASK) >> RADEON_TEX_VSIZE_SHIFT) + 1;
 		break;
 	case R200_PP_TXPITCH_0:
 	case R200_PP_TXPITCH_1:
@@ -331,7 +376,7 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 	case R200_PP_TXPITCH_4:
 	case R200_PP_TXPITCH_5:
 		i = (reg - R200_PP_TXPITCH_0) / 32;
-		track->textures[i].pitch = ib_chunk->kdata[idx] + 32;
+		track->textures[i].pitch = idx_value + 32;
 		break;
 	case R200_PP_TXFILTER_0:
 	case R200_PP_TXFILTER_1:
@@ -340,12 +385,12 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 	case R200_PP_TXFILTER_4:
 	case R200_PP_TXFILTER_5:
 		i = (reg - R200_PP_TXFILTER_0) / 32;
-		track->textures[i].num_levels = ((ib_chunk->kdata[idx] & R200_MAX_MIP_LEVEL_MASK)
+		track->textures[i].num_levels = ((idx_value & R200_MAX_MIP_LEVEL_MASK)
 						 >> R200_MAX_MIP_LEVEL_SHIFT);
-		tmp = (ib_chunk->kdata[idx] >> 23) & 0x7;
+		tmp = (idx_value >> 23) & 0x7;
 		if (tmp == 2 || tmp == 6)
 			track->textures[i].roundup_w = false;
-		tmp = (ib_chunk->kdata[idx] >> 27) & 0x7;
+		tmp = (idx_value >> 27) & 0x7;
 		if (tmp == 2 || tmp == 6)
 			track->textures[i].roundup_h = false;
 		break;
@@ -364,21 +409,24 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 	case R200_PP_TXFORMAT_X_4:
 	case R200_PP_TXFORMAT_X_5:
 		i = (reg - R200_PP_TXFORMAT_X_0) / 32;
-		track->textures[i].txdepth = ib_chunk->kdata[idx] & 0x7;
-		tmp = (ib_chunk->kdata[idx] >> 16) & 0x3;
+		track->textures[i].txdepth = idx_value & 0x7;
+		tmp = (idx_value >> 16) & 0x3;
 		/* 2D, 3D, CUBE */
 		switch (tmp) {
 		case 0:
 		case 5:
 		case 6:
 		case 7:
+			/* 1D/2D */
 			track->textures[i].tex_coord_type = 0;
 			break;
 		case 1:
-			track->textures[i].tex_coord_type = 1;
+			/* CUBE */
+			track->textures[i].tex_coord_type = 2;
 			break;
 		case 2:
-			track->textures[i].tex_coord_type = 2;
+			/* 3D */
+			track->textures[i].tex_coord_type = 1;
 			break;
 		}
 		break;
@@ -389,20 +437,19 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 	case R200_PP_TXFORMAT_4:
 	case R200_PP_TXFORMAT_5:
 		i = (reg - R200_PP_TXFORMAT_0) / 32;
-		if (ib_chunk->kdata[idx] & R200_TXFORMAT_NON_POWER2) {
+		if (idx_value & R200_TXFORMAT_NON_POWER2) {
 			track->textures[i].use_pitch = 1;
 		} else {
 			track->textures[i].use_pitch = 0;
-			track->textures[i].width = 1 << ((ib_chunk->kdata[idx] >> RADEON_TXFORMAT_WIDTH_SHIFT) & RADEON_TXFORMAT_WIDTH_MASK);
-			track->textures[i].height = 1 << ((ib_chunk->kdata[idx] >> RADEON_TXFORMAT_HEIGHT_SHIFT) & RADEON_TXFORMAT_HEIGHT_MASK);
+			track->textures[i].width = 1 << ((idx_value >> RADEON_TXFORMAT_WIDTH_SHIFT) & RADEON_TXFORMAT_WIDTH_MASK);
+			track->textures[i].height = 1 << ((idx_value >> RADEON_TXFORMAT_HEIGHT_SHIFT) & RADEON_TXFORMAT_HEIGHT_MASK);
 		}
-		switch ((ib_chunk->kdata[idx] & RADEON_TXFORMAT_FORMAT_MASK)) {
+		switch ((idx_value & RADEON_TXFORMAT_FORMAT_MASK)) {
 		case R200_TXFORMAT_I8:
 		case R200_TXFORMAT_RGB332:
 		case R200_TXFORMAT_Y8:
 			track->textures[i].cpp = 1;
 			break;
-		case R200_TXFORMAT_DXT1:
 		case R200_TXFORMAT_AI88:
 		case R200_TXFORMAT_ARGB1555:
 		case R200_TXFORMAT_RGB565:
@@ -419,13 +466,20 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 		case R200_TXFORMAT_ABGR8888:
 		case R200_TXFORMAT_BGR111110:
 		case R200_TXFORMAT_LDVDU8888:
+			track->textures[i].cpp = 4;
+			break;
+		case R200_TXFORMAT_DXT1:
+			track->textures[i].cpp = 1;
+			track->textures[i].compress_format = R100_TRACK_COMP_DXT1;
+			break;
 		case R200_TXFORMAT_DXT23:
 		case R200_TXFORMAT_DXT45:
-			track->textures[i].cpp = 4;
+			track->textures[i].cpp = 1;
+			track->textures[i].compress_format = R100_TRACK_COMP_DXT1;
 			break;
 		}
-		track->textures[i].cube_info[4].width = 1 << ((ib_chunk->kdata[idx] >> 16) & 0xf);
-		track->textures[i].cube_info[4].height = 1 << ((ib_chunk->kdata[idx] >> 20) & 0xf);
+		track->textures[i].cube_info[4].width = 1 << ((idx_value >> 16) & 0xf);
+		track->textures[i].cube_info[4].height = 1 << ((idx_value >> 20) & 0xf);
 		break;
 	case R200_PP_CUBIC_FACES_0:
 	case R200_PP_CUBIC_FACES_1:
@@ -433,7 +487,7 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 	case R200_PP_CUBIC_FACES_3:
 	case R200_PP_CUBIC_FACES_4:
 	case R200_PP_CUBIC_FACES_5:
-		tmp = ib_chunk->kdata[idx];
+		tmp = idx_value;
 		i = (reg - R200_PP_CUBIC_FACES_0) / 32;
 		for (face = 0; face < 4; face++) {
 			track->textures[i].cube_info[face].width = 1 << ((tmp >> (face * 8)) & 0xf);
@@ -448,9 +502,8 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 	return 0;
 }
 
-int r200_init(struct radeon_device *rdev)
+void r200_set_safe_registers(struct radeon_device *rdev)
 {
 	rdev->config.r100.reg_safe_bm = r200_reg_safe_bm;
 	rdev->config.r100.reg_safe_bm_size = ARRAY_SIZE(r200_reg_safe_bm);
-	return 0;
 }
diff --git a/drivers/gpu/drm/radeon/r300.c b/drivers/gpu/drm/radeon/r300.c
index bb151ecdf8fc..4cef90cd74e5 100644
--- a/drivers/gpu/drm/radeon/r300.c
+++ b/drivers/gpu/drm/radeon/r300.c
@@ -33,43 +33,24 @@
 #include "radeon_drm.h"
 #include "r100_track.h"
 #include "r300d.h"
-
+#include "rv350d.h"
 #include "r300_reg_safe.h"
 
-/* r300,r350,rv350,rv370,rv380 depends on : */
-void r100_hdp_reset(struct radeon_device *rdev);
-int r100_cp_reset(struct radeon_device *rdev);
-int r100_rb2d_reset(struct radeon_device *rdev);
-int r100_cp_init(struct radeon_device *rdev, unsigned ring_size);
-int r100_pci_gart_enable(struct radeon_device *rdev);
-void r100_mc_setup(struct radeon_device *rdev);
-void r100_mc_disable_clients(struct radeon_device *rdev);
-int r100_gui_wait_for_idle(struct radeon_device *rdev);
-int r100_cs_packet_parse(struct radeon_cs_parser *p,
-			 struct radeon_cs_packet *pkt,
-			 unsigned idx);
-int r100_cs_packet_parse_vline(struct radeon_cs_parser *p);
-int r100_cs_parse_packet0(struct radeon_cs_parser *p,
-			  struct radeon_cs_packet *pkt,
-			  const unsigned *auth, unsigned n,
-			  radeon_packet0_check_t check);
-int r100_cs_track_check_pkt3_indx_buffer(struct radeon_cs_parser *p,
-					 struct radeon_cs_packet *pkt,
-					 struct radeon_object *robj);
-
-/* This files gather functions specifics to:
- * r300,r350,rv350,rv370,rv380
+/* This files gather functions specifics to: r300,r350,rv350,rv370,rv380
  *
- * Some of these functions might be used by newer ASICs.
+ * GPU Errata:
+ * - HOST_PATH_CNTL: r300 family seems to dislike write to HOST_PATH_CNTL
+ *   using MMIO to flush host path read cache, this lead to HARDLOCKUP.
+ *   However, scheduling such write to the ring seems harmless, i suspect
+ *   the CP read collide with the flush somehow, or maybe the MC, hard to
+ *   tell. (Jerome Glisse)
  */
-void r300_gpu_init(struct radeon_device *rdev);
-int r300_mc_wait_for_idle(struct radeon_device *rdev);
-int rv370_debugfs_pcie_gart_info_init(struct radeon_device *rdev);
-
 
 /*
  * rv370,rv380 PCIE GART
  */
+static int rv370_debugfs_pcie_gart_info_init(struct radeon_device *rdev);
+
 void rv370_pcie_gart_tlb_flush(struct radeon_device *rdev)
 {
 	uint32_t tmp;
@@ -136,18 +117,19 @@ int rv370_pcie_gart_enable(struct radeon_device *rdev)
 	r = radeon_gart_table_vram_pin(rdev);
 	if (r)
 		return r;
+	radeon_gart_restore(rdev);
 	/* discard memory request outside of configured range */
 	tmp = RADEON_PCIE_TX_GART_UNMAPPED_ACCESS_DISCARD;
 	WREG32_PCIE(RADEON_PCIE_TX_GART_CNTL, tmp);
-	WREG32_PCIE(RADEON_PCIE_TX_GART_START_LO, rdev->mc.gtt_location);
-	tmp = rdev->mc.gtt_location + rdev->mc.gtt_size - 4096;
+	WREG32_PCIE(RADEON_PCIE_TX_GART_START_LO, rdev->mc.gtt_start);
+	tmp = rdev->mc.gtt_end & ~RADEON_GPU_PAGE_MASK;
 	WREG32_PCIE(RADEON_PCIE_TX_GART_END_LO, tmp);
 	WREG32_PCIE(RADEON_PCIE_TX_GART_START_HI, 0);
 	WREG32_PCIE(RADEON_PCIE_TX_GART_END_HI, 0);
 	table_addr = rdev->gart.table_addr;
 	WREG32_PCIE(RADEON_PCIE_TX_GART_BASE, table_addr);
 	/* FIXME: setup default page */
-	WREG32_PCIE(RADEON_PCIE_TX_DISCARD_RD_ADDR_LO, rdev->mc.vram_location);
+	WREG32_PCIE(RADEON_PCIE_TX_DISCARD_RD_ADDR_LO, rdev->mc.vram_start);
 	WREG32_PCIE(RADEON_PCIE_TX_DISCARD_RD_ADDR_HI, 0);
 	/* Clear error */
 	WREG32_PCIE(0x18, 0);
@@ -164,14 +146,19 @@ int rv370_pcie_gart_enable(struct radeon_device *rdev)
 
 void rv370_pcie_gart_disable(struct radeon_device *rdev)
 {
-	uint32_t tmp;
+	u32 tmp;
+	int r;
 
 	tmp = RREG32_PCIE(RADEON_PCIE_TX_GART_CNTL);
 	tmp |= RADEON_PCIE_TX_GART_UNMAPPED_ACCESS_DISCARD;
 	WREG32_PCIE(RADEON_PCIE_TX_GART_CNTL, tmp & ~RADEON_PCIE_TX_GART_EN);
 	if (rdev->gart.table.vram.robj) {
-		radeon_object_kunmap(rdev->gart.table.vram.robj);
-		radeon_object_unpin(rdev->gart.table.vram.robj);
+		r = radeon_bo_reserve(rdev->gart.table.vram.robj, false);
+		if (likely(r == 0)) {
+			radeon_bo_kunmap(rdev->gart.table.vram.robj);
+			radeon_bo_unpin(rdev->gart.table.vram.robj);
+			radeon_bo_unreserve(rdev->gart.table.vram.robj);
+		}
 	}
 }
 
@@ -182,77 +169,31 @@ void rv370_pcie_gart_fini(struct radeon_device *rdev)
 	radeon_gart_fini(rdev);
 }
 
-/*
- * MC
- */
-int r300_mc_init(struct radeon_device *rdev)
-{
-	int r;
-
-	if (r100_debugfs_rbbm_init(rdev)) {
-		DRM_ERROR("Failed to register debugfs file for RBBM !\n");
-	}
-
-	r300_gpu_init(rdev);
-	r100_pci_gart_disable(rdev);
-	if (rdev->flags & RADEON_IS_PCIE) {
-		rv370_pcie_gart_disable(rdev);
-	}
-
-	/* Setup GPU memory space */
-	rdev->mc.vram_location = 0xFFFFFFFFUL;
-	rdev->mc.gtt_location = 0xFFFFFFFFUL;
-	if (rdev->flags & RADEON_IS_AGP) {
-		r = radeon_agp_init(rdev);
-		if (r) {
-			printk(KERN_WARNING "[drm] Disabling AGP\n");
-			rdev->flags &= ~RADEON_IS_AGP;
-			rdev->mc.gtt_size = radeon_gart_size * 1024 * 1024;
-		} else {
-			rdev->mc.gtt_location = rdev->mc.agp_base;
-		}
-	}
-	r = radeon_mc_setup(rdev);
-	if (r) {
-		return r;
-	}
-
-	/* Program GPU memory space */
-	r100_mc_disable_clients(rdev);
-	if (r300_mc_wait_for_idle(rdev)) {
-		printk(KERN_WARNING "Failed to wait MC idle while "
-		       "programming pipes. Bad things might happen.\n");
-	}
-	r100_mc_setup(rdev);
-	return 0;
-}
-
-void r300_mc_fini(struct radeon_device *rdev)
-{
-}
-
-
-/*
- * Fence emission
- */
 void r300_fence_ring_emit(struct radeon_device *rdev,
 			  struct radeon_fence *fence)
 {
 	/* Who ever call radeon_fence_emit should call ring_lock and ask
 	 * for enough space (today caller are ib schedule and buffer move) */
 	/* Write SC register so SC & US assert idle */
-	radeon_ring_write(rdev, PACKET0(0x43E0, 0));
+	radeon_ring_write(rdev, PACKET0(R300_RE_SCISSORS_TL, 0));
 	radeon_ring_write(rdev, 0);
-	radeon_ring_write(rdev, PACKET0(0x43E4, 0));
+	radeon_ring_write(rdev, PACKET0(R300_RE_SCISSORS_BR, 0));
 	radeon_ring_write(rdev, 0);
 	/* Flush 3D cache */
-	radeon_ring_write(rdev, PACKET0(0x4E4C, 0));
-	radeon_ring_write(rdev, (2 << 0));
-	radeon_ring_write(rdev, PACKET0(0x4F18, 0));
-	radeon_ring_write(rdev, (1 << 0));
+	radeon_ring_write(rdev, PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));
+	radeon_ring_write(rdev, R300_RB3D_DC_FLUSH);
+	radeon_ring_write(rdev, PACKET0(R300_RB3D_ZCACHE_CTLSTAT, 0));
+	radeon_ring_write(rdev, R300_ZC_FLUSH);
 	/* Wait until IDLE & CLEAN */
-	radeon_ring_write(rdev, PACKET0(0x1720, 0));
-	radeon_ring_write(rdev, (1 << 17) | (1 << 16)  | (1 << 9));
+	radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0));
+	radeon_ring_write(rdev, (RADEON_WAIT_3D_IDLECLEAN |
+				 RADEON_WAIT_2D_IDLECLEAN |
+				 RADEON_WAIT_DMA_GUI_IDLE));
+	radeon_ring_write(rdev, PACKET0(RADEON_HOST_PATH_CNTL, 0));
+	radeon_ring_write(rdev, rdev->config.r300.hdp_cntl |
+				RADEON_HDP_READ_BUFFER_INVALIDATE);
+	radeon_ring_write(rdev, PACKET0(RADEON_HOST_PATH_CNTL, 0));
+	radeon_ring_write(rdev, rdev->config.r300.hdp_cntl);
 	/* Emit fence sequence & fire IRQ */
 	radeon_ring_write(rdev, PACKET0(rdev->fence_drv.scratch_reg, 0));
 	radeon_ring_write(rdev, fence->seq);
@@ -260,54 +201,6 @@ void r300_fence_ring_emit(struct radeon_device *rdev,
 	radeon_ring_write(rdev, RADEON_SW_INT_FIRE);
 }
 
-
-/*
- * Global GPU functions
- */
-int r300_copy_dma(struct radeon_device *rdev,
-		  uint64_t src_offset,
-		  uint64_t dst_offset,
-		  unsigned num_pages,
-		  struct radeon_fence *fence)
-{
-	uint32_t size;
-	uint32_t cur_size;
-	int i, num_loops;
-	int r = 0;
-
-	/* radeon pitch is /64 */
-	size = num_pages << PAGE_SHIFT;
-	num_loops = DIV_ROUND_UP(size, 0x1FFFFF);
-	r = radeon_ring_lock(rdev, num_loops * 4 + 64);
-	if (r) {
-		DRM_ERROR("radeon: moving bo (%d).\n", r);
-		return r;
-	}
-	/* Must wait for 2D idle & clean before DMA or hangs might happen */
-	radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0 ));
-	radeon_ring_write(rdev, (1 << 16));
-	for (i = 0; i < num_loops; i++) {
-		cur_size = size;
-		if (cur_size > 0x1FFFFF) {
-			cur_size = 0x1FFFFF;
-		}
-		size -= cur_size;
-		radeon_ring_write(rdev, PACKET0(0x720, 2));
-		radeon_ring_write(rdev, src_offset);
-		radeon_ring_write(rdev, dst_offset);
-		radeon_ring_write(rdev, cur_size | (1 << 31) | (1 << 30));
-		src_offset += cur_size;
-		dst_offset += cur_size;
-	}
-	radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0));
-	radeon_ring_write(rdev, RADEON_WAIT_DMA_GUI_IDLE);
-	if (fence) {
-		r = radeon_fence_emit(rdev, fence);
-	}
-	radeon_ring_unlock_commit(rdev);
-	return r;
-}
-
 void r300_ring_start(struct radeon_device *rdev)
 {
 	unsigned gb_tile_config;
@@ -347,8 +240,8 @@ void r300_ring_start(struct radeon_device *rdev)
 	radeon_ring_write(rdev,
 			  RADEON_WAIT_2D_IDLECLEAN |
 			  RADEON_WAIT_3D_IDLECLEAN);
-	radeon_ring_write(rdev, PACKET0(0x170C, 0));
-	radeon_ring_write(rdev, 1 << 31);
+	radeon_ring_write(rdev, PACKET0(R300_DST_PIPE_CONFIG, 0));
+	radeon_ring_write(rdev, R300_PIPE_AUTO_CONFIG);
 	radeon_ring_write(rdev, PACKET0(R300_GB_SELECT, 0));
 	radeon_ring_write(rdev, 0);
 	radeon_ring_write(rdev, PACKET0(R300_GB_ENABLE, 0));
@@ -415,8 +308,8 @@ int r300_mc_wait_for_idle(struct radeon_device *rdev)
 
 	for (i = 0; i < rdev->usec_timeout; i++) {
 		/* read MC_STATUS */
-		tmp = RREG32(0x0150);
-		if (tmp & (1 << 4)) {
+		tmp = RREG32(RADEON_MC_STATUS);
+		if (tmp & R300_MC_IDLE) {
 			return 0;
 		}
 		DRM_UDELAY(1);
@@ -461,8 +354,8 @@ void r300_gpu_init(struct radeon_device *rdev)
 		       "programming pipes. Bad things might happen.\n");
 	}
 
-	tmp = RREG32(0x170C);
-	WREG32(0x170C, tmp | (1 << 31));
+	tmp = RREG32(R300_DST_PIPE_CONFIG);
+	WREG32(R300_DST_PIPE_CONFIG, tmp | R300_PIPE_AUTO_CONFIG);
 
 	WREG32(R300_RB2D_DSTCACHE_MODE,
 	       R300_DC_AUTOFLUSH_ENABLE |
@@ -503,8 +396,8 @@ int r300_ga_reset(struct radeon_device *rdev)
 			/* GA still busy soft reset it */
 			WREG32(0x429C, 0x200);
 			WREG32(R300_VAP_PVS_STATE_FLUSH_REG, 0);
-			WREG32(0x43E0, 0);
-			WREG32(0x43E4, 0);
+			WREG32(R300_RE_SCISSORS_TL, 0);
+			WREG32(R300_RE_SCISSORS_BR, 0);
 			WREG32(0x24AC, 0);
 		}
 		/* Wait to prevent race in RBBM_STATUS */
@@ -554,7 +447,7 @@ int r300_gpu_reset(struct radeon_device *rdev)
 	}
 	/* Check if GPU is idle */
 	status = RREG32(RADEON_RBBM_STATUS);
-	if (status & (1 << 31)) {
+	if (status & RADEON_RBBM_ACTIVE) {
 		DRM_ERROR("Failed to reset GPU (RBBM_STATUS=0x%08X)\n", status);
 		return -1;
 	}
@@ -566,27 +459,30 @@ int r300_gpu_reset(struct radeon_device *rdev)
 /*
  * r300,r350,rv350,rv380 VRAM info
  */
-void r300_vram_info(struct radeon_device *rdev)
+void r300_mc_init(struct radeon_device *rdev)
 {
-	uint32_t tmp;
+	u64 base;
+	u32 tmp;
 
 	/* DDR for all card after R300 & IGP */
 	rdev->mc.vram_is_ddr = true;
 	tmp = RREG32(RADEON_MEM_CNTL);
-	if (tmp & R300_MEM_NUM_CHANNELS_MASK) {
-		rdev->mc.vram_width = 128;
-	} else {
-		rdev->mc.vram_width = 64;
+	tmp &= R300_MEM_NUM_CHANNELS_MASK;
+	switch (tmp) {
+	case 0: rdev->mc.vram_width = 64; break;
+	case 1: rdev->mc.vram_width = 128; break;
+	case 2: rdev->mc.vram_width = 256; break;
+	default:  rdev->mc.vram_width = 128; break;
 	}
-
 	r100_vram_init_sizes(rdev);
+	base = rdev->mc.aper_base;
+	if (rdev->flags & RADEON_IS_IGP)
+		base = (RREG32(RADEON_NB_TOM) & 0xffff) << 16;
+	radeon_vram_location(rdev, &rdev->mc, base);
+	if (!(rdev->flags & RADEON_IS_AGP))
+		radeon_gtt_location(rdev, &rdev->mc);
 }
 
-
-/*
- * PCIE Lanes
- */
-
 void rv370_set_pcie_lanes(struct radeon_device *rdev, int lanes)
 {
 	uint32_t link_width_cntl, mask;
@@ -646,10 +542,40 @@ void rv370_set_pcie_lanes(struct radeon_device *rdev, int lanes)
 
 }
 
+int rv370_get_pcie_lanes(struct radeon_device *rdev)
+{
+	u32 link_width_cntl;
+
+	if (rdev->flags & RADEON_IS_IGP)
+		return 0;
+
+	if (!(rdev->flags & RADEON_IS_PCIE))
+		return 0;
+
+	/* FIXME wait for idle */
+
+	if (rdev->family < CHIP_R600)
+		link_width_cntl = RREG32_PCIE(RADEON_PCIE_LC_LINK_WIDTH_CNTL);
+	else
+		link_width_cntl = RREG32_PCIE_P(RADEON_PCIE_LC_LINK_WIDTH_CNTL);
+
+	switch ((link_width_cntl & RADEON_PCIE_LC_LINK_WIDTH_RD_MASK) >> RADEON_PCIE_LC_LINK_WIDTH_RD_SHIFT) {
+	case RADEON_PCIE_LC_LINK_WIDTH_X0:
+		return 0;
+	case RADEON_PCIE_LC_LINK_WIDTH_X1:
+		return 1;
+	case RADEON_PCIE_LC_LINK_WIDTH_X2:
+		return 2;
+	case RADEON_PCIE_LC_LINK_WIDTH_X4:
+		return 4;
+	case RADEON_PCIE_LC_LINK_WIDTH_X8:
+		return 8;
+	case RADEON_PCIE_LC_LINK_WIDTH_X16:
+	default:
+		return 16;
+	}
+}
 
-/*
- * Debugfs info
- */
 #if defined(CONFIG_DEBUG_FS)
 static int rv370_debugfs_pcie_gart_info(struct seq_file *m, void *data)
 {
@@ -680,7 +606,7 @@ static struct drm_info_list rv370_pcie_gart_info_list[] = {
 };
 #endif
 
-int rv370_debugfs_pcie_gart_info_init(struct radeon_device *rdev)
+static int rv370_debugfs_pcie_gart_info_init(struct radeon_device *rdev)
 {
 #if defined(CONFIG_DEBUG_FS)
 	return radeon_debugfs_add_files(rdev, rv370_pcie_gart_info_list, 1);
@@ -689,25 +615,22 @@ int rv370_debugfs_pcie_gart_info_init(struct radeon_device *rdev)
 #endif
 }
 
-
-/*
- * CS functions
- */
 static int r300_packet0_check(struct radeon_cs_parser *p,
 		struct radeon_cs_packet *pkt,
 		unsigned idx, unsigned reg)
 {
-	struct radeon_cs_chunk *ib_chunk;
 	struct radeon_cs_reloc *reloc;
 	struct r100_cs_track *track;
 	volatile uint32_t *ib;
 	uint32_t tmp, tile_flags = 0;
 	unsigned i;
 	int r;
+	u32 idx_value;
 
 	ib = p->ib->ptr;
-	ib_chunk = &p->chunks[p->chunk_ib_idx];
 	track = (struct r100_cs_track *)p->track;
+	idx_value = radeon_get_ib_value(p, idx);
+
 	switch(reg) {
 	case AVIVO_D1MODE_VLINE_START_END:
 	case RADEON_CRTC_GUI_TRIG_VLINE:
@@ -738,8 +661,8 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 			return r;
 		}
 		track->cb[i].robj = reloc->robj;
-		track->cb[i].offset = ib_chunk->kdata[idx];
-		ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset);
+		track->cb[i].offset = idx_value;
+		ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
 		break;
 	case R300_ZB_DEPTHOFFSET:
 		r = r100_cs_packet_next_reloc(p, &reloc);
@@ -750,8 +673,8 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 			return r;
 		}
 		track->zb.robj = reloc->robj;
-		track->zb.offset = ib_chunk->kdata[idx];
-		ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset);
+		track->zb.offset = idx_value;
+		ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
 		break;
 	case R300_TX_OFFSET_0:
 	case R300_TX_OFFSET_0+4:
@@ -777,32 +700,42 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 			r100_cs_dump_packet(p, pkt);
 			return r;
 		}
-		ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset);
+
+		if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO)
+			tile_flags |= R300_TXO_MACRO_TILE;
+		if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO)
+			tile_flags |= R300_TXO_MICRO_TILE;
+		else if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO_SQUARE)
+			tile_flags |= R300_TXO_MICRO_TILE_SQUARE;
+
+		tmp = idx_value + ((u32)reloc->lobj.gpu_offset);
+		tmp |= tile_flags;
+		ib[idx] = tmp;
 		track->textures[i].robj = reloc->robj;
 		break;
 	/* Tracked registers */
 	case 0x2084:
 		/* VAP_VF_CNTL */
-		track->vap_vf_cntl = ib_chunk->kdata[idx];
+		track->vap_vf_cntl = idx_value;
 		break;
 	case 0x20B4:
 		/* VAP_VTX_SIZE */
-		track->vtx_size = ib_chunk->kdata[idx] & 0x7F;
+		track->vtx_size = idx_value & 0x7F;
 		break;
 	case 0x2134:
 		/* VAP_VF_MAX_VTX_INDX */
-		track->max_indx = ib_chunk->kdata[idx] & 0x00FFFFFFUL;
+		track->max_indx = idx_value & 0x00FFFFFFUL;
 		break;
 	case 0x43E4:
 		/* SC_SCISSOR1 */
-		track->maxy = ((ib_chunk->kdata[idx] >> 13) & 0x1FFF) + 1;
+		track->maxy = ((idx_value >> 13) & 0x1FFF) + 1;
 		if (p->rdev->family < CHIP_RV515) {
 			track->maxy -= 1440;
 		}
 		break;
 	case 0x4E00:
 		/* RB3D_CCTL */
-		track->num_cb = ((ib_chunk->kdata[idx] >> 5) & 0x3) + 1;
+		track->num_cb = ((idx_value >> 5) & 0x3) + 1;
 		break;
 	case 0x4E38:
 	case 0x4E3C:
@@ -824,14 +757,16 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 			tile_flags |= R300_COLOR_TILE_ENABLE;
 		if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO)
 			tile_flags |= R300_COLOR_MICROTILE_ENABLE;
+		else if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO_SQUARE)
+			tile_flags |= R300_COLOR_MICROTILE_SQUARE_ENABLE;
 
-		tmp = ib_chunk->kdata[idx] & ~(0x7 << 16);
+		tmp = idx_value & ~(0x7 << 16);
 		tmp |= tile_flags;
 		ib[idx] = tmp;
 
 		i = (reg - 0x4E38) >> 2;
-		track->cb[i].pitch = ib_chunk->kdata[idx] & 0x3FFE;
-		switch (((ib_chunk->kdata[idx] >> 21) & 0xF)) {
+		track->cb[i].pitch = idx_value & 0x3FFE;
+		switch (((idx_value >> 21) & 0xF)) {
 		case 9:
 		case 11:
 		case 12:
@@ -854,13 +789,13 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 			break;
 		default:
 			DRM_ERROR("Invalid color buffer format (%d) !\n",
-				  ((ib_chunk->kdata[idx] >> 21) & 0xF));
+				  ((idx_value >> 21) & 0xF));
 			return -EINVAL;
 		}
 		break;
 	case 0x4F00:
 		/* ZB_CNTL */
-		if (ib_chunk->kdata[idx] & 2) {
+		if (idx_value & 2) {
 			track->z_enabled = true;
 		} else {
 			track->z_enabled = false;
@@ -868,7 +803,7 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 		break;
 	case 0x4F10:
 		/* ZB_FORMAT */
-		switch ((ib_chunk->kdata[idx] & 0xF)) {
+		switch ((idx_value & 0xF)) {
 		case 0:
 		case 1:
 			track->zb.cpp = 2;
@@ -878,7 +813,7 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 			break;
 		default:
 			DRM_ERROR("Invalid z buffer format (%d) !\n",
-				  (ib_chunk->kdata[idx] & 0xF));
+				  (idx_value & 0xF));
 			return -EINVAL;
 		}
 		break;
@@ -895,19 +830,21 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 		if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO)
 			tile_flags |= R300_DEPTHMACROTILE_ENABLE;
 		if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO)
-			tile_flags |= R300_DEPTHMICROTILE_TILED;;
+			tile_flags |= R300_DEPTHMICROTILE_TILED;
+		else if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO_SQUARE)
+			tile_flags |= R300_DEPTHMICROTILE_TILED_SQUARE;
 
-		tmp = ib_chunk->kdata[idx] & ~(0x7 << 16);
+		tmp = idx_value & ~(0x7 << 16);
 		tmp |= tile_flags;
 		ib[idx] = tmp;
 
-		track->zb.pitch = ib_chunk->kdata[idx] & 0x3FFC;
+		track->zb.pitch = idx_value & 0x3FFC;
 		break;
 	case 0x4104:
 		for (i = 0; i < 16; i++) {
 			bool enabled;
 
-			enabled = !!(ib_chunk->kdata[idx] & (1 << i));
+			enabled = !!(idx_value & (1 << i));
 			track->textures[i].enabled = enabled;
 		}
 		break;
@@ -929,9 +866,9 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 	case 0x44FC:
 		/* TX_FORMAT1_[0-15] */
 		i = (reg - 0x44C0) >> 2;
-		tmp = (ib_chunk->kdata[idx] >> 25) & 0x3;
+		tmp = (idx_value >> 25) & 0x3;
 		track->textures[i].tex_coord_type = tmp;
-		switch ((ib_chunk->kdata[idx] & 0x1F)) {
+		switch ((idx_value & 0x1F)) {
 		case R300_TX_FORMAT_X8:
 		case R300_TX_FORMAT_Y4X4:
 		case R300_TX_FORMAT_Z3Y3X2:
@@ -943,7 +880,6 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 		case R300_TX_FORMAT_Z6Y5X5:
 		case R300_TX_FORMAT_W4Z4Y4X4:
 		case R300_TX_FORMAT_W1Z5Y5X5:
-		case R300_TX_FORMAT_DXT1:
 		case R300_TX_FORMAT_D3DMFT_CxV8U8:
 		case R300_TX_FORMAT_B8G8_B8G8:
 		case R300_TX_FORMAT_G8R8_G8B8:
@@ -957,8 +893,6 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 		case 0x17:
 		case R300_TX_FORMAT_FL_I32:
 		case 0x1e:
-		case R300_TX_FORMAT_DXT3:
-		case R300_TX_FORMAT_DXT5:
 			track->textures[i].cpp = 4;
 			break;
 		case R300_TX_FORMAT_W16Z16Y16X16:
@@ -969,9 +903,26 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 		case R300_TX_FORMAT_FL_R32G32B32A32:
 			track->textures[i].cpp = 16;
 			break;
+		case R300_TX_FORMAT_DXT1:
+			track->textures[i].cpp = 1;
+			track->textures[i].compress_format = R100_TRACK_COMP_DXT1;
+			break;
+		case R300_TX_FORMAT_ATI2N:
+			if (p->rdev->family < CHIP_R420) {
+				DRM_ERROR("Invalid texture format %u\n",
+					  (idx_value & 0x1F));
+				return -EINVAL;
+			}
+			/* The same rules apply as for DXT3/5. */
+			/* Pass through. */
+		case R300_TX_FORMAT_DXT3:
+		case R300_TX_FORMAT_DXT5:
+			track->textures[i].cpp = 1;
+			track->textures[i].compress_format = R100_TRACK_COMP_DXT35;
+			break;
 		default:
 			DRM_ERROR("Invalid texture format %u\n",
-				  (ib_chunk->kdata[idx] & 0x1F));
+				  (idx_value & 0x1F));
 			return -EINVAL;
 			break;
 		}
@@ -994,11 +945,11 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 	case 0x443C:
 		/* TX_FILTER0_[0-15] */
 		i = (reg - 0x4400) >> 2;
-		tmp = ib_chunk->kdata[idx] & 0x7;
+		tmp = idx_value & 0x7;
 		if (tmp == 2 || tmp == 4 || tmp == 6) {
 			track->textures[i].roundup_w = false;
 		}
-		tmp = (ib_chunk->kdata[idx] >> 3) & 0x7;
+		tmp = (idx_value >> 3) & 0x7;
 		if (tmp == 2 || tmp == 4 || tmp == 6) {
 			track->textures[i].roundup_h = false;
 		}
@@ -1021,13 +972,23 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 	case 0x453C:
 		/* TX_FORMAT2_[0-15] */
 		i = (reg - 0x4500) >> 2;
-		tmp = ib_chunk->kdata[idx] & 0x3FFF;
+		tmp = idx_value & 0x3FFF;
 		track->textures[i].pitch = tmp + 1;
 		if (p->rdev->family >= CHIP_RV515) {
-			tmp = ((ib_chunk->kdata[idx] >> 15) & 1) << 11;
+			tmp = ((idx_value >> 15) & 1) << 11;
 			track->textures[i].width_11 = tmp;
-			tmp = ((ib_chunk->kdata[idx] >> 16) & 1) << 11;
+			tmp = ((idx_value >> 16) & 1) << 11;
 			track->textures[i].height_11 = tmp;
+
+			/* ATI1N */
+			if (idx_value & (1 << 14)) {
+				/* The same rules apply as for DXT1. */
+				track->textures[i].compress_format =
+					R100_TRACK_COMP_DXT1;
+			}
+		} else if (idx_value & (1 << 14)) {
+			DRM_ERROR("Forbidden bit TXFORMAT_MSB\n");
+			return -EINVAL;
 		}
 		break;
 	case 0x4480:
@@ -1048,15 +1009,15 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 	case 0x44BC:
 		/* TX_FORMAT0_[0-15] */
 		i = (reg - 0x4480) >> 2;
-		tmp = ib_chunk->kdata[idx] & 0x7FF;
+		tmp = idx_value & 0x7FF;
 		track->textures[i].width = tmp + 1;
-		tmp = (ib_chunk->kdata[idx] >> 11) & 0x7FF;
+		tmp = (idx_value >> 11) & 0x7FF;
 		track->textures[i].height = tmp + 1;
-		tmp = (ib_chunk->kdata[idx] >> 26) & 0xF;
+		tmp = (idx_value >> 26) & 0xF;
 		track->textures[i].num_levels = tmp;
-		tmp = ib_chunk->kdata[idx] & (1 << 31);
+		tmp = idx_value & (1 << 31);
 		track->textures[i].use_pitch = !!tmp;
-		tmp = (ib_chunk->kdata[idx] >> 22) & 0xF;
+		tmp = (idx_value >> 22) & 0xF;
 		track->textures[i].txdepth = tmp;
 		break;
 	case R300_ZB_ZPASS_ADDR:
@@ -1067,7 +1028,19 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 			r100_cs_dump_packet(p, pkt);
 			return r;
 		}
-		ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset);
+		ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
+		break;
+	case 0x4e0c:
+		/* RB3D_COLOR_CHANNEL_MASK */
+		track->color_channel_mask = idx_value;
+		break;
+	case 0x4d1c:
+		/* ZB_BW_CNTL */
+		track->fastfill = !!(idx_value & (1 << 2));
+		break;
+	case 0x4e04:
+		/* RB3D_BLENDCNTL */
+		track->blend_read_enable = !!(idx_value & (1 << 2));
 		break;
 	case 0x4be8:
 		/* valid register only on RV530 */
@@ -1085,60 +1058,20 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 static int r300_packet3_check(struct radeon_cs_parser *p,
 			      struct radeon_cs_packet *pkt)
 {
-	struct radeon_cs_chunk *ib_chunk;
-
 	struct radeon_cs_reloc *reloc;
 	struct r100_cs_track *track;
 	volatile uint32_t *ib;
 	unsigned idx;
-	unsigned i, c;
 	int r;
 
 	ib = p->ib->ptr;
-	ib_chunk = &p->chunks[p->chunk_ib_idx];
 	idx = pkt->idx + 1;
 	track = (struct r100_cs_track *)p->track;
 	switch(pkt->opcode) {
 	case PACKET3_3D_LOAD_VBPNTR:
-		c = ib_chunk->kdata[idx++] & 0x1F;
-		track->num_arrays = c;
-		for (i = 0; i < (c - 1); i+=2, idx+=3) {
-			r = r100_cs_packet_next_reloc(p, &reloc);
-			if (r) {
-				DRM_ERROR("No reloc for packet3 %d\n",
-					  pkt->opcode);
-				r100_cs_dump_packet(p, pkt);
-				return r;
-			}
-			ib[idx+1] = ib_chunk->kdata[idx+1] + ((u32)reloc->lobj.gpu_offset);
-			track->arrays[i + 0].robj = reloc->robj;
-			track->arrays[i + 0].esize = ib_chunk->kdata[idx] >> 8;
-			track->arrays[i + 0].esize &= 0x7F;
-			r = r100_cs_packet_next_reloc(p, &reloc);
-			if (r) {
-				DRM_ERROR("No reloc for packet3 %d\n",
-					  pkt->opcode);
-				r100_cs_dump_packet(p, pkt);
-				return r;
-			}
-			ib[idx+2] = ib_chunk->kdata[idx+2] + ((u32)reloc->lobj.gpu_offset);
-			track->arrays[i + 1].robj = reloc->robj;
-			track->arrays[i + 1].esize = ib_chunk->kdata[idx] >> 24;
-			track->arrays[i + 1].esize &= 0x7F;
-		}
-		if (c & 1) {
-			r = r100_cs_packet_next_reloc(p, &reloc);
-			if (r) {
-				DRM_ERROR("No reloc for packet3 %d\n",
-					  pkt->opcode);
-				r100_cs_dump_packet(p, pkt);
-				return r;
-			}
-			ib[idx+1] = ib_chunk->kdata[idx+1] + ((u32)reloc->lobj.gpu_offset);
-			track->arrays[i + 0].robj = reloc->robj;
-			track->arrays[i + 0].esize = ib_chunk->kdata[idx] >> 8;
-			track->arrays[i + 0].esize &= 0x7F;
-		}
+		r = r100_packet3_load_vbpntr(p, pkt, idx);
+		if (r)
+			return r;
 		break;
 	case PACKET3_INDX_BUFFER:
 		r = r100_cs_packet_next_reloc(p, &reloc);
@@ -1147,7 +1080,7 @@ static int r300_packet3_check(struct radeon_cs_parser *p,
 			r100_cs_dump_packet(p, pkt);
 			return r;
 		}
-		ib[idx+1] = ib_chunk->kdata[idx+1] + ((u32)reloc->lobj.gpu_offset);
+		ib[idx+1] = radeon_get_ib_value(p, idx + 1) + ((u32)reloc->lobj.gpu_offset);
 		r = r100_cs_track_check_pkt3_indx_buffer(p, pkt, reloc->robj);
 		if (r) {
 			return r;
@@ -1158,11 +1091,11 @@ static int r300_packet3_check(struct radeon_cs_parser *p,
 		/* Number of dwords is vtx_size * (num_vertices - 1)
 		 * PRIM_WALK must be equal to 3 vertex data in embedded
 		 * in cmd stream */
-		if (((ib_chunk->kdata[idx+1] >> 4) & 0x3) != 3) {
+		if (((radeon_get_ib_value(p, idx + 1) >> 4) & 0x3) != 3) {
 			DRM_ERROR("PRIM_WALK must be 3 for IMMD draw\n");
 			return -EINVAL;
 		}
-		track->vap_vf_cntl = ib_chunk->kdata[idx+1];
+		track->vap_vf_cntl = radeon_get_ib_value(p, idx + 1);
 		track->immd_dwords = pkt->count - 1;
 		r = r100_cs_track_check(p->rdev, track);
 		if (r) {
@@ -1173,11 +1106,11 @@ static int r300_packet3_check(struct radeon_cs_parser *p,
 		/* Number of dwords is vtx_size * (num_vertices - 1)
 		 * PRIM_WALK must be equal to 3 vertex data in embedded
 		 * in cmd stream */
-		if (((ib_chunk->kdata[idx] >> 4) & 0x3) != 3) {
+		if (((radeon_get_ib_value(p, idx) >> 4) & 0x3) != 3) {
 			DRM_ERROR("PRIM_WALK must be 3 for IMMD draw\n");
 			return -EINVAL;
 		}
-		track->vap_vf_cntl = ib_chunk->kdata[idx];
+		track->vap_vf_cntl = radeon_get_ib_value(p, idx);
 		track->immd_dwords = pkt->count;
 		r = r100_cs_track_check(p->rdev, track);
 		if (r) {
@@ -1185,28 +1118,28 @@ static int r300_packet3_check(struct radeon_cs_parser *p,
 		}
 		break;
 	case PACKET3_3D_DRAW_VBUF:
-		track->vap_vf_cntl = ib_chunk->kdata[idx + 1];
+		track->vap_vf_cntl = radeon_get_ib_value(p, idx + 1);
 		r = r100_cs_track_check(p->rdev, track);
 		if (r) {
 			return r;
 		}
 		break;
 	case PACKET3_3D_DRAW_VBUF_2:
-		track->vap_vf_cntl = ib_chunk->kdata[idx];
+		track->vap_vf_cntl = radeon_get_ib_value(p, idx);
 		r = r100_cs_track_check(p->rdev, track);
 		if (r) {
 			return r;
 		}
 		break;
 	case PACKET3_3D_DRAW_INDX:
-		track->vap_vf_cntl = ib_chunk->kdata[idx + 1];
+		track->vap_vf_cntl = radeon_get_ib_value(p, idx + 1);
 		r = r100_cs_track_check(p->rdev, track);
 		if (r) {
 			return r;
 		}
 		break;
 	case PACKET3_3D_DRAW_INDX_2:
-		track->vap_vf_cntl = ib_chunk->kdata[idx];
+		track->vap_vf_cntl = radeon_get_ib_value(p, idx);
 		r = r100_cs_track_check(p->rdev, track);
 		if (r) {
 			return r;
@@ -1265,12 +1198,6 @@ void r300_set_reg_safe(struct radeon_device *rdev)
 	rdev->config.r300.reg_safe_bm_size = ARRAY_SIZE(r300_reg_safe_bm);
 }
 
-int r300_init(struct radeon_device *rdev)
-{
-	r300_set_reg_safe(rdev);
-	return 0;
-}
-
 void r300_mc_program(struct radeon_device *rdev)
 {
 	struct r100_mc_save save;
@@ -1304,3 +1231,212 @@ void r300_mc_program(struct radeon_device *rdev)
 		S_000148_MC_FB_TOP(rdev->mc.vram_end >> 16));
 	r100_mc_resume(rdev, &save);
 }
+
+void r300_clock_startup(struct radeon_device *rdev)
+{
+	u32 tmp;
+
+	if (radeon_dynclks != -1 && radeon_dynclks)
+		radeon_legacy_set_clock_gating(rdev, 1);
+	/* We need to force on some of the block */
+	tmp = RREG32_PLL(R_00000D_SCLK_CNTL);
+	tmp |= S_00000D_FORCE_CP(1) | S_00000D_FORCE_VIP(1);
+	if ((rdev->family == CHIP_RV350) || (rdev->family == CHIP_RV380))
+		tmp |= S_00000D_FORCE_VAP(1);
+	WREG32_PLL(R_00000D_SCLK_CNTL, tmp);
+}
+
+static int r300_startup(struct radeon_device *rdev)
+{
+	int r;
+
+	/* set common regs */
+	r100_set_common_regs(rdev);
+	/* program mc */
+	r300_mc_program(rdev);
+	/* Resume clock */
+	r300_clock_startup(rdev);
+	/* Initialize GPU configuration (# pipes, ...) */
+	r300_gpu_init(rdev);
+	/* Initialize GART (initialize after TTM so we can allocate
+	 * memory through TTM but finalize after TTM) */
+	if (rdev->flags & RADEON_IS_PCIE) {
+		r = rv370_pcie_gart_enable(rdev);
+		if (r)
+			return r;
+	}
+
+	if (rdev->family == CHIP_R300 ||
+	    rdev->family == CHIP_R350 ||
+	    rdev->family == CHIP_RV350)
+		r100_enable_bm(rdev);
+
+	if (rdev->flags & RADEON_IS_PCI) {
+		r = r100_pci_gart_enable(rdev);
+		if (r)
+			return r;
+	}
+	/* Enable IRQ */
+	r100_irq_set(rdev);
+	rdev->config.r300.hdp_cntl = RREG32(RADEON_HOST_PATH_CNTL);
+	/* 1M ring buffer */
+	r = r100_cp_init(rdev, 1024 * 1024);
+	if (r) {
+		dev_err(rdev->dev, "failled initializing CP (%d).\n", r);
+		return r;
+	}
+	r = r100_wb_init(rdev);
+	if (r)
+		dev_err(rdev->dev, "failled initializing WB (%d).\n", r);
+	r = r100_ib_init(rdev);
+	if (r) {
+		dev_err(rdev->dev, "failled initializing IB (%d).\n", r);
+		return r;
+	}
+	return 0;
+}
+
+int r300_resume(struct radeon_device *rdev)
+{
+	/* Make sur GART are not working */
+	if (rdev->flags & RADEON_IS_PCIE)
+		rv370_pcie_gart_disable(rdev);
+	if (rdev->flags & RADEON_IS_PCI)
+		r100_pci_gart_disable(rdev);
+	/* Resume clock before doing reset */
+	r300_clock_startup(rdev);
+	/* Reset gpu before posting otherwise ATOM will enter infinite loop */
+	if (radeon_gpu_reset(rdev)) {
+		dev_warn(rdev->dev, "GPU reset failed ! (0xE40=0x%08X, 0x7C0=0x%08X)\n",
+			RREG32(R_000E40_RBBM_STATUS),
+			RREG32(R_0007C0_CP_STAT));
+	}
+	/* post */
+	radeon_combios_asic_init(rdev->ddev);
+	/* Resume clock after posting */
+	r300_clock_startup(rdev);
+	/* Initialize surface registers */
+	radeon_surface_init(rdev);
+	return r300_startup(rdev);
+}
+
+int r300_suspend(struct radeon_device *rdev)
+{
+	r100_cp_disable(rdev);
+	r100_wb_disable(rdev);
+	r100_irq_disable(rdev);
+	if (rdev->flags & RADEON_IS_PCIE)
+		rv370_pcie_gart_disable(rdev);
+	if (rdev->flags & RADEON_IS_PCI)
+		r100_pci_gart_disable(rdev);
+	return 0;
+}
+
+void r300_fini(struct radeon_device *rdev)
+{
+	r100_cp_fini(rdev);
+	r100_wb_fini(rdev);
+	r100_ib_fini(rdev);
+	radeon_gem_fini(rdev);
+	if (rdev->flags & RADEON_IS_PCIE)
+		rv370_pcie_gart_fini(rdev);
+	if (rdev->flags & RADEON_IS_PCI)
+		r100_pci_gart_fini(rdev);
+	radeon_agp_fini(rdev);
+	radeon_irq_kms_fini(rdev);
+	radeon_fence_driver_fini(rdev);
+	radeon_bo_fini(rdev);
+	radeon_atombios_fini(rdev);
+	kfree(rdev->bios);
+	rdev->bios = NULL;
+}
+
+int r300_init(struct radeon_device *rdev)
+{
+	int r;
+
+	/* Disable VGA */
+	r100_vga_render_disable(rdev);
+	/* Initialize scratch registers */
+	radeon_scratch_init(rdev);
+	/* Initialize surface registers */
+	radeon_surface_init(rdev);
+	/* TODO: disable VGA need to use VGA request */
+	/* BIOS*/
+	if (!radeon_get_bios(rdev)) {
+		if (ASIC_IS_AVIVO(rdev))
+			return -EINVAL;
+	}
+	if (rdev->is_atom_bios) {
+		dev_err(rdev->dev, "Expecting combios for RS400/RS480 GPU\n");
+		return -EINVAL;
+	} else {
+		r = radeon_combios_init(rdev);
+		if (r)
+			return r;
+	}
+	/* Reset gpu before posting otherwise ATOM will enter infinite loop */
+	if (radeon_gpu_reset(rdev)) {
+		dev_warn(rdev->dev,
+			"GPU reset failed ! (0xE40=0x%08X, 0x7C0=0x%08X)\n",
+			RREG32(R_000E40_RBBM_STATUS),
+			RREG32(R_0007C0_CP_STAT));
+	}
+	/* check if cards are posted or not */
+	if (radeon_boot_test_post_card(rdev) == false)
+		return -EINVAL;
+	/* Set asic errata */
+	r300_errata(rdev);
+	/* Initialize clocks */
+	radeon_get_clock_info(rdev->ddev);
+	/* Initialize power management */
+	radeon_pm_init(rdev);
+	/* initialize AGP */
+	if (rdev->flags & RADEON_IS_AGP) {
+		r = radeon_agp_init(rdev);
+		if (r) {
+			radeon_agp_disable(rdev);
+		}
+	}
+	/* initialize memory controller */
+	r300_mc_init(rdev);
+	/* Fence driver */
+	r = radeon_fence_driver_init(rdev);
+	if (r)
+		return r;
+	r = radeon_irq_kms_init(rdev);
+	if (r)
+		return r;
+	/* Memory manager */
+	r = radeon_bo_init(rdev);
+	if (r)
+		return r;
+	if (rdev->flags & RADEON_IS_PCIE) {
+		r = rv370_pcie_gart_init(rdev);
+		if (r)
+			return r;
+	}
+	if (rdev->flags & RADEON_IS_PCI) {
+		r = r100_pci_gart_init(rdev);
+		if (r)
+			return r;
+	}
+	r300_set_reg_safe(rdev);
+	rdev->accel_working = true;
+	r = r300_startup(rdev);
+	if (r) {
+		/* Somethings want wront with the accel init stop accel */
+		dev_err(rdev->dev, "Disabling GPU acceleration\n");
+		r100_cp_fini(rdev);
+		r100_wb_fini(rdev);
+		r100_ib_fini(rdev);
+		radeon_irq_kms_fini(rdev);
+		if (rdev->flags & RADEON_IS_PCIE)
+			rv370_pcie_gart_fini(rdev);
+		if (rdev->flags & RADEON_IS_PCI)
+			r100_pci_gart_fini(rdev);
+		radeon_agp_fini(rdev);
+		rdev->accel_working = false;
+	}
+	return 0;
+}
diff --git a/drivers/gpu/drm/radeon/r300_cmdbuf.c b/drivers/gpu/drm/radeon/r300_cmdbuf.c
index cb2e470f97d4..ea46d558e8f3 100644
--- a/drivers/gpu/drm/radeon/r300_cmdbuf.c
+++ b/drivers/gpu/drm/radeon/r300_cmdbuf.c
@@ -33,6 +33,7 @@
 
 #include "drmP.h"
 #include "drm.h"
+#include "drm_buffer.h"
 #include "radeon_drm.h"
 #include "radeon_drv.h"
 #include "r300_reg.h"
@@ -299,46 +300,42 @@ static __inline__ int r300_emit_carefully_checked_packet0(drm_radeon_private_t *
 	int reg;
 	int sz;
 	int i;
-	int values[64];
+	u32 *value;
 	RING_LOCALS;
 
 	sz = header.packet0.count;
 	reg = (header.packet0.reghi << 8) | header.packet0.reglo;
 
 	if ((sz > 64) || (sz < 0)) {
-		DRM_ERROR
-		    ("Cannot emit more than 64 values at a time (reg=%04x sz=%d)\n",
-		     reg, sz);
+		DRM_ERROR("Cannot emit more than 64 values at a time (reg=%04x sz=%d)\n",
+			 reg, sz);
 		return -EINVAL;
 	}
+
 	for (i = 0; i < sz; i++) {
-		values[i] = ((int *)cmdbuf->buf)[i];
 		switch (r300_reg_flags[(reg >> 2) + i]) {
 		case MARK_SAFE:
 			break;
 		case MARK_CHECK_OFFSET:
-			if (!radeon_check_offset(dev_priv, (u32) values[i])) {
-				DRM_ERROR
-				    ("Offset failed range check (reg=%04x sz=%d)\n",
-				     reg, sz);
+			value = drm_buffer_pointer_to_dword(cmdbuf->buffer, i);
+			if (!radeon_check_offset(dev_priv, *value)) {
+				DRM_ERROR("Offset failed range check (reg=%04x sz=%d)\n",
+					 reg, sz);
 				return -EINVAL;
 			}
 			break;
 		default:
 			DRM_ERROR("Register %04x failed check as flag=%02x\n",
-				  reg + i * 4, r300_reg_flags[(reg >> 2) + i]);
+				reg + i * 4, r300_reg_flags[(reg >> 2) + i]);
 			return -EINVAL;
 		}
 	}
 
 	BEGIN_RING(1 + sz);
 	OUT_RING(CP_PACKET0(reg, sz - 1));
-	OUT_RING_TABLE(values, sz);
+	OUT_RING_DRM_BUFFER(cmdbuf->buffer, sz);
 	ADVANCE_RING();
 
-	cmdbuf->buf += sz * 4;
-	cmdbuf->bufsz -= sz * 4;
-
 	return 0;
 }
 
@@ -362,7 +359,7 @@ static __inline__ int r300_emit_packet0(drm_radeon_private_t *dev_priv,
 	if (!sz)
 		return 0;
 
-	if (sz * 4 > cmdbuf->bufsz)
+	if (sz * 4 > drm_buffer_unprocessed(cmdbuf->buffer))
 		return -EINVAL;
 
 	if (reg + sz * 4 >= 0x10000) {
@@ -380,12 +377,9 @@ static __inline__ int r300_emit_packet0(drm_radeon_private_t *dev_priv,
 
 	BEGIN_RING(1 + sz);
 	OUT_RING(CP_PACKET0(reg, sz - 1));
-	OUT_RING_TABLE((int *)cmdbuf->buf, sz);
+	OUT_RING_DRM_BUFFER(cmdbuf->buffer, sz);
 	ADVANCE_RING();
 
-	cmdbuf->buf += sz * 4;
-	cmdbuf->bufsz -= sz * 4;
-
 	return 0;
 }
 
@@ -407,7 +401,7 @@ static __inline__ int r300_emit_vpu(drm_radeon_private_t *dev_priv,
 
 	if (!sz)
 		return 0;
-	if (sz * 16 > cmdbuf->bufsz)
+	if (sz * 16 > drm_buffer_unprocessed(cmdbuf->buffer))
 		return -EINVAL;
 
 	/* VAP is very sensitive so we purge cache before we program it
@@ -426,7 +420,7 @@ static __inline__ int r300_emit_vpu(drm_radeon_private_t *dev_priv,
 	BEGIN_RING(3 + sz * 4);
 	OUT_RING_REG(R300_VAP_PVS_UPLOAD_ADDRESS, addr);
 	OUT_RING(CP_PACKET0_TABLE(R300_VAP_PVS_UPLOAD_DATA, sz * 4 - 1));
-	OUT_RING_TABLE((int *)cmdbuf->buf, sz * 4);
+	OUT_RING_DRM_BUFFER(cmdbuf->buffer, sz * 4);
 	ADVANCE_RING();
 
 	BEGIN_RING(2);
@@ -434,9 +428,6 @@ static __inline__ int r300_emit_vpu(drm_radeon_private_t *dev_priv,
 	OUT_RING(0);
 	ADVANCE_RING();
 
-	cmdbuf->buf += sz * 16;
-	cmdbuf->bufsz -= sz * 16;
-
 	return 0;
 }
 
@@ -449,14 +440,14 @@ static __inline__ int r300_emit_clear(drm_radeon_private_t *dev_priv,
 {
 	RING_LOCALS;
 
-	if (8 * 4 > cmdbuf->bufsz)
+	if (8 * 4 > drm_buffer_unprocessed(cmdbuf->buffer))
 		return -EINVAL;
 
 	BEGIN_RING(10);
 	OUT_RING(CP_PACKET3(R200_3D_DRAW_IMMD_2, 8));
 	OUT_RING(R300_PRIM_TYPE_POINT | R300_PRIM_WALK_RING |
 		 (1 << R300_PRIM_NUM_VERTICES_SHIFT));
-	OUT_RING_TABLE((int *)cmdbuf->buf, 8);
+	OUT_RING_DRM_BUFFER(cmdbuf->buffer, 8);
 	ADVANCE_RING();
 
 	BEGIN_RING(4);
@@ -468,9 +459,6 @@ static __inline__ int r300_emit_clear(drm_radeon_private_t *dev_priv,
 	/* set flush flag */
 	dev_priv->track_flush |= RADEON_FLUSH_EMITED;
 
-	cmdbuf->buf += 8 * 4;
-	cmdbuf->bufsz -= 8 * 4;
-
 	return 0;
 }
 
@@ -480,28 +468,29 @@ static __inline__ int r300_emit_3d_load_vbpntr(drm_radeon_private_t *dev_priv,
 {
 	int count, i, k;
 #define MAX_ARRAY_PACKET  64
-	u32 payload[MAX_ARRAY_PACKET];
+	u32 *data;
 	u32 narrays;
 	RING_LOCALS;
 
-	count = (header >> 16) & 0x3fff;
+	count = (header & RADEON_CP_PACKET_COUNT_MASK) >> 16;
 
 	if ((count + 1) > MAX_ARRAY_PACKET) {
 		DRM_ERROR("Too large payload in 3D_LOAD_VBPNTR (count=%d)\n",
 			  count);
 		return -EINVAL;
 	}
-	memset(payload, 0, MAX_ARRAY_PACKET * 4);
-	memcpy(payload, cmdbuf->buf + 4, (count + 1) * 4);
-
 	/* carefully check packet contents */
 
-	narrays = payload[0];
+	/* We have already read the header so advance the buffer. */
+	drm_buffer_advance(cmdbuf->buffer, 4);
+
+	narrays = *(u32 *)drm_buffer_pointer_to_dword(cmdbuf->buffer, 0);
 	k = 0;
 	i = 1;
 	while ((k < narrays) && (i < (count + 1))) {
 		i++;		/* skip attribute field */
-		if (!radeon_check_offset(dev_priv, payload[i])) {
+		data = drm_buffer_pointer_to_dword(cmdbuf->buffer, i);
+		if (!radeon_check_offset(dev_priv, *data)) {
 			DRM_ERROR
 			    ("Offset failed range check (k=%d i=%d) while processing 3D_LOAD_VBPNTR packet.\n",
 			     k, i);
@@ -512,7 +501,8 @@ static __inline__ int r300_emit_3d_load_vbpntr(drm_radeon_private_t *dev_priv,
 		if (k == narrays)
 			break;
 		/* have one more to process, they come in pairs */
-		if (!radeon_check_offset(dev_priv, payload[i])) {
+		data = drm_buffer_pointer_to_dword(cmdbuf->buffer, i);
+		if (!radeon_check_offset(dev_priv, *data)) {
 			DRM_ERROR
 			    ("Offset failed range check (k=%d i=%d) while processing 3D_LOAD_VBPNTR packet.\n",
 			     k, i);
@@ -533,30 +523,30 @@ static __inline__ int r300_emit_3d_load_vbpntr(drm_radeon_private_t *dev_priv,
 
 	BEGIN_RING(count + 2);
 	OUT_RING(header);
-	OUT_RING_TABLE(payload, count + 1);
+	OUT_RING_DRM_BUFFER(cmdbuf->buffer, count + 1);
 	ADVANCE_RING();
 
-	cmdbuf->buf += (count + 2) * 4;
-	cmdbuf->bufsz -= (count + 2) * 4;
-
 	return 0;
 }
 
 static __inline__ int r300_emit_bitblt_multi(drm_radeon_private_t *dev_priv,
 					     drm_radeon_kcmd_buffer_t *cmdbuf)
 {
-	u32 *cmd = (u32 *) cmdbuf->buf;
+	u32 *cmd = drm_buffer_pointer_to_dword(cmdbuf->buffer, 0);
 	int count, ret;
 	RING_LOCALS;
 
-	count=(cmd[0]>>16) & 0x3fff;
 
-	if (cmd[0] & 0x8000) {
-		u32 offset;
+	count = (*cmd & RADEON_CP_PACKET_COUNT_MASK) >> 16;
 
-		if (cmd[1] & (RADEON_GMC_SRC_PITCH_OFFSET_CNTL
+	if (*cmd & 0x8000) {
+		u32 offset;
+		u32 *cmd1 = drm_buffer_pointer_to_dword(cmdbuf->buffer, 1);
+		if (*cmd1 & (RADEON_GMC_SRC_PITCH_OFFSET_CNTL
 			      | RADEON_GMC_DST_PITCH_OFFSET_CNTL)) {
-			offset = cmd[2] << 10;
+
+			u32 *cmd2 = drm_buffer_pointer_to_dword(cmdbuf->buffer, 2);
+			offset = *cmd2 << 10;
 			ret = !radeon_check_offset(dev_priv, offset);
 			if (ret) {
 				DRM_ERROR("Invalid bitblt first offset is %08X\n", offset);
@@ -564,9 +554,10 @@ static __inline__ int r300_emit_bitblt_multi(drm_radeon_private_t *dev_priv,
 			}
 		}
 
-		if ((cmd[1] & RADEON_GMC_SRC_PITCH_OFFSET_CNTL) &&
-		    (cmd[1] & RADEON_GMC_DST_PITCH_OFFSET_CNTL)) {
-			offset = cmd[3] << 10;
+		if ((*cmd1 & RADEON_GMC_SRC_PITCH_OFFSET_CNTL) &&
+		    (*cmd1 & RADEON_GMC_DST_PITCH_OFFSET_CNTL)) {
+			u32 *cmd3 = drm_buffer_pointer_to_dword(cmdbuf->buffer, 3);
+			offset = *cmd3 << 10;
 			ret = !radeon_check_offset(dev_priv, offset);
 			if (ret) {
 				DRM_ERROR("Invalid bitblt second offset is %08X\n", offset);
@@ -577,28 +568,25 @@ static __inline__ int r300_emit_bitblt_multi(drm_radeon_private_t *dev_priv,
 	}
 
 	BEGIN_RING(count+2);
-	OUT_RING(cmd[0]);
-	OUT_RING_TABLE((int *)(cmdbuf->buf + 4), count + 1);
+	OUT_RING_DRM_BUFFER(cmdbuf->buffer, count + 2);
 	ADVANCE_RING();
 
-	cmdbuf->buf += (count+2)*4;
-	cmdbuf->bufsz -= (count+2)*4;
-
 	return 0;
 }
 
 static __inline__ int r300_emit_draw_indx_2(drm_radeon_private_t *dev_priv,
 					    drm_radeon_kcmd_buffer_t *cmdbuf)
 {
-	u32 *cmd;
+	u32 *cmd = drm_buffer_pointer_to_dword(cmdbuf->buffer, 0);
+	u32 *cmd1 = drm_buffer_pointer_to_dword(cmdbuf->buffer, 1);
 	int count;
 	int expected_count;
 	RING_LOCALS;
 
-	cmd = (u32 *) cmdbuf->buf;
-	count = (cmd[0]>>16) & 0x3fff;
-	expected_count = cmd[1] >> 16;
-	if (!(cmd[1] & R300_VAP_VF_CNTL__INDEX_SIZE_32bit))
+	count = (*cmd & RADEON_CP_PACKET_COUNT_MASK) >> 16;
+
+	expected_count = *cmd1 >> 16;
+	if (!(*cmd1 & R300_VAP_VF_CNTL__INDEX_SIZE_32bit))
 		expected_count = (expected_count+1)/2;
 
 	if (count && count != expected_count) {
@@ -608,55 +596,53 @@ static __inline__ int r300_emit_draw_indx_2(drm_radeon_private_t *dev_priv,
 	}
 
 	BEGIN_RING(count+2);
-	OUT_RING(cmd[0]);
-	OUT_RING_TABLE((int *)(cmdbuf->buf + 4), count + 1);
+	OUT_RING_DRM_BUFFER(cmdbuf->buffer, count + 2);
 	ADVANCE_RING();
 
-	cmdbuf->buf += (count+2)*4;
-	cmdbuf->bufsz -= (count+2)*4;
-
 	if (!count) {
-		drm_r300_cmd_header_t header;
+		drm_r300_cmd_header_t stack_header, *header;
+		u32 *cmd1, *cmd2, *cmd3;
 
-		if (cmdbuf->bufsz < 4*4 + sizeof(header)) {
+		if (drm_buffer_unprocessed(cmdbuf->buffer)
+				< 4*4 + sizeof(stack_header)) {
 			DRM_ERROR("3D_DRAW_INDX_2: expect subsequent INDX_BUFFER, but stream is too short.\n");
 			return -EINVAL;
 		}
 
-		header.u = *(unsigned int *)cmdbuf->buf;
+		header = drm_buffer_read_object(cmdbuf->buffer,
+				sizeof(stack_header), &stack_header);
 
-		cmdbuf->buf += sizeof(header);
-		cmdbuf->bufsz -= sizeof(header);
-		cmd = (u32 *) cmdbuf->buf;
+		cmd = drm_buffer_pointer_to_dword(cmdbuf->buffer, 0);
+		cmd1 = drm_buffer_pointer_to_dword(cmdbuf->buffer, 1);
+		cmd2 = drm_buffer_pointer_to_dword(cmdbuf->buffer, 2);
+		cmd3 = drm_buffer_pointer_to_dword(cmdbuf->buffer, 3);
 
-		if (header.header.cmd_type != R300_CMD_PACKET3 ||
-		    header.packet3.packet != R300_CMD_PACKET3_RAW ||
-		    cmd[0] != CP_PACKET3(RADEON_CP_INDX_BUFFER, 2)) {
+		if (header->header.cmd_type != R300_CMD_PACKET3 ||
+		    header->packet3.packet != R300_CMD_PACKET3_RAW ||
+		    *cmd != CP_PACKET3(RADEON_CP_INDX_BUFFER, 2)) {
 			DRM_ERROR("3D_DRAW_INDX_2: expect subsequent INDX_BUFFER.\n");
 			return -EINVAL;
 		}
 
-		if ((cmd[1] & 0x8000ffff) != 0x80000810) {
-			DRM_ERROR("Invalid indx_buffer reg address %08X\n", cmd[1]);
+		if ((*cmd1 & 0x8000ffff) != 0x80000810) {
+			DRM_ERROR("Invalid indx_buffer reg address %08X\n",
+					*cmd1);
 			return -EINVAL;
 		}
-		if (!radeon_check_offset(dev_priv, cmd[2])) {
-			DRM_ERROR("Invalid indx_buffer offset is %08X\n", cmd[2]);
+		if (!radeon_check_offset(dev_priv, *cmd2)) {
+			DRM_ERROR("Invalid indx_buffer offset is %08X\n",
+					*cmd2);
 			return -EINVAL;
 		}
-		if (cmd[3] != expected_count) {
+		if (*cmd3 != expected_count) {
 			DRM_ERROR("INDX_BUFFER: buffer size %i, expected %i\n",
-				cmd[3], expected_count);
+				*cmd3, expected_count);
 			return -EINVAL;
 		}
 
 		BEGIN_RING(4);
-		OUT_RING(cmd[0]);
-		OUT_RING_TABLE((int *)(cmdbuf->buf + 4), 3);
+		OUT_RING_DRM_BUFFER(cmdbuf->buffer, 4);
 		ADVANCE_RING();
-
-		cmdbuf->buf += 4*4;
-		cmdbuf->bufsz -= 4*4;
 	}
 
 	return 0;
@@ -665,39 +651,39 @@ static __inline__ int r300_emit_draw_indx_2(drm_radeon_private_t *dev_priv,
 static __inline__ int r300_emit_raw_packet3(drm_radeon_private_t *dev_priv,
 					    drm_radeon_kcmd_buffer_t *cmdbuf)
 {
-	u32 header;
+	u32 *header;
 	int count;
 	RING_LOCALS;
 
-	if (4 > cmdbuf->bufsz)
+	if (4 > drm_buffer_unprocessed(cmdbuf->buffer))
 		return -EINVAL;
 
 	/* Fixme !! This simply emits a packet without much checking.
 	   We need to be smarter. */
 
 	/* obtain first word - actual packet3 header */
-	header = *(u32 *) cmdbuf->buf;
+	header = drm_buffer_pointer_to_dword(cmdbuf->buffer, 0);
 
 	/* Is it packet 3 ? */
-	if ((header >> 30) != 0x3) {
-		DRM_ERROR("Not a packet3 header (0x%08x)\n", header);
+	if ((*header >> 30) != 0x3) {
+		DRM_ERROR("Not a packet3 header (0x%08x)\n", *header);
 		return -EINVAL;
 	}
 
-	count = (header >> 16) & 0x3fff;
+	count = (*header >> 16) & 0x3fff;
 
 	/* Check again now that we know how much data to expect */
-	if ((count + 2) * 4 > cmdbuf->bufsz) {
+	if ((count + 2) * 4 > drm_buffer_unprocessed(cmdbuf->buffer)) {
 		DRM_ERROR
 		    ("Expected packet3 of length %d but have only %d bytes left\n",
-		     (count + 2) * 4, cmdbuf->bufsz);
+		     (count + 2) * 4, drm_buffer_unprocessed(cmdbuf->buffer));
 		return -EINVAL;
 	}
 
 	/* Is it a packet type we know about ? */
-	switch (header & 0xff00) {
+	switch (*header & 0xff00) {
 	case RADEON_3D_LOAD_VBPNTR:	/* load vertex array pointers */
-		return r300_emit_3d_load_vbpntr(dev_priv, cmdbuf, header);
+		return r300_emit_3d_load_vbpntr(dev_priv, cmdbuf, *header);
 
 	case RADEON_CNTL_BITBLT_MULTI:
 		return r300_emit_bitblt_multi(dev_priv, cmdbuf);
@@ -723,18 +709,14 @@ static __inline__ int r300_emit_raw_packet3(drm_radeon_private_t *dev_priv,
 		/* these packets are safe */
 		break;
 	default:
-		DRM_ERROR("Unknown packet3 header (0x%08x)\n", header);
+		DRM_ERROR("Unknown packet3 header (0x%08x)\n", *header);
 		return -EINVAL;
 	}
 
 	BEGIN_RING(count + 2);
-	OUT_RING(header);
-	OUT_RING_TABLE((int *)(cmdbuf->buf + 4), count + 1);
+	OUT_RING_DRM_BUFFER(cmdbuf->buffer, count + 2);
 	ADVANCE_RING();
 
-	cmdbuf->buf += (count + 2) * 4;
-	cmdbuf->bufsz -= (count + 2) * 4;
-
 	return 0;
 }
 
@@ -748,8 +730,7 @@ static __inline__ int r300_emit_packet3(drm_radeon_private_t *dev_priv,
 {
 	int n;
 	int ret;
-	char *orig_buf = cmdbuf->buf;
-	int orig_bufsz = cmdbuf->bufsz;
+	int orig_iter = cmdbuf->buffer->iterator;
 
 	/* This is a do-while-loop so that we run the interior at least once,
 	 * even if cmdbuf->nbox is 0. Compare r300_emit_cliprects for rationale.
@@ -761,8 +742,7 @@ static __inline__ int r300_emit_packet3(drm_radeon_private_t *dev_priv,
 			if (ret)
 				return ret;
 
-			cmdbuf->buf = orig_buf;
-			cmdbuf->bufsz = orig_bufsz;
+			cmdbuf->buffer->iterator = orig_iter;
 		}
 
 		switch (header.packet3.packet) {
@@ -785,9 +765,9 @@ static __inline__ int r300_emit_packet3(drm_radeon_private_t *dev_priv,
 			break;
 
 		default:
-			DRM_ERROR("bad packet3 type %i at %p\n",
+			DRM_ERROR("bad packet3 type %i at byte %d\n",
 				  header.packet3.packet,
-				  cmdbuf->buf - sizeof(header));
+				  cmdbuf->buffer->iterator - (int)sizeof(header));
 			return -EINVAL;
 		}
 
@@ -923,12 +903,13 @@ static int r300_scratch(drm_radeon_private_t *dev_priv,
 			drm_r300_cmd_header_t header)
 {
 	u32 *ref_age_base;
-	u32 i, buf_idx, h_pending;
-	u64 ptr_addr;
+	u32 i, *buf_idx, h_pending;
+	u64 *ptr_addr;
+	u64 stack_ptr_addr;
 	RING_LOCALS;
 
-	if (cmdbuf->bufsz <
-	    (sizeof(u64) + header.scratch.n_bufs * sizeof(buf_idx))) {
+	if (drm_buffer_unprocessed(cmdbuf->buffer) <
+	    (sizeof(u64) + header.scratch.n_bufs * sizeof(*buf_idx))) {
 		return -EINVAL;
 	}
 
@@ -938,36 +919,35 @@ static int r300_scratch(drm_radeon_private_t *dev_priv,
 
 	dev_priv->scratch_ages[header.scratch.reg]++;
 
-	ptr_addr = get_unaligned((u64 *)cmdbuf->buf);
-	ref_age_base = (u32 *)(unsigned long)ptr_addr;
-
-	cmdbuf->buf += sizeof(u64);
-	cmdbuf->bufsz -= sizeof(u64);
+	ptr_addr = drm_buffer_read_object(cmdbuf->buffer,
+			sizeof(stack_ptr_addr), &stack_ptr_addr);
+	ref_age_base = (u32 *)(unsigned long)*ptr_addr;
 
 	for (i=0; i < header.scratch.n_bufs; i++) {
-		buf_idx = *(u32 *)cmdbuf->buf;
-		buf_idx *= 2; /* 8 bytes per buf */
+		buf_idx = drm_buffer_pointer_to_dword(cmdbuf->buffer, 0);
+		*buf_idx *= 2; /* 8 bytes per buf */
 
-		if (DRM_COPY_TO_USER(ref_age_base + buf_idx, &dev_priv->scratch_ages[header.scratch.reg], sizeof(u32))) {
+		if (DRM_COPY_TO_USER(ref_age_base + *buf_idx,
+				&dev_priv->scratch_ages[header.scratch.reg],
+				sizeof(u32)))
 			return -EINVAL;
-		}
 
-		if (DRM_COPY_FROM_USER(&h_pending, ref_age_base + buf_idx + 1, sizeof(u32))) {
+		if (DRM_COPY_FROM_USER(&h_pending,
+				ref_age_base + *buf_idx + 1,
+				sizeof(u32)))
 			return -EINVAL;
-		}
 
-		if (h_pending == 0) {
+		if (h_pending == 0)
 			return -EINVAL;
-		}
 
 		h_pending--;
 
-		if (DRM_COPY_TO_USER(ref_age_base + buf_idx + 1, &h_pending, sizeof(u32))) {
+		if (DRM_COPY_TO_USER(ref_age_base + *buf_idx + 1,
+					&h_pending,
+					sizeof(u32)))
 			return -EINVAL;
-		}
 
-		cmdbuf->buf += sizeof(buf_idx);
-		cmdbuf->bufsz -= sizeof(buf_idx);
+		drm_buffer_advance(cmdbuf->buffer, sizeof(*buf_idx));
 	}
 
 	BEGIN_RING(2);
@@ -990,7 +970,7 @@ static inline int r300_emit_r500fp(drm_radeon_private_t *dev_priv,
 	int sz;
 	int addr;
 	int type;
-	int clamp;
+	int isclamp;
 	int stride;
 	RING_LOCALS;
 
@@ -999,29 +979,26 @@ static inline int r300_emit_r500fp(drm_radeon_private_t *dev_priv,
 	addr = ((header.r500fp.adrhi_flags & 1) << 8) | header.r500fp.adrlo;
 
 	type = !!(header.r500fp.adrhi_flags & R500FP_CONSTANT_TYPE);
-	clamp = !!(header.r500fp.adrhi_flags & R500FP_CONSTANT_CLAMP);
+	isclamp = !!(header.r500fp.adrhi_flags & R500FP_CONSTANT_CLAMP);
 
 	addr |= (type << 16);
-	addr |= (clamp << 17);
+	addr |= (isclamp << 17);
 
 	stride = type ? 4 : 6;
 
 	DRM_DEBUG("r500fp %d %d type: %d\n", sz, addr, type);
 	if (!sz)
 		return 0;
-	if (sz * stride * 4 > cmdbuf->bufsz)
+	if (sz * stride * 4 > drm_buffer_unprocessed(cmdbuf->buffer))
 		return -EINVAL;
 
 	BEGIN_RING(3 + sz * stride);
 	OUT_RING_REG(R500_GA_US_VECTOR_INDEX, addr);
 	OUT_RING(CP_PACKET0_TABLE(R500_GA_US_VECTOR_DATA, sz * stride - 1));
-	OUT_RING_TABLE((int *)cmdbuf->buf, sz * stride);
+	OUT_RING_DRM_BUFFER(cmdbuf->buffer, sz * stride);
 
 	ADVANCE_RING();
 
-	cmdbuf->buf += sz * stride * 4;
-	cmdbuf->bufsz -= sz * stride * 4;
-
 	return 0;
 }
 
@@ -1053,19 +1030,18 @@ int r300_do_cp_cmdbuf(struct drm_device *dev,
 			goto cleanup;
 	}
 
-	while (cmdbuf->bufsz >= sizeof(drm_r300_cmd_header_t)) {
+	while (drm_buffer_unprocessed(cmdbuf->buffer)
+			>= sizeof(drm_r300_cmd_header_t)) {
 		int idx;
-		drm_r300_cmd_header_t header;
-
-		header.u = *(unsigned int *)cmdbuf->buf;
+		drm_r300_cmd_header_t *header, stack_header;
 
-		cmdbuf->buf += sizeof(header);
-		cmdbuf->bufsz -= sizeof(header);
+		header = drm_buffer_read_object(cmdbuf->buffer,
+				sizeof(stack_header), &stack_header);
 
-		switch (header.header.cmd_type) {
+		switch (header->header.cmd_type) {
 		case R300_CMD_PACKET0:
 			DRM_DEBUG("R300_CMD_PACKET0\n");
-			ret = r300_emit_packet0(dev_priv, cmdbuf, header);
+			ret = r300_emit_packet0(dev_priv, cmdbuf, *header);
 			if (ret) {
 				DRM_ERROR("r300_emit_packet0 failed\n");
 				goto cleanup;
@@ -1074,7 +1050,7 @@ int r300_do_cp_cmdbuf(struct drm_device *dev,
 
 		case R300_CMD_VPU:
 			DRM_DEBUG("R300_CMD_VPU\n");
-			ret = r300_emit_vpu(dev_priv, cmdbuf, header);
+			ret = r300_emit_vpu(dev_priv, cmdbuf, *header);
 			if (ret) {
 				DRM_ERROR("r300_emit_vpu failed\n");
 				goto cleanup;
@@ -1083,7 +1059,7 @@ int r300_do_cp_cmdbuf(struct drm_device *dev,
 
 		case R300_CMD_PACKET3:
 			DRM_DEBUG("R300_CMD_PACKET3\n");
-			ret = r300_emit_packet3(dev_priv, cmdbuf, header);
+			ret = r300_emit_packet3(dev_priv, cmdbuf, *header);
 			if (ret) {
 				DRM_ERROR("r300_emit_packet3 failed\n");
 				goto cleanup;
@@ -1117,8 +1093,8 @@ int r300_do_cp_cmdbuf(struct drm_device *dev,
 				int i;
 				RING_LOCALS;
 
-				BEGIN_RING(header.delay.count);
-				for (i = 0; i < header.delay.count; i++)
+				BEGIN_RING(header->delay.count);
+				for (i = 0; i < header->delay.count; i++)
 					OUT_RING(RADEON_CP_PACKET2);
 				ADVANCE_RING();
 			}
@@ -1126,7 +1102,7 @@ int r300_do_cp_cmdbuf(struct drm_device *dev,
 
 		case R300_CMD_DMA_DISCARD:
 			DRM_DEBUG("RADEON_CMD_DMA_DISCARD\n");
-			idx = header.dma.buf_idx;
+			idx = header->dma.buf_idx;
 			if (idx < 0 || idx >= dma->buf_count) {
 				DRM_ERROR("buffer index %d (of %d max)\n",
 					  idx, dma->buf_count - 1);
@@ -1149,12 +1125,12 @@ int r300_do_cp_cmdbuf(struct drm_device *dev,
 
 		case R300_CMD_WAIT:
 			DRM_DEBUG("R300_CMD_WAIT\n");
-			r300_cmd_wait(dev_priv, header);
+			r300_cmd_wait(dev_priv, *header);
 			break;
 
 		case R300_CMD_SCRATCH:
 			DRM_DEBUG("R300_CMD_SCRATCH\n");
-			ret = r300_scratch(dev_priv, cmdbuf, header);
+			ret = r300_scratch(dev_priv, cmdbuf, *header);
 			if (ret) {
 				DRM_ERROR("r300_scratch failed\n");
 				goto cleanup;
@@ -1168,16 +1144,16 @@ int r300_do_cp_cmdbuf(struct drm_device *dev,
 				goto cleanup;
 			}
 			DRM_DEBUG("R300_CMD_R500FP\n");
-			ret = r300_emit_r500fp(dev_priv, cmdbuf, header);
+			ret = r300_emit_r500fp(dev_priv, cmdbuf, *header);
 			if (ret) {
 				DRM_ERROR("r300_emit_r500fp failed\n");
 				goto cleanup;
 			}
 			break;
 		default:
-			DRM_ERROR("bad cmd_type %i at %p\n",
-				  header.header.cmd_type,
-				  cmdbuf->buf - sizeof(header));
+			DRM_ERROR("bad cmd_type %i at byte %d\n",
+				  header->header.cmd_type,
+				  cmdbuf->buffer->iterator - (int)sizeof(*header));
 			ret = -EINVAL;
 			goto cleanup;
 		}
diff --git a/drivers/gpu/drm/radeon/r300_reg.h b/drivers/gpu/drm/radeon/r300_reg.h
index 4b7afef35a65..1a0d5362cd79 100644
--- a/drivers/gpu/drm/radeon/r300_reg.h
+++ b/drivers/gpu/drm/radeon/r300_reg.h
@@ -900,6 +900,7 @@
 #	define R300_TX_FORMAT_FL_I32		    0x1B
 #	define R300_TX_FORMAT_FL_I32A32		    0x1C
 #	define R300_TX_FORMAT_FL_R32G32B32A32	    0x1D
+#	define R300_TX_FORMAT_ATI2N		    0x1F
 	/* alpha modes, convenience mostly */
 	/* if you have alpha, pick constant appropriate to the
 	   number of channels (1 for I8, 2 for I8A8, 4 for R8G8B8A8, etc */
@@ -951,6 +952,7 @@
 #       define R300_TXO_ENDIAN_HALFDW_SWAP       (3 << 0)
 #       define R300_TXO_MACRO_TILE               (1 << 2)
 #       define R300_TXO_MICRO_TILE               (1 << 3)
+#       define R300_TXO_MICRO_TILE_SQUARE        (2 << 3)
 #       define R300_TXO_OFFSET_MASK              0xffffffe0
 #       define R300_TXO_OFFSET_SHIFT             5
 	/* END: Guess from R200 */
@@ -1359,6 +1361,7 @@
 #       define R300_COLORPITCH_MASK              0x00001FF8 /* GUESS */
 #       define R300_COLOR_TILE_ENABLE            (1 << 16) /* GUESS */
 #       define R300_COLOR_MICROTILE_ENABLE       (1 << 17) /* GUESS */
+#       define R300_COLOR_MICROTILE_SQUARE_ENABLE (2 << 17)
 #       define R300_COLOR_ENDIAN_NO_SWAP         (0 << 18) /* GUESS */
 #       define R300_COLOR_ENDIAN_WORD_SWAP       (1 << 18) /* GUESS */
 #       define R300_COLOR_ENDIAN_DWORD_SWAP      (2 << 18) /* GUESS */
diff --git a/drivers/gpu/drm/radeon/r300d.h b/drivers/gpu/drm/radeon/r300d.h
index d4fa3eb1074f..4c73114f0de9 100644
--- a/drivers/gpu/drm/radeon/r300d.h
+++ b/drivers/gpu/drm/radeon/r300d.h
@@ -96,6 +96,211 @@
 #define   S_000170_AGP_BASE_ADDR(x)                    (((x) & 0xFFFFFFFF) << 0)
 #define   G_000170_AGP_BASE_ADDR(x)                    (((x) >> 0) & 0xFFFFFFFF)
 #define   C_000170_AGP_BASE_ADDR                       0x00000000
+#define R_0007C0_CP_STAT                             0x0007C0
+#define   S_0007C0_MRU_BUSY(x)                         (((x) & 0x1) << 0)
+#define   G_0007C0_MRU_BUSY(x)                         (((x) >> 0) & 0x1)
+#define   C_0007C0_MRU_BUSY                            0xFFFFFFFE
+#define   S_0007C0_MWU_BUSY(x)                         (((x) & 0x1) << 1)
+#define   G_0007C0_MWU_BUSY(x)                         (((x) >> 1) & 0x1)
+#define   C_0007C0_MWU_BUSY                            0xFFFFFFFD
+#define   S_0007C0_RSIU_BUSY(x)                        (((x) & 0x1) << 2)
+#define   G_0007C0_RSIU_BUSY(x)                        (((x) >> 2) & 0x1)
+#define   C_0007C0_RSIU_BUSY                           0xFFFFFFFB
+#define   S_0007C0_RCIU_BUSY(x)                        (((x) & 0x1) << 3)
+#define   G_0007C0_RCIU_BUSY(x)                        (((x) >> 3) & 0x1)
+#define   C_0007C0_RCIU_BUSY                           0xFFFFFFF7
+#define   S_0007C0_CSF_PRIMARY_BUSY(x)                 (((x) & 0x1) << 9)
+#define   G_0007C0_CSF_PRIMARY_BUSY(x)                 (((x) >> 9) & 0x1)
+#define   C_0007C0_CSF_PRIMARY_BUSY                    0xFFFFFDFF
+#define   S_0007C0_CSF_INDIRECT_BUSY(x)                (((x) & 0x1) << 10)
+#define   G_0007C0_CSF_INDIRECT_BUSY(x)                (((x) >> 10) & 0x1)
+#define   C_0007C0_CSF_INDIRECT_BUSY                   0xFFFFFBFF
+#define   S_0007C0_CSQ_PRIMARY_BUSY(x)                 (((x) & 0x1) << 11)
+#define   G_0007C0_CSQ_PRIMARY_BUSY(x)                 (((x) >> 11) & 0x1)
+#define   C_0007C0_CSQ_PRIMARY_BUSY                    0xFFFFF7FF
+#define   S_0007C0_CSQ_INDIRECT_BUSY(x)                (((x) & 0x1) << 12)
+#define   G_0007C0_CSQ_INDIRECT_BUSY(x)                (((x) >> 12) & 0x1)
+#define   C_0007C0_CSQ_INDIRECT_BUSY                   0xFFFFEFFF
+#define   S_0007C0_CSI_BUSY(x)                         (((x) & 0x1) << 13)
+#define   G_0007C0_CSI_BUSY(x)                         (((x) >> 13) & 0x1)
+#define   C_0007C0_CSI_BUSY                            0xFFFFDFFF
+#define   S_0007C0_CSF_INDIRECT2_BUSY(x)               (((x) & 0x1) << 14)
+#define   G_0007C0_CSF_INDIRECT2_BUSY(x)               (((x) >> 14) & 0x1)
+#define   C_0007C0_CSF_INDIRECT2_BUSY                  0xFFFFBFFF
+#define   S_0007C0_CSQ_INDIRECT2_BUSY(x)               (((x) & 0x1) << 15)
+#define   G_0007C0_CSQ_INDIRECT2_BUSY(x)               (((x) >> 15) & 0x1)
+#define   C_0007C0_CSQ_INDIRECT2_BUSY                  0xFFFF7FFF
+#define   S_0007C0_GUIDMA_BUSY(x)                      (((x) & 0x1) << 28)
+#define   G_0007C0_GUIDMA_BUSY(x)                      (((x) >> 28) & 0x1)
+#define   C_0007C0_GUIDMA_BUSY                         0xEFFFFFFF
+#define   S_0007C0_VIDDMA_BUSY(x)                      (((x) & 0x1) << 29)
+#define   G_0007C0_VIDDMA_BUSY(x)                      (((x) >> 29) & 0x1)
+#define   C_0007C0_VIDDMA_BUSY                         0xDFFFFFFF
+#define   S_0007C0_CMDSTRM_BUSY(x)                     (((x) & 0x1) << 30)
+#define   G_0007C0_CMDSTRM_BUSY(x)                     (((x) >> 30) & 0x1)
+#define   C_0007C0_CMDSTRM_BUSY                        0xBFFFFFFF
+#define   S_0007C0_CP_BUSY(x)                          (((x) & 0x1) << 31)
+#define   G_0007C0_CP_BUSY(x)                          (((x) >> 31) & 0x1)
+#define   C_0007C0_CP_BUSY                             0x7FFFFFFF
+#define R_000E40_RBBM_STATUS                         0x000E40
+#define   S_000E40_CMDFIFO_AVAIL(x)                    (((x) & 0x7F) << 0)
+#define   G_000E40_CMDFIFO_AVAIL(x)                    (((x) >> 0) & 0x7F)
+#define   C_000E40_CMDFIFO_AVAIL                       0xFFFFFF80
+#define   S_000E40_HIRQ_ON_RBB(x)                      (((x) & 0x1) << 8)
+#define   G_000E40_HIRQ_ON_RBB(x)                      (((x) >> 8) & 0x1)
+#define   C_000E40_HIRQ_ON_RBB                         0xFFFFFEFF
+#define   S_000E40_CPRQ_ON_RBB(x)                      (((x) & 0x1) << 9)
+#define   G_000E40_CPRQ_ON_RBB(x)                      (((x) >> 9) & 0x1)
+#define   C_000E40_CPRQ_ON_RBB                         0xFFFFFDFF
+#define   S_000E40_CFRQ_ON_RBB(x)                      (((x) & 0x1) << 10)
+#define   G_000E40_CFRQ_ON_RBB(x)                      (((x) >> 10) & 0x1)
+#define   C_000E40_CFRQ_ON_RBB                         0xFFFFFBFF
+#define   S_000E40_HIRQ_IN_RTBUF(x)                    (((x) & 0x1) << 11)
+#define   G_000E40_HIRQ_IN_RTBUF(x)                    (((x) >> 11) & 0x1)
+#define   C_000E40_HIRQ_IN_RTBUF                       0xFFFFF7FF
+#define   S_000E40_CPRQ_IN_RTBUF(x)                    (((x) & 0x1) << 12)
+#define   G_000E40_CPRQ_IN_RTBUF(x)                    (((x) >> 12) & 0x1)
+#define   C_000E40_CPRQ_IN_RTBUF                       0xFFFFEFFF
+#define   S_000E40_CFRQ_IN_RTBUF(x)                    (((x) & 0x1) << 13)
+#define   G_000E40_CFRQ_IN_RTBUF(x)                    (((x) >> 13) & 0x1)
+#define   C_000E40_CFRQ_IN_RTBUF                       0xFFFFDFFF
+#define   S_000E40_CF_PIPE_BUSY(x)                     (((x) & 0x1) << 14)
+#define   G_000E40_CF_PIPE_BUSY(x)                     (((x) >> 14) & 0x1)
+#define   C_000E40_CF_PIPE_BUSY                        0xFFFFBFFF
+#define   S_000E40_ENG_EV_BUSY(x)                      (((x) & 0x1) << 15)
+#define   G_000E40_ENG_EV_BUSY(x)                      (((x) >> 15) & 0x1)
+#define   C_000E40_ENG_EV_BUSY                         0xFFFF7FFF
+#define   S_000E40_CP_CMDSTRM_BUSY(x)                  (((x) & 0x1) << 16)
+#define   G_000E40_CP_CMDSTRM_BUSY(x)                  (((x) >> 16) & 0x1)
+#define   C_000E40_CP_CMDSTRM_BUSY                     0xFFFEFFFF
+#define   S_000E40_E2_BUSY(x)                          (((x) & 0x1) << 17)
+#define   G_000E40_E2_BUSY(x)                          (((x) >> 17) & 0x1)
+#define   C_000E40_E2_BUSY                             0xFFFDFFFF
+#define   S_000E40_RB2D_BUSY(x)                        (((x) & 0x1) << 18)
+#define   G_000E40_RB2D_BUSY(x)                        (((x) >> 18) & 0x1)
+#define   C_000E40_RB2D_BUSY                           0xFFFBFFFF
+#define   S_000E40_RB3D_BUSY(x)                        (((x) & 0x1) << 19)
+#define   G_000E40_RB3D_BUSY(x)                        (((x) >> 19) & 0x1)
+#define   C_000E40_RB3D_BUSY                           0xFFF7FFFF
+#define   S_000E40_VAP_BUSY(x)                         (((x) & 0x1) << 20)
+#define   G_000E40_VAP_BUSY(x)                         (((x) >> 20) & 0x1)
+#define   C_000E40_VAP_BUSY                            0xFFEFFFFF
+#define   S_000E40_RE_BUSY(x)                          (((x) & 0x1) << 21)
+#define   G_000E40_RE_BUSY(x)                          (((x) >> 21) & 0x1)
+#define   C_000E40_RE_BUSY                             0xFFDFFFFF
+#define   S_000E40_TAM_BUSY(x)                         (((x) & 0x1) << 22)
+#define   G_000E40_TAM_BUSY(x)                         (((x) >> 22) & 0x1)
+#define   C_000E40_TAM_BUSY                            0xFFBFFFFF
+#define   S_000E40_TDM_BUSY(x)                         (((x) & 0x1) << 23)
+#define   G_000E40_TDM_BUSY(x)                         (((x) >> 23) & 0x1)
+#define   C_000E40_TDM_BUSY                            0xFF7FFFFF
+#define   S_000E40_PB_BUSY(x)                          (((x) & 0x1) << 24)
+#define   G_000E40_PB_BUSY(x)                          (((x) >> 24) & 0x1)
+#define   C_000E40_PB_BUSY                             0xFEFFFFFF
+#define   S_000E40_TIM_BUSY(x)                         (((x) & 0x1) << 25)
+#define   G_000E40_TIM_BUSY(x)                         (((x) >> 25) & 0x1)
+#define   C_000E40_TIM_BUSY                            0xFDFFFFFF
+#define   S_000E40_GA_BUSY(x)                          (((x) & 0x1) << 26)
+#define   G_000E40_GA_BUSY(x)                          (((x) >> 26) & 0x1)
+#define   C_000E40_GA_BUSY                             0xFBFFFFFF
+#define   S_000E40_CBA2D_BUSY(x)                       (((x) & 0x1) << 27)
+#define   G_000E40_CBA2D_BUSY(x)                       (((x) >> 27) & 0x1)
+#define   C_000E40_CBA2D_BUSY                          0xF7FFFFFF
+#define   S_000E40_GUI_ACTIVE(x)                       (((x) & 0x1) << 31)
+#define   G_000E40_GUI_ACTIVE(x)                       (((x) >> 31) & 0x1)
+#define   C_000E40_GUI_ACTIVE                          0x7FFFFFFF
 
 
+#define R_00000D_SCLK_CNTL                           0x00000D
+#define   S_00000D_SCLK_SRC_SEL(x)                     (((x) & 0x7) << 0)
+#define   G_00000D_SCLK_SRC_SEL(x)                     (((x) >> 0) & 0x7)
+#define   C_00000D_SCLK_SRC_SEL                        0xFFFFFFF8
+#define   S_00000D_CP_MAX_DYN_STOP_LAT(x)              (((x) & 0x1) << 3)
+#define   G_00000D_CP_MAX_DYN_STOP_LAT(x)              (((x) >> 3) & 0x1)
+#define   C_00000D_CP_MAX_DYN_STOP_LAT                 0xFFFFFFF7
+#define   S_00000D_HDP_MAX_DYN_STOP_LAT(x)             (((x) & 0x1) << 4)
+#define   G_00000D_HDP_MAX_DYN_STOP_LAT(x)             (((x) >> 4) & 0x1)
+#define   C_00000D_HDP_MAX_DYN_STOP_LAT                0xFFFFFFEF
+#define   S_00000D_TV_MAX_DYN_STOP_LAT(x)              (((x) & 0x1) << 5)
+#define   G_00000D_TV_MAX_DYN_STOP_LAT(x)              (((x) >> 5) & 0x1)
+#define   C_00000D_TV_MAX_DYN_STOP_LAT                 0xFFFFFFDF
+#define   S_00000D_E2_MAX_DYN_STOP_LAT(x)              (((x) & 0x1) << 6)
+#define   G_00000D_E2_MAX_DYN_STOP_LAT(x)              (((x) >> 6) & 0x1)
+#define   C_00000D_E2_MAX_DYN_STOP_LAT                 0xFFFFFFBF
+#define   S_00000D_SE_MAX_DYN_STOP_LAT(x)              (((x) & 0x1) << 7)
+#define   G_00000D_SE_MAX_DYN_STOP_LAT(x)              (((x) >> 7) & 0x1)
+#define   C_00000D_SE_MAX_DYN_STOP_LAT                 0xFFFFFF7F
+#define   S_00000D_IDCT_MAX_DYN_STOP_LAT(x)            (((x) & 0x1) << 8)
+#define   G_00000D_IDCT_MAX_DYN_STOP_LAT(x)            (((x) >> 8) & 0x1)
+#define   C_00000D_IDCT_MAX_DYN_STOP_LAT               0xFFFFFEFF
+#define   S_00000D_VIP_MAX_DYN_STOP_LAT(x)             (((x) & 0x1) << 9)
+#define   G_00000D_VIP_MAX_DYN_STOP_LAT(x)             (((x) >> 9) & 0x1)
+#define   C_00000D_VIP_MAX_DYN_STOP_LAT                0xFFFFFDFF
+#define   S_00000D_RE_MAX_DYN_STOP_LAT(x)              (((x) & 0x1) << 10)
+#define   G_00000D_RE_MAX_DYN_STOP_LAT(x)              (((x) >> 10) & 0x1)
+#define   C_00000D_RE_MAX_DYN_STOP_LAT                 0xFFFFFBFF
+#define   S_00000D_PB_MAX_DYN_STOP_LAT(x)              (((x) & 0x1) << 11)
+#define   G_00000D_PB_MAX_DYN_STOP_LAT(x)              (((x) >> 11) & 0x1)
+#define   C_00000D_PB_MAX_DYN_STOP_LAT                 0xFFFFF7FF
+#define   S_00000D_TAM_MAX_DYN_STOP_LAT(x)             (((x) & 0x1) << 12)
+#define   G_00000D_TAM_MAX_DYN_STOP_LAT(x)             (((x) >> 12) & 0x1)
+#define   C_00000D_TAM_MAX_DYN_STOP_LAT                0xFFFFEFFF
+#define   S_00000D_TDM_MAX_DYN_STOP_LAT(x)             (((x) & 0x1) << 13)
+#define   G_00000D_TDM_MAX_DYN_STOP_LAT(x)             (((x) >> 13) & 0x1)
+#define   C_00000D_TDM_MAX_DYN_STOP_LAT                0xFFFFDFFF
+#define   S_00000D_RB_MAX_DYN_STOP_LAT(x)              (((x) & 0x1) << 14)
+#define   G_00000D_RB_MAX_DYN_STOP_LAT(x)              (((x) >> 14) & 0x1)
+#define   C_00000D_RB_MAX_DYN_STOP_LAT                 0xFFFFBFFF
+#define   S_00000D_FORCE_DISP2(x)                      (((x) & 0x1) << 15)
+#define   G_00000D_FORCE_DISP2(x)                      (((x) >> 15) & 0x1)
+#define   C_00000D_FORCE_DISP2                         0xFFFF7FFF
+#define   S_00000D_FORCE_CP(x)                         (((x) & 0x1) << 16)
+#define   G_00000D_FORCE_CP(x)                         (((x) >> 16) & 0x1)
+#define   C_00000D_FORCE_CP                            0xFFFEFFFF
+#define   S_00000D_FORCE_HDP(x)                        (((x) & 0x1) << 17)
+#define   G_00000D_FORCE_HDP(x)                        (((x) >> 17) & 0x1)
+#define   C_00000D_FORCE_HDP                           0xFFFDFFFF
+#define   S_00000D_FORCE_DISP1(x)                      (((x) & 0x1) << 18)
+#define   G_00000D_FORCE_DISP1(x)                      (((x) >> 18) & 0x1)
+#define   C_00000D_FORCE_DISP1                         0xFFFBFFFF
+#define   S_00000D_FORCE_TOP(x)                        (((x) & 0x1) << 19)
+#define   G_00000D_FORCE_TOP(x)                        (((x) >> 19) & 0x1)
+#define   C_00000D_FORCE_TOP                           0xFFF7FFFF
+#define   S_00000D_FORCE_E2(x)                         (((x) & 0x1) << 20)
+#define   G_00000D_FORCE_E2(x)                         (((x) >> 20) & 0x1)
+#define   C_00000D_FORCE_E2                            0xFFEFFFFF
+#define   S_00000D_FORCE_SE(x)                         (((x) & 0x1) << 21)
+#define   G_00000D_FORCE_SE(x)                         (((x) >> 21) & 0x1)
+#define   C_00000D_FORCE_SE                            0xFFDFFFFF
+#define   S_00000D_FORCE_IDCT(x)                       (((x) & 0x1) << 22)
+#define   G_00000D_FORCE_IDCT(x)                       (((x) >> 22) & 0x1)
+#define   C_00000D_FORCE_IDCT                          0xFFBFFFFF
+#define   S_00000D_FORCE_VIP(x)                        (((x) & 0x1) << 23)
+#define   G_00000D_FORCE_VIP(x)                        (((x) >> 23) & 0x1)
+#define   C_00000D_FORCE_VIP                           0xFF7FFFFF
+#define   S_00000D_FORCE_RE(x)                         (((x) & 0x1) << 24)
+#define   G_00000D_FORCE_RE(x)                         (((x) >> 24) & 0x1)
+#define   C_00000D_FORCE_RE                            0xFEFFFFFF
+#define   S_00000D_FORCE_PB(x)                         (((x) & 0x1) << 25)
+#define   G_00000D_FORCE_PB(x)                         (((x) >> 25) & 0x1)
+#define   C_00000D_FORCE_PB                            0xFDFFFFFF
+#define   S_00000D_FORCE_TAM(x)                        (((x) & 0x1) << 26)
+#define   G_00000D_FORCE_TAM(x)                        (((x) >> 26) & 0x1)
+#define   C_00000D_FORCE_TAM                           0xFBFFFFFF
+#define   S_00000D_FORCE_TDM(x)                        (((x) & 0x1) << 27)
+#define   G_00000D_FORCE_TDM(x)                        (((x) >> 27) & 0x1)
+#define   C_00000D_FORCE_TDM                           0xF7FFFFFF
+#define   S_00000D_FORCE_RB(x)                         (((x) & 0x1) << 28)
+#define   G_00000D_FORCE_RB(x)                         (((x) >> 28) & 0x1)
+#define   C_00000D_FORCE_RB                            0xEFFFFFFF
+#define   S_00000D_FORCE_TV_SCLK(x)                    (((x) & 0x1) << 29)
+#define   G_00000D_FORCE_TV_SCLK(x)                    (((x) >> 29) & 0x1)
+#define   C_00000D_FORCE_TV_SCLK                       0xDFFFFFFF
+#define   S_00000D_FORCE_SUBPIC(x)                     (((x) & 0x1) << 30)
+#define   G_00000D_FORCE_SUBPIC(x)                     (((x) >> 30) & 0x1)
+#define   C_00000D_FORCE_SUBPIC                        0xBFFFFFFF
+#define   S_00000D_FORCE_OV0(x)                        (((x) & 0x1) << 31)
+#define   G_00000D_FORCE_OV0(x)                        (((x) >> 31) & 0x1)
+#define   C_00000D_FORCE_OV0                           0x7FFFFFFF
+
 #endif
diff --git a/drivers/gpu/drm/radeon/r420.c b/drivers/gpu/drm/radeon/r420.c
index 49a2fdc57d27..c7593b8f58ee 100644
--- a/drivers/gpu/drm/radeon/r420.c
+++ b/drivers/gpu/drm/radeon/r420.c
@@ -30,30 +30,14 @@
 #include "radeon_reg.h"
 #include "radeon.h"
 #include "atom.h"
+#include "r100d.h"
 #include "r420d.h"
+#include "r420_reg_safe.h"
 
-int r420_mc_init(struct radeon_device *rdev)
+static void r420_set_reg_safe(struct radeon_device *rdev)
 {
-	int r;
-
-	/* Setup GPU memory space */
-	rdev->mc.vram_location = 0xFFFFFFFFUL;
-	rdev->mc.gtt_location = 0xFFFFFFFFUL;
-	if (rdev->flags & RADEON_IS_AGP) {
-		r = radeon_agp_init(rdev);
-		if (r) {
-			printk(KERN_WARNING "[drm] Disabling AGP\n");
-			rdev->flags &= ~RADEON_IS_AGP;
-			rdev->mc.gtt_size = radeon_gart_size * 1024 * 1024;
-		} else {
-			rdev->mc.gtt_location = rdev->mc.agp_base;
-		}
-	}
-	r = radeon_mc_setup(rdev);
-	if (r) {
-		return r;
-	}
-	return 0;
+	rdev->config.r300.reg_safe_bm = r420_reg_safe_bm;
+	rdev->config.r300.reg_safe_bm_size = ARRAY_SIZE(r420_reg_safe_bm);
 }
 
 void r420_pipes_init(struct radeon_device *rdev)
@@ -63,7 +47,8 @@ void r420_pipes_init(struct radeon_device *rdev)
 	unsigned num_pipes;
 
 	/* GA_ENHANCE workaround TCL deadlock issue */
-	WREG32(0x4274, (1 << 0) | (1 << 1) | (1 << 2) | (1 << 3));
+	WREG32(R300_GA_ENHANCE, R300_GA_DEADLOCK_CNTL | R300_GA_FASTSYNC_CNTL |
+	       (1 << 2) | (1 << 3));
 	/* add idle wait as per freedesktop.org bug 24041 */
 	if (r100_gui_wait_for_idle(rdev)) {
 		printk(KERN_WARNING "Failed to wait GUI idle while "
@@ -91,17 +76,17 @@ void r420_pipes_init(struct radeon_device *rdev)
 		tmp = (7 << 1);
 		break;
 	}
-	WREG32(0x42C8, (1 << num_pipes) - 1);
+	WREG32(R500_SU_REG_DEST, (1 << num_pipes) - 1);
 	/* Sub pixel 1/12 so we can have 4K rendering according to doc */
-	tmp |= (1 << 4) | (1 << 0);
-	WREG32(0x4018, tmp);
+	tmp |= R300_TILE_SIZE_16 | R300_ENABLE_TILING;
+	WREG32(R300_GB_TILE_CONFIG, tmp);
 	if (r100_gui_wait_for_idle(rdev)) {
 		printk(KERN_WARNING "Failed to wait GUI idle while "
 		       "programming pipes. Bad things might happen.\n");
 	}
 
-	tmp = RREG32(0x170C);
-	WREG32(0x170C, tmp | (1 << 31));
+	tmp = RREG32(R300_DST_PIPE_CONFIG);
+	WREG32(R300_DST_PIPE_CONFIG, tmp | R300_PIPE_AUTO_CONFIG);
 
 	WREG32(R300_RB2D_DSTCACHE_MODE,
 	       RREG32(R300_RB2D_DSTCACHE_MODE) |
@@ -155,6 +140,9 @@ static void r420_debugfs(struct radeon_device *rdev)
 static void r420_clock_resume(struct radeon_device *rdev)
 {
 	u32 sclk_cntl;
+
+	if (radeon_dynclks != -1 && radeon_dynclks)
+		radeon_atom_set_clock_gating(rdev, 1);
 	sclk_cntl = RREG32_PLL(R_00000D_SCLK_CNTL);
 	sclk_cntl |= S_00000D_FORCE_CP(1) | S_00000D_FORCE_VIP(1);
 	if (rdev->family == CHIP_R420)
@@ -162,11 +150,44 @@ static void r420_clock_resume(struct radeon_device *rdev)
 	WREG32_PLL(R_00000D_SCLK_CNTL, sclk_cntl);
 }
 
+static void r420_cp_errata_init(struct radeon_device *rdev)
+{
+	/* RV410 and R420 can lock up if CP DMA to host memory happens
+	 * while the 2D engine is busy.
+	 *
+	 * The proper workaround is to queue a RESYNC at the beginning
+	 * of the CP init, apparently.
+	 */
+	radeon_scratch_get(rdev, &rdev->config.r300.resync_scratch);
+	radeon_ring_lock(rdev, 8);
+	radeon_ring_write(rdev, PACKET0(R300_CP_RESYNC_ADDR, 1));
+	radeon_ring_write(rdev, rdev->config.r300.resync_scratch);
+	radeon_ring_write(rdev, 0xDEADBEEF);
+	radeon_ring_unlock_commit(rdev);
+}
+
+static void r420_cp_errata_fini(struct radeon_device *rdev)
+{
+	/* Catch the RESYNC we dispatched all the way back,
+	 * at the very beginning of the CP init.
+	 */
+	radeon_ring_lock(rdev, 8);
+	radeon_ring_write(rdev, PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));
+	radeon_ring_write(rdev, R300_RB3D_DC_FINISH);
+	radeon_ring_unlock_commit(rdev);
+	radeon_scratch_free(rdev, rdev->config.r300.resync_scratch);
+}
+
 static int r420_startup(struct radeon_device *rdev)
 {
 	int r;
 
+	/* set common regs */
+	r100_set_common_regs(rdev);
+	/* program mc */
 	r300_mc_program(rdev);
+	/* Resume clock */
+	r420_clock_resume(rdev);
 	/* Initialize GART (initialize after TTM so we can allocate
 	 * memory through TTM but finalize after TTM) */
 	if (rdev->flags & RADEON_IS_PCIE) {
@@ -181,14 +202,15 @@ static int r420_startup(struct radeon_device *rdev)
 	}
 	r420_pipes_init(rdev);
 	/* Enable IRQ */
-	rdev->irq.sw_int = true;
 	r100_irq_set(rdev);
+	rdev->config.r300.hdp_cntl = RREG32(RADEON_HOST_PATH_CNTL);
 	/* 1M ring buffer */
 	r = r100_cp_init(rdev, 1024 * 1024);
 	if (r) {
 		dev_err(rdev->dev, "failled initializing CP (%d).\n", r);
 		return r;
 	}
+	r420_cp_errata_init(rdev);
 	r = r100_wb_init(rdev);
 	if (r) {
 		dev_err(rdev->dev, "failled initializing WB (%d).\n", r);
@@ -224,12 +246,14 @@ int r420_resume(struct radeon_device *rdev)
 	}
 	/* Resume clock after posting */
 	r420_clock_resume(rdev);
-
+	/* Initialize surface registers */
+	radeon_surface_init(rdev);
 	return r420_startup(rdev);
 }
 
 int r420_suspend(struct radeon_device *rdev)
 {
+	r420_cp_errata_fini(rdev);
 	r100_cp_disable(rdev);
 	r100_wb_disable(rdev);
 	r100_irq_disable(rdev);
@@ -253,7 +277,7 @@ void r420_fini(struct radeon_device *rdev)
 	radeon_agp_fini(rdev);
 	radeon_irq_kms_fini(rdev);
 	radeon_fence_driver_fini(rdev);
-	radeon_object_fini(rdev);
+	radeon_bo_fini(rdev);
 	if (rdev->is_atom_bios) {
 		radeon_atombios_fini(rdev);
 	} else {
@@ -267,7 +291,6 @@ int r420_init(struct radeon_device *rdev)
 {
 	int r;
 
-	rdev->new_init_path = true;
 	/* Initialize scratch registers */
 	radeon_scratch_init(rdev);
 	/* Initialize surface registers */
@@ -297,23 +320,22 @@ int r420_init(struct radeon_device *rdev)
 			RREG32(R_0007C0_CP_STAT));
 	}
 	/* check if cards are posted or not */
-	if (!radeon_card_posted(rdev) && rdev->bios) {
-		DRM_INFO("GPU not posted. posting now...\n");
-		if (rdev->is_atom_bios) {
-			atom_asic_init(rdev->mode_info.atom_context);
-		} else {
-			radeon_combios_asic_init(rdev->ddev);
-		}
-	}
+	if (radeon_boot_test_post_card(rdev) == false)
+		return -EINVAL;
+
 	/* Initialize clocks */
 	radeon_get_clock_info(rdev->ddev);
-	/* Get vram informations */
-	r300_vram_info(rdev);
-	/* Initialize memory controller (also test AGP) */
-	r = r420_mc_init(rdev);
-	if (r) {
-		return r;
+	/* Initialize power management */
+	radeon_pm_init(rdev);
+	/* initialize AGP */
+	if (rdev->flags & RADEON_IS_AGP) {
+		r = radeon_agp_init(rdev);
+		if (r) {
+			radeon_agp_disable(rdev);
+		}
 	}
+	/* initialize memory controller */
+	r300_mc_init(rdev);
 	r420_debugfs(rdev);
 	/* Fence driver */
 	r = radeon_fence_driver_init(rdev);
@@ -325,10 +347,13 @@ int r420_init(struct radeon_device *rdev)
 		return r;
 	}
 	/* Memory manager */
-	r = radeon_object_init(rdev);
+	r = radeon_bo_init(rdev);
 	if (r) {
 		return r;
 	}
+	if (rdev->family == CHIP_R420)
+		r100_enable_bm(rdev);
+
 	if (rdev->flags & RADEON_IS_PCIE) {
 		r = rv370_pcie_gart_init(rdev);
 		if (r)
@@ -339,22 +364,21 @@ int r420_init(struct radeon_device *rdev)
 		if (r)
 			return r;
 	}
-	r300_set_reg_safe(rdev);
+	r420_set_reg_safe(rdev);
 	rdev->accel_working = true;
 	r = r420_startup(rdev);
 	if (r) {
 		/* Somethings want wront with the accel init stop accel */
 		dev_err(rdev->dev, "Disabling GPU acceleration\n");
-		r420_suspend(rdev);
 		r100_cp_fini(rdev);
 		r100_wb_fini(rdev);
 		r100_ib_fini(rdev);
+		radeon_irq_kms_fini(rdev);
 		if (rdev->flags & RADEON_IS_PCIE)
 			rv370_pcie_gart_fini(rdev);
 		if (rdev->flags & RADEON_IS_PCI)
 			r100_pci_gart_fini(rdev);
 		radeon_agp_fini(rdev);
-		radeon_irq_kms_fini(rdev);
 		rdev->accel_working = false;
 	}
 	return 0;
diff --git a/drivers/gpu/drm/radeon/r420d.h b/drivers/gpu/drm/radeon/r420d.h
index a48a7db1e2aa..fc78d31a0b4a 100644
--- a/drivers/gpu/drm/radeon/r420d.h
+++ b/drivers/gpu/drm/radeon/r420d.h
@@ -212,9 +212,9 @@
 #define   S_00000D_FORCE_E2(x)                         (((x) & 0x1) << 20)
 #define   G_00000D_FORCE_E2(x)                         (((x) >> 20) & 0x1)
 #define   C_00000D_FORCE_E2                            0xFFEFFFFF
-#define   S_00000D_FORCE_SE(x)                         (((x) & 0x1) << 21)
-#define   G_00000D_FORCE_SE(x)                         (((x) >> 21) & 0x1)
-#define   C_00000D_FORCE_SE                            0xFFDFFFFF
+#define   S_00000D_FORCE_VAP(x)                        (((x) & 0x1) << 21)
+#define   G_00000D_FORCE_VAP(x)                        (((x) >> 21) & 0x1)
+#define   C_00000D_FORCE_VAP                           0xFFDFFFFF
 #define   S_00000D_FORCE_IDCT(x)                       (((x) & 0x1) << 22)
 #define   G_00000D_FORCE_IDCT(x)                       (((x) >> 22) & 0x1)
 #define   C_00000D_FORCE_IDCT                          0xFFBFFFFF
@@ -224,24 +224,24 @@
 #define   S_00000D_FORCE_RE(x)                         (((x) & 0x1) << 24)
 #define   G_00000D_FORCE_RE(x)                         (((x) >> 24) & 0x1)
 #define   C_00000D_FORCE_RE                            0xFEFFFFFF
-#define   S_00000D_FORCE_PB(x)                         (((x) & 0x1) << 25)
-#define   G_00000D_FORCE_PB(x)                         (((x) >> 25) & 0x1)
-#define   C_00000D_FORCE_PB                            0xFDFFFFFF
+#define   S_00000D_FORCE_SR(x)                         (((x) & 0x1) << 25)
+#define   G_00000D_FORCE_SR(x)                         (((x) >> 25) & 0x1)
+#define   C_00000D_FORCE_SR                            0xFDFFFFFF
 #define   S_00000D_FORCE_PX(x)                         (((x) & 0x1) << 26)
 #define   G_00000D_FORCE_PX(x)                         (((x) >> 26) & 0x1)
 #define   C_00000D_FORCE_PX                            0xFBFFFFFF
 #define   S_00000D_FORCE_TX(x)                         (((x) & 0x1) << 27)
 #define   G_00000D_FORCE_TX(x)                         (((x) >> 27) & 0x1)
 #define   C_00000D_FORCE_TX                            0xF7FFFFFF
-#define   S_00000D_FORCE_RB(x)                         (((x) & 0x1) << 28)
-#define   G_00000D_FORCE_RB(x)                         (((x) >> 28) & 0x1)
-#define   C_00000D_FORCE_RB                            0xEFFFFFFF
+#define   S_00000D_FORCE_US(x)                         (((x) & 0x1) << 28)
+#define   G_00000D_FORCE_US(x)                         (((x) >> 28) & 0x1)
+#define   C_00000D_FORCE_US                            0xEFFFFFFF
 #define   S_00000D_FORCE_TV_SCLK(x)                    (((x) & 0x1) << 29)
 #define   G_00000D_FORCE_TV_SCLK(x)                    (((x) >> 29) & 0x1)
 #define   C_00000D_FORCE_TV_SCLK                       0xDFFFFFFF
-#define   S_00000D_FORCE_SUBPIC(x)                     (((x) & 0x1) << 30)
-#define   G_00000D_FORCE_SUBPIC(x)                     (((x) >> 30) & 0x1)
-#define   C_00000D_FORCE_SUBPIC                        0xBFFFFFFF
+#define   S_00000D_FORCE_SU(x)                         (((x) & 0x1) << 30)
+#define   G_00000D_FORCE_SU(x)                         (((x) >> 30) & 0x1)
+#define   C_00000D_FORCE_SU                            0xBFFFFFFF
 #define   S_00000D_FORCE_OV0(x)                        (((x) & 0x1) << 31)
 #define   G_00000D_FORCE_OV0(x)                        (((x) >> 31) & 0x1)
 #define   C_00000D_FORCE_OV0                           0x7FFFFFFF
diff --git a/drivers/gpu/drm/radeon/r500_reg.h b/drivers/gpu/drm/radeon/r500_reg.h
index e1d5e0331e19..0cf2ad2a5585 100644
--- a/drivers/gpu/drm/radeon/r500_reg.h
+++ b/drivers/gpu/drm/radeon/r500_reg.h
@@ -384,9 +384,16 @@
 #       define AVIVO_D1GRPH_TILED                               (1 << 20)
 #       define AVIVO_D1GRPH_MACRO_ADDRESS_MODE                  (1 << 21)
 
+/* The R7xx *_HIGH surface regs are backwards; the D1 regs are in the D2
+ * block and vice versa.  This applies to GRPH, CUR, etc.
+ */
 #define AVIVO_D1GRPH_LUT_SEL                                    0x6108
 #define AVIVO_D1GRPH_PRIMARY_SURFACE_ADDRESS                    0x6110
+#define R700_D1GRPH_PRIMARY_SURFACE_ADDRESS_HIGH                0x6914
+#define R700_D2GRPH_PRIMARY_SURFACE_ADDRESS_HIGH                0x6114
 #define AVIVO_D1GRPH_SECONDARY_SURFACE_ADDRESS                  0x6118
+#define R700_D1GRPH_SECONDARY_SURFACE_ADDRESS_HIGH              0x691c
+#define R700_D2GRPH_SECONDARY_SURFACE_ADDRESS_HIGH              0x611c
 #define AVIVO_D1GRPH_PITCH                                      0x6120
 #define AVIVO_D1GRPH_SURFACE_OFFSET_X                           0x6124
 #define AVIVO_D1GRPH_SURFACE_OFFSET_Y                           0x6128
@@ -404,6 +411,8 @@
 #       define AVIVO_D1CURSOR_MODE_MASK         (3 << 8)
 #       define AVIVO_D1CURSOR_MODE_24BPP        2
 #define AVIVO_D1CUR_SURFACE_ADDRESS             0x6408
+#define R700_D1CUR_SURFACE_ADDRESS_HIGH         0x6c0c
+#define R700_D2CUR_SURFACE_ADDRESS_HIGH         0x640c
 #define AVIVO_D1CUR_SIZE                        0x6410
 #define AVIVO_D1CUR_POSITION                    0x6414
 #define AVIVO_D1CUR_HOT_SPOT                    0x6418
@@ -445,6 +454,8 @@
 #define AVIVO_D1MODE_VBLANK_STATUS              0x6534
 #       define AVIVO_VBLANK_ACK                 (1 << 4)
 #define AVIVO_D1MODE_VLINE_START_END            0x6538
+#define AVIVO_D1MODE_VLINE_STATUS               0x653c
+#       define AVIVO_D1MODE_VLINE_STAT          (1 << 12)
 #define AVIVO_DxMODE_INT_MASK                   0x6540
 #       define AVIVO_D1MODE_INT_MASK            (1 << 0)
 #       define AVIVO_D2MODE_INT_MASK            (1 << 8)
@@ -502,6 +513,7 @@
 
 #define AVIVO_D2MODE_VBLANK_STATUS              0x6d34
 #define AVIVO_D2MODE_VLINE_START_END            0x6d38
+#define AVIVO_D2MODE_VLINE_STATUS               0x6d3c
 #define AVIVO_D2MODE_VIEWPORT_START             0x6d80
 #define AVIVO_D2MODE_VIEWPORT_SIZE              0x6d84
 #define AVIVO_D2MODE_EXT_OVERSCAN_LEFT_RIGHT    0x6d88
@@ -704,53 +716,63 @@
 
 #define AVIVO_DVOA_BIT_DEPTH_CONTROL			0x7988
 
-#define AVIVO_GPIO_0                        0x7e30
-#define AVIVO_GPIO_1                        0x7e40
-#define AVIVO_GPIO_2                        0x7e50
-#define AVIVO_GPIO_3                        0x7e60
-
+#define AVIVO_DC_GPIO_HPD_A                 0x7e94
 #define AVIVO_DC_GPIO_HPD_Y                 0x7e9c
 
-#define AVIVO_I2C_STATUS					0x7d30
-#	define AVIVO_I2C_STATUS_DONE				(1 << 0)
-#	define AVIVO_I2C_STATUS_NACK				(1 << 1)
-#	define AVIVO_I2C_STATUS_HALT				(1 << 2)
-#	define AVIVO_I2C_STATUS_GO				(1 << 3)
-#	define AVIVO_I2C_STATUS_MASK				0x7
-/* If radeon_mm_i2c is to be believed, this is HALT, NACK, and maybe
- * DONE? */
-#	define AVIVO_I2C_STATUS_CMD_RESET			0x7
-#	define AVIVO_I2C_STATUS_CMD_WAIT			(1 << 3)
-#define AVIVO_I2C_STOP						0x7d34
-#define AVIVO_I2C_START_CNTL				0x7d38
-#	define AVIVO_I2C_START						(1 << 8)
-#	define AVIVO_I2C_CONNECTOR0					(0 << 16)
-#	define AVIVO_I2C_CONNECTOR1					(1 << 16)
-#define R520_I2C_START (1<<0)
-#define R520_I2C_STOP (1<<1)
-#define R520_I2C_RX (1<<2)
-#define R520_I2C_EN (1<<8)
-#define R520_I2C_DDC1 (0<<16)
-#define R520_I2C_DDC2 (1<<16)
-#define R520_I2C_DDC3 (2<<16)
-#define R520_I2C_DDC_MASK (3<<16)
-#define AVIVO_I2C_CONTROL2					0x7d3c
-#	define AVIVO_I2C_7D3C_SIZE_SHIFT			8
-#	define AVIVO_I2C_7D3C_SIZE_MASK				(0xf << 8)
-#define AVIVO_I2C_CONTROL3						0x7d40
-/* Reading is done 4 bytes at a time: read the bottom 8 bits from
- * 7d44, four times in a row.
- * Writing is a little more complex.  First write DATA with
- * 0xnnnnnnzz, then 0xnnnnnnyy, where nnnnnn is some non-deterministic
- * magic number, zz is, I think, the slave address, and yy is the byte
- * you want to write. */
-#define AVIVO_I2C_DATA						0x7d44
-#define R520_I2C_ADDR_COUNT_MASK (0x7)
-#define R520_I2C_DATA_COUNT_SHIFT (8)
-#define R520_I2C_DATA_COUNT_MASK (0xF00)
-#define AVIVO_I2C_CNTL						0x7d50
-#	define AVIVO_I2C_EN							(1 << 0)
-#	define AVIVO_I2C_RESET						(1 << 8)
+#define AVIVO_DC_I2C_STATUS1				0x7d30
+#	define AVIVO_DC_I2C_DONE			(1 << 0)
+#	define AVIVO_DC_I2C_NACK			(1 << 1)
+#	define AVIVO_DC_I2C_HALT			(1 << 2)
+#	define AVIVO_DC_I2C_GO			        (1 << 3)
+#define AVIVO_DC_I2C_RESET 				0x7d34
+#	define AVIVO_DC_I2C_SOFT_RESET			(1 << 0)
+#	define AVIVO_DC_I2C_ABORT			(1 << 8)
+#define AVIVO_DC_I2C_CONTROL1 				0x7d38
+#	define AVIVO_DC_I2C_START			(1 << 0)
+#	define AVIVO_DC_I2C_STOP			(1 << 1)
+#	define AVIVO_DC_I2C_RECEIVE			(1 << 2)
+#	define AVIVO_DC_I2C_EN			        (1 << 8)
+#	define AVIVO_DC_I2C_PIN_SELECT(x)		((x) << 16)
+#	define AVIVO_SEL_DDC1			        0
+#	define AVIVO_SEL_DDC2			        1
+#	define AVIVO_SEL_DDC3			        2
+#define AVIVO_DC_I2C_CONTROL2 				0x7d3c
+#	define AVIVO_DC_I2C_ADDR_COUNT(x)		((x) << 0)
+#	define AVIVO_DC_I2C_DATA_COUNT(x)		((x) << 8)
+#define AVIVO_DC_I2C_CONTROL3 				0x7d40
+#	define AVIVO_DC_I2C_DATA_DRIVE_EN		(1 << 0)
+#	define AVIVO_DC_I2C_DATA_DRIVE_SEL		(1 << 1)
+#	define AVIVO_DC_I2C_CLK_DRIVE_EN		(1 << 7)
+#	define AVIVO_DC_I2C_RD_INTRA_BYTE_DELAY(x)      ((x) << 8)
+#	define AVIVO_DC_I2C_WR_INTRA_BYTE_DELAY(x)	((x) << 16)
+#	define AVIVO_DC_I2C_TIME_LIMIT(x)		((x) << 24)
+#define AVIVO_DC_I2C_DATA 				0x7d44
+#define AVIVO_DC_I2C_INTERRUPT_CONTROL 			0x7d48
+#	define AVIVO_DC_I2C_INTERRUPT_STATUS		(1 << 0)
+#	define AVIVO_DC_I2C_INTERRUPT_AK		(1 << 8)
+#	define AVIVO_DC_I2C_INTERRUPT_ENABLE		(1 << 16)
+#define AVIVO_DC_I2C_ARBITRATION 			0x7d50
+#	define AVIVO_DC_I2C_SW_WANTS_TO_USE_I2C		(1 << 0)
+#	define AVIVO_DC_I2C_SW_CAN_USE_I2C		(1 << 1)
+#	define AVIVO_DC_I2C_SW_DONE_USING_I2C		(1 << 8)
+#	define AVIVO_DC_I2C_HW_NEEDS_I2C		(1 << 9)
+#	define AVIVO_DC_I2C_ABORT_HDCP_I2C		(1 << 16)
+#	define AVIVO_DC_I2C_HW_USING_I2C		(1 << 17)
+
+#define AVIVO_DC_GPIO_DDC1_MASK 		        0x7e40
+#define AVIVO_DC_GPIO_DDC1_A 		                0x7e44
+#define AVIVO_DC_GPIO_DDC1_EN 		                0x7e48
+#define AVIVO_DC_GPIO_DDC1_Y 		                0x7e4c
+
+#define AVIVO_DC_GPIO_DDC2_MASK 		        0x7e50
+#define AVIVO_DC_GPIO_DDC2_A 		                0x7e54
+#define AVIVO_DC_GPIO_DDC2_EN 		                0x7e58
+#define AVIVO_DC_GPIO_DDC2_Y 		                0x7e5c
+
+#define AVIVO_DC_GPIO_DDC3_MASK 		        0x7e60
+#define AVIVO_DC_GPIO_DDC3_A 		                0x7e64
+#define AVIVO_DC_GPIO_DDC3_EN 		                0x7e68
+#define AVIVO_DC_GPIO_DDC3_Y 		                0x7e6c
 
 #define AVIVO_DISP_INTERRUPT_STATUS                             0x7edc
 #       define AVIVO_D1_VBLANK_INTERRUPT                        (1 << 4)
diff --git a/drivers/gpu/drm/radeon/r520.c b/drivers/gpu/drm/radeon/r520.c
index d4b0b9d2e39b..2b8a5dd13516 100644
--- a/drivers/gpu/drm/radeon/r520.c
+++ b/drivers/gpu/drm/radeon/r520.c
@@ -26,108 +26,13 @@
  *          Jerome Glisse
  */
 #include "drmP.h"
-#include "radeon_reg.h"
 #include "radeon.h"
+#include "atom.h"
+#include "r520d.h"
 
-/* r520,rv530,rv560,rv570,r580 depends on : */
-void r100_hdp_reset(struct radeon_device *rdev);
-void r420_pipes_init(struct radeon_device *rdev);
-void rs600_mc_disable_clients(struct radeon_device *rdev);
-void rs600_disable_vga(struct radeon_device *rdev);
-int rv515_debugfs_pipes_info_init(struct radeon_device *rdev);
-int rv515_debugfs_ga_info_init(struct radeon_device *rdev);
+/* This files gather functions specifics to: r520,rv530,rv560,rv570,r580 */
 
-/* This files gather functions specifics to:
- * r520,rv530,rv560,rv570,r580
- *
- * Some of these functions might be used by newer ASICs.
- */
-void r520_gpu_init(struct radeon_device *rdev);
-int r520_mc_wait_for_idle(struct radeon_device *rdev);
-
-
-/*
- * MC
- */
-int r520_mc_init(struct radeon_device *rdev)
-{
-	uint32_t tmp;
-	int r;
-
-	if (r100_debugfs_rbbm_init(rdev)) {
-		DRM_ERROR("Failed to register debugfs file for RBBM !\n");
-	}
-	if (rv515_debugfs_pipes_info_init(rdev)) {
-		DRM_ERROR("Failed to register debugfs file for pipes !\n");
-	}
-	if (rv515_debugfs_ga_info_init(rdev)) {
-		DRM_ERROR("Failed to register debugfs file for pipes !\n");
-	}
-
-	r520_gpu_init(rdev);
-	rv370_pcie_gart_disable(rdev);
-
-	/* Setup GPU memory space */
-	rdev->mc.vram_location = 0xFFFFFFFFUL;
-	rdev->mc.gtt_location = 0xFFFFFFFFUL;
-	if (rdev->flags & RADEON_IS_AGP) {
-		r = radeon_agp_init(rdev);
-		if (r) {
-			printk(KERN_WARNING "[drm] Disabling AGP\n");
-			rdev->flags &= ~RADEON_IS_AGP;
-			rdev->mc.gtt_size = radeon_gart_size * 1024 * 1024;
-		} else {
-			rdev->mc.gtt_location = rdev->mc.agp_base;
-		}
-	}
-	r = radeon_mc_setup(rdev);
-	if (r) {
-		return r;
-	}
-
-	/* Program GPU memory space */
-	rs600_mc_disable_clients(rdev);
-	if (r520_mc_wait_for_idle(rdev)) {
-		printk(KERN_WARNING "Failed to wait MC idle while "
-		       "programming pipes. Bad things might happen.\n");
-	}
-	/* Write VRAM size in case we are limiting it */
-	WREG32(RADEON_CONFIG_MEMSIZE, rdev->mc.real_vram_size);
-	tmp = rdev->mc.vram_location + rdev->mc.mc_vram_size - 1;
-	tmp = REG_SET(R520_MC_FB_TOP, tmp >> 16);
-	tmp |= REG_SET(R520_MC_FB_START, rdev->mc.vram_location >> 16);
-	WREG32_MC(R520_MC_FB_LOCATION, tmp);
-	WREG32(RS690_HDP_FB_LOCATION, rdev->mc.vram_location >> 16);
-	WREG32(0x310, rdev->mc.vram_location);
-	if (rdev->flags & RADEON_IS_AGP) {
-		tmp = rdev->mc.gtt_location + rdev->mc.gtt_size - 1;
-		tmp = REG_SET(R520_MC_AGP_TOP, tmp >> 16);
-		tmp |= REG_SET(R520_MC_AGP_START, rdev->mc.gtt_location >> 16);
-		WREG32_MC(R520_MC_AGP_LOCATION, tmp);
-		WREG32_MC(R520_MC_AGP_BASE, rdev->mc.agp_base);
-		WREG32_MC(R520_MC_AGP_BASE_2, 0);
-	} else {
-		WREG32_MC(R520_MC_AGP_LOCATION, 0x0FFFFFFF);
-		WREG32_MC(R520_MC_AGP_BASE, 0);
-		WREG32_MC(R520_MC_AGP_BASE_2, 0);
-	}
-	return 0;
-}
-
-void r520_mc_fini(struct radeon_device *rdev)
-{
-}
-
-
-/*
- * Global GPU functions
- */
-void r520_errata(struct radeon_device *rdev)
-{
-	rdev->pll_errata = 0;
-}
-
-int r520_mc_wait_for_idle(struct radeon_device *rdev)
+static int r520_mc_wait_for_idle(struct radeon_device *rdev)
 {
 	unsigned i;
 	uint32_t tmp;
@@ -143,12 +48,12 @@ int r520_mc_wait_for_idle(struct radeon_device *rdev)
 	return -1;
 }
 
-void r520_gpu_init(struct radeon_device *rdev)
+static void r520_gpu_init(struct radeon_device *rdev)
 {
 	unsigned pipe_select_current, gb_pipe_select, tmp;
 
 	r100_hdp_reset(rdev);
-	rs600_disable_vga(rdev);
+	rv515_vga_render_disable(rdev);
 	/*
 	 * DST_PIPE_CONFIG		0x170C
 	 * GB_TILE_CONFIG		0x4018
@@ -186,10 +91,6 @@ void r520_gpu_init(struct radeon_device *rdev)
 	}
 }
 
-
-/*
- * VRAM info
- */
 static void r520_vram_get_type(struct radeon_device *rdev)
 {
 	uint32_t tmp;
@@ -218,13 +119,15 @@ static void r520_vram_get_type(struct radeon_device *rdev)
 		rdev->mc.vram_width *= 2;
 }
 
-void r520_vram_info(struct radeon_device *rdev)
+void r520_mc_init(struct radeon_device *rdev)
 {
 	fixed20_12 a;
 
 	r520_vram_get_type(rdev);
-
 	r100_vram_init_sizes(rdev);
+	radeon_vram_location(rdev, &rdev->mc, 0);
+	if (!(rdev->flags & RADEON_IS_AGP))
+		radeon_gtt_location(rdev, &rdev->mc);
 	/* FIXME: we should enforce default clock in case GPU is not in
 	 * default setup
 	 */
@@ -233,7 +136,176 @@ void r520_vram_info(struct radeon_device *rdev)
 	rdev->pm.sclk.full = rfixed_div(rdev->pm.sclk, a);
 }
 
-void r520_bandwidth_update(struct radeon_device *rdev)
+void r520_mc_program(struct radeon_device *rdev)
 {
-	rv515_bandwidth_avivo_update(rdev);
+	struct rv515_mc_save save;
+
+	/* Stops all mc clients */
+	rv515_mc_stop(rdev, &save);
+
+	/* Wait for mc idle */
+	if (r520_mc_wait_for_idle(rdev))
+		dev_warn(rdev->dev, "Wait MC idle timeout before updating MC.\n");
+	/* Write VRAM size in case we are limiting it */
+	WREG32(R_0000F8_CONFIG_MEMSIZE, rdev->mc.real_vram_size);
+	/* Program MC, should be a 32bits limited address space */
+	WREG32_MC(R_000004_MC_FB_LOCATION,
+			S_000004_MC_FB_START(rdev->mc.vram_start >> 16) |
+			S_000004_MC_FB_TOP(rdev->mc.vram_end >> 16));
+	WREG32(R_000134_HDP_FB_LOCATION,
+		S_000134_HDP_FB_START(rdev->mc.vram_start >> 16));
+	if (rdev->flags & RADEON_IS_AGP) {
+		WREG32_MC(R_000005_MC_AGP_LOCATION,
+			S_000005_MC_AGP_START(rdev->mc.gtt_start >> 16) |
+			S_000005_MC_AGP_TOP(rdev->mc.gtt_end >> 16));
+		WREG32_MC(R_000006_AGP_BASE, lower_32_bits(rdev->mc.agp_base));
+		WREG32_MC(R_000007_AGP_BASE_2,
+			S_000007_AGP_BASE_ADDR_2(upper_32_bits(rdev->mc.agp_base)));
+	} else {
+		WREG32_MC(R_000005_MC_AGP_LOCATION, 0xFFFFFFFF);
+		WREG32_MC(R_000006_AGP_BASE, 0);
+		WREG32_MC(R_000007_AGP_BASE_2, 0);
+	}
+
+	rv515_mc_resume(rdev, &save);
+}
+
+static int r520_startup(struct radeon_device *rdev)
+{
+	int r;
+
+	r520_mc_program(rdev);
+	/* Resume clock */
+	rv515_clock_startup(rdev);
+	/* Initialize GPU configuration (# pipes, ...) */
+	r520_gpu_init(rdev);
+	/* Initialize GART (initialize after TTM so we can allocate
+	 * memory through TTM but finalize after TTM) */
+	if (rdev->flags & RADEON_IS_PCIE) {
+		r = rv370_pcie_gart_enable(rdev);
+		if (r)
+			return r;
+	}
+	/* Enable IRQ */
+	rs600_irq_set(rdev);
+	rdev->config.r300.hdp_cntl = RREG32(RADEON_HOST_PATH_CNTL);
+	/* 1M ring buffer */
+	r = r100_cp_init(rdev, 1024 * 1024);
+	if (r) {
+		dev_err(rdev->dev, "failled initializing CP (%d).\n", r);
+		return r;
+	}
+	r = r100_wb_init(rdev);
+	if (r)
+		dev_err(rdev->dev, "failled initializing WB (%d).\n", r);
+	r = r100_ib_init(rdev);
+	if (r) {
+		dev_err(rdev->dev, "failled initializing IB (%d).\n", r);
+		return r;
+	}
+	return 0;
+}
+
+int r520_resume(struct radeon_device *rdev)
+{
+	/* Make sur GART are not working */
+	if (rdev->flags & RADEON_IS_PCIE)
+		rv370_pcie_gart_disable(rdev);
+	/* Resume clock before doing reset */
+	rv515_clock_startup(rdev);
+	/* Reset gpu before posting otherwise ATOM will enter infinite loop */
+	if (radeon_gpu_reset(rdev)) {
+		dev_warn(rdev->dev, "GPU reset failed ! (0xE40=0x%08X, 0x7C0=0x%08X)\n",
+			RREG32(R_000E40_RBBM_STATUS),
+			RREG32(R_0007C0_CP_STAT));
+	}
+	/* post */
+	atom_asic_init(rdev->mode_info.atom_context);
+	/* Resume clock after posting */
+	rv515_clock_startup(rdev);
+	/* Initialize surface registers */
+	radeon_surface_init(rdev);
+	return r520_startup(rdev);
+}
+
+int r520_init(struct radeon_device *rdev)
+{
+	int r;
+
+	/* Initialize scratch registers */
+	radeon_scratch_init(rdev);
+	/* Initialize surface registers */
+	radeon_surface_init(rdev);
+	/* TODO: disable VGA need to use VGA request */
+	/* BIOS*/
+	if (!radeon_get_bios(rdev)) {
+		if (ASIC_IS_AVIVO(rdev))
+			return -EINVAL;
+	}
+	if (rdev->is_atom_bios) {
+		r = radeon_atombios_init(rdev);
+		if (r)
+			return r;
+	} else {
+		dev_err(rdev->dev, "Expecting atombios for RV515 GPU\n");
+		return -EINVAL;
+	}
+	/* Reset gpu before posting otherwise ATOM will enter infinite loop */
+	if (radeon_gpu_reset(rdev)) {
+		dev_warn(rdev->dev,
+			"GPU reset failed ! (0xE40=0x%08X, 0x7C0=0x%08X)\n",
+			RREG32(R_000E40_RBBM_STATUS),
+			RREG32(R_0007C0_CP_STAT));
+	}
+	/* check if cards are posted or not */
+	if (radeon_boot_test_post_card(rdev) == false)
+		return -EINVAL;
+
+	if (!radeon_card_posted(rdev) && rdev->bios) {
+		DRM_INFO("GPU not posted. posting now...\n");
+		atom_asic_init(rdev->mode_info.atom_context);
+	}
+	/* Initialize clocks */
+	radeon_get_clock_info(rdev->ddev);
+	/* Initialize power management */
+	radeon_pm_init(rdev);
+	/* initialize AGP */
+	if (rdev->flags & RADEON_IS_AGP) {
+		r = radeon_agp_init(rdev);
+		if (r) {
+			radeon_agp_disable(rdev);
+		}
+	}
+	/* initialize memory controller */
+	r520_mc_init(rdev);
+	rv515_debugfs(rdev);
+	/* Fence driver */
+	r = radeon_fence_driver_init(rdev);
+	if (r)
+		return r;
+	r = radeon_irq_kms_init(rdev);
+	if (r)
+		return r;
+	/* Memory manager */
+	r = radeon_bo_init(rdev);
+	if (r)
+		return r;
+	r = rv370_pcie_gart_init(rdev);
+	if (r)
+		return r;
+	rv515_set_safe_registers(rdev);
+	rdev->accel_working = true;
+	r = r520_startup(rdev);
+	if (r) {
+		/* Somethings want wront with the accel init stop accel */
+		dev_err(rdev->dev, "Disabling GPU acceleration\n");
+		r100_cp_fini(rdev);
+		r100_wb_fini(rdev);
+		r100_ib_fini(rdev);
+		radeon_irq_kms_fini(rdev);
+		rv370_pcie_gart_fini(rdev);
+		radeon_agp_fini(rdev);
+		rdev->accel_working = false;
+	}
+	return 0;
 }
diff --git a/drivers/gpu/drm/radeon/r520d.h b/drivers/gpu/drm/radeon/r520d.h
new file mode 100644
index 000000000000..61af61f644bc
--- /dev/null
+++ b/drivers/gpu/drm/radeon/r520d.h
@@ -0,0 +1,187 @@
+/*
+ * Copyright 2008 Advanced Micro Devices, Inc.
+ * Copyright 2008 Red Hat Inc.
+ * Copyright 2009 Jerome Glisse.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Dave Airlie
+ *          Alex Deucher
+ *          Jerome Glisse
+ */
+#ifndef __R520D_H__
+#define __R520D_H__
+
+/* Registers */
+#define R_0000F8_CONFIG_MEMSIZE                      0x0000F8
+#define   S_0000F8_CONFIG_MEMSIZE(x)                   (((x) & 0xFFFFFFFF) << 0)
+#define   G_0000F8_CONFIG_MEMSIZE(x)                   (((x) >> 0) & 0xFFFFFFFF)
+#define   C_0000F8_CONFIG_MEMSIZE                      0x00000000
+#define R_000134_HDP_FB_LOCATION                     0x000134
+#define   S_000134_HDP_FB_START(x)                     (((x) & 0xFFFF) << 0)
+#define   G_000134_HDP_FB_START(x)                     (((x) >> 0) & 0xFFFF)
+#define   C_000134_HDP_FB_START                        0xFFFF0000
+#define R_0007C0_CP_STAT                             0x0007C0
+#define   S_0007C0_MRU_BUSY(x)                         (((x) & 0x1) << 0)
+#define   G_0007C0_MRU_BUSY(x)                         (((x) >> 0) & 0x1)
+#define   C_0007C0_MRU_BUSY                            0xFFFFFFFE
+#define   S_0007C0_MWU_BUSY(x)                         (((x) & 0x1) << 1)
+#define   G_0007C0_MWU_BUSY(x)                         (((x) >> 1) & 0x1)
+#define   C_0007C0_MWU_BUSY                            0xFFFFFFFD
+#define   S_0007C0_RSIU_BUSY(x)                        (((x) & 0x1) << 2)
+#define   G_0007C0_RSIU_BUSY(x)                        (((x) >> 2) & 0x1)
+#define   C_0007C0_RSIU_BUSY                           0xFFFFFFFB
+#define   S_0007C0_RCIU_BUSY(x)                        (((x) & 0x1) << 3)
+#define   G_0007C0_RCIU_BUSY(x)                        (((x) >> 3) & 0x1)
+#define   C_0007C0_RCIU_BUSY                           0xFFFFFFF7
+#define   S_0007C0_CSF_PRIMARY_BUSY(x)                 (((x) & 0x1) << 9)
+#define   G_0007C0_CSF_PRIMARY_BUSY(x)                 (((x) >> 9) & 0x1)
+#define   C_0007C0_CSF_PRIMARY_BUSY                    0xFFFFFDFF
+#define   S_0007C0_CSF_INDIRECT_BUSY(x)                (((x) & 0x1) << 10)
+#define   G_0007C0_CSF_INDIRECT_BUSY(x)                (((x) >> 10) & 0x1)
+#define   C_0007C0_CSF_INDIRECT_BUSY                   0xFFFFFBFF
+#define   S_0007C0_CSQ_PRIMARY_BUSY(x)                 (((x) & 0x1) << 11)
+#define   G_0007C0_CSQ_PRIMARY_BUSY(x)                 (((x) >> 11) & 0x1)
+#define   C_0007C0_CSQ_PRIMARY_BUSY                    0xFFFFF7FF
+#define   S_0007C0_CSQ_INDIRECT_BUSY(x)                (((x) & 0x1) << 12)
+#define   G_0007C0_CSQ_INDIRECT_BUSY(x)                (((x) >> 12) & 0x1)
+#define   C_0007C0_CSQ_INDIRECT_BUSY                   0xFFFFEFFF
+#define   S_0007C0_CSI_BUSY(x)                         (((x) & 0x1) << 13)
+#define   G_0007C0_CSI_BUSY(x)                         (((x) >> 13) & 0x1)
+#define   C_0007C0_CSI_BUSY                            0xFFFFDFFF
+#define   S_0007C0_CSF_INDIRECT2_BUSY(x)               (((x) & 0x1) << 14)
+#define   G_0007C0_CSF_INDIRECT2_BUSY(x)               (((x) >> 14) & 0x1)
+#define   C_0007C0_CSF_INDIRECT2_BUSY                  0xFFFFBFFF
+#define   S_0007C0_CSQ_INDIRECT2_BUSY(x)               (((x) & 0x1) << 15)
+#define   G_0007C0_CSQ_INDIRECT2_BUSY(x)               (((x) >> 15) & 0x1)
+#define   C_0007C0_CSQ_INDIRECT2_BUSY                  0xFFFF7FFF
+#define   S_0007C0_GUIDMA_BUSY(x)                      (((x) & 0x1) << 28)
+#define   G_0007C0_GUIDMA_BUSY(x)                      (((x) >> 28) & 0x1)
+#define   C_0007C0_GUIDMA_BUSY                         0xEFFFFFFF
+#define   S_0007C0_VIDDMA_BUSY(x)                      (((x) & 0x1) << 29)
+#define   G_0007C0_VIDDMA_BUSY(x)                      (((x) >> 29) & 0x1)
+#define   C_0007C0_VIDDMA_BUSY                         0xDFFFFFFF
+#define   S_0007C0_CMDSTRM_BUSY(x)                     (((x) & 0x1) << 30)
+#define   G_0007C0_CMDSTRM_BUSY(x)                     (((x) >> 30) & 0x1)
+#define   C_0007C0_CMDSTRM_BUSY                        0xBFFFFFFF
+#define   S_0007C0_CP_BUSY(x)                          (((x) & 0x1) << 31)
+#define   G_0007C0_CP_BUSY(x)                          (((x) >> 31) & 0x1)
+#define   C_0007C0_CP_BUSY                             0x7FFFFFFF
+#define R_000E40_RBBM_STATUS                         0x000E40
+#define   S_000E40_CMDFIFO_AVAIL(x)                    (((x) & 0x7F) << 0)
+#define   G_000E40_CMDFIFO_AVAIL(x)                    (((x) >> 0) & 0x7F)
+#define   C_000E40_CMDFIFO_AVAIL                       0xFFFFFF80
+#define   S_000E40_HIRQ_ON_RBB(x)                      (((x) & 0x1) << 8)
+#define   G_000E40_HIRQ_ON_RBB(x)                      (((x) >> 8) & 0x1)
+#define   C_000E40_HIRQ_ON_RBB                         0xFFFFFEFF
+#define   S_000E40_CPRQ_ON_RBB(x)                      (((x) & 0x1) << 9)
+#define   G_000E40_CPRQ_ON_RBB(x)                      (((x) >> 9) & 0x1)
+#define   C_000E40_CPRQ_ON_RBB                         0xFFFFFDFF
+#define   S_000E40_CFRQ_ON_RBB(x)                      (((x) & 0x1) << 10)
+#define   G_000E40_CFRQ_ON_RBB(x)                      (((x) >> 10) & 0x1)
+#define   C_000E40_CFRQ_ON_RBB                         0xFFFFFBFF
+#define   S_000E40_HIRQ_IN_RTBUF(x)                    (((x) & 0x1) << 11)
+#define   G_000E40_HIRQ_IN_RTBUF(x)                    (((x) >> 11) & 0x1)
+#define   C_000E40_HIRQ_IN_RTBUF                       0xFFFFF7FF
+#define   S_000E40_CPRQ_IN_RTBUF(x)                    (((x) & 0x1) << 12)
+#define   G_000E40_CPRQ_IN_RTBUF(x)                    (((x) >> 12) & 0x1)
+#define   C_000E40_CPRQ_IN_RTBUF                       0xFFFFEFFF
+#define   S_000E40_CFRQ_IN_RTBUF(x)                    (((x) & 0x1) << 13)
+#define   G_000E40_CFRQ_IN_RTBUF(x)                    (((x) >> 13) & 0x1)
+#define   C_000E40_CFRQ_IN_RTBUF                       0xFFFFDFFF
+#define   S_000E40_CF_PIPE_BUSY(x)                     (((x) & 0x1) << 14)
+#define   G_000E40_CF_PIPE_BUSY(x)                     (((x) >> 14) & 0x1)
+#define   C_000E40_CF_PIPE_BUSY                        0xFFFFBFFF
+#define   S_000E40_ENG_EV_BUSY(x)                      (((x) & 0x1) << 15)
+#define   G_000E40_ENG_EV_BUSY(x)                      (((x) >> 15) & 0x1)
+#define   C_000E40_ENG_EV_BUSY                         0xFFFF7FFF
+#define   S_000E40_CP_CMDSTRM_BUSY(x)                  (((x) & 0x1) << 16)
+#define   G_000E40_CP_CMDSTRM_BUSY(x)                  (((x) >> 16) & 0x1)
+#define   C_000E40_CP_CMDSTRM_BUSY                     0xFFFEFFFF
+#define   S_000E40_E2_BUSY(x)                          (((x) & 0x1) << 17)
+#define   G_000E40_E2_BUSY(x)                          (((x) >> 17) & 0x1)
+#define   C_000E40_E2_BUSY                             0xFFFDFFFF
+#define   S_000E40_RB2D_BUSY(x)                        (((x) & 0x1) << 18)
+#define   G_000E40_RB2D_BUSY(x)                        (((x) >> 18) & 0x1)
+#define   C_000E40_RB2D_BUSY                           0xFFFBFFFF
+#define   S_000E40_RB3D_BUSY(x)                        (((x) & 0x1) << 19)
+#define   G_000E40_RB3D_BUSY(x)                        (((x) >> 19) & 0x1)
+#define   C_000E40_RB3D_BUSY                           0xFFF7FFFF
+#define   S_000E40_VAP_BUSY(x)                         (((x) & 0x1) << 20)
+#define   G_000E40_VAP_BUSY(x)                         (((x) >> 20) & 0x1)
+#define   C_000E40_VAP_BUSY                            0xFFEFFFFF
+#define   S_000E40_RE_BUSY(x)                          (((x) & 0x1) << 21)
+#define   G_000E40_RE_BUSY(x)                          (((x) >> 21) & 0x1)
+#define   C_000E40_RE_BUSY                             0xFFDFFFFF
+#define   S_000E40_TAM_BUSY(x)                         (((x) & 0x1) << 22)
+#define   G_000E40_TAM_BUSY(x)                         (((x) >> 22) & 0x1)
+#define   C_000E40_TAM_BUSY                            0xFFBFFFFF
+#define   S_000E40_TDM_BUSY(x)                         (((x) & 0x1) << 23)
+#define   G_000E40_TDM_BUSY(x)                         (((x) >> 23) & 0x1)
+#define   C_000E40_TDM_BUSY                            0xFF7FFFFF
+#define   S_000E40_PB_BUSY(x)                          (((x) & 0x1) << 24)
+#define   G_000E40_PB_BUSY(x)                          (((x) >> 24) & 0x1)
+#define   C_000E40_PB_BUSY                             0xFEFFFFFF
+#define   S_000E40_TIM_BUSY(x)                         (((x) & 0x1) << 25)
+#define   G_000E40_TIM_BUSY(x)                         (((x) >> 25) & 0x1)
+#define   C_000E40_TIM_BUSY                            0xFDFFFFFF
+#define   S_000E40_GA_BUSY(x)                          (((x) & 0x1) << 26)
+#define   G_000E40_GA_BUSY(x)                          (((x) >> 26) & 0x1)
+#define   C_000E40_GA_BUSY                             0xFBFFFFFF
+#define   S_000E40_CBA2D_BUSY(x)                       (((x) & 0x1) << 27)
+#define   G_000E40_CBA2D_BUSY(x)                       (((x) >> 27) & 0x1)
+#define   C_000E40_CBA2D_BUSY                          0xF7FFFFFF
+#define   S_000E40_RBBM_HIBUSY(x)                      (((x) & 0x1) << 28)
+#define   G_000E40_RBBM_HIBUSY(x)                      (((x) >> 28) & 0x1)
+#define   C_000E40_RBBM_HIBUSY                         0xEFFFFFFF
+#define   S_000E40_SKID_CFBUSY(x)                      (((x) & 0x1) << 29)
+#define   G_000E40_SKID_CFBUSY(x)                      (((x) >> 29) & 0x1)
+#define   C_000E40_SKID_CFBUSY                         0xDFFFFFFF
+#define   S_000E40_VAP_VF_BUSY(x)                      (((x) & 0x1) << 30)
+#define   G_000E40_VAP_VF_BUSY(x)                      (((x) >> 30) & 0x1)
+#define   C_000E40_VAP_VF_BUSY                         0xBFFFFFFF
+#define   S_000E40_GUI_ACTIVE(x)                       (((x) & 0x1) << 31)
+#define   G_000E40_GUI_ACTIVE(x)                       (((x) >> 31) & 0x1)
+#define   C_000E40_GUI_ACTIVE                          0x7FFFFFFF
+
+
+#define R_000004_MC_FB_LOCATION                      0x000004
+#define   S_000004_MC_FB_START(x)                      (((x) & 0xFFFF) << 0)
+#define   G_000004_MC_FB_START(x)                      (((x) >> 0) & 0xFFFF)
+#define   C_000004_MC_FB_START                         0xFFFF0000
+#define   S_000004_MC_FB_TOP(x)                        (((x) & 0xFFFF) << 16)
+#define   G_000004_MC_FB_TOP(x)                        (((x) >> 16) & 0xFFFF)
+#define   C_000004_MC_FB_TOP                           0x0000FFFF
+#define R_000005_MC_AGP_LOCATION                     0x000005
+#define   S_000005_MC_AGP_START(x)                     (((x) & 0xFFFF) << 0)
+#define   G_000005_MC_AGP_START(x)                     (((x) >> 0) & 0xFFFF)
+#define   C_000005_MC_AGP_START                        0xFFFF0000
+#define   S_000005_MC_AGP_TOP(x)                       (((x) & 0xFFFF) << 16)
+#define   G_000005_MC_AGP_TOP(x)                       (((x) >> 16) & 0xFFFF)
+#define   C_000005_MC_AGP_TOP                          0x0000FFFF
+#define R_000006_AGP_BASE                            0x000006
+#define   S_000006_AGP_BASE_ADDR(x)                    (((x) & 0xFFFFFFFF) << 0)
+#define   G_000006_AGP_BASE_ADDR(x)                    (((x) >> 0) & 0xFFFFFFFF)
+#define   C_000006_AGP_BASE_ADDR                       0x00000000
+#define R_000007_AGP_BASE_2                          0x000007
+#define   S_000007_AGP_BASE_ADDR_2(x)                  (((x) & 0xF) << 0)
+#define   G_000007_AGP_BASE_ADDR_2(x)                  (((x) >> 0) & 0xF)
+#define   C_000007_AGP_BASE_ADDR_2                     0xFFFFFFF0
+
+#endif
diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c
index eab31c1d6df1..c52290197292 100644
--- a/drivers/gpu/drm/radeon/r600.c
+++ b/drivers/gpu/drm/radeon/r600.c
@@ -33,13 +33,15 @@
 #include "radeon.h"
 #include "radeon_mode.h"
 #include "r600d.h"
-#include "avivod.h"
 #include "atom.h"
+#include "avivod.h"
 
 #define PFP_UCODE_SIZE 576
 #define PM4_UCODE_SIZE 1792
+#define RLC_UCODE_SIZE 768
 #define R700_PFP_UCODE_SIZE 848
 #define R700_PM4_UCODE_SIZE 1360
+#define R700_RLC_UCODE_SIZE 1024
 
 /* Firmware Names */
 MODULE_FIRMWARE("radeon/R600_pfp.bin");
@@ -62,39 +64,303 @@ MODULE_FIRMWARE("radeon/RV730_pfp.bin");
 MODULE_FIRMWARE("radeon/RV730_me.bin");
 MODULE_FIRMWARE("radeon/RV710_pfp.bin");
 MODULE_FIRMWARE("radeon/RV710_me.bin");
+MODULE_FIRMWARE("radeon/R600_rlc.bin");
+MODULE_FIRMWARE("radeon/R700_rlc.bin");
 
 int r600_debugfs_mc_info_init(struct radeon_device *rdev);
 
-/* This files gather functions specifics to:
- * r600,rv610,rv630,rv620,rv635,rv670
- *
- * Some of these functions might be used by newer ASICs.
- */
+/* r600,rv610,rv630,rv620,rv635,rv670 */
 int r600_mc_wait_for_idle(struct radeon_device *rdev);
 void r600_gpu_init(struct radeon_device *rdev);
 void r600_fini(struct radeon_device *rdev);
 
+/* hpd for digital panel detect/disconnect */
+bool r600_hpd_sense(struct radeon_device *rdev, enum radeon_hpd_id hpd)
+{
+	bool connected = false;
 
-/*
- * R600 PCIE GART
- */
-int r600_gart_clear_page(struct radeon_device *rdev, int i)
+	if (ASIC_IS_DCE3(rdev)) {
+		switch (hpd) {
+		case RADEON_HPD_1:
+			if (RREG32(DC_HPD1_INT_STATUS) & DC_HPDx_SENSE)
+				connected = true;
+			break;
+		case RADEON_HPD_2:
+			if (RREG32(DC_HPD2_INT_STATUS) & DC_HPDx_SENSE)
+				connected = true;
+			break;
+		case RADEON_HPD_3:
+			if (RREG32(DC_HPD3_INT_STATUS) & DC_HPDx_SENSE)
+				connected = true;
+			break;
+		case RADEON_HPD_4:
+			if (RREG32(DC_HPD4_INT_STATUS) & DC_HPDx_SENSE)
+				connected = true;
+			break;
+			/* DCE 3.2 */
+		case RADEON_HPD_5:
+			if (RREG32(DC_HPD5_INT_STATUS) & DC_HPDx_SENSE)
+				connected = true;
+			break;
+		case RADEON_HPD_6:
+			if (RREG32(DC_HPD6_INT_STATUS) & DC_HPDx_SENSE)
+				connected = true;
+			break;
+		default:
+			break;
+		}
+	} else {
+		switch (hpd) {
+		case RADEON_HPD_1:
+			if (RREG32(DC_HOT_PLUG_DETECT1_INT_STATUS) & DC_HOT_PLUG_DETECTx_SENSE)
+				connected = true;
+			break;
+		case RADEON_HPD_2:
+			if (RREG32(DC_HOT_PLUG_DETECT2_INT_STATUS) & DC_HOT_PLUG_DETECTx_SENSE)
+				connected = true;
+			break;
+		case RADEON_HPD_3:
+			if (RREG32(DC_HOT_PLUG_DETECT3_INT_STATUS) & DC_HOT_PLUG_DETECTx_SENSE)
+				connected = true;
+			break;
+		default:
+			break;
+		}
+	}
+	return connected;
+}
+
+void r600_hpd_set_polarity(struct radeon_device *rdev,
+			   enum radeon_hpd_id hpd)
+{
+	u32 tmp;
+	bool connected = r600_hpd_sense(rdev, hpd);
+
+	if (ASIC_IS_DCE3(rdev)) {
+		switch (hpd) {
+		case RADEON_HPD_1:
+			tmp = RREG32(DC_HPD1_INT_CONTROL);
+			if (connected)
+				tmp &= ~DC_HPDx_INT_POLARITY;
+			else
+				tmp |= DC_HPDx_INT_POLARITY;
+			WREG32(DC_HPD1_INT_CONTROL, tmp);
+			break;
+		case RADEON_HPD_2:
+			tmp = RREG32(DC_HPD2_INT_CONTROL);
+			if (connected)
+				tmp &= ~DC_HPDx_INT_POLARITY;
+			else
+				tmp |= DC_HPDx_INT_POLARITY;
+			WREG32(DC_HPD2_INT_CONTROL, tmp);
+			break;
+		case RADEON_HPD_3:
+			tmp = RREG32(DC_HPD3_INT_CONTROL);
+			if (connected)
+				tmp &= ~DC_HPDx_INT_POLARITY;
+			else
+				tmp |= DC_HPDx_INT_POLARITY;
+			WREG32(DC_HPD3_INT_CONTROL, tmp);
+			break;
+		case RADEON_HPD_4:
+			tmp = RREG32(DC_HPD4_INT_CONTROL);
+			if (connected)
+				tmp &= ~DC_HPDx_INT_POLARITY;
+			else
+				tmp |= DC_HPDx_INT_POLARITY;
+			WREG32(DC_HPD4_INT_CONTROL, tmp);
+			break;
+		case RADEON_HPD_5:
+			tmp = RREG32(DC_HPD5_INT_CONTROL);
+			if (connected)
+				tmp &= ~DC_HPDx_INT_POLARITY;
+			else
+				tmp |= DC_HPDx_INT_POLARITY;
+			WREG32(DC_HPD5_INT_CONTROL, tmp);
+			break;
+			/* DCE 3.2 */
+		case RADEON_HPD_6:
+			tmp = RREG32(DC_HPD6_INT_CONTROL);
+			if (connected)
+				tmp &= ~DC_HPDx_INT_POLARITY;
+			else
+				tmp |= DC_HPDx_INT_POLARITY;
+			WREG32(DC_HPD6_INT_CONTROL, tmp);
+			break;
+		default:
+			break;
+		}
+	} else {
+		switch (hpd) {
+		case RADEON_HPD_1:
+			tmp = RREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL);
+			if (connected)
+				tmp &= ~DC_HOT_PLUG_DETECTx_INT_POLARITY;
+			else
+				tmp |= DC_HOT_PLUG_DETECTx_INT_POLARITY;
+			WREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL, tmp);
+			break;
+		case RADEON_HPD_2:
+			tmp = RREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL);
+			if (connected)
+				tmp &= ~DC_HOT_PLUG_DETECTx_INT_POLARITY;
+			else
+				tmp |= DC_HOT_PLUG_DETECTx_INT_POLARITY;
+			WREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL, tmp);
+			break;
+		case RADEON_HPD_3:
+			tmp = RREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL);
+			if (connected)
+				tmp &= ~DC_HOT_PLUG_DETECTx_INT_POLARITY;
+			else
+				tmp |= DC_HOT_PLUG_DETECTx_INT_POLARITY;
+			WREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL, tmp);
+			break;
+		default:
+			break;
+		}
+	}
+}
+
+void r600_hpd_init(struct radeon_device *rdev)
 {
-	void __iomem *ptr = (void *)rdev->gart.table.vram.ptr;
-	u64 pte;
+	struct drm_device *dev = rdev->ddev;
+	struct drm_connector *connector;
+
+	if (ASIC_IS_DCE3(rdev)) {
+		u32 tmp = DC_HPDx_CONNECTION_TIMER(0x9c4) | DC_HPDx_RX_INT_TIMER(0xfa);
+		if (ASIC_IS_DCE32(rdev))
+			tmp |= DC_HPDx_EN;
+
+		list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
+			struct radeon_connector *radeon_connector = to_radeon_connector(connector);
+			switch (radeon_connector->hpd.hpd) {
+			case RADEON_HPD_1:
+				WREG32(DC_HPD1_CONTROL, tmp);
+				rdev->irq.hpd[0] = true;
+				break;
+			case RADEON_HPD_2:
+				WREG32(DC_HPD2_CONTROL, tmp);
+				rdev->irq.hpd[1] = true;
+				break;
+			case RADEON_HPD_3:
+				WREG32(DC_HPD3_CONTROL, tmp);
+				rdev->irq.hpd[2] = true;
+				break;
+			case RADEON_HPD_4:
+				WREG32(DC_HPD4_CONTROL, tmp);
+				rdev->irq.hpd[3] = true;
+				break;
+				/* DCE 3.2 */
+			case RADEON_HPD_5:
+				WREG32(DC_HPD5_CONTROL, tmp);
+				rdev->irq.hpd[4] = true;
+				break;
+			case RADEON_HPD_6:
+				WREG32(DC_HPD6_CONTROL, tmp);
+				rdev->irq.hpd[5] = true;
+				break;
+			default:
+				break;
+			}
+		}
+	} else {
+		list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
+			struct radeon_connector *radeon_connector = to_radeon_connector(connector);
+			switch (radeon_connector->hpd.hpd) {
+			case RADEON_HPD_1:
+				WREG32(DC_HOT_PLUG_DETECT1_CONTROL, DC_HOT_PLUG_DETECTx_EN);
+				rdev->irq.hpd[0] = true;
+				break;
+			case RADEON_HPD_2:
+				WREG32(DC_HOT_PLUG_DETECT2_CONTROL, DC_HOT_PLUG_DETECTx_EN);
+				rdev->irq.hpd[1] = true;
+				break;
+			case RADEON_HPD_3:
+				WREG32(DC_HOT_PLUG_DETECT3_CONTROL, DC_HOT_PLUG_DETECTx_EN);
+				rdev->irq.hpd[2] = true;
+				break;
+			default:
+				break;
+			}
+		}
+	}
+	if (rdev->irq.installed)
+		r600_irq_set(rdev);
+}
 
-	if (i < 0 || i > rdev->gart.num_gpu_pages)
-		return -EINVAL;
-	pte = 0;
-	writeq(pte, ((void __iomem *)ptr) + (i * 8));
-	return 0;
+void r600_hpd_fini(struct radeon_device *rdev)
+{
+	struct drm_device *dev = rdev->ddev;
+	struct drm_connector *connector;
+
+	if (ASIC_IS_DCE3(rdev)) {
+		list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
+			struct radeon_connector *radeon_connector = to_radeon_connector(connector);
+			switch (radeon_connector->hpd.hpd) {
+			case RADEON_HPD_1:
+				WREG32(DC_HPD1_CONTROL, 0);
+				rdev->irq.hpd[0] = false;
+				break;
+			case RADEON_HPD_2:
+				WREG32(DC_HPD2_CONTROL, 0);
+				rdev->irq.hpd[1] = false;
+				break;
+			case RADEON_HPD_3:
+				WREG32(DC_HPD3_CONTROL, 0);
+				rdev->irq.hpd[2] = false;
+				break;
+			case RADEON_HPD_4:
+				WREG32(DC_HPD4_CONTROL, 0);
+				rdev->irq.hpd[3] = false;
+				break;
+				/* DCE 3.2 */
+			case RADEON_HPD_5:
+				WREG32(DC_HPD5_CONTROL, 0);
+				rdev->irq.hpd[4] = false;
+				break;
+			case RADEON_HPD_6:
+				WREG32(DC_HPD6_CONTROL, 0);
+				rdev->irq.hpd[5] = false;
+				break;
+			default:
+				break;
+			}
+		}
+	} else {
+		list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
+			struct radeon_connector *radeon_connector = to_radeon_connector(connector);
+			switch (radeon_connector->hpd.hpd) {
+			case RADEON_HPD_1:
+				WREG32(DC_HOT_PLUG_DETECT1_CONTROL, 0);
+				rdev->irq.hpd[0] = false;
+				break;
+			case RADEON_HPD_2:
+				WREG32(DC_HOT_PLUG_DETECT2_CONTROL, 0);
+				rdev->irq.hpd[1] = false;
+				break;
+			case RADEON_HPD_3:
+				WREG32(DC_HOT_PLUG_DETECT3_CONTROL, 0);
+				rdev->irq.hpd[2] = false;
+				break;
+			default:
+				break;
+			}
+		}
+	}
 }
 
+/*
+ * R600 PCIE GART
+ */
 void r600_pcie_gart_tlb_flush(struct radeon_device *rdev)
 {
 	unsigned i;
 	u32 tmp;
 
+	/* flush hdp cache so updates hit vram */
+	WREG32(R_005480_HDP_MEM_COHERENCY_FLUSH_CNTL, 0x1);
+
 	WREG32(VM_CONTEXT0_INVALIDATION_LOW_ADDR, rdev->mc.gtt_start >> 12);
 	WREG32(VM_CONTEXT0_INVALIDATION_HIGH_ADDR, (rdev->mc.gtt_end - 1) >> 12);
 	WREG32(VM_CONTEXT0_REQUEST_RESPONSE, REQUEST_TYPE(1));
@@ -141,6 +407,7 @@ int r600_pcie_gart_enable(struct radeon_device *rdev)
 	r = radeon_gart_table_vram_pin(rdev);
 	if (r)
 		return r;
+	radeon_gart_restore(rdev);
 
 	/* Setup L2 cache */
 	WREG32(VM_L2_CNTL, ENABLE_L2_CACHE | ENABLE_L2_FRAGMENT_PROCESSING |
@@ -168,7 +435,7 @@ int r600_pcie_gart_enable(struct radeon_device *rdev)
 	WREG32(MC_VM_L1_TLB_MCB_RD_SEM_CNTL, tmp | ENABLE_SEMAPHORE_MODE);
 	WREG32(MC_VM_L1_TLB_MCB_WR_SEM_CNTL, tmp | ENABLE_SEMAPHORE_MODE);
 	WREG32(VM_CONTEXT0_PAGE_TABLE_START_ADDR, rdev->mc.gtt_start >> 12);
-	WREG32(VM_CONTEXT0_PAGE_TABLE_END_ADDR, (rdev->mc.gtt_end - 1) >> 12);
+	WREG32(VM_CONTEXT0_PAGE_TABLE_END_ADDR, rdev->mc.gtt_end >> 12);
 	WREG32(VM_CONTEXT0_PAGE_TABLE_BASE_ADDR, rdev->gart.table_addr >> 12);
 	WREG32(VM_CONTEXT0_CNTL, ENABLE_CONTEXT | PAGE_TABLE_DEPTH(0) |
 				RANGE_PROTECTION_FAULT_ENABLE_DEFAULT);
@@ -185,7 +452,7 @@ int r600_pcie_gart_enable(struct radeon_device *rdev)
 void r600_pcie_gart_disable(struct radeon_device *rdev)
 {
 	u32 tmp;
-	int i;
+	int i, r;
 
 	/* Disable all tables */
 	for (i = 0; i < 7; i++)
@@ -213,8 +480,12 @@ void r600_pcie_gart_disable(struct radeon_device *rdev)
 	WREG32(MC_VM_L1_TLB_MCB_RD_HDP_CNTL, tmp);
 	WREG32(MC_VM_L1_TLB_MCB_WR_HDP_CNTL, tmp);
 	if (rdev->gart.table.vram.robj) {
-		radeon_object_kunmap(rdev->gart.table.vram.robj);
-		radeon_object_unpin(rdev->gart.table.vram.robj);
+		r = radeon_bo_reserve(rdev->gart.table.vram.robj, false);
+		if (likely(r == 0)) {
+			radeon_bo_kunmap(rdev->gart.table.vram.robj);
+			radeon_bo_unpin(rdev->gart.table.vram.robj);
+			radeon_bo_unreserve(rdev->gart.table.vram.robj);
+		}
 	}
 }
 
@@ -225,6 +496,40 @@ void r600_pcie_gart_fini(struct radeon_device *rdev)
 	radeon_gart_fini(rdev);
 }
 
+void r600_agp_enable(struct radeon_device *rdev)
+{
+	u32 tmp;
+	int i;
+
+	/* Setup L2 cache */
+	WREG32(VM_L2_CNTL, ENABLE_L2_CACHE | ENABLE_L2_FRAGMENT_PROCESSING |
+				ENABLE_L2_PTE_CACHE_LRU_UPDATE_BY_WRITE |
+				EFFECTIVE_L2_QUEUE_SIZE(7));
+	WREG32(VM_L2_CNTL2, 0);
+	WREG32(VM_L2_CNTL3, BANK_SELECT_0(0) | BANK_SELECT_1(1));
+	/* Setup TLB control */
+	tmp = ENABLE_L1_TLB | ENABLE_L1_FRAGMENT_PROCESSING |
+		SYSTEM_ACCESS_MODE_NOT_IN_SYS |
+		EFFECTIVE_L1_TLB_SIZE(5) | EFFECTIVE_L1_QUEUE_SIZE(5) |
+		ENABLE_WAIT_L2_QUERY;
+	WREG32(MC_VM_L1_TLB_MCB_RD_SYS_CNTL, tmp);
+	WREG32(MC_VM_L1_TLB_MCB_WR_SYS_CNTL, tmp);
+	WREG32(MC_VM_L1_TLB_MCB_RD_HDP_CNTL, tmp | ENABLE_L1_STRICT_ORDERING);
+	WREG32(MC_VM_L1_TLB_MCB_WR_HDP_CNTL, tmp);
+	WREG32(MC_VM_L1_TLB_MCD_RD_A_CNTL, tmp);
+	WREG32(MC_VM_L1_TLB_MCD_WR_A_CNTL, tmp);
+	WREG32(MC_VM_L1_TLB_MCD_RD_B_CNTL, tmp);
+	WREG32(MC_VM_L1_TLB_MCD_WR_B_CNTL, tmp);
+	WREG32(MC_VM_L1_TLB_MCB_RD_GFX_CNTL, tmp);
+	WREG32(MC_VM_L1_TLB_MCB_WR_GFX_CNTL, tmp);
+	WREG32(MC_VM_L1_TLB_MCB_RD_PDMA_CNTL, tmp);
+	WREG32(MC_VM_L1_TLB_MCB_WR_PDMA_CNTL, tmp);
+	WREG32(MC_VM_L1_TLB_MCB_RD_SEM_CNTL, tmp | ENABLE_SEMAPHORE_MODE);
+	WREG32(MC_VM_L1_TLB_MCB_WR_SEM_CNTL, tmp | ENABLE_SEMAPHORE_MODE);
+	for (i = 0; i < 7; i++)
+		WREG32(VM_CONTEXT0_CNTL + (i * 4), 0);
+}
+
 int r600_mc_wait_for_idle(struct radeon_device *rdev)
 {
 	unsigned i;
@@ -240,14 +545,9 @@ int r600_mc_wait_for_idle(struct radeon_device *rdev)
 	return -1;
 }
 
-static void r600_mc_resume(struct radeon_device *rdev)
+static void r600_mc_program(struct radeon_device *rdev)
 {
-	u32 d1vga_control, d2vga_control;
-	u32 vga_render_control, vga_hdp_control;
-	u32 d1crtc_control, d2crtc_control;
-	u32 new_d1grph_primary, new_d1grph_secondary;
-	u32 new_d2grph_primary, new_d2grph_secondary;
-	u64 old_vram_start;
+	struct rv515_mc_save save;
 	u32 tmp;
 	int i, j;
 
@@ -261,99 +561,125 @@ static void r600_mc_resume(struct radeon_device *rdev)
 	}
 	WREG32(HDP_REG_COHERENCY_FLUSH_CNTL, 0);
 
-	d1vga_control = RREG32(D1VGA_CONTROL);
-	d2vga_control = RREG32(D2VGA_CONTROL);
-	vga_render_control = RREG32(VGA_RENDER_CONTROL);
-	vga_hdp_control = RREG32(VGA_HDP_CONTROL);
-	d1crtc_control = RREG32(D1CRTC_CONTROL);
-	d2crtc_control = RREG32(D2CRTC_CONTROL);
-	old_vram_start = (u64)(RREG32(MC_VM_FB_LOCATION) & 0xFFFF) << 24;
-	new_d1grph_primary = RREG32(D1GRPH_PRIMARY_SURFACE_ADDRESS);
-	new_d1grph_secondary = RREG32(D1GRPH_SECONDARY_SURFACE_ADDRESS);
-	new_d1grph_primary += rdev->mc.vram_start - old_vram_start;
-	new_d1grph_secondary += rdev->mc.vram_start - old_vram_start;
-	new_d2grph_primary = RREG32(D2GRPH_PRIMARY_SURFACE_ADDRESS);
-	new_d2grph_secondary = RREG32(D2GRPH_SECONDARY_SURFACE_ADDRESS);
-	new_d2grph_primary += rdev->mc.vram_start - old_vram_start;
-	new_d2grph_secondary += rdev->mc.vram_start - old_vram_start;
-
-	/* Stop all video */
-	WREG32(D1VGA_CONTROL, 0);
-	WREG32(D2VGA_CONTROL, 0);
-	WREG32(VGA_RENDER_CONTROL, 0);
-	WREG32(D1CRTC_UPDATE_LOCK, 1);
-	WREG32(D2CRTC_UPDATE_LOCK, 1);
-	WREG32(D1CRTC_CONTROL, 0);
-	WREG32(D2CRTC_CONTROL, 0);
-	WREG32(D1CRTC_UPDATE_LOCK, 0);
-	WREG32(D2CRTC_UPDATE_LOCK, 0);
-
-	mdelay(1);
+	rv515_mc_stop(rdev, &save);
 	if (r600_mc_wait_for_idle(rdev)) {
-		printk(KERN_WARNING "[drm] MC not idle !\n");
+		dev_warn(rdev->dev, "Wait for MC idle timedout !\n");
 	}
-
-	/* Lockout access through VGA aperture*/
+	/* Lockout access through VGA aperture (doesn't exist before R600) */
 	WREG32(VGA_HDP_CONTROL, VGA_MEMORY_DISABLE);
-
 	/* Update configuration */
-	WREG32(MC_VM_SYSTEM_APERTURE_LOW_ADDR, rdev->mc.vram_start >> 12);
-	WREG32(MC_VM_SYSTEM_APERTURE_HIGH_ADDR, (rdev->mc.vram_end - 1) >> 12);
+	if (rdev->flags & RADEON_IS_AGP) {
+		if (rdev->mc.vram_start < rdev->mc.gtt_start) {
+			/* VRAM before AGP */
+			WREG32(MC_VM_SYSTEM_APERTURE_LOW_ADDR,
+				rdev->mc.vram_start >> 12);
+			WREG32(MC_VM_SYSTEM_APERTURE_HIGH_ADDR,
+				rdev->mc.gtt_end >> 12);
+		} else {
+			/* VRAM after AGP */
+			WREG32(MC_VM_SYSTEM_APERTURE_LOW_ADDR,
+				rdev->mc.gtt_start >> 12);
+			WREG32(MC_VM_SYSTEM_APERTURE_HIGH_ADDR,
+				rdev->mc.vram_end >> 12);
+		}
+	} else {
+		WREG32(MC_VM_SYSTEM_APERTURE_LOW_ADDR, rdev->mc.vram_start >> 12);
+		WREG32(MC_VM_SYSTEM_APERTURE_HIGH_ADDR, rdev->mc.vram_end >> 12);
+	}
 	WREG32(MC_VM_SYSTEM_APERTURE_DEFAULT_ADDR, 0);
-	tmp = (((rdev->mc.vram_end - 1) >> 24) & 0xFFFF) << 16;
+	tmp = ((rdev->mc.vram_end >> 24) & 0xFFFF) << 16;
 	tmp |= ((rdev->mc.vram_start >> 24) & 0xFFFF);
 	WREG32(MC_VM_FB_LOCATION, tmp);
 	WREG32(HDP_NONSURFACE_BASE, (rdev->mc.vram_start >> 8));
 	WREG32(HDP_NONSURFACE_INFO, (2 << 7));
-	WREG32(HDP_NONSURFACE_SIZE, (rdev->mc.mc_vram_size - 1) | 0x3FF);
+	WREG32(HDP_NONSURFACE_SIZE, rdev->mc.mc_vram_size | 0x3FF);
 	if (rdev->flags & RADEON_IS_AGP) {
-		WREG32(MC_VM_AGP_TOP, (rdev->mc.gtt_end - 1) >> 16);
-		WREG32(MC_VM_AGP_BOT, rdev->mc.gtt_start >> 16);
+		WREG32(MC_VM_AGP_TOP, rdev->mc.gtt_end >> 22);
+		WREG32(MC_VM_AGP_BOT, rdev->mc.gtt_start >> 22);
 		WREG32(MC_VM_AGP_BASE, rdev->mc.agp_base >> 22);
 	} else {
 		WREG32(MC_VM_AGP_BASE, 0);
 		WREG32(MC_VM_AGP_TOP, 0x0FFFFFFF);
 		WREG32(MC_VM_AGP_BOT, 0x0FFFFFFF);
 	}
-	WREG32(D1GRPH_PRIMARY_SURFACE_ADDRESS, new_d1grph_primary);
-	WREG32(D1GRPH_SECONDARY_SURFACE_ADDRESS, new_d1grph_secondary);
-	WREG32(D2GRPH_PRIMARY_SURFACE_ADDRESS, new_d2grph_primary);
-	WREG32(D2GRPH_SECONDARY_SURFACE_ADDRESS, new_d2grph_secondary);
-	WREG32(VGA_MEMORY_BASE_ADDRESS, rdev->mc.vram_start);
-
-	/* Unlock host access */
-	WREG32(VGA_HDP_CONTROL, vga_hdp_control);
-
-	mdelay(1);
 	if (r600_mc_wait_for_idle(rdev)) {
-		printk(KERN_WARNING "[drm] MC not idle !\n");
+		dev_warn(rdev->dev, "Wait for MC idle timedout !\n");
 	}
-
-	/* Restore video state */
-	WREG32(D1CRTC_UPDATE_LOCK, 1);
-	WREG32(D2CRTC_UPDATE_LOCK, 1);
-	WREG32(D1CRTC_CONTROL, d1crtc_control);
-	WREG32(D2CRTC_CONTROL, d2crtc_control);
-	WREG32(D1CRTC_UPDATE_LOCK, 0);
-	WREG32(D2CRTC_UPDATE_LOCK, 0);
-	WREG32(D1VGA_CONTROL, d1vga_control);
-	WREG32(D2VGA_CONTROL, d2vga_control);
-	WREG32(VGA_RENDER_CONTROL, vga_render_control);
-
+	rv515_mc_resume(rdev, &save);
 	/* we need to own VRAM, so turn off the VGA renderer here
 	 * to stop it overwriting our objects */
-	radeon_avivo_vga_render_disable(rdev);
+	rv515_vga_render_disable(rdev);
+}
+
+/**
+ * r600_vram_gtt_location - try to find VRAM & GTT location
+ * @rdev: radeon device structure holding all necessary informations
+ * @mc: memory controller structure holding memory informations
+ *
+ * Function will place try to place VRAM at same place as in CPU (PCI)
+ * address space as some GPU seems to have issue when we reprogram at
+ * different address space.
+ *
+ * If there is not enough space to fit the unvisible VRAM after the
+ * aperture then we limit the VRAM size to the aperture.
+ *
+ * If we are using AGP then place VRAM adjacent to AGP aperture are we need
+ * them to be in one from GPU point of view so that we can program GPU to
+ * catch access outside them (weird GPU policy see ??).
+ *
+ * This function will never fails, worst case are limiting VRAM or GTT.
+ *
+ * Note: GTT start, end, size should be initialized before calling this
+ * function on AGP platform.
+ */
+void r600_vram_gtt_location(struct radeon_device *rdev, struct radeon_mc *mc)
+{
+	u64 size_bf, size_af;
+
+	if (mc->mc_vram_size > 0xE0000000) {
+		/* leave room for at least 512M GTT */
+		dev_warn(rdev->dev, "limiting VRAM\n");
+		mc->real_vram_size = 0xE0000000;
+		mc->mc_vram_size = 0xE0000000;
+	}
+	if (rdev->flags & RADEON_IS_AGP) {
+		size_bf = mc->gtt_start;
+		size_af = 0xFFFFFFFF - mc->gtt_end + 1;
+		if (size_bf > size_af) {
+			if (mc->mc_vram_size > size_bf) {
+				dev_warn(rdev->dev, "limiting VRAM\n");
+				mc->real_vram_size = size_bf;
+				mc->mc_vram_size = size_bf;
+			}
+			mc->vram_start = mc->gtt_start - mc->mc_vram_size;
+		} else {
+			if (mc->mc_vram_size > size_af) {
+				dev_warn(rdev->dev, "limiting VRAM\n");
+				mc->real_vram_size = size_af;
+				mc->mc_vram_size = size_af;
+			}
+			mc->vram_start = mc->gtt_end;
+		}
+		mc->vram_end = mc->vram_start + mc->mc_vram_size - 1;
+		dev_info(rdev->dev, "VRAM: %lluM 0x%08llX - 0x%08llX (%lluM used)\n",
+				mc->mc_vram_size >> 20, mc->vram_start,
+				mc->vram_end, mc->real_vram_size >> 20);
+	} else {
+		u64 base = 0;
+		if (rdev->flags & RADEON_IS_IGP)
+			base = (RREG32(MC_VM_FB_LOCATION) & 0xFFFF) << 24;
+		radeon_vram_location(rdev, &rdev->mc, base);
+		radeon_gtt_location(rdev, mc);
+	}
 }
 
 int r600_mc_init(struct radeon_device *rdev)
 {
 	fixed20_12 a;
 	u32 tmp;
-	int chansize;
-	int r;
+	int chansize, numchan;
 
 	/* Get VRAM informations */
-	rdev->mc.vram_width = 128;
 	rdev->mc.vram_is_ddr = true;
 	tmp = RREG32(RAMCFG);
 	if (tmp & CHANSIZE_OVERRIDE) {
@@ -363,90 +689,44 @@ int r600_mc_init(struct radeon_device *rdev)
 	} else {
 		chansize = 32;
 	}
-	if (rdev->family == CHIP_R600) {
-		rdev->mc.vram_width = 8 * chansize;
-	} else if (rdev->family == CHIP_RV670) {
-		rdev->mc.vram_width = 4 * chansize;
-	} else if ((rdev->family == CHIP_RV610) ||
-			(rdev->family == CHIP_RV620)) {
-		rdev->mc.vram_width = chansize;
-	} else if ((rdev->family == CHIP_RV630) ||
-			(rdev->family == CHIP_RV635)) {
-		rdev->mc.vram_width = 2 * chansize;
+	tmp = RREG32(CHMAP);
+	switch ((tmp & NOOFCHAN_MASK) >> NOOFCHAN_SHIFT) {
+	case 0:
+	default:
+		numchan = 1;
+		break;
+	case 1:
+		numchan = 2;
+		break;
+	case 2:
+		numchan = 4;
+		break;
+	case 3:
+		numchan = 8;
+		break;
 	}
+	rdev->mc.vram_width = numchan * chansize;
 	/* Could aper size report 0 ? */
 	rdev->mc.aper_base = drm_get_resource_start(rdev->ddev, 0);
 	rdev->mc.aper_size = drm_get_resource_len(rdev->ddev, 0);
 	/* Setup GPU memory space */
 	rdev->mc.mc_vram_size = RREG32(CONFIG_MEMSIZE);
 	rdev->mc.real_vram_size = RREG32(CONFIG_MEMSIZE);
-	if (rdev->flags & RADEON_IS_AGP) {
-		r = radeon_agp_init(rdev);
-		if (r)
-			return r;
-		/* gtt_size is setup by radeon_agp_init */
-		rdev->mc.gtt_location = rdev->mc.agp_base;
-		tmp = 0xFFFFFFFFUL - rdev->mc.agp_base - rdev->mc.gtt_size;
-		/* Try to put vram before or after AGP because we
-		 * we want SYSTEM_APERTURE to cover both VRAM and
-		 * AGP so that GPU can catch out of VRAM/AGP access
-		 */
-		if (rdev->mc.gtt_location > rdev->mc.mc_vram_size) {
-			/* Enought place before */
-			rdev->mc.vram_location = rdev->mc.gtt_location -
-							rdev->mc.mc_vram_size;
-		} else if (tmp > rdev->mc.mc_vram_size) {
-			/* Enought place after */
-			rdev->mc.vram_location = rdev->mc.gtt_location +
-							rdev->mc.gtt_size;
-		} else {
-			/* Try to setup VRAM then AGP might not
-			 * not work on some card
-			 */
-			rdev->mc.vram_location = 0x00000000UL;
-			rdev->mc.gtt_location = rdev->mc.mc_vram_size;
-		}
-	} else {
-		if (rdev->family == CHIP_RS780 || rdev->family == CHIP_RS880) {
-			rdev->mc.vram_location = (RREG32(MC_VM_FB_LOCATION) &
-								0xFFFF) << 24;
-			rdev->mc.gtt_size = radeon_gart_size * 1024 * 1024;
-			tmp = rdev->mc.vram_location + rdev->mc.mc_vram_size;
-			if ((0xFFFFFFFFUL - tmp) >= rdev->mc.gtt_size) {
-				/* Enough place after vram */
-				rdev->mc.gtt_location = tmp;
-			} else if (rdev->mc.vram_location >= rdev->mc.gtt_size) {
-				/* Enough place before vram */
-				rdev->mc.gtt_location = 0;
-			} else {
-				/* Not enough place after or before shrink
-				 * gart size
-				 */
-				if (rdev->mc.vram_location > (0xFFFFFFFFUL - tmp)) {
-					rdev->mc.gtt_location = 0;
-					rdev->mc.gtt_size = rdev->mc.vram_location;
-				} else {
-					rdev->mc.gtt_location = tmp;
-					rdev->mc.gtt_size = 0xFFFFFFFFUL - tmp;
-				}
-			}
-			rdev->mc.gtt_location = rdev->mc.mc_vram_size;
-		} else {
-			rdev->mc.vram_location = 0x00000000UL;
-			rdev->mc.gtt_location = rdev->mc.mc_vram_size;
-			rdev->mc.gtt_size = radeon_gart_size * 1024 * 1024;
-		}
+	rdev->mc.visible_vram_size = rdev->mc.aper_size;
+	/* FIXME remove this once we support unmappable VRAM */
+	if (rdev->mc.mc_vram_size > rdev->mc.aper_size) {
+		rdev->mc.mc_vram_size = rdev->mc.aper_size;
+		rdev->mc.real_vram_size = rdev->mc.aper_size;
 	}
-	rdev->mc.vram_start = rdev->mc.vram_location;
-	rdev->mc.vram_end = rdev->mc.vram_location + rdev->mc.mc_vram_size;
-	rdev->mc.gtt_start = rdev->mc.gtt_location;
-	rdev->mc.gtt_end = rdev->mc.gtt_location + rdev->mc.gtt_size;
+	r600_vram_gtt_location(rdev, &rdev->mc);
 	/* FIXME: we should enforce default clock in case GPU is not in
 	 * default setup
 	 */
 	a.full = rfixed_const(100);
 	rdev->pm.sclk.full = rfixed_const(rdev->clock.default_sclk);
 	rdev->pm.sclk.full = rfixed_div(rdev->pm.sclk, a);
+	if (rdev->flags & RADEON_IS_IGP)
+		rdev->mc.igp_sideport_enabled = radeon_atombios_sideport_present(rdev);
 	return 0;
 }
 
@@ -456,6 +736,7 @@ int r600_mc_init(struct radeon_device *rdev)
  */
 int r600_gpu_soft_reset(struct radeon_device *rdev)
 {
+	struct rv515_mc_save save;
 	u32 grbm_busy_mask = S_008010_VC_BUSY(1) | S_008010_VGT_BUSY_NO_DMA(1) |
 				S_008010_VGT_BUSY(1) | S_008010_TA03_BUSY(1) |
 				S_008010_TC_BUSY(1) | S_008010_SX_BUSY(1) |
@@ -473,13 +754,25 @@ int r600_gpu_soft_reset(struct radeon_device *rdev)
 			S_008014_CB0_BUSY(1) | S_008014_CB1_BUSY(1) |
 			S_008014_CB2_BUSY(1) | S_008014_CB3_BUSY(1);
 	u32 srbm_reset = 0;
+	u32 tmp;
 
+	dev_info(rdev->dev, "GPU softreset \n");
+	dev_info(rdev->dev, "  R_008010_GRBM_STATUS=0x%08X\n",
+		RREG32(R_008010_GRBM_STATUS));
+	dev_info(rdev->dev, "  R_008014_GRBM_STATUS2=0x%08X\n",
+		RREG32(R_008014_GRBM_STATUS2));
+	dev_info(rdev->dev, "  R_000E50_SRBM_STATUS=0x%08X\n",
+		RREG32(R_000E50_SRBM_STATUS));
+	rv515_mc_stop(rdev, &save);
+	if (r600_mc_wait_for_idle(rdev)) {
+		dev_warn(rdev->dev, "Wait for MC idle timedout !\n");
+	}
 	/* Disable CP parsing/prefetching */
 	WREG32(R_0086D8_CP_ME_CNTL, S_0086D8_CP_ME_HALT(0xff));
 	/* Check if any of the rendering block is busy and reset it */
 	if ((RREG32(R_008010_GRBM_STATUS) & grbm_busy_mask) ||
 	    (RREG32(R_008014_GRBM_STATUS2) & grbm2_busy_mask)) {
-		WREG32(R_008020_GRBM_SOFT_RESET, S_008020_SOFT_RESET_CR(1) |
+		tmp = S_008020_SOFT_RESET_CR(1) |
 			S_008020_SOFT_RESET_DB(1) |
 			S_008020_SOFT_RESET_CB(1) |
 			S_008020_SOFT_RESET_PA(1) |
@@ -491,14 +784,18 @@ int r600_gpu_soft_reset(struct radeon_device *rdev)
 			S_008020_SOFT_RESET_TC(1) |
 			S_008020_SOFT_RESET_TA(1) |
 			S_008020_SOFT_RESET_VC(1) |
-			S_008020_SOFT_RESET_VGT(1));
+			S_008020_SOFT_RESET_VGT(1);
+		dev_info(rdev->dev, "  R_008020_GRBM_SOFT_RESET=0x%08X\n", tmp);
+		WREG32(R_008020_GRBM_SOFT_RESET, tmp);
 		(void)RREG32(R_008020_GRBM_SOFT_RESET);
 		udelay(50);
 		WREG32(R_008020_GRBM_SOFT_RESET, 0);
 		(void)RREG32(R_008020_GRBM_SOFT_RESET);
 	}
 	/* Reset CP (we always reset CP) */
-	WREG32(R_008020_GRBM_SOFT_RESET, S_008020_SOFT_RESET_CP(1));
+	tmp = S_008020_SOFT_RESET_CP(1);
+	dev_info(rdev->dev, "R_008020_GRBM_SOFT_RESET=0x%08X\n", tmp);
+	WREG32(R_008020_GRBM_SOFT_RESET, tmp);
 	(void)RREG32(R_008020_GRBM_SOFT_RESET);
 	udelay(50);
 	WREG32(R_008020_GRBM_SOFT_RESET, 0);
@@ -526,6 +823,14 @@ int r600_gpu_soft_reset(struct radeon_device *rdev)
 		srbm_reset |= S_000E60_SOFT_RESET_RLC(1);
 	if (G_000E50_SEM_BUSY(RREG32(R_000E50_SRBM_STATUS)))
 		srbm_reset |= S_000E60_SOFT_RESET_SEM(1);
+	if (G_000E50_BIF_BUSY(RREG32(R_000E50_SRBM_STATUS)))
+		srbm_reset |= S_000E60_SOFT_RESET_BIF(1);
+	dev_info(rdev->dev, "  R_000E60_SRBM_SOFT_RESET=0x%08X\n", srbm_reset);
+	WREG32(R_000E60_SRBM_SOFT_RESET, srbm_reset);
+	(void)RREG32(R_000E60_SRBM_SOFT_RESET);
+	udelay(50);
+	WREG32(R_000E60_SRBM_SOFT_RESET, 0);
+	(void)RREG32(R_000E60_SRBM_SOFT_RESET);
 	WREG32(R_000E60_SRBM_SOFT_RESET, srbm_reset);
 	(void)RREG32(R_000E60_SRBM_SOFT_RESET);
 	udelay(50);
@@ -533,6 +838,17 @@ int r600_gpu_soft_reset(struct radeon_device *rdev)
 	(void)RREG32(R_000E60_SRBM_SOFT_RESET);
 	/* Wait a little for things to settle down */
 	udelay(50);
+	dev_info(rdev->dev, "  R_008010_GRBM_STATUS=0x%08X\n",
+		RREG32(R_008010_GRBM_STATUS));
+	dev_info(rdev->dev, "  R_008014_GRBM_STATUS2=0x%08X\n",
+		RREG32(R_008014_GRBM_STATUS2));
+	dev_info(rdev->dev, "  R_000E50_SRBM_STATUS=0x%08X\n",
+		RREG32(R_000E50_SRBM_STATUS));
+	/* After reset we need to reinit the asic as GPU often endup in an
+	 * incoherent state.
+	 */
+	atom_asic_init(rdev->mode_info.atom_context);
+	rv515_mc_resume(rdev, &save);
 	return 0;
 }
 
@@ -665,6 +981,9 @@ void r600_gpu_init(struct radeon_device *rdev)
 {
 	u32 tiling_config;
 	u32 ramcfg;
+	u32 backend_map;
+	u32 cc_rb_backend_disable;
+	u32 cc_gc_shader_pipe_config;
 	u32 tmp;
 	int i, j;
 	u32 sq_config;
@@ -774,8 +1093,11 @@ void r600_gpu_init(struct radeon_device *rdev)
 	default:
 		break;
 	}
+	rdev->config.r600.tiling_npipes = rdev->config.r600.max_tile_pipes;
+	rdev->config.r600.tiling_nbanks = 4 << ((ramcfg & NOOFBANK_MASK) >> NOOFBANK_SHIFT);
 	tiling_config |= BANK_TILING((ramcfg & NOOFBANK_MASK) >> NOOFBANK_SHIFT);
 	tiling_config |= GROUP_SIZE(0);
+	rdev->config.r600.tiling_group_size = 256;
 	tmp = (ramcfg & NOOFROWS_MASK) >> NOOFROWS_SHIFT;
 	if (tmp > 3) {
 		tiling_config |= ROW_TILING(3);
@@ -785,24 +1107,33 @@ void r600_gpu_init(struct radeon_device *rdev)
 		tiling_config |= SAMPLE_SPLIT(tmp);
 	}
 	tiling_config |= BANK_SWAPS(1);
-	tmp = r600_get_tile_pipe_to_backend_map(rdev->config.r600.max_tile_pipes,
-						rdev->config.r600.max_backends,
-						(0xff << rdev->config.r600.max_backends) & 0xff);
-	tiling_config |= BACKEND_MAP(tmp);
+
+	cc_rb_backend_disable = RREG32(CC_RB_BACKEND_DISABLE) & 0x00ff0000;
+	cc_rb_backend_disable |=
+		BACKEND_DISABLE((R6XX_MAX_BACKENDS_MASK << rdev->config.r600.max_backends) & R6XX_MAX_BACKENDS_MASK);
+
+	cc_gc_shader_pipe_config = RREG32(CC_GC_SHADER_PIPE_CONFIG) & 0xffffff00;
+	cc_gc_shader_pipe_config |=
+		INACTIVE_QD_PIPES((R6XX_MAX_PIPES_MASK << rdev->config.r600.max_pipes) & R6XX_MAX_PIPES_MASK);
+	cc_gc_shader_pipe_config |=
+		INACTIVE_SIMDS((R6XX_MAX_SIMDS_MASK << rdev->config.r600.max_simds) & R6XX_MAX_SIMDS_MASK);
+
+	backend_map = r600_get_tile_pipe_to_backend_map(rdev->config.r600.max_tile_pipes,
+							(R6XX_MAX_BACKENDS -
+							 r600_count_pipe_bits((cc_rb_backend_disable &
+									       R6XX_MAX_BACKENDS_MASK) >> 16)),
+							(cc_rb_backend_disable >> 16));
+
+	tiling_config |= BACKEND_MAP(backend_map);
 	WREG32(GB_TILING_CONFIG, tiling_config);
 	WREG32(DCP_TILING_CONFIG, tiling_config & 0xffff);
 	WREG32(HDP_TILING_CONFIG, tiling_config & 0xffff);
 
-	tmp = BACKEND_DISABLE((R6XX_MAX_BACKENDS_MASK << rdev->config.r600.max_backends) & R6XX_MAX_BACKENDS_MASK);
-	WREG32(CC_RB_BACKEND_DISABLE, tmp);
-
 	/* Setup pipes */
-	tmp = INACTIVE_QD_PIPES((R6XX_MAX_PIPES_MASK << rdev->config.r600.max_pipes) & R6XX_MAX_PIPES_MASK);
-	tmp |= INACTIVE_SIMDS((R6XX_MAX_SIMDS_MASK << rdev->config.r600.max_simds) & R6XX_MAX_SIMDS_MASK);
-	WREG32(CC_GC_SHADER_PIPE_CONFIG, tmp);
-	WREG32(GC_USER_SHADER_PIPE_CONFIG, tmp);
+	WREG32(CC_RB_BACKEND_DISABLE, cc_rb_backend_disable);
+	WREG32(CC_GC_SHADER_PIPE_CONFIG, cc_gc_shader_pipe_config);
 
-	tmp = R6XX_MAX_BACKENDS - r600_count_pipe_bits(tmp & INACTIVE_QD_PIPES_MASK);
+	tmp = R6XX_MAX_PIPES - r600_count_pipe_bits((cc_gc_shader_pipe_config & INACTIVE_QD_PIPES_MASK) >> 8);
 	WREG32(VGT_OUT_DEALLOC_CNTL, (tmp * 4) & DEALLOC_DIST_MASK);
 	WREG32(VGT_VERTEX_REUSE_BLOCK_CNTL, ((tmp * 4) - 2) & VTX_REUSE_DEPTH_MASK);
 
@@ -826,7 +1157,8 @@ void r600_gpu_init(struct radeon_device *rdev)
 	    ((rdev->family) == CHIP_RV630) ||
 	    ((rdev->family) == CHIP_RV610) ||
 	    ((rdev->family) == CHIP_RV620) ||
-	    ((rdev->family) == CHIP_RS780)) {
+	    ((rdev->family) == CHIP_RS780) ||
+	    ((rdev->family) == CHIP_RS880)) {
 		WREG32(DB_DEBUG, PREZ_MUST_WAIT_FOR_POSTZ_DONE);
 	} else {
 		WREG32(DB_DEBUG, 0);
@@ -843,7 +1175,8 @@ void r600_gpu_init(struct radeon_device *rdev)
 	tmp = RREG32(SQ_MS_FIFO_SIZES);
 	if (((rdev->family) == CHIP_RV610) ||
 	    ((rdev->family) == CHIP_RV620) ||
-	    ((rdev->family) == CHIP_RS780)) {
+	    ((rdev->family) == CHIP_RS780) ||
+	    ((rdev->family) == CHIP_RS880)) {
 		tmp = (CACHE_FIFO_SIZE(0xa) |
 		       FETCH_FIFO_HIWATER(0xa) |
 		       DONE_FIFO_HIWATER(0xe0) |
@@ -886,7 +1219,8 @@ void r600_gpu_init(struct radeon_device *rdev)
 					    NUM_ES_STACK_ENTRIES(0));
 	} else if (((rdev->family) == CHIP_RV610) ||
 		   ((rdev->family) == CHIP_RV620) ||
-		   ((rdev->family) == CHIP_RS780)) {
+		   ((rdev->family) == CHIP_RS780) ||
+		   ((rdev->family) == CHIP_RS880)) {
 		/* no vertex cache */
 		sq_config &= ~VC_ENABLE;
 
@@ -943,7 +1277,8 @@ void r600_gpu_init(struct radeon_device *rdev)
 
 	if (((rdev->family) == CHIP_RV610) ||
 	    ((rdev->family) == CHIP_RV620) ||
-	    ((rdev->family) == CHIP_RS780)) {
+	    ((rdev->family) == CHIP_RS780) ||
+	    ((rdev->family) == CHIP_RS880)) {
 		WREG32(VGT_CACHE_INVALIDATION, CACHE_INVALIDATION(TC_ONLY));
 	} else {
 		WREG32(VGT_CACHE_INVALIDATION, CACHE_INVALIDATION(VC_AND_TC));
@@ -969,8 +1304,9 @@ void r600_gpu_init(struct radeon_device *rdev)
 	tmp = rdev->config.r600.max_pipes * 16;
 	switch (rdev->family) {
 	case CHIP_RV610:
-	case CHIP_RS780:
 	case CHIP_RV620:
+	case CHIP_RS780:
+	case CHIP_RS880:
 		tmp += 32;
 		break;
 	case CHIP_RV670:
@@ -1011,8 +1347,9 @@ void r600_gpu_init(struct radeon_device *rdev)
 
 	switch (rdev->family) {
 	case CHIP_RV610:
-	case CHIP_RS780:
 	case CHIP_RV620:
+	case CHIP_RS780:
+	case CHIP_RS880:
 		tmp = TC_L2_SIZE(8);
 		break;
 	case CHIP_RV630:
@@ -1063,7 +1400,6 @@ void r600_pciep_wreg(struct radeon_device *rdev, u32 reg, u32 v)
 	(void)RREG32(PCIE_PORT_DATA);
 }
 
-
 /*
  * CP & Ring
  */
@@ -1072,11 +1408,12 @@ void r600_cp_stop(struct radeon_device *rdev)
 	WREG32(R_0086D8_CP_ME_CNTL, S_0086D8_CP_ME_HALT(1));
 }
 
-int r600_cp_init_microcode(struct radeon_device *rdev)
+int r600_init_microcode(struct radeon_device *rdev)
 {
 	struct platform_device *pdev;
 	const char *chip_name;
-	size_t pfp_req_size, me_req_size;
+	const char *rlc_chip_name;
+	size_t pfp_req_size, me_req_size, rlc_req_size;
 	char fw_name[30];
 	int err;
 
@@ -1090,30 +1427,62 @@ int r600_cp_init_microcode(struct radeon_device *rdev)
 	}
 
 	switch (rdev->family) {
-	case CHIP_R600: chip_name = "R600"; break;
-	case CHIP_RV610: chip_name = "RV610"; break;
-	case CHIP_RV630: chip_name = "RV630"; break;
-	case CHIP_RV620: chip_name = "RV620"; break;
-	case CHIP_RV635: chip_name = "RV635"; break;
-	case CHIP_RV670: chip_name = "RV670"; break;
+	case CHIP_R600:
+		chip_name = "R600";
+		rlc_chip_name = "R600";
+		break;
+	case CHIP_RV610:
+		chip_name = "RV610";
+		rlc_chip_name = "R600";
+		break;
+	case CHIP_RV630:
+		chip_name = "RV630";
+		rlc_chip_name = "R600";
+		break;
+	case CHIP_RV620:
+		chip_name = "RV620";
+		rlc_chip_name = "R600";
+		break;
+	case CHIP_RV635:
+		chip_name = "RV635";
+		rlc_chip_name = "R600";
+		break;
+	case CHIP_RV670:
+		chip_name = "RV670";
+		rlc_chip_name = "R600";
+		break;
 	case CHIP_RS780:
-	case CHIP_RS880: chip_name = "RS780"; break;
-	case CHIP_RV770: chip_name = "RV770"; break;
+	case CHIP_RS880:
+		chip_name = "RS780";
+		rlc_chip_name = "R600";
+		break;
+	case CHIP_RV770:
+		chip_name = "RV770";
+		rlc_chip_name = "R700";
+		break;
 	case CHIP_RV730:
-	case CHIP_RV740: chip_name = "RV730"; break;
-	case CHIP_RV710: chip_name = "RV710"; break;
+	case CHIP_RV740:
+		chip_name = "RV730";
+		rlc_chip_name = "R700";
+		break;
+	case CHIP_RV710:
+		chip_name = "RV710";
+		rlc_chip_name = "R700";
+		break;
 	default: BUG();
 	}
 
 	if (rdev->family >= CHIP_RV770) {
 		pfp_req_size = R700_PFP_UCODE_SIZE * 4;
 		me_req_size = R700_PM4_UCODE_SIZE * 4;
+		rlc_req_size = R700_RLC_UCODE_SIZE * 4;
 	} else {
 		pfp_req_size = PFP_UCODE_SIZE * 4;
 		me_req_size = PM4_UCODE_SIZE * 12;
+		rlc_req_size = RLC_UCODE_SIZE * 4;
 	}
 
-	DRM_INFO("Loading %s CP Microcode\n", chip_name);
+	DRM_INFO("Loading %s Microcode\n", chip_name);
 
 	snprintf(fw_name, sizeof(fw_name), "radeon/%s_pfp.bin", chip_name);
 	err = request_firmware(&rdev->pfp_fw, fw_name, &pdev->dev);
@@ -1137,6 +1506,18 @@ int r600_cp_init_microcode(struct radeon_device *rdev)
 		       rdev->me_fw->size, fw_name);
 		err = -EINVAL;
 	}
+
+	snprintf(fw_name, sizeof(fw_name), "radeon/%s_rlc.bin", rlc_chip_name);
+	err = request_firmware(&rdev->rlc_fw, fw_name, &pdev->dev);
+	if (err)
+		goto out;
+	if (rdev->rlc_fw->size != rlc_req_size) {
+		printk(KERN_ERR
+		       "r600_rlc: Bogus length %zu in firmware \"%s\"\n",
+		       rdev->rlc_fw->size, fw_name);
+		err = -EINVAL;
+	}
+
 out:
 	platform_device_unregister(pdev);
 
@@ -1149,6 +1530,8 @@ out:
 		rdev->pfp_fw = NULL;
 		release_firmware(rdev->me_fw);
 		rdev->me_fw = NULL;
+		release_firmware(rdev->rlc_fw);
+		rdev->rlc_fw = NULL;
 	}
 	return err;
 }
@@ -1234,19 +1617,17 @@ int r600_cp_resume(struct radeon_device *rdev)
 
 	/* Set ring buffer size */
 	rb_bufsz = drm_order(rdev->cp.ring_size / 8);
+	tmp = RB_NO_UPDATE | (drm_order(RADEON_GPU_PAGE_SIZE/8) << 8) | rb_bufsz;
 #ifdef __BIG_ENDIAN
-	WREG32(CP_RB_CNTL, BUF_SWAP_32BIT | RB_NO_UPDATE |
-		(drm_order(4096/8) << 8) | rb_bufsz);
-#else
-	WREG32(CP_RB_CNTL, RB_NO_UPDATE | (drm_order(4096/8) << 8) | rb_bufsz);
+	tmp |= BUF_SWAP_32BIT;
 #endif
+	WREG32(CP_RB_CNTL, tmp);
 	WREG32(CP_SEM_WAIT_TIMER, 0x4);
 
 	/* Set the write pointer delay */
 	WREG32(CP_RB_WPTR_DELAY, 0);
 
 	/* Initialize the ring buffer's read and write pointers */
-	tmp = RREG32(CP_RB_CNTL);
 	WREG32(CP_RB_CNTL, tmp | RB_RPTR_WR_ENA);
 	WREG32(CP_RB_RPTR_WR, 0);
 	WREG32(CP_RB_WPTR, 0);
@@ -1288,6 +1669,12 @@ void r600_ring_init(struct radeon_device *rdev, unsigned ring_size)
 	rdev->cp.align_mask = 16 - 1;
 }
 
+void r600_cp_fini(struct radeon_device *rdev)
+{
+	r600_cp_stop(rdev);
+	radeon_ring_fini(rdev);
+}
+
 
 /*
  * GPU scratch registers helpers function.
@@ -1343,32 +1730,60 @@ int r600_ring_test(struct radeon_device *rdev)
 	return r;
 }
 
-/*
- * Writeback
- */
-int r600_wb_init(struct radeon_device *rdev)
+void r600_wb_disable(struct radeon_device *rdev)
+{
+	int r;
+
+	WREG32(SCRATCH_UMSK, 0);
+	if (rdev->wb.wb_obj) {
+		r = radeon_bo_reserve(rdev->wb.wb_obj, false);
+		if (unlikely(r != 0))
+			return;
+		radeon_bo_kunmap(rdev->wb.wb_obj);
+		radeon_bo_unpin(rdev->wb.wb_obj);
+		radeon_bo_unreserve(rdev->wb.wb_obj);
+	}
+}
+
+void r600_wb_fini(struct radeon_device *rdev)
+{
+	r600_wb_disable(rdev);
+	if (rdev->wb.wb_obj) {
+		radeon_bo_unref(&rdev->wb.wb_obj);
+		rdev->wb.wb = NULL;
+		rdev->wb.wb_obj = NULL;
+	}
+}
+
+int r600_wb_enable(struct radeon_device *rdev)
 {
 	int r;
 
 	if (rdev->wb.wb_obj == NULL) {
-		r = radeon_object_create(rdev, NULL, 4096,
-					 true,
-					 RADEON_GEM_DOMAIN_GTT,
-					 false, &rdev->wb.wb_obj);
+		r = radeon_bo_create(rdev, NULL, RADEON_GPU_PAGE_SIZE, true,
+				RADEON_GEM_DOMAIN_GTT, &rdev->wb.wb_obj);
 		if (r) {
-			DRM_ERROR("radeon: failed to create WB buffer (%d).\n", r);
+			dev_warn(rdev->dev, "(%d) create WB bo failed\n", r);
 			return r;
 		}
-		r = radeon_object_pin(rdev->wb.wb_obj,
-				      RADEON_GEM_DOMAIN_GTT,
-				      &rdev->wb.gpu_addr);
+		r = radeon_bo_reserve(rdev->wb.wb_obj, false);
+		if (unlikely(r != 0)) {
+			r600_wb_fini(rdev);
+			return r;
+		}
+		r = radeon_bo_pin(rdev->wb.wb_obj, RADEON_GEM_DOMAIN_GTT,
+				&rdev->wb.gpu_addr);
 		if (r) {
-			DRM_ERROR("radeon: failed to pin WB buffer (%d).\n", r);
+			radeon_bo_unreserve(rdev->wb.wb_obj);
+			dev_warn(rdev->dev, "(%d) pin WB bo failed\n", r);
+			r600_wb_fini(rdev);
 			return r;
 		}
-		r = radeon_object_kmap(rdev->wb.wb_obj, (void **)&rdev->wb.wb);
+		r = radeon_bo_kmap(rdev->wb.wb_obj, (void **)&rdev->wb.wb);
+		radeon_bo_unreserve(rdev->wb.wb_obj);
 		if (r) {
-			DRM_ERROR("radeon: failed to map WB buffer (%d).\n", r);
+			dev_warn(rdev->dev, "(%d) map WB bo failed\n", r);
+			r600_wb_fini(rdev);
 			return r;
 		}
 	}
@@ -1379,59 +1794,44 @@ int r600_wb_init(struct radeon_device *rdev)
 	return 0;
 }
 
-void r600_wb_fini(struct radeon_device *rdev)
-{
-	if (rdev->wb.wb_obj) {
-		radeon_object_kunmap(rdev->wb.wb_obj);
-		radeon_object_unpin(rdev->wb.wb_obj);
-		radeon_object_unref(&rdev->wb.wb_obj);
-		rdev->wb.wb = NULL;
-		rdev->wb.wb_obj = NULL;
-	}
-}
-
-
-/*
- * CS
- */
 void r600_fence_ring_emit(struct radeon_device *rdev,
 			  struct radeon_fence *fence)
 {
+	/* Also consider EVENT_WRITE_EOP.  it handles the interrupts + timestamps + events */
+
+	radeon_ring_write(rdev, PACKET3(PACKET3_EVENT_WRITE, 0));
+	radeon_ring_write(rdev, CACHE_FLUSH_AND_INV_EVENT);
+	/* wait for 3D idle clean */
+	radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONFIG_REG, 1));
+	radeon_ring_write(rdev, (WAIT_UNTIL - PACKET3_SET_CONFIG_REG_OFFSET) >> 2);
+	radeon_ring_write(rdev, WAIT_3D_IDLE_bit | WAIT_3D_IDLECLEAN_bit);
 	/* Emit fence sequence & fire IRQ */
 	radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONFIG_REG, 1));
 	radeon_ring_write(rdev, ((rdev->fence_drv.scratch_reg - PACKET3_SET_CONFIG_REG_OFFSET) >> 2));
 	radeon_ring_write(rdev, fence->seq);
-}
-
-int r600_copy_dma(struct radeon_device *rdev,
-		  uint64_t src_offset,
-		  uint64_t dst_offset,
-		  unsigned num_pages,
-		  struct radeon_fence *fence)
-{
-	/* FIXME: implement */
-	return 0;
+	/* CP_INTERRUPT packet 3 no longer exists, use packet 0 */
+	radeon_ring_write(rdev, PACKET0(CP_INT_STATUS, 0));
+	radeon_ring_write(rdev, RB_INT_STAT);
 }
 
 int r600_copy_blit(struct radeon_device *rdev,
 		   uint64_t src_offset, uint64_t dst_offset,
 		   unsigned num_pages, struct radeon_fence *fence)
 {
-	r600_blit_prepare_copy(rdev, num_pages * 4096);
-	r600_kms_blit_copy(rdev, src_offset, dst_offset, num_pages * 4096);
-	r600_blit_done_copy(rdev, fence);
-	return 0;
-}
-
-int r600_irq_process(struct radeon_device *rdev)
-{
-	/* FIXME: implement */
-	return 0;
-}
+	int r;
 
-int r600_irq_set(struct radeon_device *rdev)
-{
-	/* FIXME: implement */
+	mutex_lock(&rdev->r600_blit.mutex);
+	rdev->r600_blit.vb_ib = NULL;
+	r = r600_blit_prepare_copy(rdev, num_pages * RADEON_GPU_PAGE_SIZE);
+	if (r) {
+		if (rdev->r600_blit.vb_ib)
+			radeon_ib_free(rdev, &rdev->r600_blit.vb_ib);
+		mutex_unlock(&rdev->r600_blit.mutex);
+		return r;
+	}
+	r600_kms_blit_copy(rdev, src_offset, dst_offset, num_pages * RADEON_GPU_PAGE_SIZE);
+	r600_blit_done_copy(rdev, fence);
+	mutex_unlock(&rdev->r600_blit.mutex);
 	return 0;
 }
 
@@ -1470,19 +1870,50 @@ int r600_startup(struct radeon_device *rdev)
 {
 	int r;
 
-	r600_gpu_reset(rdev);
-	r600_mc_resume(rdev);
-	r = r600_pcie_gart_enable(rdev);
-	if (r)
-		return r;
-	r600_gpu_init(rdev);
+	if (!rdev->me_fw || !rdev->pfp_fw || !rdev->rlc_fw) {
+		r = r600_init_microcode(rdev);
+		if (r) {
+			DRM_ERROR("Failed to load firmware!\n");
+			return r;
+		}
+	}
 
-	r = radeon_object_pin(rdev->r600_blit.shader_obj, RADEON_GEM_DOMAIN_VRAM,
-			      &rdev->r600_blit.shader_gpu_addr);
+	r600_mc_program(rdev);
+	if (rdev->flags & RADEON_IS_AGP) {
+		r600_agp_enable(rdev);
+	} else {
+		r = r600_pcie_gart_enable(rdev);
+		if (r)
+			return r;
+	}
+	r600_gpu_init(rdev);
+	r = r600_blit_init(rdev);
+	if (r) {
+		r600_blit_fini(rdev);
+		rdev->asic->copy = NULL;
+		dev_warn(rdev->dev, "failed blitter (%d) falling back to memcpy\n", r);
+	}
+	/* pin copy shader into vram */
+	if (rdev->r600_blit.shader_obj) {
+		r = radeon_bo_reserve(rdev->r600_blit.shader_obj, false);
+		if (unlikely(r != 0))
+			return r;
+		r = radeon_bo_pin(rdev->r600_blit.shader_obj, RADEON_GEM_DOMAIN_VRAM,
+				&rdev->r600_blit.shader_gpu_addr);
+		radeon_bo_unreserve(rdev->r600_blit.shader_obj);
+		if (r) {
+			dev_err(rdev->dev, "(%d) pin blit object failed\n", r);
+			return r;
+		}
+	}
+	/* Enable IRQ */
+	r = r600_irq_init(rdev);
 	if (r) {
-		DRM_ERROR("failed to pin blit object %d\n", r);
+		DRM_ERROR("radeon: IH init failed (%d).\n", r);
+		radeon_irq_kms_fini(rdev);
 		return r;
 	}
+	r600_irq_set(rdev);
 
 	r = radeon_ring_init(rdev, rdev->cp.ring_size);
 	if (r)
@@ -1493,9 +1924,8 @@ int r600_startup(struct radeon_device *rdev)
 	r = r600_cp_resume(rdev);
 	if (r)
 		return r;
-	r = r600_wb_init(rdev);
-	if (r)
-		return r;
+	/* write back buffer are not vital so don't worry about failure */
+	r600_wb_enable(rdev);
 	return 0;
 }
 
@@ -1517,15 +1947,12 @@ int r600_resume(struct radeon_device *rdev)
 {
 	int r;
 
-	if (radeon_gpu_reset(rdev)) {
-		/* FIXME: what do we want to do here ? */
-	}
+	/* Do not reset GPU before posting, on r600 hw unlike on r500 hw,
+	 * posting will perform necessary task to bring back GPU into good
+	 * shape.
+	 */
 	/* post card */
-	if (rdev->is_atom_bios) {
-		atom_asic_init(rdev->mode_info.atom_context);
-	} else {
-		radeon_combios_asic_init(rdev->ddev);
-	}
+	atom_asic_init(rdev->mode_info.atom_context);
 	/* Initialize clocks */
 	r = radeon_clocks_init(rdev);
 	if (r) {
@@ -1538,24 +1965,40 @@ int r600_resume(struct radeon_device *rdev)
 		return r;
 	}
 
-	r = radeon_ib_test(rdev);
+	r = r600_ib_test(rdev);
 	if (r) {
 		DRM_ERROR("radeon: failled testing IB (%d).\n", r);
 		return r;
 	}
+
+	r = r600_audio_init(rdev);
+	if (r) {
+		DRM_ERROR("radeon: audio resume failed\n");
+		return r;
+	}
+
 	return r;
 }
 
-
 int r600_suspend(struct radeon_device *rdev)
 {
+	int r;
+
+	r600_audio_fini(rdev);
 	/* FIXME: we should wait for ring to be empty */
 	r600_cp_stop(rdev);
 	rdev->cp.ready = false;
-
+	r600_irq_suspend(rdev);
+	r600_wb_disable(rdev);
 	r600_pcie_gart_disable(rdev);
 	/* unpin shaders bo */
-	radeon_object_unpin(rdev->r600_blit.shader_obj);
+	if (rdev->r600_blit.shader_obj) {
+		r = radeon_bo_reserve(rdev->r600_blit.shader_obj, false);
+		if (!r) {
+			radeon_bo_unpin(rdev->r600_blit.shader_obj);
+			radeon_bo_unreserve(rdev->r600_blit.shader_obj);
+		}
+	}
 	return 0;
 }
 
@@ -1569,7 +2012,6 @@ int r600_init(struct radeon_device *rdev)
 {
 	int r;
 
-	rdev->new_init_path = true;
 	r = radeon_dummy_page_init(rdev);
 	if (r)
 		return r;
@@ -1586,13 +2028,19 @@ int r600_init(struct radeon_device *rdev)
 			return -EINVAL;
 	}
 	/* Must be an ATOMBIOS */
-	if (!rdev->is_atom_bios)
+	if (!rdev->is_atom_bios) {
+		dev_err(rdev->dev, "Expecting atombios for R600 GPU\n");
 		return -EINVAL;
+	}
 	r = radeon_atombios_init(rdev);
 	if (r)
 		return r;
 	/* Post card if necessary */
-	if (!r600_card_posted(rdev) && rdev->bios) {
+	if (!r600_card_posted(rdev)) {
+		if (!rdev->bios) {
+			dev_err(rdev->dev, "Card not posted and no BIOS - ignoring\n");
+			return -EINVAL;
+		}
 		DRM_INFO("GPU not posted. posting now...\n");
 		atom_asic_init(rdev->mode_info.atom_context);
 	}
@@ -1600,95 +2048,90 @@ int r600_init(struct radeon_device *rdev)
 	r600_scratch_init(rdev);
 	/* Initialize surface registers */
 	radeon_surface_init(rdev);
+	/* Initialize clocks */
 	radeon_get_clock_info(rdev->ddev);
 	r = radeon_clocks_init(rdev);
 	if (r)
 		return r;
+	/* Initialize power management */
+	radeon_pm_init(rdev);
 	/* Fence driver */
 	r = radeon_fence_driver_init(rdev);
 	if (r)
 		return r;
+	if (rdev->flags & RADEON_IS_AGP) {
+		r = radeon_agp_init(rdev);
+		if (r)
+			radeon_agp_disable(rdev);
+	}
 	r = r600_mc_init(rdev);
-	if (r) {
-		if (rdev->flags & RADEON_IS_AGP) {
-			/* Retry with disabling AGP */
-			r600_fini(rdev);
-			rdev->flags &= ~RADEON_IS_AGP;
-			return r600_init(rdev);
-		}
+	if (r)
 		return r;
-	}
 	/* Memory manager */
-	r = radeon_object_init(rdev);
+	r = radeon_bo_init(rdev);
+	if (r)
+		return r;
+
+	r = radeon_irq_kms_init(rdev);
 	if (r)
 		return r;
+
 	rdev->cp.ring_obj = NULL;
 	r600_ring_init(rdev, 1024 * 1024);
 
-	if (!rdev->me_fw || !rdev->pfp_fw) {
-		r = r600_cp_init_microcode(rdev);
-		if (r) {
-			DRM_ERROR("Failed to load firmware!\n");
-			return r;
-		}
-	}
+	rdev->ih.ring_obj = NULL;
+	r600_ih_ring_init(rdev, 64 * 1024);
 
 	r = r600_pcie_gart_init(rdev);
 	if (r)
 		return r;
 
 	rdev->accel_working = true;
-	r = r600_blit_init(rdev);
-	if (r) {
-		DRM_ERROR("radeon: failled blitter (%d).\n", r);
-		return r;
-	}
-
 	r = r600_startup(rdev);
 	if (r) {
-		if (rdev->flags & RADEON_IS_AGP) {
-			/* Retry with disabling AGP */
-			r600_fini(rdev);
-			rdev->flags &= ~RADEON_IS_AGP;
-			return r600_init(rdev);
-		}
+		dev_err(rdev->dev, "disabling GPU acceleration\n");
+		r600_cp_fini(rdev);
+		r600_wb_fini(rdev);
+		r600_irq_fini(rdev);
+		radeon_irq_kms_fini(rdev);
+		r600_pcie_gart_fini(rdev);
 		rdev->accel_working = false;
 	}
 	if (rdev->accel_working) {
 		r = radeon_ib_pool_init(rdev);
 		if (r) {
-			DRM_ERROR("radeon: failled initializing IB pool (%d).\n", r);
-			rdev->accel_working = false;
-		}
-		r = radeon_ib_test(rdev);
-		if (r) {
-			DRM_ERROR("radeon: failled testing IB (%d).\n", r);
+			dev_err(rdev->dev, "IB initialization failed (%d).\n", r);
 			rdev->accel_working = false;
+		} else {
+			r = r600_ib_test(rdev);
+			if (r) {
+				dev_err(rdev->dev, "IB test failed (%d).\n", r);
+				rdev->accel_working = false;
+			}
 		}
 	}
+
+	r = r600_audio_init(rdev);
+	if (r)
+		return r; /* TODO error handling */
 	return 0;
 }
 
 void r600_fini(struct radeon_device *rdev)
 {
-	/* Suspend operations */
-	r600_suspend(rdev);
-
+	r600_audio_fini(rdev);
 	r600_blit_fini(rdev);
-	radeon_ring_fini(rdev);
+	r600_cp_fini(rdev);
+	r600_wb_fini(rdev);
+	r600_irq_fini(rdev);
+	radeon_irq_kms_fini(rdev);
 	r600_pcie_gart_fini(rdev);
+	radeon_agp_fini(rdev);
 	radeon_gem_fini(rdev);
 	radeon_fence_driver_fini(rdev);
 	radeon_clocks_fini(rdev);
-#if __OS_HAS_AGP
-	if (rdev->flags & RADEON_IS_AGP)
-		radeon_agp_fini(rdev);
-#endif
-	radeon_object_fini(rdev);
-	if (rdev->is_atom_bios)
-		radeon_atombios_fini(rdev);
-	else
-		radeon_combios_fini(rdev);
+	radeon_bo_fini(rdev);
+	radeon_atombios_fini(rdev);
 	kfree(rdev->bios);
 	rdev->bios = NULL;
 	radeon_dummy_page_fini(rdev);
@@ -1773,8 +2216,670 @@ int r600_ib_test(struct radeon_device *rdev)
 	return r;
 }
 
+/*
+ * Interrupts
+ *
+ * Interrupts use a ring buffer on r6xx/r7xx hardware.  It works pretty
+ * the same as the CP ring buffer, but in reverse.  Rather than the CPU
+ * writing to the ring and the GPU consuming, the GPU writes to the ring
+ * and host consumes.  As the host irq handler processes interrupts, it
+ * increments the rptr.  When the rptr catches up with the wptr, all the
+ * current interrupts have been processed.
+ */
+
+void r600_ih_ring_init(struct radeon_device *rdev, unsigned ring_size)
+{
+	u32 rb_bufsz;
+
+	/* Align ring size */
+	rb_bufsz = drm_order(ring_size / 4);
+	ring_size = (1 << rb_bufsz) * 4;
+	rdev->ih.ring_size = ring_size;
+	rdev->ih.ptr_mask = rdev->ih.ring_size - 1;
+	rdev->ih.rptr = 0;
+}
+
+static int r600_ih_ring_alloc(struct radeon_device *rdev)
+{
+	int r;
+
+	/* Allocate ring buffer */
+	if (rdev->ih.ring_obj == NULL) {
+		r = radeon_bo_create(rdev, NULL, rdev->ih.ring_size,
+				     true,
+				     RADEON_GEM_DOMAIN_GTT,
+				     &rdev->ih.ring_obj);
+		if (r) {
+			DRM_ERROR("radeon: failed to create ih ring buffer (%d).\n", r);
+			return r;
+		}
+		r = radeon_bo_reserve(rdev->ih.ring_obj, false);
+		if (unlikely(r != 0))
+			return r;
+		r = radeon_bo_pin(rdev->ih.ring_obj,
+				  RADEON_GEM_DOMAIN_GTT,
+				  &rdev->ih.gpu_addr);
+		if (r) {
+			radeon_bo_unreserve(rdev->ih.ring_obj);
+			DRM_ERROR("radeon: failed to pin ih ring buffer (%d).\n", r);
+			return r;
+		}
+		r = radeon_bo_kmap(rdev->ih.ring_obj,
+				   (void **)&rdev->ih.ring);
+		radeon_bo_unreserve(rdev->ih.ring_obj);
+		if (r) {
+			DRM_ERROR("radeon: failed to map ih ring buffer (%d).\n", r);
+			return r;
+		}
+	}
+	return 0;
+}
+
+static void r600_ih_ring_fini(struct radeon_device *rdev)
+{
+	int r;
+	if (rdev->ih.ring_obj) {
+		r = radeon_bo_reserve(rdev->ih.ring_obj, false);
+		if (likely(r == 0)) {
+			radeon_bo_kunmap(rdev->ih.ring_obj);
+			radeon_bo_unpin(rdev->ih.ring_obj);
+			radeon_bo_unreserve(rdev->ih.ring_obj);
+		}
+		radeon_bo_unref(&rdev->ih.ring_obj);
+		rdev->ih.ring = NULL;
+		rdev->ih.ring_obj = NULL;
+	}
+}
+
+static void r600_rlc_stop(struct radeon_device *rdev)
+{
+
+	if (rdev->family >= CHIP_RV770) {
+		/* r7xx asics need to soft reset RLC before halting */
+		WREG32(SRBM_SOFT_RESET, SOFT_RESET_RLC);
+		RREG32(SRBM_SOFT_RESET);
+		udelay(15000);
+		WREG32(SRBM_SOFT_RESET, 0);
+		RREG32(SRBM_SOFT_RESET);
+	}
+
+	WREG32(RLC_CNTL, 0);
+}
+
+static void r600_rlc_start(struct radeon_device *rdev)
+{
+	WREG32(RLC_CNTL, RLC_ENABLE);
+}
+
+static int r600_rlc_init(struct radeon_device *rdev)
+{
+	u32 i;
+	const __be32 *fw_data;
+
+	if (!rdev->rlc_fw)
+		return -EINVAL;
+
+	r600_rlc_stop(rdev);
+
+	WREG32(RLC_HB_BASE, 0);
+	WREG32(RLC_HB_CNTL, 0);
+	WREG32(RLC_HB_RPTR, 0);
+	WREG32(RLC_HB_WPTR, 0);
+	WREG32(RLC_HB_WPTR_LSB_ADDR, 0);
+	WREG32(RLC_HB_WPTR_MSB_ADDR, 0);
+	WREG32(RLC_MC_CNTL, 0);
+	WREG32(RLC_UCODE_CNTL, 0);
+
+	fw_data = (const __be32 *)rdev->rlc_fw->data;
+	if (rdev->family >= CHIP_RV770) {
+		for (i = 0; i < R700_RLC_UCODE_SIZE; i++) {
+			WREG32(RLC_UCODE_ADDR, i);
+			WREG32(RLC_UCODE_DATA, be32_to_cpup(fw_data++));
+		}
+	} else {
+		for (i = 0; i < RLC_UCODE_SIZE; i++) {
+			WREG32(RLC_UCODE_ADDR, i);
+			WREG32(RLC_UCODE_DATA, be32_to_cpup(fw_data++));
+		}
+	}
+	WREG32(RLC_UCODE_ADDR, 0);
+
+	r600_rlc_start(rdev);
+
+	return 0;
+}
+
+static void r600_enable_interrupts(struct radeon_device *rdev)
+{
+	u32 ih_cntl = RREG32(IH_CNTL);
+	u32 ih_rb_cntl = RREG32(IH_RB_CNTL);
+
+	ih_cntl |= ENABLE_INTR;
+	ih_rb_cntl |= IH_RB_ENABLE;
+	WREG32(IH_CNTL, ih_cntl);
+	WREG32(IH_RB_CNTL, ih_rb_cntl);
+	rdev->ih.enabled = true;
+}
+
+static void r600_disable_interrupts(struct radeon_device *rdev)
+{
+	u32 ih_rb_cntl = RREG32(IH_RB_CNTL);
+	u32 ih_cntl = RREG32(IH_CNTL);
+
+	ih_rb_cntl &= ~IH_RB_ENABLE;
+	ih_cntl &= ~ENABLE_INTR;
+	WREG32(IH_RB_CNTL, ih_rb_cntl);
+	WREG32(IH_CNTL, ih_cntl);
+	/* set rptr, wptr to 0 */
+	WREG32(IH_RB_RPTR, 0);
+	WREG32(IH_RB_WPTR, 0);
+	rdev->ih.enabled = false;
+	rdev->ih.wptr = 0;
+	rdev->ih.rptr = 0;
+}
+
+static void r600_disable_interrupt_state(struct radeon_device *rdev)
+{
+	u32 tmp;
+
+	WREG32(CP_INT_CNTL, 0);
+	WREG32(GRBM_INT_CNTL, 0);
+	WREG32(DxMODE_INT_MASK, 0);
+	if (ASIC_IS_DCE3(rdev)) {
+		WREG32(DCE3_DACA_AUTODETECT_INT_CONTROL, 0);
+		WREG32(DCE3_DACB_AUTODETECT_INT_CONTROL, 0);
+		tmp = RREG32(DC_HPD1_INT_CONTROL) & DC_HPDx_INT_POLARITY;
+		WREG32(DC_HPD1_INT_CONTROL, tmp);
+		tmp = RREG32(DC_HPD2_INT_CONTROL) & DC_HPDx_INT_POLARITY;
+		WREG32(DC_HPD2_INT_CONTROL, tmp);
+		tmp = RREG32(DC_HPD3_INT_CONTROL) & DC_HPDx_INT_POLARITY;
+		WREG32(DC_HPD3_INT_CONTROL, tmp);
+		tmp = RREG32(DC_HPD4_INT_CONTROL) & DC_HPDx_INT_POLARITY;
+		WREG32(DC_HPD4_INT_CONTROL, tmp);
+		if (ASIC_IS_DCE32(rdev)) {
+			tmp = RREG32(DC_HPD5_INT_CONTROL) & DC_HPDx_INT_POLARITY;
+			WREG32(DC_HPD5_INT_CONTROL, 0);
+			tmp = RREG32(DC_HPD6_INT_CONTROL) & DC_HPDx_INT_POLARITY;
+			WREG32(DC_HPD6_INT_CONTROL, 0);
+		}
+	} else {
+		WREG32(DACA_AUTODETECT_INT_CONTROL, 0);
+		WREG32(DACB_AUTODETECT_INT_CONTROL, 0);
+		tmp = RREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL) & DC_HOT_PLUG_DETECTx_INT_POLARITY;
+		WREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL, 0);
+		tmp = RREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL) & DC_HOT_PLUG_DETECTx_INT_POLARITY;
+		WREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL, 0);
+		tmp = RREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL) & DC_HOT_PLUG_DETECTx_INT_POLARITY;
+		WREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL, 0);
+	}
+}
+
+int r600_irq_init(struct radeon_device *rdev)
+{
+	int ret = 0;
+	int rb_bufsz;
+	u32 interrupt_cntl, ih_cntl, ih_rb_cntl;
+
+	/* allocate ring */
+	ret = r600_ih_ring_alloc(rdev);
+	if (ret)
+		return ret;
+
+	/* disable irqs */
+	r600_disable_interrupts(rdev);
+
+	/* init rlc */
+	ret = r600_rlc_init(rdev);
+	if (ret) {
+		r600_ih_ring_fini(rdev);
+		return ret;
+	}
+
+	/* setup interrupt control */
+	/* set dummy read address to ring address */
+	WREG32(INTERRUPT_CNTL2, rdev->ih.gpu_addr >> 8);
+	interrupt_cntl = RREG32(INTERRUPT_CNTL);
+	/* IH_DUMMY_RD_OVERRIDE=0 - dummy read disabled with msi, enabled without msi
+	 * IH_DUMMY_RD_OVERRIDE=1 - dummy read controlled by IH_DUMMY_RD_EN
+	 */
+	interrupt_cntl &= ~IH_DUMMY_RD_OVERRIDE;
+	/* IH_REQ_NONSNOOP_EN=1 if ring is in non-cacheable memory, e.g., vram */
+	interrupt_cntl &= ~IH_REQ_NONSNOOP_EN;
+	WREG32(INTERRUPT_CNTL, interrupt_cntl);
+
+	WREG32(IH_RB_BASE, rdev->ih.gpu_addr >> 8);
+	rb_bufsz = drm_order(rdev->ih.ring_size / 4);
+
+	ih_rb_cntl = (IH_WPTR_OVERFLOW_ENABLE |
+		      IH_WPTR_OVERFLOW_CLEAR |
+		      (rb_bufsz << 1));
+	/* WPTR writeback, not yet */
+	/*ih_rb_cntl |= IH_WPTR_WRITEBACK_ENABLE;*/
+	WREG32(IH_RB_WPTR_ADDR_LO, 0);
+	WREG32(IH_RB_WPTR_ADDR_HI, 0);
+
+	WREG32(IH_RB_CNTL, ih_rb_cntl);
+
+	/* set rptr, wptr to 0 */
+	WREG32(IH_RB_RPTR, 0);
+	WREG32(IH_RB_WPTR, 0);
+
+	/* Default settings for IH_CNTL (disabled at first) */
+	ih_cntl = MC_WRREQ_CREDIT(0x10) | MC_WR_CLEAN_CNT(0x10);
+	/* RPTR_REARM only works if msi's are enabled */
+	if (rdev->msi_enabled)
+		ih_cntl |= RPTR_REARM;
+
+#ifdef __BIG_ENDIAN
+	ih_cntl |= IH_MC_SWAP(IH_MC_SWAP_32BIT);
+#endif
+	WREG32(IH_CNTL, ih_cntl);
+
+	/* force the active interrupt state to all disabled */
+	r600_disable_interrupt_state(rdev);
+
+	/* enable irqs */
+	r600_enable_interrupts(rdev);
+
+	return ret;
+}
+
+void r600_irq_suspend(struct radeon_device *rdev)
+{
+	r600_disable_interrupts(rdev);
+	r600_rlc_stop(rdev);
+}
+
+void r600_irq_fini(struct radeon_device *rdev)
+{
+	r600_irq_suspend(rdev);
+	r600_ih_ring_fini(rdev);
+}
+
+int r600_irq_set(struct radeon_device *rdev)
+{
+	u32 cp_int_cntl = CNTX_BUSY_INT_ENABLE | CNTX_EMPTY_INT_ENABLE;
+	u32 mode_int = 0;
+	u32 hpd1, hpd2, hpd3, hpd4 = 0, hpd5 = 0, hpd6 = 0;
+
+	if (!rdev->irq.installed) {
+		WARN(1, "Can't enable IRQ/MSI because no handler is installed.\n");
+		return -EINVAL;
+	}
+	/* don't enable anything if the ih is disabled */
+	if (!rdev->ih.enabled) {
+		r600_disable_interrupts(rdev);
+		/* force the active interrupt state to all disabled */
+		r600_disable_interrupt_state(rdev);
+		return 0;
+	}
+
+	if (ASIC_IS_DCE3(rdev)) {
+		hpd1 = RREG32(DC_HPD1_INT_CONTROL) & ~DC_HPDx_INT_EN;
+		hpd2 = RREG32(DC_HPD2_INT_CONTROL) & ~DC_HPDx_INT_EN;
+		hpd3 = RREG32(DC_HPD3_INT_CONTROL) & ~DC_HPDx_INT_EN;
+		hpd4 = RREG32(DC_HPD4_INT_CONTROL) & ~DC_HPDx_INT_EN;
+		if (ASIC_IS_DCE32(rdev)) {
+			hpd5 = RREG32(DC_HPD5_INT_CONTROL) & ~DC_HPDx_INT_EN;
+			hpd6 = RREG32(DC_HPD6_INT_CONTROL) & ~DC_HPDx_INT_EN;
+		}
+	} else {
+		hpd1 = RREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL) & ~DC_HPDx_INT_EN;
+		hpd2 = RREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL) & ~DC_HPDx_INT_EN;
+		hpd3 = RREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL) & ~DC_HPDx_INT_EN;
+	}
+
+	if (rdev->irq.sw_int) {
+		DRM_DEBUG("r600_irq_set: sw int\n");
+		cp_int_cntl |= RB_INT_ENABLE;
+	}
+	if (rdev->irq.crtc_vblank_int[0]) {
+		DRM_DEBUG("r600_irq_set: vblank 0\n");
+		mode_int |= D1MODE_VBLANK_INT_MASK;
+	}
+	if (rdev->irq.crtc_vblank_int[1]) {
+		DRM_DEBUG("r600_irq_set: vblank 1\n");
+		mode_int |= D2MODE_VBLANK_INT_MASK;
+	}
+	if (rdev->irq.hpd[0]) {
+		DRM_DEBUG("r600_irq_set: hpd 1\n");
+		hpd1 |= DC_HPDx_INT_EN;
+	}
+	if (rdev->irq.hpd[1]) {
+		DRM_DEBUG("r600_irq_set: hpd 2\n");
+		hpd2 |= DC_HPDx_INT_EN;
+	}
+	if (rdev->irq.hpd[2]) {
+		DRM_DEBUG("r600_irq_set: hpd 3\n");
+		hpd3 |= DC_HPDx_INT_EN;
+	}
+	if (rdev->irq.hpd[3]) {
+		DRM_DEBUG("r600_irq_set: hpd 4\n");
+		hpd4 |= DC_HPDx_INT_EN;
+	}
+	if (rdev->irq.hpd[4]) {
+		DRM_DEBUG("r600_irq_set: hpd 5\n");
+		hpd5 |= DC_HPDx_INT_EN;
+	}
+	if (rdev->irq.hpd[5]) {
+		DRM_DEBUG("r600_irq_set: hpd 6\n");
+		hpd6 |= DC_HPDx_INT_EN;
+	}
+
+	WREG32(CP_INT_CNTL, cp_int_cntl);
+	WREG32(DxMODE_INT_MASK, mode_int);
+	if (ASIC_IS_DCE3(rdev)) {
+		WREG32(DC_HPD1_INT_CONTROL, hpd1);
+		WREG32(DC_HPD2_INT_CONTROL, hpd2);
+		WREG32(DC_HPD3_INT_CONTROL, hpd3);
+		WREG32(DC_HPD4_INT_CONTROL, hpd4);
+		if (ASIC_IS_DCE32(rdev)) {
+			WREG32(DC_HPD5_INT_CONTROL, hpd5);
+			WREG32(DC_HPD6_INT_CONTROL, hpd6);
+		}
+	} else {
+		WREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL, hpd1);
+		WREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL, hpd2);
+		WREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL, hpd3);
+	}
+
+	return 0;
+}
+
+static inline void r600_irq_ack(struct radeon_device *rdev,
+				u32 *disp_int,
+				u32 *disp_int_cont,
+				u32 *disp_int_cont2)
+{
+	u32 tmp;
+
+	if (ASIC_IS_DCE3(rdev)) {
+		*disp_int = RREG32(DCE3_DISP_INTERRUPT_STATUS);
+		*disp_int_cont = RREG32(DCE3_DISP_INTERRUPT_STATUS_CONTINUE);
+		*disp_int_cont2 = RREG32(DCE3_DISP_INTERRUPT_STATUS_CONTINUE2);
+	} else {
+		*disp_int = RREG32(DISP_INTERRUPT_STATUS);
+		*disp_int_cont = RREG32(DISP_INTERRUPT_STATUS_CONTINUE);
+		*disp_int_cont2 = 0;
+	}
+
+	if (*disp_int & LB_D1_VBLANK_INTERRUPT)
+		WREG32(D1MODE_VBLANK_STATUS, DxMODE_VBLANK_ACK);
+	if (*disp_int & LB_D1_VLINE_INTERRUPT)
+		WREG32(D1MODE_VLINE_STATUS, DxMODE_VLINE_ACK);
+	if (*disp_int & LB_D2_VBLANK_INTERRUPT)
+		WREG32(D2MODE_VBLANK_STATUS, DxMODE_VBLANK_ACK);
+	if (*disp_int & LB_D2_VLINE_INTERRUPT)
+		WREG32(D2MODE_VLINE_STATUS, DxMODE_VLINE_ACK);
+	if (*disp_int & DC_HPD1_INTERRUPT) {
+		if (ASIC_IS_DCE3(rdev)) {
+			tmp = RREG32(DC_HPD1_INT_CONTROL);
+			tmp |= DC_HPDx_INT_ACK;
+			WREG32(DC_HPD1_INT_CONTROL, tmp);
+		} else {
+			tmp = RREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL);
+			tmp |= DC_HPDx_INT_ACK;
+			WREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL, tmp);
+		}
+	}
+	if (*disp_int & DC_HPD2_INTERRUPT) {
+		if (ASIC_IS_DCE3(rdev)) {
+			tmp = RREG32(DC_HPD2_INT_CONTROL);
+			tmp |= DC_HPDx_INT_ACK;
+			WREG32(DC_HPD2_INT_CONTROL, tmp);
+		} else {
+			tmp = RREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL);
+			tmp |= DC_HPDx_INT_ACK;
+			WREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL, tmp);
+		}
+	}
+	if (*disp_int_cont & DC_HPD3_INTERRUPT) {
+		if (ASIC_IS_DCE3(rdev)) {
+			tmp = RREG32(DC_HPD3_INT_CONTROL);
+			tmp |= DC_HPDx_INT_ACK;
+			WREG32(DC_HPD3_INT_CONTROL, tmp);
+		} else {
+			tmp = RREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL);
+			tmp |= DC_HPDx_INT_ACK;
+			WREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL, tmp);
+		}
+	}
+	if (*disp_int_cont & DC_HPD4_INTERRUPT) {
+		tmp = RREG32(DC_HPD4_INT_CONTROL);
+		tmp |= DC_HPDx_INT_ACK;
+		WREG32(DC_HPD4_INT_CONTROL, tmp);
+	}
+	if (ASIC_IS_DCE32(rdev)) {
+		if (*disp_int_cont2 & DC_HPD5_INTERRUPT) {
+			tmp = RREG32(DC_HPD5_INT_CONTROL);
+			tmp |= DC_HPDx_INT_ACK;
+			WREG32(DC_HPD5_INT_CONTROL, tmp);
+		}
+		if (*disp_int_cont2 & DC_HPD6_INTERRUPT) {
+			tmp = RREG32(DC_HPD5_INT_CONTROL);
+			tmp |= DC_HPDx_INT_ACK;
+			WREG32(DC_HPD6_INT_CONTROL, tmp);
+		}
+	}
+}
+
+void r600_irq_disable(struct radeon_device *rdev)
+{
+	u32 disp_int, disp_int_cont, disp_int_cont2;
+
+	r600_disable_interrupts(rdev);
+	/* Wait and acknowledge irq */
+	mdelay(1);
+	r600_irq_ack(rdev, &disp_int, &disp_int_cont, &disp_int_cont2);
+	r600_disable_interrupt_state(rdev);
+}
+
+static inline u32 r600_get_ih_wptr(struct radeon_device *rdev)
+{
+	u32 wptr, tmp;
 
+	/* XXX use writeback */
+	wptr = RREG32(IH_RB_WPTR);
 
+	if (wptr & RB_OVERFLOW) {
+		/* When a ring buffer overflow happen start parsing interrupt
+		 * from the last not overwritten vector (wptr + 16). Hopefully
+		 * this should allow us to catchup.
+		 */
+		dev_warn(rdev->dev, "IH ring buffer overflow (0x%08X, %d, %d)\n",
+			wptr, rdev->ih.rptr, (wptr + 16) + rdev->ih.ptr_mask);
+		rdev->ih.rptr = (wptr + 16) & rdev->ih.ptr_mask;
+		tmp = RREG32(IH_RB_CNTL);
+		tmp |= IH_WPTR_OVERFLOW_CLEAR;
+		WREG32(IH_RB_CNTL, tmp);
+	}
+	return (wptr & rdev->ih.ptr_mask);
+}
+
+/*        r600 IV Ring
+ * Each IV ring entry is 128 bits:
+ * [7:0]    - interrupt source id
+ * [31:8]   - reserved
+ * [59:32]  - interrupt source data
+ * [127:60]  - reserved
+ *
+ * The basic interrupt vector entries
+ * are decoded as follows:
+ * src_id  src_data  description
+ *      1         0  D1 Vblank
+ *      1         1  D1 Vline
+ *      5         0  D2 Vblank
+ *      5         1  D2 Vline
+ *     19         0  FP Hot plug detection A
+ *     19         1  FP Hot plug detection B
+ *     19         2  DAC A auto-detection
+ *     19         3  DAC B auto-detection
+ *    176         -  CP_INT RB
+ *    177         -  CP_INT IB1
+ *    178         -  CP_INT IB2
+ *    181         -  EOP Interrupt
+ *    233         -  GUI Idle
+ *
+ * Note, these are based on r600 and may need to be
+ * adjusted or added to on newer asics
+ */
+
+int r600_irq_process(struct radeon_device *rdev)
+{
+	u32 wptr = r600_get_ih_wptr(rdev);
+	u32 rptr = rdev->ih.rptr;
+	u32 src_id, src_data;
+	u32 ring_index, disp_int, disp_int_cont, disp_int_cont2;
+	unsigned long flags;
+	bool queue_hotplug = false;
+
+	DRM_DEBUG("r600_irq_process start: rptr %d, wptr %d\n", rptr, wptr);
+	if (!rdev->ih.enabled)
+		return IRQ_NONE;
+
+	spin_lock_irqsave(&rdev->ih.lock, flags);
+
+	if (rptr == wptr) {
+		spin_unlock_irqrestore(&rdev->ih.lock, flags);
+		return IRQ_NONE;
+	}
+	if (rdev->shutdown) {
+		spin_unlock_irqrestore(&rdev->ih.lock, flags);
+		return IRQ_NONE;
+	}
+
+restart_ih:
+	/* display interrupts */
+	r600_irq_ack(rdev, &disp_int, &disp_int_cont, &disp_int_cont2);
+
+	rdev->ih.wptr = wptr;
+	while (rptr != wptr) {
+		/* wptr/rptr are in bytes! */
+		ring_index = rptr / 4;
+		src_id =  rdev->ih.ring[ring_index] & 0xff;
+		src_data = rdev->ih.ring[ring_index + 1] & 0xfffffff;
+
+		switch (src_id) {
+		case 1: /* D1 vblank/vline */
+			switch (src_data) {
+			case 0: /* D1 vblank */
+				if (disp_int & LB_D1_VBLANK_INTERRUPT) {
+					drm_handle_vblank(rdev->ddev, 0);
+					wake_up(&rdev->irq.vblank_queue);
+					disp_int &= ~LB_D1_VBLANK_INTERRUPT;
+					DRM_DEBUG("IH: D1 vblank\n");
+				}
+				break;
+			case 1: /* D1 vline */
+				if (disp_int & LB_D1_VLINE_INTERRUPT) {
+					disp_int &= ~LB_D1_VLINE_INTERRUPT;
+					DRM_DEBUG("IH: D1 vline\n");
+				}
+				break;
+			default:
+				DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data);
+				break;
+			}
+			break;
+		case 5: /* D2 vblank/vline */
+			switch (src_data) {
+			case 0: /* D2 vblank */
+				if (disp_int & LB_D2_VBLANK_INTERRUPT) {
+					drm_handle_vblank(rdev->ddev, 1);
+					wake_up(&rdev->irq.vblank_queue);
+					disp_int &= ~LB_D2_VBLANK_INTERRUPT;
+					DRM_DEBUG("IH: D2 vblank\n");
+				}
+				break;
+			case 1: /* D1 vline */
+				if (disp_int & LB_D2_VLINE_INTERRUPT) {
+					disp_int &= ~LB_D2_VLINE_INTERRUPT;
+					DRM_DEBUG("IH: D2 vline\n");
+				}
+				break;
+			default:
+				DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data);
+				break;
+			}
+			break;
+		case 19: /* HPD/DAC hotplug */
+			switch (src_data) {
+			case 0:
+				if (disp_int & DC_HPD1_INTERRUPT) {
+					disp_int &= ~DC_HPD1_INTERRUPT;
+					queue_hotplug = true;
+					DRM_DEBUG("IH: HPD1\n");
+				}
+				break;
+			case 1:
+				if (disp_int & DC_HPD2_INTERRUPT) {
+					disp_int &= ~DC_HPD2_INTERRUPT;
+					queue_hotplug = true;
+					DRM_DEBUG("IH: HPD2\n");
+				}
+				break;
+			case 4:
+				if (disp_int_cont & DC_HPD3_INTERRUPT) {
+					disp_int_cont &= ~DC_HPD3_INTERRUPT;
+					queue_hotplug = true;
+					DRM_DEBUG("IH: HPD3\n");
+				}
+				break;
+			case 5:
+				if (disp_int_cont & DC_HPD4_INTERRUPT) {
+					disp_int_cont &= ~DC_HPD4_INTERRUPT;
+					queue_hotplug = true;
+					DRM_DEBUG("IH: HPD4\n");
+				}
+				break;
+			case 10:
+				if (disp_int_cont2 & DC_HPD5_INTERRUPT) {
+					disp_int_cont &= ~DC_HPD5_INTERRUPT;
+					queue_hotplug = true;
+					DRM_DEBUG("IH: HPD5\n");
+				}
+				break;
+			case 12:
+				if (disp_int_cont2 & DC_HPD6_INTERRUPT) {
+					disp_int_cont &= ~DC_HPD6_INTERRUPT;
+					queue_hotplug = true;
+					DRM_DEBUG("IH: HPD6\n");
+				}
+				break;
+			default:
+				DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data);
+				break;
+			}
+			break;
+		case 176: /* CP_INT in ring buffer */
+		case 177: /* CP_INT in IB1 */
+		case 178: /* CP_INT in IB2 */
+			DRM_DEBUG("IH: CP int: 0x%08x\n", src_data);
+			radeon_fence_process(rdev);
+			break;
+		case 181: /* CP EOP event */
+			DRM_DEBUG("IH: CP EOP\n");
+			break;
+		default:
+			DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data);
+			break;
+		}
+
+		/* wptr/rptr are in bytes! */
+		rptr += 16;
+		rptr &= rdev->ih.ptr_mask;
+	}
+	/* make sure wptr hasn't changed while processing */
+	wptr = r600_get_ih_wptr(rdev);
+	if (wptr != rdev->ih.wptr)
+		goto restart_ih;
+	if (queue_hotplug)
+		queue_work(rdev->wq, &rdev->hotplug_work);
+	rdev->ih.rptr = rptr;
+	WREG32(IH_RB_RPTR, rdev->ih.rptr);
+	spin_unlock_irqrestore(&rdev->ih.lock, flags);
+	return IRQ_HANDLED;
+}
 
 /*
  * Debugfs info
@@ -1786,21 +2891,21 @@ static int r600_debugfs_cp_ring_info(struct seq_file *m, void *data)
 	struct drm_info_node *node = (struct drm_info_node *) m->private;
 	struct drm_device *dev = node->minor->dev;
 	struct radeon_device *rdev = dev->dev_private;
-	uint32_t rdp, wdp;
 	unsigned count, i, j;
 
 	radeon_ring_free_size(rdev);
-	rdp = RREG32(CP_RB_RPTR);
-	wdp = RREG32(CP_RB_WPTR);
-	count = (rdp + rdev->cp.ring_size - wdp) & rdev->cp.ptr_mask;
+	count = (rdev->cp.ring_size / 4) - rdev->cp.ring_free_dw;
 	seq_printf(m, "CP_STAT 0x%08x\n", RREG32(CP_STAT));
-	seq_printf(m, "CP_RB_WPTR 0x%08x\n", wdp);
-	seq_printf(m, "CP_RB_RPTR 0x%08x\n", rdp);
+	seq_printf(m, "CP_RB_WPTR 0x%08x\n", RREG32(CP_RB_WPTR));
+	seq_printf(m, "CP_RB_RPTR 0x%08x\n", RREG32(CP_RB_RPTR));
+	seq_printf(m, "driver's copy of the CP_RB_WPTR 0x%08x\n", rdev->cp.wptr);
+	seq_printf(m, "driver's copy of the CP_RB_RPTR 0x%08x\n", rdev->cp.rptr);
 	seq_printf(m, "%u free dwords in ring\n", rdev->cp.ring_free_dw);
 	seq_printf(m, "%u dwords in ring\n", count);
+	i = rdev->cp.rptr;
 	for (j = 0; j <= count; j++) {
-		i = (rdp + j) & rdev->cp.ptr_mask;
 		seq_printf(m, "r[%04d]=0x%08x\n", i, rdev->cp.ring[i]);
+		i = (i + 1) & rdev->cp.ptr_mask;
 	}
 	return 0;
 }
@@ -1830,3 +2935,18 @@ int r600_debugfs_mc_info_init(struct radeon_device *rdev)
 	return 0;
 #endif
 }
+
+/**
+ * r600_ioctl_wait_idle - flush host path cache on wait idle ioctl
+ * rdev: radeon device structure
+ * bo: buffer object struct which userspace is waiting for idle
+ *
+ * Some R6XX/R7XX doesn't seems to take into account HDP flush performed
+ * through ring buffer, this leads to corruption in rendering, see
+ * http://bugzilla.kernel.org/show_bug.cgi?id=15186 to avoid this we
+ * directly perform HDP flush by writing register through MMIO.
+ */
+void r600_ioctl_wait_idle(struct radeon_device *rdev, struct radeon_bo *bo)
+{
+	WREG32(R_005480_HDP_MEM_COHERENCY_FLUSH_CNTL, 0x1);
+}
diff --git a/drivers/gpu/drm/radeon/r600_audio.c b/drivers/gpu/drm/radeon/r600_audio.c
new file mode 100644
index 000000000000..db928016d034
--- /dev/null
+++ b/drivers/gpu/drm/radeon/r600_audio.c
@@ -0,0 +1,275 @@
+/*
+ * Copyright 2008 Advanced Micro Devices, Inc.
+ * Copyright 2008 Red Hat Inc.
+ * Copyright 2009 Christian König.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Christian König
+ */
+#include "drmP.h"
+#include "radeon.h"
+#include "radeon_reg.h"
+#include "atom.h"
+
+#define AUDIO_TIMER_INTERVALL 100 /* 1/10 sekund should be enough */
+
+/*
+ * check if the chipset is supported
+ */
+static int r600_audio_chipset_supported(struct radeon_device *rdev)
+{
+	return rdev->family >= CHIP_R600
+		|| rdev->family == CHIP_RS600
+		|| rdev->family == CHIP_RS690
+		|| rdev->family == CHIP_RS740;
+}
+
+/*
+ * current number of channels
+ */
+static int r600_audio_channels(struct radeon_device *rdev)
+{
+	return (RREG32(R600_AUDIO_RATE_BPS_CHANNEL) & 0x7) + 1;
+}
+
+/*
+ * current bits per sample
+ */
+static int r600_audio_bits_per_sample(struct radeon_device *rdev)
+{
+	uint32_t value = (RREG32(R600_AUDIO_RATE_BPS_CHANNEL) & 0xF0) >> 4;
+	switch (value) {
+	case 0x0: return  8;
+	case 0x1: return 16;
+	case 0x2: return 20;
+	case 0x3: return 24;
+	case 0x4: return 32;
+	}
+
+	DRM_ERROR("Unknown bits per sample 0x%x using 16 instead.\n", (int)value);
+
+	return 16;
+}
+
+/*
+ * current sampling rate in HZ
+ */
+static int r600_audio_rate(struct radeon_device *rdev)
+{
+	uint32_t value = RREG32(R600_AUDIO_RATE_BPS_CHANNEL);
+	uint32_t result;
+
+	if (value & 0x4000)
+		result = 44100;
+	else
+		result = 48000;
+
+	result *= ((value >> 11) & 0x7) + 1;
+	result /= ((value >> 8) & 0x7) + 1;
+
+	return result;
+}
+
+/*
+ * iec 60958 status bits
+ */
+static uint8_t r600_audio_status_bits(struct radeon_device *rdev)
+{
+	return RREG32(R600_AUDIO_STATUS_BITS) & 0xff;
+}
+
+/*
+ * iec 60958 category code
+ */
+static uint8_t r600_audio_category_code(struct radeon_device *rdev)
+{
+	return (RREG32(R600_AUDIO_STATUS_BITS) >> 8) & 0xff;
+}
+
+/*
+ * update all hdmi interfaces with current audio parameters
+ */
+static void r600_audio_update_hdmi(unsigned long param)
+{
+	struct radeon_device *rdev = (struct radeon_device *)param;
+	struct drm_device *dev = rdev->ddev;
+
+	int channels = r600_audio_channels(rdev);
+	int rate = r600_audio_rate(rdev);
+	int bps = r600_audio_bits_per_sample(rdev);
+	uint8_t status_bits = r600_audio_status_bits(rdev);
+	uint8_t category_code = r600_audio_category_code(rdev);
+
+	struct drm_encoder *encoder;
+	int changes = 0;
+
+	changes |= channels != rdev->audio_channels;
+	changes |= rate != rdev->audio_rate;
+	changes |= bps != rdev->audio_bits_per_sample;
+	changes |= status_bits != rdev->audio_status_bits;
+	changes |= category_code != rdev->audio_category_code;
+
+	if (changes) {
+		rdev->audio_channels = channels;
+		rdev->audio_rate = rate;
+		rdev->audio_bits_per_sample = bps;
+		rdev->audio_status_bits = status_bits;
+		rdev->audio_category_code = category_code;
+	}
+
+	list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) {
+		if (changes || r600_hdmi_buffer_status_changed(encoder))
+			r600_hdmi_update_audio_settings(
+				encoder, channels,
+				rate, bps, status_bits,
+				category_code);
+	}
+
+	mod_timer(&rdev->audio_timer,
+		jiffies + msecs_to_jiffies(AUDIO_TIMER_INTERVALL));
+}
+
+/*
+ * turn on/off audio engine
+ */
+static void r600_audio_engine_enable(struct radeon_device *rdev, bool enable)
+{
+	DRM_INFO("%s audio support", enable ? "Enabling" : "Disabling");
+	WREG32_P(R600_AUDIO_ENABLE, enable ? 0x81000000 : 0x0, ~0x81000000);
+}
+
+/*
+ * initialize the audio vars and register the update timer
+ */
+int r600_audio_init(struct radeon_device *rdev)
+{
+	if (!radeon_audio || !r600_audio_chipset_supported(rdev))
+		return 0;
+
+	r600_audio_engine_enable(rdev, true);
+
+	rdev->audio_channels = -1;
+	rdev->audio_rate = -1;
+	rdev->audio_bits_per_sample = -1;
+	rdev->audio_status_bits = 0;
+	rdev->audio_category_code = 0;
+
+	setup_timer(
+		&rdev->audio_timer,
+		r600_audio_update_hdmi,
+		(unsigned long)rdev);
+
+	mod_timer(&rdev->audio_timer, jiffies + 1);
+
+	return 0;
+}
+
+/*
+ * determin how the encoders and audio interface is wired together
+ */
+int r600_audio_tmds_index(struct drm_encoder *encoder)
+{
+	struct drm_device *dev = encoder->dev;
+	struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
+	struct drm_encoder *other;
+
+	switch (radeon_encoder->encoder_id) {
+	case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_TMDS1:
+	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY:
+	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1:
+		return 0;
+
+	case ENCODER_OBJECT_ID_INTERNAL_LVTM1:
+		/* special case check if an TMDS1 is present */
+		list_for_each_entry(other, &dev->mode_config.encoder_list, head) {
+			if (to_radeon_encoder(other)->encoder_id ==
+				ENCODER_OBJECT_ID_INTERNAL_TMDS1)
+				return 1;
+		}
+		return 0;
+
+	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2:
+	case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_LVTMA:
+		return 1;
+
+	default:
+		DRM_ERROR("Unsupported encoder type 0x%02X\n",
+			  radeon_encoder->encoder_id);
+		return -1;
+	}
+}
+
+/*
+ * atach the audio codec to the clock source of the encoder
+ */
+void r600_audio_set_clock(struct drm_encoder *encoder, int clock)
+{
+	struct drm_device *dev = encoder->dev;
+	struct radeon_device *rdev = dev->dev_private;
+	struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
+	int base_rate = 48000;
+
+	switch (radeon_encoder->encoder_id) {
+	case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_TMDS1:
+	case ENCODER_OBJECT_ID_INTERNAL_LVTM1:
+		WREG32_P(R600_AUDIO_TIMING, 0, ~0x301);
+		break;
+
+	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY:
+	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1:
+	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2:
+	case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_LVTMA:
+		WREG32_P(R600_AUDIO_TIMING, 0x100, ~0x301);
+		break;
+
+	default:
+		DRM_ERROR("Unsupported encoder type 0x%02X\n",
+			  radeon_encoder->encoder_id);
+		return;
+	}
+
+	switch (r600_audio_tmds_index(encoder)) {
+	case 0:
+		WREG32(R600_AUDIO_PLL1_MUL, base_rate*50);
+		WREG32(R600_AUDIO_PLL1_DIV, clock*100);
+		WREG32(R600_AUDIO_CLK_SRCSEL, 0);
+		break;
+
+	case 1:
+		WREG32(R600_AUDIO_PLL2_MUL, base_rate*50);
+		WREG32(R600_AUDIO_PLL2_DIV, clock*100);
+		WREG32(R600_AUDIO_CLK_SRCSEL, 1);
+		break;
+	}
+}
+
+/*
+ * release the audio timer
+ * TODO: How to do this correctly on SMP systems?
+ */
+void r600_audio_fini(struct radeon_device *rdev)
+{
+	if (!radeon_audio || !r600_audio_chipset_supported(rdev))
+		return;
+
+	del_timer(&rdev->audio_timer);
+
+	r600_audio_engine_enable(rdev, false);
+}
diff --git a/drivers/gpu/drm/radeon/r600_blit.c b/drivers/gpu/drm/radeon/r600_blit.c
index d988eece0187..f4fb88ece2bb 100644
--- a/drivers/gpu/drm/radeon/r600_blit.c
+++ b/drivers/gpu/drm/radeon/r600_blit.c
@@ -49,7 +49,7 @@ set_render_target(drm_radeon_private_t *dev_priv, int format, int w, int h, u64
 	RING_LOCALS;
 	DRM_DEBUG("\n");
 
-	h = (h + 7) & ~7;
+	h = ALIGN(h, 8);
 	if (h < 8)
 		h = 8;
 
@@ -774,11 +774,10 @@ r600_blit_swap(struct drm_device *dev,
 {
 	drm_radeon_private_t *dev_priv = dev->dev_private;
 	int cb_format, tex_format;
+	int sx2, sy2, dx2, dy2;
 	u64 vb_addr;
 	u32 *vb;
 
-	vb = r600_nomm_get_vb_ptr(dev);
-
 	if ((dev_priv->blit_vb->used + 48) > dev_priv->blit_vb->total) {
 
 		r600_nomm_put_vb(dev);
@@ -787,19 +786,13 @@ r600_blit_swap(struct drm_device *dev,
 			return;
 
 		set_shaders(dev);
-		vb = r600_nomm_get_vb_ptr(dev);
 	}
+	vb = r600_nomm_get_vb_ptr(dev);
 
-	if (cpp == 4) {
-		cb_format = COLOR_8_8_8_8;
-		tex_format = FMT_8_8_8_8;
-	} else if (cpp == 2) {
-		cb_format = COLOR_5_6_5;
-		tex_format = FMT_5_6_5;
-	} else {
-		cb_format = COLOR_8;
-		tex_format = FMT_8;
-	}
+	sx2 = sx + w;
+	sy2 = sy + h;
+	dx2 = dx + w;
+	dy2 = dy + h;
 
 	vb[0] = i2f(dx);
 	vb[1] = i2f(dy);
@@ -807,31 +800,46 @@ r600_blit_swap(struct drm_device *dev,
 	vb[3] = i2f(sy);
 
 	vb[4] = i2f(dx);
-	vb[5] = i2f(dy + h);
+	vb[5] = i2f(dy2);
 	vb[6] = i2f(sx);
-	vb[7] = i2f(sy + h);
+	vb[7] = i2f(sy2);
+
+	vb[8] = i2f(dx2);
+	vb[9] = i2f(dy2);
+	vb[10] = i2f(sx2);
+	vb[11] = i2f(sy2);
 
-	vb[8] = i2f(dx + w);
-	vb[9] = i2f(dy + h);
-	vb[10] = i2f(sx + w);
-	vb[11] = i2f(sy + h);
+	switch(cpp) {
+	case 4:
+		cb_format = COLOR_8_8_8_8;
+		tex_format = FMT_8_8_8_8;
+		break;
+	case 2:
+		cb_format = COLOR_5_6_5;
+		tex_format = FMT_5_6_5;
+		break;
+	default:
+		cb_format = COLOR_8;
+		tex_format = FMT_8;
+		break;
+	}
 
 	/* src */
 	set_tex_resource(dev_priv, tex_format,
 			 src_pitch / cpp,
-			 sy + h, src_pitch / cpp,
+			 sy2, src_pitch / cpp,
 			 src_gpu_addr);
 
 	cp_set_surface_sync(dev_priv,
-			    R600_TC_ACTION_ENA, (src_pitch * (sy + h)), src_gpu_addr);
+			    R600_TC_ACTION_ENA, src_pitch * sy2, src_gpu_addr);
 
 	/* dst */
 	set_render_target(dev_priv, cb_format,
-			  dst_pitch / cpp, dy + h,
+			  dst_pitch / cpp, dy2,
 			  dst_gpu_addr);
 
 	/* scissors */
-	set_scissors(dev_priv, dx, dy, dx + w, dy + h);
+	set_scissors(dev_priv, dx, dy, dx2, dy2);
 
 	/* Vertex buffer setup */
 	vb_addr = dev_priv->gart_buffers_offset +
@@ -844,7 +852,7 @@ r600_blit_swap(struct drm_device *dev,
 
 	cp_set_surface_sync(dev_priv,
 			    R600_CB_ACTION_ENA | R600_CB0_DEST_BASE_ENA,
-			    dst_pitch * (dy + h), dst_gpu_addr);
+			    dst_pitch * dy2, dst_gpu_addr);
 
 	dev_priv->blit_vb->used += 12 * 4;
 }
diff --git a/drivers/gpu/drm/radeon/r600_blit_kms.c b/drivers/gpu/drm/radeon/r600_blit_kms.c
index acae33e2ad51..f6c6c77db7e0 100644
--- a/drivers/gpu/drm/radeon/r600_blit_kms.c
+++ b/drivers/gpu/drm/radeon/r600_blit_kms.c
@@ -25,7 +25,7 @@ set_render_target(struct radeon_device *rdev, int format,
 	u32 cb_color_info;
 	int pitch, slice;
 
-	h = (h + 7) & ~7;
+	h = ALIGN(h, 8);
 	if (h < 8)
 		h = 8;
 
@@ -368,7 +368,7 @@ set_default_state(struct radeon_device *rdev)
 	if ((rdev->family == CHIP_RV610) ||
 	    (rdev->family == CHIP_RV620) ||
 	    (rdev->family == CHIP_RS780) ||
-	    (rdev->family == CHIP_RS780) ||
+	    (rdev->family == CHIP_RS880) ||
 	    (rdev->family == CHIP_RV710))
 		sq_config = 0;
 	else
@@ -396,15 +396,13 @@ set_default_state(struct radeon_device *rdev)
 				    NUM_ES_STACK_ENTRIES(num_es_stack_entries));
 
 	/* emit an IB pointing at default state */
-	dwords = (rdev->r600_blit.state_len + 0xf) & ~0xf;
+	dwords = ALIGN(rdev->r600_blit.state_len, 0x10);
 	gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.state_offset;
 	radeon_ring_write(rdev, PACKET3(PACKET3_INDIRECT_BUFFER, 2));
 	radeon_ring_write(rdev, gpu_addr & 0xFFFFFFFC);
 	radeon_ring_write(rdev, upper_32_bits(gpu_addr) & 0xFF);
 	radeon_ring_write(rdev, dwords);
 
-	radeon_ring_write(rdev, PACKET3(PACKET3_EVENT_WRITE, 0));
-	radeon_ring_write(rdev, CACHE_FLUSH_AND_INV_EVENT);
 	/* SQ config */
 	radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONFIG_REG, 6));
 	radeon_ring_write(rdev, (SQ_CONFIG - PACKET3_SET_CONFIG_REG_OFFSET) >> 2);
@@ -449,6 +447,7 @@ int r600_blit_init(struct radeon_device *rdev)
 	u32 packet2s[16];
 	int num_packet2s = 0;
 
+	mutex_init(&rdev->r600_blit.mutex);
 	rdev->r600_blit.state_offset = 0;
 
 	if (rdev->family >= CHIP_RV770)
@@ -473,9 +472,8 @@ int r600_blit_init(struct radeon_device *rdev)
 	obj_size += r6xx_ps_size * 4;
 	obj_size = ALIGN(obj_size, 256);
 
-	r = radeon_object_create(rdev, NULL, obj_size,
-				 true, RADEON_GEM_DOMAIN_VRAM,
-				 false, &rdev->r600_blit.shader_obj);
+	r = radeon_bo_create(rdev, NULL, obj_size, true, RADEON_GEM_DOMAIN_VRAM,
+				&rdev->r600_blit.shader_obj);
 	if (r) {
 		DRM_ERROR("r600 failed to allocate shader\n");
 		return r;
@@ -485,12 +483,14 @@ int r600_blit_init(struct radeon_device *rdev)
 		  obj_size,
 		  rdev->r600_blit.vs_offset, rdev->r600_blit.ps_offset);
 
-	r = radeon_object_kmap(rdev->r600_blit.shader_obj, &ptr);
+	r = radeon_bo_reserve(rdev->r600_blit.shader_obj, false);
+	if (unlikely(r != 0))
+		return r;
+	r = radeon_bo_kmap(rdev->r600_blit.shader_obj, &ptr);
 	if (r) {
 		DRM_ERROR("failed to map blit object %d\n", r);
 		return r;
 	}
-
 	if (rdev->family >= CHIP_RV770)
 		memcpy_toio(ptr + rdev->r600_blit.state_offset,
 			    r7xx_default_state, rdev->r600_blit.state_len * 4);
@@ -500,19 +500,28 @@ int r600_blit_init(struct radeon_device *rdev)
 	if (num_packet2s)
 		memcpy_toio(ptr + rdev->r600_blit.state_offset + (rdev->r600_blit.state_len * 4),
 			    packet2s, num_packet2s * 4);
-
-
 	memcpy(ptr + rdev->r600_blit.vs_offset, r6xx_vs, r6xx_vs_size * 4);
 	memcpy(ptr + rdev->r600_blit.ps_offset, r6xx_ps, r6xx_ps_size * 4);
-
-	radeon_object_kunmap(rdev->r600_blit.shader_obj);
+	radeon_bo_kunmap(rdev->r600_blit.shader_obj);
+	radeon_bo_unreserve(rdev->r600_blit.shader_obj);
 	return 0;
 }
 
 void r600_blit_fini(struct radeon_device *rdev)
 {
-	radeon_object_unpin(rdev->r600_blit.shader_obj);
-	radeon_object_unref(&rdev->r600_blit.shader_obj);
+	int r;
+
+	if (rdev->r600_blit.shader_obj == NULL)
+		return;
+	/* If we can't reserve the bo, unref should be enough to destroy
+	 * it when it becomes idle.
+	 */
+	r = radeon_bo_reserve(rdev->r600_blit.shader_obj, false);
+	if (!r) {
+		radeon_bo_unpin(rdev->r600_blit.shader_obj);
+		radeon_bo_unreserve(rdev->r600_blit.shader_obj);
+	}
+	radeon_bo_unref(&rdev->r600_blit.shader_obj);
 }
 
 int r600_vb_ib_get(struct radeon_device *rdev)
@@ -532,9 +541,6 @@ int r600_vb_ib_get(struct radeon_device *rdev)
 void r600_vb_ib_put(struct radeon_device *rdev)
 {
 	radeon_fence_emit(rdev, rdev->r600_blit.vb_ib->fence);
-	mutex_lock(&rdev->ib_pool.mutex);
-	list_add_tail(&rdev->r600_blit.vb_ib->list, &rdev->ib_pool.scheduled_ibs);
-	mutex_unlock(&rdev->ib_pool.mutex);
 	radeon_ib_free(rdev, &rdev->r600_blit.vb_ib);
 }
 
@@ -547,7 +553,8 @@ int r600_blit_prepare_copy(struct radeon_device *rdev, int size_bytes)
 	int dwords_per_loop = 76, num_loops;
 
 	r = r600_vb_ib_get(rdev);
-	WARN_ON(r);
+	if (r)
+		return r;
 
 	/* set_render_target emits 2 extra dwords on rv6xx */
 	if (rdev->family > CHIP_R600 && rdev->family < CHIP_RV770)
@@ -569,11 +576,12 @@ int r600_blit_prepare_copy(struct radeon_device *rdev, int size_bytes)
 	ring_size = num_loops * dwords_per_loop;
 	/* set default  + shaders */
 	ring_size += 40; /* shaders + def state */
-	ring_size += 3; /* fence emit for VB IB */
+	ring_size += 10; /* fence emit for VB IB */
 	ring_size += 5; /* done copy */
-	ring_size += 3; /* fence emit for done copy */
+	ring_size += 10; /* fence emit for done copy */
 	r = radeon_ring_lock(rdev, ring_size);
-	WARN_ON(r);
+	if (r)
+		return r;
 
 	set_default_state(rdev); /* 14 */
 	set_shaders(rdev); /* 26 */
@@ -584,13 +592,6 @@ void r600_blit_done_copy(struct radeon_device *rdev, struct radeon_fence *fence)
 {
 	int r;
 
-	radeon_ring_write(rdev, PACKET3(PACKET3_EVENT_WRITE, 0));
-	radeon_ring_write(rdev, CACHE_FLUSH_AND_INV_EVENT);
-	/* wait for 3D idle clean */
-	radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONFIG_REG, 1));
-	radeon_ring_write(rdev, (WAIT_UNTIL - PACKET3_SET_CONFIG_REG_OFFSET) >> 2);
-	radeon_ring_write(rdev, WAIT_3D_IDLE_bit | WAIT_3D_IDLECLEAN_bit);
-
 	if (rdev->r600_blit.vb_ib)
 		r600_vb_ib_put(rdev);
 
diff --git a/drivers/gpu/drm/radeon/r600_blit_shaders.c b/drivers/gpu/drm/radeon/r600_blit_shaders.c
index d745e815c2e8..a112c59f9d82 100644
--- a/drivers/gpu/drm/radeon/r600_blit_shaders.c
+++ b/drivers/gpu/drm/radeon/r600_blit_shaders.c
@@ -9,11 +9,6 @@ const u32 r6xx_default_state[] =
 	0xc0012800,
 	0x80000000,
 	0x80000000,
-	0xc0004600,
-	0x00000016,
-	0xc0016800,
-	0x00000010,
-	0x00028000,
 	0xc0016800,
 	0x00000010,
 	0x00008000,
@@ -531,11 +526,6 @@ const u32 r7xx_default_state[] =
 	0xc0012800,
 	0x80000000,
 	0x80000000,
-	0xc0004600,
-	0x00000016,
-	0xc0016800,
-	0x00000010,
-	0x00028000,
 	0xc0016800,
 	0x00000010,
 	0x00008000,
diff --git a/drivers/gpu/drm/radeon/r600_cp.c b/drivers/gpu/drm/radeon/r600_cp.c
index 6d5a711c2e91..40416c068d9f 100644
--- a/drivers/gpu/drm/radeon/r600_cp.c
+++ b/drivers/gpu/drm/radeon/r600_cp.c
@@ -734,8 +734,8 @@ static void r600_gfx_init(struct drm_device *dev,
 	u32 hdp_host_path_cntl;
 	u32 backend_map;
 	u32 gb_tiling_config = 0;
-	u32 cc_rb_backend_disable = 0;
-	u32 cc_gc_shader_pipe_config = 0;
+	u32 cc_rb_backend_disable;
+	u32 cc_gc_shader_pipe_config;
 	u32 ramcfg;
 
 	/* setup chip specs */
@@ -857,29 +857,44 @@ static void r600_gfx_init(struct drm_device *dev,
 
 	gb_tiling_config |= R600_BANK_SWAPS(1);
 
-	backend_map = r600_get_tile_pipe_to_backend_map(dev_priv->r600_max_tile_pipes,
-							dev_priv->r600_max_backends,
-							(0xff << dev_priv->r600_max_backends) & 0xff);
-	gb_tiling_config |= R600_BACKEND_MAP(backend_map);
+	cc_rb_backend_disable = RADEON_READ(R600_CC_RB_BACKEND_DISABLE) & 0x00ff0000;
+	cc_rb_backend_disable |=
+		R600_BACKEND_DISABLE((R6XX_MAX_BACKENDS_MASK << dev_priv->r600_max_backends) & R6XX_MAX_BACKENDS_MASK);
 
-	cc_gc_shader_pipe_config =
+	cc_gc_shader_pipe_config = RADEON_READ(R600_CC_GC_SHADER_PIPE_CONFIG) & 0xffffff00;
+	cc_gc_shader_pipe_config |=
 		R600_INACTIVE_QD_PIPES((R6XX_MAX_PIPES_MASK << dev_priv->r600_max_pipes) & R6XX_MAX_PIPES_MASK);
 	cc_gc_shader_pipe_config |=
 		R600_INACTIVE_SIMDS((R6XX_MAX_SIMDS_MASK << dev_priv->r600_max_simds) & R6XX_MAX_SIMDS_MASK);
 
-	cc_rb_backend_disable =
-		R600_BACKEND_DISABLE((R6XX_MAX_BACKENDS_MASK << dev_priv->r600_max_backends) & R6XX_MAX_BACKENDS_MASK);
+	backend_map = r600_get_tile_pipe_to_backend_map(dev_priv->r600_max_tile_pipes,
+							(R6XX_MAX_BACKENDS -
+							 r600_count_pipe_bits((cc_rb_backend_disable &
+									       R6XX_MAX_BACKENDS_MASK) >> 16)),
+							(cc_rb_backend_disable >> 16));
+	gb_tiling_config |= R600_BACKEND_MAP(backend_map);
 
 	RADEON_WRITE(R600_GB_TILING_CONFIG,      gb_tiling_config);
 	RADEON_WRITE(R600_DCP_TILING_CONFIG,    (gb_tiling_config & 0xffff));
 	RADEON_WRITE(R600_HDP_TILING_CONFIG,    (gb_tiling_config & 0xffff));
+	if (gb_tiling_config & 0xc0) {
+		dev_priv->r600_group_size = 512;
+	} else {
+		dev_priv->r600_group_size = 256;
+	}
+	dev_priv->r600_npipes = 1 << ((gb_tiling_config >> 1) & 0x7);
+	if (gb_tiling_config & 0x30) {
+		dev_priv->r600_nbanks = 8;
+	} else {
+		dev_priv->r600_nbanks = 4;
+	}
 
 	RADEON_WRITE(R600_CC_RB_BACKEND_DISABLE,      cc_rb_backend_disable);
 	RADEON_WRITE(R600_CC_GC_SHADER_PIPE_CONFIG,   cc_gc_shader_pipe_config);
 	RADEON_WRITE(R600_GC_USER_SHADER_PIPE_CONFIG, cc_gc_shader_pipe_config);
 
 	num_qd_pipes =
-		R6XX_MAX_BACKENDS - r600_count_pipe_bits(cc_gc_shader_pipe_config & R600_INACTIVE_QD_PIPES_MASK);
+		R6XX_MAX_PIPES - r600_count_pipe_bits((cc_gc_shader_pipe_config & R600_INACTIVE_QD_PIPES_MASK) >> 8);
 	RADEON_WRITE(R600_VGT_OUT_DEALLOC_CNTL, (num_qd_pipes * 4) & R600_DEALLOC_DIST_MASK);
 	RADEON_WRITE(R600_VGT_VERTEX_REUSE_BLOCK_CNTL, ((num_qd_pipes * 4) - 2) & R600_VTX_REUSE_DEPTH_MASK);
 
@@ -1151,7 +1166,8 @@ static void r600_gfx_init(struct drm_device *dev,
 
 }
 
-static u32 r700_get_tile_pipe_to_backend_map(u32 num_tile_pipes,
+static u32 r700_get_tile_pipe_to_backend_map(drm_radeon_private_t *dev_priv,
+					     u32 num_tile_pipes,
 					     u32 num_backends,
 					     u32 backend_disable_mask)
 {
@@ -1162,6 +1178,7 @@ static u32 r700_get_tile_pipe_to_backend_map(u32 num_tile_pipes,
 	u32 swizzle_pipe[R7XX_MAX_PIPES];
 	u32 cur_backend;
 	u32 i;
+	bool force_no_swizzle;
 
 	if (num_tile_pipes > R7XX_MAX_PIPES)
 		num_tile_pipes = R7XX_MAX_PIPES;
@@ -1191,6 +1208,18 @@ static u32 r700_get_tile_pipe_to_backend_map(u32 num_tile_pipes,
 	if (enabled_backends_count != num_backends)
 		num_backends = enabled_backends_count;
 
+	switch (dev_priv->flags & RADEON_FAMILY_MASK) {
+	case CHIP_RV770:
+	case CHIP_RV730:
+		force_no_swizzle = false;
+		break;
+	case CHIP_RV710:
+	case CHIP_RV740:
+	default:
+		force_no_swizzle = true;
+		break;
+	}
+
 	memset((uint8_t *)&swizzle_pipe[0], 0, sizeof(u32) * R7XX_MAX_PIPES);
 	switch (num_tile_pipes) {
 	case 1:
@@ -1201,49 +1230,100 @@ static u32 r700_get_tile_pipe_to_backend_map(u32 num_tile_pipes,
 		swizzle_pipe[1] = 1;
 		break;
 	case 3:
-		swizzle_pipe[0] = 0;
-		swizzle_pipe[1] = 2;
-		swizzle_pipe[2] = 1;
+		if (force_no_swizzle) {
+			swizzle_pipe[0] = 0;
+			swizzle_pipe[1] = 1;
+			swizzle_pipe[2] = 2;
+		} else {
+			swizzle_pipe[0] = 0;
+			swizzle_pipe[1] = 2;
+			swizzle_pipe[2] = 1;
+		}
 		break;
 	case 4:
-		swizzle_pipe[0] = 0;
-		swizzle_pipe[1] = 2;
-		swizzle_pipe[2] = 3;
-		swizzle_pipe[3] = 1;
+		if (force_no_swizzle) {
+			swizzle_pipe[0] = 0;
+			swizzle_pipe[1] = 1;
+			swizzle_pipe[2] = 2;
+			swizzle_pipe[3] = 3;
+		} else {
+			swizzle_pipe[0] = 0;
+			swizzle_pipe[1] = 2;
+			swizzle_pipe[2] = 3;
+			swizzle_pipe[3] = 1;
+		}
 		break;
 	case 5:
-		swizzle_pipe[0] = 0;
-		swizzle_pipe[1] = 2;
-		swizzle_pipe[2] = 4;
-		swizzle_pipe[3] = 1;
-		swizzle_pipe[4] = 3;
+		if (force_no_swizzle) {
+			swizzle_pipe[0] = 0;
+			swizzle_pipe[1] = 1;
+			swizzle_pipe[2] = 2;
+			swizzle_pipe[3] = 3;
+			swizzle_pipe[4] = 4;
+		} else {
+			swizzle_pipe[0] = 0;
+			swizzle_pipe[1] = 2;
+			swizzle_pipe[2] = 4;
+			swizzle_pipe[3] = 1;
+			swizzle_pipe[4] = 3;
+		}
 		break;
 	case 6:
-		swizzle_pipe[0] = 0;
-		swizzle_pipe[1] = 2;
-		swizzle_pipe[2] = 4;
-		swizzle_pipe[3] = 5;
-		swizzle_pipe[4] = 3;
-		swizzle_pipe[5] = 1;
+		if (force_no_swizzle) {
+			swizzle_pipe[0] = 0;
+			swizzle_pipe[1] = 1;
+			swizzle_pipe[2] = 2;
+			swizzle_pipe[3] = 3;
+			swizzle_pipe[4] = 4;
+			swizzle_pipe[5] = 5;
+		} else {
+			swizzle_pipe[0] = 0;
+			swizzle_pipe[1] = 2;
+			swizzle_pipe[2] = 4;
+			swizzle_pipe[3] = 5;
+			swizzle_pipe[4] = 3;
+			swizzle_pipe[5] = 1;
+		}
 		break;
 	case 7:
-		swizzle_pipe[0] = 0;
-		swizzle_pipe[1] = 2;
-		swizzle_pipe[2] = 4;
-		swizzle_pipe[3] = 6;
-		swizzle_pipe[4] = 3;
-		swizzle_pipe[5] = 1;
-		swizzle_pipe[6] = 5;
+		if (force_no_swizzle) {
+			swizzle_pipe[0] = 0;
+			swizzle_pipe[1] = 1;
+			swizzle_pipe[2] = 2;
+			swizzle_pipe[3] = 3;
+			swizzle_pipe[4] = 4;
+			swizzle_pipe[5] = 5;
+			swizzle_pipe[6] = 6;
+		} else {
+			swizzle_pipe[0] = 0;
+			swizzle_pipe[1] = 2;
+			swizzle_pipe[2] = 4;
+			swizzle_pipe[3] = 6;
+			swizzle_pipe[4] = 3;
+			swizzle_pipe[5] = 1;
+			swizzle_pipe[6] = 5;
+		}
 		break;
 	case 8:
-		swizzle_pipe[0] = 0;
-		swizzle_pipe[1] = 2;
-		swizzle_pipe[2] = 4;
-		swizzle_pipe[3] = 6;
-		swizzle_pipe[4] = 3;
-		swizzle_pipe[5] = 1;
-		swizzle_pipe[6] = 7;
-		swizzle_pipe[7] = 5;
+		if (force_no_swizzle) {
+			swizzle_pipe[0] = 0;
+			swizzle_pipe[1] = 1;
+			swizzle_pipe[2] = 2;
+			swizzle_pipe[3] = 3;
+			swizzle_pipe[4] = 4;
+			swizzle_pipe[5] = 5;
+			swizzle_pipe[6] = 6;
+			swizzle_pipe[7] = 7;
+		} else {
+			swizzle_pipe[0] = 0;
+			swizzle_pipe[1] = 2;
+			swizzle_pipe[2] = 4;
+			swizzle_pipe[3] = 6;
+			swizzle_pipe[4] = 3;
+			swizzle_pipe[5] = 1;
+			swizzle_pipe[6] = 7;
+			swizzle_pipe[7] = 5;
+		}
 		break;
 	}
 
@@ -1264,8 +1344,10 @@ static void r700_gfx_init(struct drm_device *dev,
 			  drm_radeon_private_t *dev_priv)
 {
 	int i, j, num_qd_pipes;
+	u32 ta_aux_cntl;
 	u32 sx_debug_1;
 	u32 smx_dc_ctl0;
+	u32 db_debug3;
 	u32 num_gs_verts_per_thread;
 	u32 vgt_gs_per_es;
 	u32 gs_prim_buffer_depth = 0;
@@ -1276,8 +1358,8 @@ static void r700_gfx_init(struct drm_device *dev,
 	u32 sq_dyn_gpr_size_simd_ab_0;
 	u32 backend_map;
 	u32 gb_tiling_config = 0;
-	u32 cc_rb_backend_disable = 0;
-	u32 cc_gc_shader_pipe_config = 0;
+	u32 cc_rb_backend_disable;
+	u32 cc_gc_shader_pipe_config;
 	u32 mc_arb_ramcfg;
 	u32 db_debug4;
 
@@ -1428,35 +1510,51 @@ static void r700_gfx_init(struct drm_device *dev,
 
 	gb_tiling_config |= R600_BANK_SWAPS(1);
 
-	backend_map = r700_get_tile_pipe_to_backend_map(dev_priv->r600_max_tile_pipes,
-							dev_priv->r600_max_backends,
-							(0xff << dev_priv->r600_max_backends) & 0xff);
-	gb_tiling_config |= R600_BACKEND_MAP(backend_map);
+	cc_rb_backend_disable = RADEON_READ(R600_CC_RB_BACKEND_DISABLE) & 0x00ff0000;
+	cc_rb_backend_disable |=
+		R600_BACKEND_DISABLE((R7XX_MAX_BACKENDS_MASK << dev_priv->r600_max_backends) & R7XX_MAX_BACKENDS_MASK);
 
-	cc_gc_shader_pipe_config =
+	cc_gc_shader_pipe_config = RADEON_READ(R600_CC_GC_SHADER_PIPE_CONFIG) & 0xffffff00;
+	cc_gc_shader_pipe_config |=
 		R600_INACTIVE_QD_PIPES((R7XX_MAX_PIPES_MASK << dev_priv->r600_max_pipes) & R7XX_MAX_PIPES_MASK);
 	cc_gc_shader_pipe_config |=
 		R600_INACTIVE_SIMDS((R7XX_MAX_SIMDS_MASK << dev_priv->r600_max_simds) & R7XX_MAX_SIMDS_MASK);
 
-	cc_rb_backend_disable =
-		R600_BACKEND_DISABLE((R7XX_MAX_BACKENDS_MASK << dev_priv->r600_max_backends) & R7XX_MAX_BACKENDS_MASK);
+	if ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RV740)
+		backend_map = 0x28;
+	else
+		backend_map = r700_get_tile_pipe_to_backend_map(dev_priv,
+								dev_priv->r600_max_tile_pipes,
+								(R7XX_MAX_BACKENDS -
+								 r600_count_pipe_bits((cc_rb_backend_disable &
+										       R7XX_MAX_BACKENDS_MASK) >> 16)),
+								(cc_rb_backend_disable >> 16));
+	gb_tiling_config |= R600_BACKEND_MAP(backend_map);
 
 	RADEON_WRITE(R600_GB_TILING_CONFIG,      gb_tiling_config);
 	RADEON_WRITE(R600_DCP_TILING_CONFIG,    (gb_tiling_config & 0xffff));
 	RADEON_WRITE(R600_HDP_TILING_CONFIG,    (gb_tiling_config & 0xffff));
+	if (gb_tiling_config & 0xc0) {
+		dev_priv->r600_group_size = 512;
+	} else {
+		dev_priv->r600_group_size = 256;
+	}
+	dev_priv->r600_npipes = 1 << ((gb_tiling_config >> 1) & 0x7);
+	if (gb_tiling_config & 0x30) {
+		dev_priv->r600_nbanks = 8;
+	} else {
+		dev_priv->r600_nbanks = 4;
+	}
 
 	RADEON_WRITE(R600_CC_RB_BACKEND_DISABLE,      cc_rb_backend_disable);
 	RADEON_WRITE(R600_CC_GC_SHADER_PIPE_CONFIG,   cc_gc_shader_pipe_config);
-	RADEON_WRITE(R600_GC_USER_SHADER_PIPE_CONFIG, cc_gc_shader_pipe_config);
 
 	RADEON_WRITE(R700_CC_SYS_RB_BACKEND_DISABLE, cc_rb_backend_disable);
 	RADEON_WRITE(R700_CGTS_SYS_TCC_DISABLE, 0);
 	RADEON_WRITE(R700_CGTS_TCC_DISABLE, 0);
-	RADEON_WRITE(R700_CGTS_USER_SYS_TCC_DISABLE, 0);
-	RADEON_WRITE(R700_CGTS_USER_TCC_DISABLE, 0);
 
 	num_qd_pipes =
-		R7XX_MAX_BACKENDS - r600_count_pipe_bits(cc_gc_shader_pipe_config & R600_INACTIVE_QD_PIPES_MASK);
+		R7XX_MAX_PIPES - r600_count_pipe_bits((cc_gc_shader_pipe_config & R600_INACTIVE_QD_PIPES_MASK) >> 8);
 	RADEON_WRITE(R600_VGT_OUT_DEALLOC_CNTL, (num_qd_pipes * 4) & R600_DEALLOC_DIST_MASK);
 	RADEON_WRITE(R600_VGT_VERTEX_REUSE_BLOCK_CNTL, ((num_qd_pipes * 4) - 2) & R600_VTX_REUSE_DEPTH_MASK);
 
@@ -1466,10 +1564,8 @@ static void r700_gfx_init(struct drm_device *dev,
 
 	RADEON_WRITE(R600_CP_MEQ_THRESHOLDS, R700_STQ_SPLIT(0x30));
 
-	RADEON_WRITE(R600_TA_CNTL_AUX, (R600_DISABLE_CUBE_ANISO |
-					R600_SYNC_GRADIENT |
-					R600_SYNC_WALKER |
-					R600_SYNC_ALIGNER));
+	ta_aux_cntl = RADEON_READ(R600_TA_CNTL_AUX);
+	RADEON_WRITE(R600_TA_CNTL_AUX, ta_aux_cntl | R600_DISABLE_CUBE_ANISO);
 
 	sx_debug_1 = RADEON_READ(R700_SX_DEBUG_1);
 	sx_debug_1 |= R700_ENABLE_NEW_SMX_ADDRESS;
@@ -1480,14 +1576,28 @@ static void r700_gfx_init(struct drm_device *dev,
 	smx_dc_ctl0 |= R700_CACHE_DEPTH((dev_priv->r700_sx_num_of_sets * 64) - 1);
 	RADEON_WRITE(R600_SMX_DC_CTL0, smx_dc_ctl0);
 
-	RADEON_WRITE(R700_SMX_EVENT_CTL, (R700_ES_FLUSH_CTL(4) |
-					  R700_GS_FLUSH_CTL(4) |
-					  R700_ACK_FLUSH_CTL(3) |
-					  R700_SYNC_FLUSH_CTL));
+	if ((dev_priv->flags & RADEON_FAMILY_MASK) != CHIP_RV740)
+		RADEON_WRITE(R700_SMX_EVENT_CTL, (R700_ES_FLUSH_CTL(4) |
+						  R700_GS_FLUSH_CTL(4) |
+						  R700_ACK_FLUSH_CTL(3) |
+						  R700_SYNC_FLUSH_CTL));
 
-	if ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RV770)
-		RADEON_WRITE(R700_DB_DEBUG3, R700_DB_CLK_OFF_DELAY(0x1f));
-	else {
+	db_debug3 = RADEON_READ(R700_DB_DEBUG3);
+	db_debug3 &= ~R700_DB_CLK_OFF_DELAY(0x1f);
+	switch (dev_priv->flags & RADEON_FAMILY_MASK) {
+	case CHIP_RV770:
+	case CHIP_RV740:
+		db_debug3 |= R700_DB_CLK_OFF_DELAY(0x1f);
+		break;
+	case CHIP_RV710:
+	case CHIP_RV730:
+	default:
+		db_debug3 |= R700_DB_CLK_OFF_DELAY(2);
+		break;
+	}
+	RADEON_WRITE(R700_DB_DEBUG3, db_debug3);
+
+	if ((dev_priv->flags & RADEON_FAMILY_MASK) != CHIP_RV770) {
 		db_debug4 = RADEON_READ(RV700_DB_DEBUG4);
 		db_debug4 |= RV700_DISABLE_TILE_COVERED_FOR_PS_ITER;
 		RADEON_WRITE(RV700_DB_DEBUG4, db_debug4);
@@ -1516,10 +1626,10 @@ static void r700_gfx_init(struct drm_device *dev,
 			    R600_ALU_UPDATE_FIFO_HIWATER(0x8));
 	switch (dev_priv->flags & RADEON_FAMILY_MASK) {
 	case CHIP_RV770:
-		sq_ms_fifo_sizes |= R600_FETCH_FIFO_HIWATER(0x1);
-		break;
 	case CHIP_RV730:
 	case CHIP_RV710:
+		sq_ms_fifo_sizes |= R600_FETCH_FIFO_HIWATER(0x1);
+		break;
 	case CHIP_RV740:
 	default:
 		sq_ms_fifo_sizes |= R600_FETCH_FIFO_HIWATER(0x4);
@@ -2526,3 +2636,12 @@ out:
 	mutex_unlock(&dev_priv->cs_mutex);
 	return r;
 }
+
+void r600_cs_legacy_get_tiling_conf(struct drm_device *dev, u32 *npipes, u32 *nbanks, u32 *group_size)
+{
+	struct drm_radeon_private *dev_priv = dev->dev_private;
+
+	*npipes = dev_priv->r600_npipes;
+	*nbanks = dev_priv->r600_nbanks;
+	*group_size = dev_priv->r600_group_size;
+}
diff --git a/drivers/gpu/drm/radeon/r600_cs.c b/drivers/gpu/drm/radeon/r600_cs.c
index 33b89cd8743e..cd2c63bce501 100644
--- a/drivers/gpu/drm/radeon/r600_cs.c
+++ b/drivers/gpu/drm/radeon/r600_cs.c
@@ -28,7 +28,7 @@
 #include "drmP.h"
 #include "radeon.h"
 #include "r600d.h"
-#include "avivod.h"
+#include "r600_reg_safe.h"
 
 static int r600_cs_packet_next_reloc_mm(struct radeon_cs_parser *p,
 					struct radeon_cs_reloc **cs_reloc);
@@ -36,6 +36,312 @@ static int r600_cs_packet_next_reloc_nomm(struct radeon_cs_parser *p,
 					struct radeon_cs_reloc **cs_reloc);
 typedef int (*next_reloc_t)(struct radeon_cs_parser*, struct radeon_cs_reloc**);
 static next_reloc_t r600_cs_packet_next_reloc = &r600_cs_packet_next_reloc_mm;
+extern void r600_cs_legacy_get_tiling_conf(struct drm_device *dev, u32 *npipes, u32 *nbanks, u32 *group_size);
+
+
+struct r600_cs_track {
+	/* configuration we miror so that we use same code btw kms/ums */
+	u32			group_size;
+	u32			nbanks;
+	u32			npipes;
+	/* value we track */
+	u32			nsamples;
+	u32			cb_color_base_last[8];
+	struct radeon_bo	*cb_color_bo[8];
+	u32			cb_color_bo_offset[8];
+	struct radeon_bo	*cb_color_frag_bo[8];
+	struct radeon_bo	*cb_color_tile_bo[8];
+	u32			cb_color_info[8];
+	u32			cb_color_size_idx[8];
+	u32			cb_target_mask;
+	u32			cb_shader_mask;
+	u32			cb_color_size[8];
+	u32			vgt_strmout_en;
+	u32			vgt_strmout_buffer_en;
+	u32			db_depth_control;
+	u32			db_depth_info;
+	u32			db_depth_size_idx;
+	u32			db_depth_view;
+	u32			db_depth_size;
+	u32			db_offset;
+	struct radeon_bo	*db_bo;
+};
+
+static inline int r600_bpe_from_format(u32 *bpe, u32 format)
+{
+	switch (format) {
+	case V_038004_COLOR_8:
+	case V_038004_COLOR_4_4:
+	case V_038004_COLOR_3_3_2:
+	case V_038004_FMT_1:
+		*bpe = 1;
+		break;
+	case V_038004_COLOR_16:
+	case V_038004_COLOR_16_FLOAT:
+	case V_038004_COLOR_8_8:
+	case V_038004_COLOR_5_6_5:
+	case V_038004_COLOR_6_5_5:
+	case V_038004_COLOR_1_5_5_5:
+	case V_038004_COLOR_4_4_4_4:
+	case V_038004_COLOR_5_5_5_1:
+		*bpe = 2;
+		break;
+	case V_038004_FMT_8_8_8:
+		*bpe = 3;
+		break;
+	case V_038004_COLOR_32:
+	case V_038004_COLOR_32_FLOAT:
+	case V_038004_COLOR_16_16:
+	case V_038004_COLOR_16_16_FLOAT:
+	case V_038004_COLOR_8_24:
+	case V_038004_COLOR_8_24_FLOAT:
+	case V_038004_COLOR_24_8:
+	case V_038004_COLOR_24_8_FLOAT:
+	case V_038004_COLOR_10_11_11:
+	case V_038004_COLOR_10_11_11_FLOAT:
+	case V_038004_COLOR_11_11_10:
+	case V_038004_COLOR_11_11_10_FLOAT:
+	case V_038004_COLOR_2_10_10_10:
+	case V_038004_COLOR_8_8_8_8:
+	case V_038004_COLOR_10_10_10_2:
+	case V_038004_FMT_5_9_9_9_SHAREDEXP:
+	case V_038004_FMT_32_AS_8:
+	case V_038004_FMT_32_AS_8_8:
+		*bpe = 4;
+		break;
+	case V_038004_COLOR_X24_8_32_FLOAT:
+	case V_038004_COLOR_32_32:
+	case V_038004_COLOR_32_32_FLOAT:
+	case V_038004_COLOR_16_16_16_16:
+	case V_038004_COLOR_16_16_16_16_FLOAT:
+		*bpe = 8;
+		break;
+	case V_038004_FMT_16_16_16:
+	case V_038004_FMT_16_16_16_FLOAT:
+		*bpe = 6;
+		break;
+	case V_038004_FMT_32_32_32:
+	case V_038004_FMT_32_32_32_FLOAT:
+		*bpe = 12;
+		break;
+	case V_038004_COLOR_32_32_32_32:
+	case V_038004_COLOR_32_32_32_32_FLOAT:
+		*bpe = 16;
+		break;
+	case V_038004_FMT_GB_GR:
+	case V_038004_FMT_BG_RG:
+	case V_038004_COLOR_INVALID:
+		*bpe = 16;
+		return -EINVAL;
+	}
+	return 0;
+}
+
+static void r600_cs_track_init(struct r600_cs_track *track)
+{
+	int i;
+
+	for (i = 0; i < 8; i++) {
+		track->cb_color_base_last[i] = 0;
+		track->cb_color_size[i] = 0;
+		track->cb_color_size_idx[i] = 0;
+		track->cb_color_info[i] = 0;
+		track->cb_color_bo[i] = NULL;
+		track->cb_color_bo_offset[i] = 0xFFFFFFFF;
+	}
+	track->cb_target_mask = 0xFFFFFFFF;
+	track->cb_shader_mask = 0xFFFFFFFF;
+	track->db_bo = NULL;
+	/* assume the biggest format and that htile is enabled */
+	track->db_depth_info = 7 | (1 << 25);
+	track->db_depth_view = 0xFFFFC000;
+	track->db_depth_size = 0xFFFFFFFF;
+	track->db_depth_size_idx = 0;
+	track->db_depth_control = 0xFFFFFFFF;
+}
+
+static inline int r600_cs_track_validate_cb(struct radeon_cs_parser *p, int i)
+{
+	struct r600_cs_track *track = p->track;
+	u32 bpe = 0, pitch, slice_tile_max, size, tmp, height;
+	volatile u32 *ib = p->ib->ptr;
+
+	if (G_0280A0_TILE_MODE(track->cb_color_info[i])) {
+		dev_warn(p->dev, "FMASK or CMASK buffer are not supported by this kernel\n");
+		return -EINVAL;
+	}
+	size = radeon_bo_size(track->cb_color_bo[i]);
+	if (r600_bpe_from_format(&bpe, G_0280A0_FORMAT(track->cb_color_info[i]))) {
+		dev_warn(p->dev, "%s:%d cb invalid format %d for %d (0x%08X)\n",
+			 __func__, __LINE__, G_0280A0_FORMAT(track->cb_color_info[i]),
+			i, track->cb_color_info[i]);
+		return -EINVAL;
+	}
+	pitch = (G_028060_PITCH_TILE_MAX(track->cb_color_size[i]) + 1) << 3;
+	slice_tile_max = G_028060_SLICE_TILE_MAX(track->cb_color_size[i]) + 1;
+	if (!pitch) {
+		dev_warn(p->dev, "%s:%d cb pitch (%d) for %d invalid (0x%08X)\n",
+			__func__, __LINE__, pitch, i, track->cb_color_size[i]);
+		return -EINVAL;
+	}
+	height = size / (pitch * bpe);
+	if (height > 8192)
+		height = 8192;
+	switch (G_0280A0_ARRAY_MODE(track->cb_color_info[i])) {
+	case V_0280A0_ARRAY_LINEAR_GENERAL:
+	case V_0280A0_ARRAY_LINEAR_ALIGNED:
+		if (pitch & 0x3f) {
+			dev_warn(p->dev, "%s:%d cb pitch (%d x %d = %d) invalid\n",
+				__func__, __LINE__, pitch, bpe, pitch * bpe);
+			return -EINVAL;
+		}
+		if ((pitch * bpe) & (track->group_size - 1)) {
+			dev_warn(p->dev, "%s:%d cb pitch (%d) invalid\n",
+				__func__, __LINE__, pitch);
+			return -EINVAL;
+		}
+		break;
+	case V_0280A0_ARRAY_1D_TILED_THIN1:
+		if ((pitch * 8 * bpe * track->nsamples) & (track->group_size - 1)) {
+			dev_warn(p->dev, "%s:%d cb pitch (%d) invalid\n",
+				__func__, __LINE__, pitch);
+			return -EINVAL;
+		}
+		height &= ~0x7;
+		if (!height)
+			height = 8;
+		break;
+	case V_0280A0_ARRAY_2D_TILED_THIN1:
+		if (pitch & ((8 * track->nbanks) - 1)) {
+			dev_warn(p->dev, "%s:%d cb pitch (%d) invalid\n",
+				__func__, __LINE__, pitch);
+			return -EINVAL;
+		}
+		tmp = pitch * 8 * bpe * track->nsamples;
+		tmp = tmp / track->nbanks;
+		if (tmp & (track->group_size - 1)) {
+			dev_warn(p->dev, "%s:%d cb pitch (%d) invalid\n",
+				__func__, __LINE__, pitch);
+			return -EINVAL;
+		}
+		height &= ~((16 * track->npipes) - 1);
+		if (!height)
+			height = 16 * track->npipes;
+		break;
+	default:
+		dev_warn(p->dev, "%s invalid tiling %d for %d (0x%08X)\n", __func__,
+			G_0280A0_ARRAY_MODE(track->cb_color_info[i]), i,
+			track->cb_color_info[i]);
+		return -EINVAL;
+	}
+	/* check offset */
+	tmp = height * pitch;
+	if ((tmp + track->cb_color_bo_offset[i]) > radeon_bo_size(track->cb_color_bo[i])) {
+		dev_warn(p->dev, "%s offset[%d] %d to big\n", __func__, i, track->cb_color_bo_offset[i]);
+		return -EINVAL;
+	}
+	/* limit max tile */
+	tmp = (height * pitch) >> 6;
+	if (tmp < slice_tile_max)
+		slice_tile_max = tmp;
+	tmp = S_028060_PITCH_TILE_MAX((pitch >> 3) - 1) |
+		S_028060_SLICE_TILE_MAX(slice_tile_max - 1);
+	ib[track->cb_color_size_idx[i]] = tmp;
+	return 0;
+}
+
+static int r600_cs_track_check(struct radeon_cs_parser *p)
+{
+	struct r600_cs_track *track = p->track;
+	u32 tmp;
+	int r, i;
+	volatile u32 *ib = p->ib->ptr;
+
+	/* on legacy kernel we don't perform advanced check */
+	if (p->rdev == NULL)
+		return 0;
+	/* we don't support out buffer yet */
+	if (track->vgt_strmout_en || track->vgt_strmout_buffer_en) {
+		dev_warn(p->dev, "this kernel doesn't support SMX output buffer\n");
+		return -EINVAL;
+	}
+	/* check that we have a cb for each enabled target, we don't check
+	 * shader_mask because it seems mesa isn't always setting it :(
+	 */
+	tmp = track->cb_target_mask;
+	for (i = 0; i < 8; i++) {
+		if ((tmp >> (i * 4)) & 0xF) {
+			/* at least one component is enabled */
+			if (track->cb_color_bo[i] == NULL) {
+				dev_warn(p->dev, "%s:%d mask 0x%08X | 0x%08X no cb for %d\n",
+					__func__, __LINE__, track->cb_target_mask, track->cb_shader_mask, i);
+				return -EINVAL;
+			}
+			/* perform rewrite of CB_COLOR[0-7]_SIZE */
+			r = r600_cs_track_validate_cb(p, i);
+			if (r)
+				return r;
+		}
+	}
+	/* Check depth buffer */
+	if (G_028800_STENCIL_ENABLE(track->db_depth_control) ||
+		G_028800_Z_ENABLE(track->db_depth_control)) {
+		u32 nviews, bpe, ntiles;
+		if (track->db_bo == NULL) {
+			dev_warn(p->dev, "z/stencil with no depth buffer\n");
+			return -EINVAL;
+		}
+		if (G_028010_TILE_SURFACE_ENABLE(track->db_depth_info)) {
+			dev_warn(p->dev, "this kernel doesn't support z/stencil htile\n");
+			return -EINVAL;
+		}
+		switch (G_028010_FORMAT(track->db_depth_info)) {
+		case V_028010_DEPTH_16:
+			bpe = 2;
+			break;
+		case V_028010_DEPTH_X8_24:
+		case V_028010_DEPTH_8_24:
+		case V_028010_DEPTH_X8_24_FLOAT:
+		case V_028010_DEPTH_8_24_FLOAT:
+		case V_028010_DEPTH_32_FLOAT:
+			bpe = 4;
+			break;
+		case V_028010_DEPTH_X24_8_32_FLOAT:
+			bpe = 8;
+			break;
+		default:
+			dev_warn(p->dev, "z/stencil with invalid format %d\n", G_028010_FORMAT(track->db_depth_info));
+			return -EINVAL;
+		}
+		if ((track->db_depth_size & 0xFFFFFC00) == 0xFFFFFC00) {
+			if (!track->db_depth_size_idx) {
+				dev_warn(p->dev, "z/stencil buffer size not set\n");
+				return -EINVAL;
+			}
+			printk_once(KERN_WARNING "You have old & broken userspace please consider updating mesa\n");
+			tmp = radeon_bo_size(track->db_bo) - track->db_offset;
+			tmp = (tmp / bpe) >> 6;
+			if (!tmp) {
+				dev_warn(p->dev, "z/stencil buffer too small (0x%08X %d %d %ld)\n",
+						track->db_depth_size, bpe, track->db_offset,
+						radeon_bo_size(track->db_bo));
+				return -EINVAL;
+			}
+			ib[track->db_depth_size_idx] = S_028000_SLICE_TILE_MAX(tmp - 1) | (track->db_depth_size & 0x3FF);
+		} else {
+			ntiles = G_028000_SLICE_TILE_MAX(track->db_depth_size) + 1;
+			nviews = G_028004_SLICE_MAX(track->db_depth_view) + 1;
+			tmp = ntiles * bpe * 64 * nviews;
+			if ((tmp + track->db_offset) > radeon_bo_size(track->db_bo)) {
+				dev_warn(p->dev, "z/stencil buffer too small (0x%08X %d %d %d -> %d have %ld)\n",
+						track->db_depth_size, ntiles, nviews, bpe, tmp + track->db_offset,
+						radeon_bo_size(track->db_bo));
+				return -EINVAL;
+			}
+		}
+	}
+	return 0;
+}
 
 /**
  * r600_cs_packet_parse() - parse cp packet and point ib index to next packet
@@ -57,7 +363,7 @@ int r600_cs_packet_parse(struct radeon_cs_parser *p,
 			  idx, ib_chunk->length_dw);
 		return -EINVAL;
 	}
-	header = ib_chunk->kdata[idx];
+	header = radeon_get_ib_value(p, idx);
 	pkt->idx = idx;
 	pkt->type = CP_PACKET_GET_TYPE(header);
 	pkt->count = CP_PACKET_GET_COUNT(header);
@@ -98,7 +404,6 @@ int r600_cs_packet_parse(struct radeon_cs_parser *p,
 static int r600_cs_packet_next_reloc_mm(struct radeon_cs_parser *p,
 					struct radeon_cs_reloc **cs_reloc)
 {
-	struct radeon_cs_chunk *ib_chunk;
 	struct radeon_cs_chunk *relocs_chunk;
 	struct radeon_cs_packet p3reloc;
 	unsigned idx;
@@ -109,7 +414,6 @@ static int r600_cs_packet_next_reloc_mm(struct radeon_cs_parser *p,
 		return -EINVAL;
 	}
 	*cs_reloc = NULL;
-	ib_chunk = &p->chunks[p->chunk_ib_idx];
 	relocs_chunk = &p->chunks[p->chunk_relocs_idx];
 	r = r600_cs_packet_parse(p, &p3reloc, p->idx);
 	if (r) {
@@ -121,7 +425,7 @@ static int r600_cs_packet_next_reloc_mm(struct radeon_cs_parser *p,
 			  p3reloc.idx);
 		return -EINVAL;
 	}
-	idx = ib_chunk->kdata[p3reloc.idx + 1];
+	idx = radeon_get_ib_value(p, p3reloc.idx + 1);
 	if (idx >= relocs_chunk->length_dw) {
 		DRM_ERROR("Relocs at %d after relocations chunk end %d !\n",
 			  idx, relocs_chunk->length_dw);
@@ -146,7 +450,6 @@ static int r600_cs_packet_next_reloc_mm(struct radeon_cs_parser *p,
 static int r600_cs_packet_next_reloc_nomm(struct radeon_cs_parser *p,
 					struct radeon_cs_reloc **cs_reloc)
 {
-	struct radeon_cs_chunk *ib_chunk;
 	struct radeon_cs_chunk *relocs_chunk;
 	struct radeon_cs_packet p3reloc;
 	unsigned idx;
@@ -157,7 +460,6 @@ static int r600_cs_packet_next_reloc_nomm(struct radeon_cs_parser *p,
 		return -EINVAL;
 	}
 	*cs_reloc = NULL;
-	ib_chunk = &p->chunks[p->chunk_ib_idx];
 	relocs_chunk = &p->chunks[p->chunk_relocs_idx];
 	r = r600_cs_packet_parse(p, &p3reloc, p->idx);
 	if (r) {
@@ -169,25 +471,170 @@ static int r600_cs_packet_next_reloc_nomm(struct radeon_cs_parser *p,
 			  p3reloc.idx);
 		return -EINVAL;
 	}
-	idx = ib_chunk->kdata[p3reloc.idx + 1];
+	idx = radeon_get_ib_value(p, p3reloc.idx + 1);
 	if (idx >= relocs_chunk->length_dw) {
 		DRM_ERROR("Relocs at %d after relocations chunk end %d !\n",
 			  idx, relocs_chunk->length_dw);
 		return -EINVAL;
 	}
-	*cs_reloc = &p->relocs[0];
+	*cs_reloc = p->relocs;
 	(*cs_reloc)->lobj.gpu_offset = (u64)relocs_chunk->kdata[idx + 3] << 32;
 	(*cs_reloc)->lobj.gpu_offset |= relocs_chunk->kdata[idx + 0];
 	return 0;
 }
 
+/**
+ * r600_cs_packet_next_is_pkt3_nop() - test if next packet is packet3 nop for reloc
+ * @parser:		parser structure holding parsing context.
+ *
+ * Check next packet is relocation packet3, do bo validation and compute
+ * GPU offset using the provided start.
+ **/
+static inline int r600_cs_packet_next_is_pkt3_nop(struct radeon_cs_parser *p)
+{
+	struct radeon_cs_packet p3reloc;
+	int r;
+
+	r = r600_cs_packet_parse(p, &p3reloc, p->idx);
+	if (r) {
+		return 0;
+	}
+	if (p3reloc.type != PACKET_TYPE3 || p3reloc.opcode != PACKET3_NOP) {
+		return 0;
+	}
+	return 1;
+}
+
+/**
+ * r600_cs_packet_next_vline() - parse userspace VLINE packet
+ * @parser:		parser structure holding parsing context.
+ *
+ * Userspace sends a special sequence for VLINE waits.
+ * PACKET0 - VLINE_START_END + value
+ * PACKET3 - WAIT_REG_MEM poll vline status reg
+ * RELOC (P3) - crtc_id in reloc.
+ *
+ * This function parses this and relocates the VLINE START END
+ * and WAIT_REG_MEM packets to the correct crtc.
+ * It also detects a switched off crtc and nulls out the
+ * wait in that case.
+ */
+static int r600_cs_packet_parse_vline(struct radeon_cs_parser *p)
+{
+	struct drm_mode_object *obj;
+	struct drm_crtc *crtc;
+	struct radeon_crtc *radeon_crtc;
+	struct radeon_cs_packet p3reloc, wait_reg_mem;
+	int crtc_id;
+	int r;
+	uint32_t header, h_idx, reg, wait_reg_mem_info;
+	volatile uint32_t *ib;
+
+	ib = p->ib->ptr;
+
+	/* parse the WAIT_REG_MEM */
+	r = r600_cs_packet_parse(p, &wait_reg_mem, p->idx);
+	if (r)
+		return r;
+
+	/* check its a WAIT_REG_MEM */
+	if (wait_reg_mem.type != PACKET_TYPE3 ||
+	    wait_reg_mem.opcode != PACKET3_WAIT_REG_MEM) {
+		DRM_ERROR("vline wait missing WAIT_REG_MEM segment\n");
+		r = -EINVAL;
+		return r;
+	}
+
+	wait_reg_mem_info = radeon_get_ib_value(p, wait_reg_mem.idx + 1);
+	/* bit 4 is reg (0) or mem (1) */
+	if (wait_reg_mem_info & 0x10) {
+		DRM_ERROR("vline WAIT_REG_MEM waiting on MEM rather than REG\n");
+		r = -EINVAL;
+		return r;
+	}
+	/* waiting for value to be equal */
+	if ((wait_reg_mem_info & 0x7) != 0x3) {
+		DRM_ERROR("vline WAIT_REG_MEM function not equal\n");
+		r = -EINVAL;
+		return r;
+	}
+	if ((radeon_get_ib_value(p, wait_reg_mem.idx + 2) << 2) != AVIVO_D1MODE_VLINE_STATUS) {
+		DRM_ERROR("vline WAIT_REG_MEM bad reg\n");
+		r = -EINVAL;
+		return r;
+	}
+
+	if (radeon_get_ib_value(p, wait_reg_mem.idx + 5) != AVIVO_D1MODE_VLINE_STAT) {
+		DRM_ERROR("vline WAIT_REG_MEM bad bit mask\n");
+		r = -EINVAL;
+		return r;
+	}
+
+	/* jump over the NOP */
+	r = r600_cs_packet_parse(p, &p3reloc, p->idx + wait_reg_mem.count + 2);
+	if (r)
+		return r;
+
+	h_idx = p->idx - 2;
+	p->idx += wait_reg_mem.count + 2;
+	p->idx += p3reloc.count + 2;
+
+	header = radeon_get_ib_value(p, h_idx);
+	crtc_id = radeon_get_ib_value(p, h_idx + 2 + 7 + 1);
+	reg = CP_PACKET0_GET_REG(header);
+	mutex_lock(&p->rdev->ddev->mode_config.mutex);
+	obj = drm_mode_object_find(p->rdev->ddev, crtc_id, DRM_MODE_OBJECT_CRTC);
+	if (!obj) {
+		DRM_ERROR("cannot find crtc %d\n", crtc_id);
+		r = -EINVAL;
+		goto out;
+	}
+	crtc = obj_to_crtc(obj);
+	radeon_crtc = to_radeon_crtc(crtc);
+	crtc_id = radeon_crtc->crtc_id;
+
+	if (!crtc->enabled) {
+		/* if the CRTC isn't enabled - we need to nop out the WAIT_REG_MEM */
+		ib[h_idx + 2] = PACKET2(0);
+		ib[h_idx + 3] = PACKET2(0);
+		ib[h_idx + 4] = PACKET2(0);
+		ib[h_idx + 5] = PACKET2(0);
+		ib[h_idx + 6] = PACKET2(0);
+		ib[h_idx + 7] = PACKET2(0);
+		ib[h_idx + 8] = PACKET2(0);
+	} else if (crtc_id == 1) {
+		switch (reg) {
+		case AVIVO_D1MODE_VLINE_START_END:
+			header &= ~R600_CP_PACKET0_REG_MASK;
+			header |= AVIVO_D2MODE_VLINE_START_END >> 2;
+			break;
+		default:
+			DRM_ERROR("unknown crtc reloc\n");
+			r = -EINVAL;
+			goto out;
+		}
+		ib[h_idx] = header;
+		ib[h_idx + 4] = AVIVO_D2MODE_VLINE_STATUS >> 2;
+	}
+out:
+	mutex_unlock(&p->rdev->ddev->mode_config.mutex);
+	return r;
+}
+
 static int r600_packet0_check(struct radeon_cs_parser *p,
 				struct radeon_cs_packet *pkt,
 				unsigned idx, unsigned reg)
 {
+	int r;
+
 	switch (reg) {
 	case AVIVO_D1MODE_VLINE_START_END:
-	case AVIVO_D2MODE_VLINE_START_END:
+		r = r600_cs_packet_parse_vline(p);
+		if (r) {
+			DRM_ERROR("No reloc for ib[%d]=0x%04X\n",
+					idx, reg);
+			return r;
+		}
 		break;
 	default:
 		printk(KERN_ERR "Forbidden register 0x%04X in cs at %d\n",
@@ -215,20 +662,351 @@ static int r600_cs_parse_packet0(struct radeon_cs_parser *p,
 	return 0;
 }
 
+/**
+ * r600_cs_check_reg() - check if register is authorized or not
+ * @parser: parser structure holding parsing context
+ * @reg: register we are testing
+ * @idx: index into the cs buffer
+ *
+ * This function will test against r600_reg_safe_bm and return 0
+ * if register is safe. If register is not flag as safe this function
+ * will test it against a list of register needind special handling.
+ */
+static inline int r600_cs_check_reg(struct radeon_cs_parser *p, u32 reg, u32 idx)
+{
+	struct r600_cs_track *track = (struct r600_cs_track *)p->track;
+	struct radeon_cs_reloc *reloc;
+	u32 last_reg = ARRAY_SIZE(r600_reg_safe_bm);
+	u32 m, i, tmp, *ib;
+	int r;
+
+	i = (reg >> 7);
+	if (i > last_reg) {
+		dev_warn(p->dev, "forbidden register 0x%08x at %d\n", reg, idx);
+		return -EINVAL;
+	}
+	m = 1 << ((reg >> 2) & 31);
+	if (!(r600_reg_safe_bm[i] & m))
+		return 0;
+	ib = p->ib->ptr;
+	switch (reg) {
+	/* force following reg to 0 in an attemp to disable out buffer
+	 * which will need us to better understand how it works to perform
+	 * security check on it (Jerome)
+	 */
+	case R_0288A8_SQ_ESGS_RING_ITEMSIZE:
+	case R_008C44_SQ_ESGS_RING_SIZE:
+	case R_0288B0_SQ_ESTMP_RING_ITEMSIZE:
+	case R_008C54_SQ_ESTMP_RING_SIZE:
+	case R_0288C0_SQ_FBUF_RING_ITEMSIZE:
+	case R_008C74_SQ_FBUF_RING_SIZE:
+	case R_0288B4_SQ_GSTMP_RING_ITEMSIZE:
+	case R_008C5C_SQ_GSTMP_RING_SIZE:
+	case R_0288AC_SQ_GSVS_RING_ITEMSIZE:
+	case R_008C4C_SQ_GSVS_RING_SIZE:
+	case R_0288BC_SQ_PSTMP_RING_ITEMSIZE:
+	case R_008C6C_SQ_PSTMP_RING_SIZE:
+	case R_0288C4_SQ_REDUC_RING_ITEMSIZE:
+	case R_008C7C_SQ_REDUC_RING_SIZE:
+	case R_0288B8_SQ_VSTMP_RING_ITEMSIZE:
+	case R_008C64_SQ_VSTMP_RING_SIZE:
+	case R_0288C8_SQ_GS_VERT_ITEMSIZE:
+		/* get value to populate the IB don't remove */
+		tmp =radeon_get_ib_value(p, idx);
+		ib[idx] = 0;
+		break;
+	case R_028800_DB_DEPTH_CONTROL:
+		track->db_depth_control = radeon_get_ib_value(p, idx);
+		break;
+	case R_028010_DB_DEPTH_INFO:
+		track->db_depth_info = radeon_get_ib_value(p, idx);
+		break;
+	case R_028004_DB_DEPTH_VIEW:
+		track->db_depth_view = radeon_get_ib_value(p, idx);
+		break;
+	case R_028000_DB_DEPTH_SIZE:
+		track->db_depth_size = radeon_get_ib_value(p, idx);
+		track->db_depth_size_idx = idx;
+		break;
+	case R_028AB0_VGT_STRMOUT_EN:
+		track->vgt_strmout_en = radeon_get_ib_value(p, idx);
+		break;
+	case R_028B20_VGT_STRMOUT_BUFFER_EN:
+		track->vgt_strmout_buffer_en = radeon_get_ib_value(p, idx);
+		break;
+	case R_028238_CB_TARGET_MASK:
+		track->cb_target_mask = radeon_get_ib_value(p, idx);
+		break;
+	case R_02823C_CB_SHADER_MASK:
+		track->cb_shader_mask = radeon_get_ib_value(p, idx);
+		break;
+	case R_028C04_PA_SC_AA_CONFIG:
+		tmp = G_028C04_MSAA_NUM_SAMPLES(radeon_get_ib_value(p, idx));
+		track->nsamples = 1 << tmp;
+		break;
+	case R_0280A0_CB_COLOR0_INFO:
+	case R_0280A4_CB_COLOR1_INFO:
+	case R_0280A8_CB_COLOR2_INFO:
+	case R_0280AC_CB_COLOR3_INFO:
+	case R_0280B0_CB_COLOR4_INFO:
+	case R_0280B4_CB_COLOR5_INFO:
+	case R_0280B8_CB_COLOR6_INFO:
+	case R_0280BC_CB_COLOR7_INFO:
+		tmp = (reg - R_0280A0_CB_COLOR0_INFO) / 4;
+		track->cb_color_info[tmp] = radeon_get_ib_value(p, idx);
+		break;
+	case R_028060_CB_COLOR0_SIZE:
+	case R_028064_CB_COLOR1_SIZE:
+	case R_028068_CB_COLOR2_SIZE:
+	case R_02806C_CB_COLOR3_SIZE:
+	case R_028070_CB_COLOR4_SIZE:
+	case R_028074_CB_COLOR5_SIZE:
+	case R_028078_CB_COLOR6_SIZE:
+	case R_02807C_CB_COLOR7_SIZE:
+		tmp = (reg - R_028060_CB_COLOR0_SIZE) / 4;
+		track->cb_color_size[tmp] = radeon_get_ib_value(p, idx);
+		track->cb_color_size_idx[tmp] = idx;
+		break;
+		/* This register were added late, there is userspace
+		 * which does provide relocation for those but set
+		 * 0 offset. In order to avoid breaking old userspace
+		 * we detect this and set address to point to last
+		 * CB_COLOR0_BASE, note that if userspace doesn't set
+		 * CB_COLOR0_BASE before this register we will report
+		 * error. Old userspace always set CB_COLOR0_BASE
+		 * before any of this.
+		 */
+	case R_0280E0_CB_COLOR0_FRAG:
+	case R_0280E4_CB_COLOR1_FRAG:
+	case R_0280E8_CB_COLOR2_FRAG:
+	case R_0280EC_CB_COLOR3_FRAG:
+	case R_0280F0_CB_COLOR4_FRAG:
+	case R_0280F4_CB_COLOR5_FRAG:
+	case R_0280F8_CB_COLOR6_FRAG:
+	case R_0280FC_CB_COLOR7_FRAG:
+		tmp = (reg - R_0280E0_CB_COLOR0_FRAG) / 4;
+		if (!r600_cs_packet_next_is_pkt3_nop(p)) {
+			if (!track->cb_color_base_last[tmp]) {
+				dev_err(p->dev, "Broken old userspace ? no cb_color0_base supplied before trying to write 0x%08X\n", reg);
+				return -EINVAL;
+			}
+			ib[idx] = track->cb_color_base_last[tmp];
+			printk_once(KERN_WARNING "You have old & broken userspace "
+					"please consider updating mesa & xf86-video-ati\n");
+			track->cb_color_frag_bo[tmp] = track->cb_color_bo[tmp];
+		} else {
+			r = r600_cs_packet_next_reloc(p, &reloc);
+			if (r) {
+				dev_err(p->dev, "bad SET_CONTEXT_REG 0x%04X\n", reg);
+				return -EINVAL;
+			}
+			ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
+			track->cb_color_frag_bo[tmp] = reloc->robj;
+		}
+		break;
+	case R_0280C0_CB_COLOR0_TILE:
+	case R_0280C4_CB_COLOR1_TILE:
+	case R_0280C8_CB_COLOR2_TILE:
+	case R_0280CC_CB_COLOR3_TILE:
+	case R_0280D0_CB_COLOR4_TILE:
+	case R_0280D4_CB_COLOR5_TILE:
+	case R_0280D8_CB_COLOR6_TILE:
+	case R_0280DC_CB_COLOR7_TILE:
+		tmp = (reg - R_0280C0_CB_COLOR0_TILE) / 4;
+		if (!r600_cs_packet_next_is_pkt3_nop(p)) {
+			if (!track->cb_color_base_last[tmp]) {
+				dev_err(p->dev, "Broken old userspace ? no cb_color0_base supplied before trying to write 0x%08X\n", reg);
+				return -EINVAL;
+			}
+			ib[idx] = track->cb_color_base_last[tmp];
+			printk_once(KERN_WARNING "You have old & broken userspace "
+					"please consider updating mesa & xf86-video-ati\n");
+			track->cb_color_tile_bo[tmp] = track->cb_color_bo[tmp];
+		} else {
+			r = r600_cs_packet_next_reloc(p, &reloc);
+			if (r) {
+				dev_err(p->dev, "bad SET_CONTEXT_REG 0x%04X\n", reg);
+				return -EINVAL;
+			}
+			ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
+			track->cb_color_tile_bo[tmp] = reloc->robj;
+		}
+		break;
+	case CB_COLOR0_BASE:
+	case CB_COLOR1_BASE:
+	case CB_COLOR2_BASE:
+	case CB_COLOR3_BASE:
+	case CB_COLOR4_BASE:
+	case CB_COLOR5_BASE:
+	case CB_COLOR6_BASE:
+	case CB_COLOR7_BASE:
+		r = r600_cs_packet_next_reloc(p, &reloc);
+		if (r) {
+			dev_warn(p->dev, "bad SET_CONTEXT_REG "
+					"0x%04X\n", reg);
+			return -EINVAL;
+		}
+		tmp = (reg - CB_COLOR0_BASE) / 4;
+		track->cb_color_bo_offset[tmp] = radeon_get_ib_value(p, idx);
+		ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
+		track->cb_color_base_last[tmp] = ib[idx];
+		track->cb_color_bo[tmp] = reloc->robj;
+		break;
+	case DB_DEPTH_BASE:
+		r = r600_cs_packet_next_reloc(p, &reloc);
+		if (r) {
+			dev_warn(p->dev, "bad SET_CONTEXT_REG "
+					"0x%04X\n", reg);
+			return -EINVAL;
+		}
+		track->db_offset = radeon_get_ib_value(p, idx);
+		ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
+		track->db_bo = reloc->robj;
+		break;
+	case DB_HTILE_DATA_BASE:
+	case SQ_PGM_START_FS:
+	case SQ_PGM_START_ES:
+	case SQ_PGM_START_VS:
+	case SQ_PGM_START_GS:
+	case SQ_PGM_START_PS:
+		r = r600_cs_packet_next_reloc(p, &reloc);
+		if (r) {
+			dev_warn(p->dev, "bad SET_CONTEXT_REG "
+					"0x%04X\n", reg);
+			return -EINVAL;
+		}
+		ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
+		break;
+	default:
+		dev_warn(p->dev, "forbidden register 0x%08x at %d\n", reg, idx);
+		return -EINVAL;
+	}
+	return 0;
+}
+
+static inline unsigned minify(unsigned size, unsigned levels)
+{
+	size = size >> levels;
+	if (size < 1)
+		size = 1;
+	return size;
+}
+
+static void r600_texture_size(unsigned nfaces, unsigned blevel, unsigned nlevels,
+				unsigned w0, unsigned h0, unsigned d0, unsigned bpe,
+				unsigned *l0_size, unsigned *mipmap_size)
+{
+	unsigned offset, i, level, face;
+	unsigned width, height, depth, rowstride, size;
+
+	w0 = minify(w0, 0);
+	h0 = minify(h0, 0);
+	d0 = minify(d0, 0);
+	for(i = 0, offset = 0, level = blevel; i < nlevels; i++, level++) {
+		width = minify(w0, i);
+		height = minify(h0, i);
+		depth = minify(d0, i);
+		for(face = 0; face < nfaces; face++) {
+			rowstride = ((width * bpe) + 255) & ~255;
+			size = height * rowstride * depth;
+			offset += size;
+			offset = (offset + 0x1f) & ~0x1f;
+		}
+	}
+	*l0_size = (((w0 * bpe) + 255) & ~255) * h0 * d0;
+	*mipmap_size = offset;
+	if (!blevel)
+		*mipmap_size -= *l0_size;
+	if (!nlevels)
+		*mipmap_size = *l0_size;
+}
+
+/**
+ * r600_check_texture_resource() - check if register is authorized or not
+ * @p: parser structure holding parsing context
+ * @idx: index into the cs buffer
+ * @texture: texture's bo structure
+ * @mipmap: mipmap's bo structure
+ *
+ * This function will check that the resource has valid field and that
+ * the texture and mipmap bo object are big enough to cover this resource.
+ */
+static inline int r600_check_texture_resource(struct radeon_cs_parser *p,  u32 idx,
+						struct radeon_bo *texture,
+						struct radeon_bo *mipmap)
+{
+	u32 nfaces, nlevels, blevel, w0, h0, d0, bpe = 0;
+	u32 word0, word1, l0_size, mipmap_size;
+
+	/* on legacy kernel we don't perform advanced check */
+	if (p->rdev == NULL)
+		return 0;
+	word0 = radeon_get_ib_value(p, idx + 0);
+	word1 = radeon_get_ib_value(p, idx + 1);
+	w0 = G_038000_TEX_WIDTH(word0) + 1;
+	h0 = G_038004_TEX_HEIGHT(word1) + 1;
+	d0 = G_038004_TEX_DEPTH(word1);
+	nfaces = 1;
+	switch (G_038000_DIM(word0)) {
+	case V_038000_SQ_TEX_DIM_1D:
+	case V_038000_SQ_TEX_DIM_2D:
+	case V_038000_SQ_TEX_DIM_3D:
+		break;
+	case V_038000_SQ_TEX_DIM_CUBEMAP:
+		nfaces = 6;
+		break;
+	case V_038000_SQ_TEX_DIM_1D_ARRAY:
+	case V_038000_SQ_TEX_DIM_2D_ARRAY:
+	case V_038000_SQ_TEX_DIM_2D_MSAA:
+	case V_038000_SQ_TEX_DIM_2D_ARRAY_MSAA:
+	default:
+		dev_warn(p->dev, "this kernel doesn't support %d texture dim\n", G_038000_DIM(word0));
+		return -EINVAL;
+	}
+	if (r600_bpe_from_format(&bpe,  G_038004_DATA_FORMAT(word1))) {
+		dev_warn(p->dev, "%s:%d texture invalid format %d\n",
+			 __func__, __LINE__, G_038004_DATA_FORMAT(word1));
+		return -EINVAL;
+	}
+	word0 = radeon_get_ib_value(p, idx + 4);
+	word1 = radeon_get_ib_value(p, idx + 5);
+	blevel = G_038010_BASE_LEVEL(word0);
+	nlevels = G_038014_LAST_LEVEL(word1);
+	r600_texture_size(nfaces, blevel, nlevels, w0, h0, d0, bpe, &l0_size, &mipmap_size);
+	/* using get ib will give us the offset into the texture bo */
+	word0 = radeon_get_ib_value(p, idx + 2);
+	if ((l0_size + word0) > radeon_bo_size(texture)) {
+		dev_warn(p->dev, "texture bo too small (%d %d %d %d -> %d have %ld)\n",
+			w0, h0, bpe, word0, l0_size, radeon_bo_size(texture));
+		return -EINVAL;
+	}
+	/* using get ib will give us the offset into the mipmap bo */
+	word0 = radeon_get_ib_value(p, idx + 3);
+	if ((mipmap_size + word0) > radeon_bo_size(mipmap)) {
+		dev_warn(p->dev, "mipmap bo too small (%d %d %d %d %d %d -> %d have %ld)\n",
+			w0, h0, bpe, blevel, nlevels, word0, mipmap_size, radeon_bo_size(texture));
+		return -EINVAL;
+	}
+	return 0;
+}
+
 static int r600_packet3_check(struct radeon_cs_parser *p,
 				struct radeon_cs_packet *pkt)
 {
-	struct radeon_cs_chunk *ib_chunk;
 	struct radeon_cs_reloc *reloc;
+	struct r600_cs_track *track;
 	volatile u32 *ib;
 	unsigned idx;
 	unsigned i;
 	unsigned start_reg, end_reg, reg;
 	int r;
+	u32 idx_value;
 
+	track = (struct r600_cs_track *)p->track;
 	ib = p->ib->ptr;
-	ib_chunk = &p->chunks[p->chunk_ib_idx];
 	idx = pkt->idx + 1;
+	idx_value = radeon_get_ib_value(p, idx);
+
 	switch (pkt->opcode) {
 	case PACKET3_START_3D_CMDBUF:
 		if (p->family >= CHIP_RV770 || pkt->count) {
@@ -259,14 +1037,24 @@ static int r600_packet3_check(struct radeon_cs_parser *p,
 			DRM_ERROR("bad DRAW_INDEX\n");
 			return -EINVAL;
 		}
-		ib[idx+0] += (u32)(reloc->lobj.gpu_offset & 0xffffffff);
-		ib[idx+1] = upper_32_bits(reloc->lobj.gpu_offset) & 0xff;
+		ib[idx+0] = idx_value + (u32)(reloc->lobj.gpu_offset & 0xffffffff);
+		ib[idx+1] += upper_32_bits(reloc->lobj.gpu_offset) & 0xff;
+		r = r600_cs_track_check(p);
+		if (r) {
+			dev_warn(p->dev, "%s:%d invalid cmd stream\n", __func__, __LINE__);
+			return r;
+		}
 		break;
 	case PACKET3_DRAW_INDEX_AUTO:
 		if (pkt->count != 1) {
 			DRM_ERROR("bad DRAW_INDEX_AUTO\n");
 			return -EINVAL;
 		}
+		r = r600_cs_track_check(p);
+		if (r) {
+			dev_warn(p->dev, "%s:%d invalid cmd stream %d\n", __func__, __LINE__, idx);
+			return r;
+		}
 		break;
 	case PACKET3_DRAW_INDEX_IMMD_BE:
 	case PACKET3_DRAW_INDEX_IMMD:
@@ -274,6 +1062,11 @@ static int r600_packet3_check(struct radeon_cs_parser *p,
 			DRM_ERROR("bad DRAW_INDEX_IMMD\n");
 			return -EINVAL;
 		}
+		r = r600_cs_track_check(p);
+		if (r) {
+			dev_warn(p->dev, "%s:%d invalid cmd stream\n", __func__, __LINE__);
+			return r;
+		}
 		break;
 	case PACKET3_WAIT_REG_MEM:
 		if (pkt->count != 5) {
@@ -281,14 +1074,14 @@ static int r600_packet3_check(struct radeon_cs_parser *p,
 			return -EINVAL;
 		}
 		/* bit 4 is reg (0) or mem (1) */
-		if (ib_chunk->kdata[idx+0] & 0x10) {
+		if (idx_value & 0x10) {
 			r = r600_cs_packet_next_reloc(p, &reloc);
 			if (r) {
 				DRM_ERROR("bad WAIT_REG_MEM\n");
 				return -EINVAL;
 			}
 			ib[idx+1] += (u32)(reloc->lobj.gpu_offset & 0xffffffff);
-			ib[idx+2] = upper_32_bits(reloc->lobj.gpu_offset) & 0xff;
+			ib[idx+2] += upper_32_bits(reloc->lobj.gpu_offset) & 0xff;
 		}
 		break;
 	case PACKET3_SURFACE_SYNC:
@@ -297,8 +1090,8 @@ static int r600_packet3_check(struct radeon_cs_parser *p,
 			return -EINVAL;
 		}
 		/* 0xffffffff/0x0 is flush all cache flag */
-		if (ib_chunk->kdata[idx+1] != 0xffffffff ||
-		    ib_chunk->kdata[idx+2] != 0) {
+		if (radeon_get_ib_value(p, idx + 1) != 0xffffffff ||
+		    radeon_get_ib_value(p, idx + 2) != 0) {
 			r = r600_cs_packet_next_reloc(p, &reloc);
 			if (r) {
 				DRM_ERROR("bad SURFACE_SYNC\n");
@@ -319,7 +1112,7 @@ static int r600_packet3_check(struct radeon_cs_parser *p,
 				return -EINVAL;
 			}
 			ib[idx+1] += (u32)(reloc->lobj.gpu_offset & 0xffffffff);
-			ib[idx+2] |= upper_32_bits(reloc->lobj.gpu_offset) & 0xff;
+			ib[idx+2] += upper_32_bits(reloc->lobj.gpu_offset) & 0xff;
 		}
 		break;
 	case PACKET3_EVENT_WRITE_EOP:
@@ -333,10 +1126,10 @@ static int r600_packet3_check(struct radeon_cs_parser *p,
 			return -EINVAL;
 		}
 		ib[idx+1] += (u32)(reloc->lobj.gpu_offset & 0xffffffff);
-		ib[idx+2] |= upper_32_bits(reloc->lobj.gpu_offset) & 0xff;
+		ib[idx+2] += upper_32_bits(reloc->lobj.gpu_offset) & 0xff;
 		break;
 	case PACKET3_SET_CONFIG_REG:
-		start_reg = (ib[idx+0] << 2) + PACKET3_SET_CONFIG_REG_OFFSET;
+		start_reg = (idx_value << 2) + PACKET3_SET_CONFIG_REG_OFFSET;
 		end_reg = 4 * pkt->count + start_reg - 4;
 		if ((start_reg < PACKET3_SET_CONFIG_REG_OFFSET) ||
 		    (start_reg >= PACKET3_SET_CONFIG_REG_END) ||
@@ -346,17 +1139,13 @@ static int r600_packet3_check(struct radeon_cs_parser *p,
 		}
 		for (i = 0; i < pkt->count; i++) {
 			reg = start_reg + (4 * i);
-			switch (reg) {
-			case CP_COHER_BASE:
-				/* use PACKET3_SURFACE_SYNC */
-				return -EINVAL;
-			default:
-				break;
-			}
+			r = r600_cs_check_reg(p, reg, idx+1+i);
+			if (r)
+				return r;
 		}
 		break;
 	case PACKET3_SET_CONTEXT_REG:
-		start_reg = (ib[idx+0] << 2) + PACKET3_SET_CONTEXT_REG_OFFSET;
+		start_reg = (idx_value << 2) + PACKET3_SET_CONTEXT_REG_OFFSET;
 		end_reg = 4 * pkt->count + start_reg - 4;
 		if ((start_reg < PACKET3_SET_CONTEXT_REG_OFFSET) ||
 		    (start_reg >= PACKET3_SET_CONTEXT_REG_END) ||
@@ -366,54 +1155,9 @@ static int r600_packet3_check(struct radeon_cs_parser *p,
 		}
 		for (i = 0; i < pkt->count; i++) {
 			reg = start_reg + (4 * i);
-			switch (reg) {
-			case DB_DEPTH_BASE:
-			case CB_COLOR0_BASE:
-			case CB_COLOR1_BASE:
-			case CB_COLOR2_BASE:
-			case CB_COLOR3_BASE:
-			case CB_COLOR4_BASE:
-			case CB_COLOR5_BASE:
-			case CB_COLOR6_BASE:
-			case CB_COLOR7_BASE:
-			case SQ_PGM_START_FS:
-			case SQ_PGM_START_ES:
-			case SQ_PGM_START_VS:
-			case SQ_PGM_START_GS:
-			case SQ_PGM_START_PS:
-				r = r600_cs_packet_next_reloc(p, &reloc);
-				if (r) {
-					DRM_ERROR("bad SET_CONTEXT_REG "
-							"0x%04X\n", reg);
-					return -EINVAL;
-				}
-				ib[idx+1+i] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
-				break;
-			case VGT_DMA_BASE:
-			case VGT_DMA_BASE_HI:
-				/* These should be handled by DRAW_INDEX packet 3 */
-			case VGT_STRMOUT_BASE_OFFSET_0:
-			case VGT_STRMOUT_BASE_OFFSET_1:
-			case VGT_STRMOUT_BASE_OFFSET_2:
-			case VGT_STRMOUT_BASE_OFFSET_3:
-			case VGT_STRMOUT_BASE_OFFSET_HI_0:
-			case VGT_STRMOUT_BASE_OFFSET_HI_1:
-			case VGT_STRMOUT_BASE_OFFSET_HI_2:
-			case VGT_STRMOUT_BASE_OFFSET_HI_3:
-			case VGT_STRMOUT_BUFFER_BASE_0:
-			case VGT_STRMOUT_BUFFER_BASE_1:
-			case VGT_STRMOUT_BUFFER_BASE_2:
-			case VGT_STRMOUT_BUFFER_BASE_3:
-			case VGT_STRMOUT_BUFFER_OFFSET_0:
-			case VGT_STRMOUT_BUFFER_OFFSET_1:
-			case VGT_STRMOUT_BUFFER_OFFSET_2:
-			case VGT_STRMOUT_BUFFER_OFFSET_3:
-				/* These should be handled by STRMOUT_BUFFER packet 3 */
-				DRM_ERROR("bad context reg: 0x%08x\n", reg);
-				return -EINVAL;
-			default:
-				break;
-			}
+			r = r600_cs_check_reg(p, reg, idx+1+i);
+			if (r)
+				return r;
 		}
 		break;
 	case PACKET3_SET_RESOURCE:
@@ -421,7 +1165,7 @@ static int r600_packet3_check(struct radeon_cs_parser *p,
 			DRM_ERROR("bad SET_RESOURCE\n");
 			return -EINVAL;
 		}
-		start_reg = (ib[idx+0] << 2) + PACKET3_SET_RESOURCE_OFFSET;
+		start_reg = (idx_value << 2) + PACKET3_SET_RESOURCE_OFFSET;
 		end_reg = 4 * pkt->count + start_reg - 4;
 		if ((start_reg < PACKET3_SET_RESOURCE_OFFSET) ||
 		    (start_reg >= PACKET3_SET_RESOURCE_END) ||
@@ -430,7 +1174,10 @@ static int r600_packet3_check(struct radeon_cs_parser *p,
 			return -EINVAL;
 		}
 		for (i = 0; i < (pkt->count / 7); i++) {
-			switch (G__SQ_VTX_CONSTANT_TYPE(ib[idx+(i*7)+6+1])) {
+			struct radeon_bo *texture, *mipmap;
+			u32 size, offset;
+
+			switch (G__SQ_VTX_CONSTANT_TYPE(radeon_get_ib_value(p, idx+(i*7)+6+1))) {
 			case SQ_TEX_VTX_VALID_TEXTURE:
 				/* tex base */
 				r = r600_cs_packet_next_reloc(p, &reloc);
@@ -439,6 +1186,7 @@ static int r600_packet3_check(struct radeon_cs_parser *p,
 					return -EINVAL;
 				}
 				ib[idx+1+(i*7)+2] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
+				texture = reloc->robj;
 				/* tex mip base */
 				r = r600_cs_packet_next_reloc(p, &reloc);
 				if (r) {
@@ -446,6 +1194,11 @@ static int r600_packet3_check(struct radeon_cs_parser *p,
 					return -EINVAL;
 				}
 				ib[idx+1+(i*7)+3] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
+				mipmap = reloc->robj;
+				r = r600_check_texture_resource(p,  idx+(i*7)+1,
+						texture, mipmap);
+				if (r)
+					return r;
 				break;
 			case SQ_TEX_VTX_VALID_BUFFER:
 				/* vtx base */
@@ -454,8 +1207,15 @@ static int r600_packet3_check(struct radeon_cs_parser *p,
 					DRM_ERROR("bad SET_RESOURCE\n");
 					return -EINVAL;
 				}
+				offset = radeon_get_ib_value(p, idx+1+(i*7)+0);
+				size = radeon_get_ib_value(p, idx+1+(i*7)+1);
+				if (p->rdev && (size + offset) > radeon_bo_size(reloc->robj)) {
+					/* force size to size of the buffer */
+					dev_warn(p->dev, "vbo resource seems too big for the bo\n");
+					ib[idx+1+(i*7)+1] = radeon_bo_size(reloc->robj);
+				}
 				ib[idx+1+(i*7)+0] += (u32)((reloc->lobj.gpu_offset) & 0xffffffff);
-				ib[idx+1+(i*7)+2] |= upper_32_bits(reloc->lobj.gpu_offset) & 0xff;
+				ib[idx+1+(i*7)+2] += upper_32_bits(reloc->lobj.gpu_offset) & 0xff;
 				break;
 			case SQ_TEX_VTX_INVALID_TEXTURE:
 			case SQ_TEX_VTX_INVALID_BUFFER:
@@ -466,7 +1226,7 @@ static int r600_packet3_check(struct radeon_cs_parser *p,
 		}
 		break;
 	case PACKET3_SET_ALU_CONST:
-		start_reg = (ib[idx+0] << 2) + PACKET3_SET_ALU_CONST_OFFSET;
+		start_reg = (idx_value << 2) + PACKET3_SET_ALU_CONST_OFFSET;
 		end_reg = 4 * pkt->count + start_reg - 4;
 		if ((start_reg < PACKET3_SET_ALU_CONST_OFFSET) ||
 		    (start_reg >= PACKET3_SET_ALU_CONST_END) ||
@@ -476,7 +1236,7 @@ static int r600_packet3_check(struct radeon_cs_parser *p,
 		}
 		break;
 	case PACKET3_SET_BOOL_CONST:
-		start_reg = (ib[idx+0] << 2) + PACKET3_SET_BOOL_CONST_OFFSET;
+		start_reg = (idx_value << 2) + PACKET3_SET_BOOL_CONST_OFFSET;
 		end_reg = 4 * pkt->count + start_reg - 4;
 		if ((start_reg < PACKET3_SET_BOOL_CONST_OFFSET) ||
 		    (start_reg >= PACKET3_SET_BOOL_CONST_END) ||
@@ -486,7 +1246,7 @@ static int r600_packet3_check(struct radeon_cs_parser *p,
 		}
 		break;
 	case PACKET3_SET_LOOP_CONST:
-		start_reg = (ib[idx+0] << 2) + PACKET3_SET_LOOP_CONST_OFFSET;
+		start_reg = (idx_value << 2) + PACKET3_SET_LOOP_CONST_OFFSET;
 		end_reg = 4 * pkt->count + start_reg - 4;
 		if ((start_reg < PACKET3_SET_LOOP_CONST_OFFSET) ||
 		    (start_reg >= PACKET3_SET_LOOP_CONST_END) ||
@@ -496,7 +1256,7 @@ static int r600_packet3_check(struct radeon_cs_parser *p,
 		}
 		break;
 	case PACKET3_SET_CTL_CONST:
-		start_reg = (ib[idx+0] << 2) + PACKET3_SET_CTL_CONST_OFFSET;
+		start_reg = (idx_value << 2) + PACKET3_SET_CTL_CONST_OFFSET;
 		end_reg = 4 * pkt->count + start_reg - 4;
 		if ((start_reg < PACKET3_SET_CTL_CONST_OFFSET) ||
 		    (start_reg >= PACKET3_SET_CTL_CONST_END) ||
@@ -510,7 +1270,7 @@ static int r600_packet3_check(struct radeon_cs_parser *p,
 			DRM_ERROR("bad SET_SAMPLER\n");
 			return -EINVAL;
 		}
-		start_reg = (ib[idx+0] << 2) + PACKET3_SET_SAMPLER_OFFSET;
+		start_reg = (idx_value << 2) + PACKET3_SET_SAMPLER_OFFSET;
 		end_reg = 4 * pkt->count + start_reg - 4;
 		if ((start_reg < PACKET3_SET_SAMPLER_OFFSET) ||
 		    (start_reg >= PACKET3_SET_SAMPLER_END) ||
@@ -541,11 +1301,31 @@ static int r600_packet3_check(struct radeon_cs_parser *p,
 int r600_cs_parse(struct radeon_cs_parser *p)
 {
 	struct radeon_cs_packet pkt;
+	struct r600_cs_track *track;
 	int r;
 
+	if (p->track == NULL) {
+		/* initialize tracker, we are in kms */
+		track = kzalloc(sizeof(*track), GFP_KERNEL);
+		if (track == NULL)
+			return -ENOMEM;
+		r600_cs_track_init(track);
+		if (p->rdev->family < CHIP_RV770) {
+			track->npipes = p->rdev->config.r600.tiling_npipes;
+			track->nbanks = p->rdev->config.r600.tiling_nbanks;
+			track->group_size = p->rdev->config.r600.tiling_group_size;
+		} else if (p->rdev->family <= CHIP_RV740) {
+			track->npipes = p->rdev->config.rv770.tiling_npipes;
+			track->nbanks = p->rdev->config.rv770.tiling_nbanks;
+			track->group_size = p->rdev->config.rv770.tiling_group_size;
+		}
+		p->track = track;
+	}
 	do {
 		r = r600_cs_packet_parse(p, &pkt, p->idx);
 		if (r) {
+			kfree(p->track);
+			p->track = NULL;
 			return r;
 		}
 		p->idx += pkt.count + 2;
@@ -560,9 +1340,13 @@ int r600_cs_parse(struct radeon_cs_parser *p)
 			break;
 		default:
 			DRM_ERROR("Unknown packet type %d !\n", pkt.type);
+			kfree(p->track);
+			p->track = NULL;
 			return -EINVAL;
 		}
 		if (r) {
+			kfree(p->track);
+			p->track = NULL;
 			return r;
 		}
 	} while (p->idx < p->chunks[p->chunk_ib_idx].length_dw);
@@ -572,6 +1356,8 @@ int r600_cs_parse(struct radeon_cs_parser *p)
 		mdelay(1);
 	}
 #endif
+	kfree(p->track);
+	p->track = NULL;
 	return 0;
 }
 
@@ -580,7 +1366,7 @@ static int r600_cs_parser_relocs_legacy(struct radeon_cs_parser *p)
 	if (p->chunk_relocs_idx == -1) {
 		return 0;
 	}
-	p->relocs = kcalloc(1, sizeof(struct radeon_cs_reloc), GFP_KERNEL);
+	p->relocs = kzalloc(sizeof(struct radeon_cs_reloc), GFP_KERNEL);
 	if (p->relocs == NULL) {
 		return -ENOMEM;
 	}
@@ -602,6 +1388,8 @@ static void r600_cs_parser_fini(struct radeon_cs_parser *parser, int error)
 	kfree(parser->relocs);
 	for (i = 0; i < parser->nchunks; i++) {
 		kfree(parser->chunks[i].kdata);
+		kfree(parser->chunks[i].kpage[0]);
+		kfree(parser->chunks[i].kpage[1]);
 	}
 	kfree(parser->chunks);
 	kfree(parser->chunks_array);
@@ -612,15 +1400,24 @@ int r600_cs_legacy(struct drm_device *dev, void *data, struct drm_file *filp,
 {
 	struct radeon_cs_parser parser;
 	struct radeon_cs_chunk *ib_chunk;
-	struct radeon_ib	fake_ib;
+	struct radeon_ib fake_ib;
+	struct r600_cs_track *track;
 	int r;
 
+	/* initialize tracker */
+	track = kzalloc(sizeof(*track), GFP_KERNEL);
+	if (track == NULL)
+		return -ENOMEM;
+	r600_cs_track_init(track);
+	r600_cs_legacy_get_tiling_conf(dev, &track->npipes, &track->nbanks, &track->group_size);
 	/* initialize parser */
 	memset(&parser, 0, sizeof(struct radeon_cs_parser));
 	parser.filp = filp;
+	parser.dev = &dev->pdev->dev;
 	parser.rdev = NULL;
 	parser.family = family;
 	parser.ib = &fake_ib;
+	parser.track = track;
 	fake_ib.ptr = ib;
 	r = radeon_cs_parser_init(&parser, data);
 	if (r) {
@@ -639,7 +1436,6 @@ int r600_cs_legacy(struct drm_device *dev, void *data, struct drm_file *filp,
 	 * uncached). */
 	ib_chunk = &parser.chunks[parser.chunk_ib_idx];
 	parser.ib->length_dw = ib_chunk->length_dw;
-	memcpy((void *)parser.ib->ptr, ib_chunk->kdata, ib_chunk->length_dw*4);
 	*l = parser.ib->length_dw;
 	r = r600_cs_parse(&parser);
 	if (r) {
@@ -647,6 +1443,12 @@ int r600_cs_legacy(struct drm_device *dev, void *data, struct drm_file *filp,
 		r600_cs_parser_fini(&parser, r);
 		return r;
 	}
+	r = radeon_cs_finish_pages(&parser);
+	if (r) {
+		DRM_ERROR("Invalid command stream !\n");
+		r600_cs_parser_fini(&parser, r);
+		return r;
+	}
 	r600_cs_parser_fini(&parser, r);
 	return r;
 }
diff --git a/drivers/gpu/drm/radeon/r600_hdmi.c b/drivers/gpu/drm/radeon/r600_hdmi.c
new file mode 100644
index 000000000000..fcc949df0e5d
--- /dev/null
+++ b/drivers/gpu/drm/radeon/r600_hdmi.c
@@ -0,0 +1,506 @@
+/*
+ * Copyright 2008 Advanced Micro Devices, Inc.
+ * Copyright 2008 Red Hat Inc.
+ * Copyright 2009 Christian König.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Christian König
+ */
+#include "drmP.h"
+#include "radeon_drm.h"
+#include "radeon.h"
+#include "atom.h"
+
+/*
+ * HDMI color format
+ */
+enum r600_hdmi_color_format {
+	RGB = 0,
+	YCC_422 = 1,
+	YCC_444 = 2
+};
+
+/*
+ * IEC60958 status bits
+ */
+enum r600_hdmi_iec_status_bits {
+	AUDIO_STATUS_DIG_ENABLE   = 0x01,
+	AUDIO_STATUS_V	    = 0x02,
+	AUDIO_STATUS_VCFG	 = 0x04,
+	AUDIO_STATUS_EMPHASIS     = 0x08,
+	AUDIO_STATUS_COPYRIGHT    = 0x10,
+	AUDIO_STATUS_NONAUDIO     = 0x20,
+	AUDIO_STATUS_PROFESSIONAL = 0x40,
+	AUDIO_STATUS_LEVEL	= 0x80
+};
+
+struct {
+	uint32_t Clock;
+
+	int N_32kHz;
+	int CTS_32kHz;
+
+	int N_44_1kHz;
+	int CTS_44_1kHz;
+
+	int N_48kHz;
+	int CTS_48kHz;
+
+} r600_hdmi_ACR[] = {
+    /*	     32kHz	  44.1kHz	48kHz    */
+    /* Clock      N     CTS      N     CTS      N     CTS */
+    {  25174,  4576,  28125,  7007,  31250,  6864,  28125 }, /*  25,20/1.001 MHz */
+    {  25200,  4096,  25200,  6272,  28000,  6144,  25200 }, /*  25.20       MHz */
+    {  27000,  4096,  27000,  6272,  30000,  6144,  27000 }, /*  27.00       MHz */
+    {  27027,  4096,  27027,  6272,  30030,  6144,  27027 }, /*  27.00*1.001 MHz */
+    {  54000,  4096,  54000,  6272,  60000,  6144,  54000 }, /*  54.00       MHz */
+    {  54054,  4096,  54054,  6272,  60060,  6144,  54054 }, /*  54.00*1.001 MHz */
+    {  74175, 11648, 210937, 17836, 234375, 11648, 140625 }, /*  74.25/1.001 MHz */
+    {  74250,  4096,  74250,  6272,  82500,  6144,  74250 }, /*  74.25       MHz */
+    { 148351, 11648, 421875,  8918, 234375,  5824, 140625 }, /* 148.50/1.001 MHz */
+    { 148500,  4096, 148500,  6272, 165000,  6144, 148500 }, /* 148.50       MHz */
+    {      0,  4096,      0,  6272,      0,  6144,      0 }  /* Other */
+};
+
+/*
+ * calculate CTS value if it's not found in the table
+ */
+static void r600_hdmi_calc_CTS(uint32_t clock, int *CTS, int N, int freq)
+{
+	if (*CTS == 0)
+		*CTS = clock*N/(128*freq)*1000;
+	DRM_DEBUG("Using ACR timing N=%d CTS=%d for frequency %d\n",
+		  N, *CTS, freq);
+}
+
+/*
+ * update the N and CTS parameters for a given pixel clock rate
+ */
+static void r600_hdmi_update_ACR(struct drm_encoder *encoder, uint32_t clock)
+{
+	struct drm_device *dev = encoder->dev;
+	struct radeon_device *rdev = dev->dev_private;
+	uint32_t offset = to_radeon_encoder(encoder)->hdmi_offset;
+	int CTS;
+	int N;
+	int i;
+
+	for (i = 0; r600_hdmi_ACR[i].Clock != clock && r600_hdmi_ACR[i].Clock != 0; i++);
+
+	CTS = r600_hdmi_ACR[i].CTS_32kHz;
+	N = r600_hdmi_ACR[i].N_32kHz;
+	r600_hdmi_calc_CTS(clock, &CTS, N, 32000);
+	WREG32(offset+R600_HDMI_32kHz_CTS, CTS << 12);
+	WREG32(offset+R600_HDMI_32kHz_N, N);
+
+	CTS = r600_hdmi_ACR[i].CTS_44_1kHz;
+	N = r600_hdmi_ACR[i].N_44_1kHz;
+	r600_hdmi_calc_CTS(clock, &CTS, N, 44100);
+	WREG32(offset+R600_HDMI_44_1kHz_CTS, CTS << 12);
+	WREG32(offset+R600_HDMI_44_1kHz_N, N);
+
+	CTS = r600_hdmi_ACR[i].CTS_48kHz;
+	N = r600_hdmi_ACR[i].N_48kHz;
+	r600_hdmi_calc_CTS(clock, &CTS, N, 48000);
+	WREG32(offset+R600_HDMI_48kHz_CTS, CTS << 12);
+	WREG32(offset+R600_HDMI_48kHz_N, N);
+}
+
+/*
+ * calculate the crc for a given info frame
+ */
+static void r600_hdmi_infoframe_checksum(uint8_t packetType,
+					 uint8_t versionNumber,
+					 uint8_t length,
+					 uint8_t *frame)
+{
+    int i;
+    frame[0] = packetType + versionNumber + length;
+    for (i = 1; i <= length; i++)
+	frame[0] += frame[i];
+    frame[0] = 0x100 - frame[0];
+}
+
+/*
+ * build a HDMI Video Info Frame
+ */
+static void r600_hdmi_videoinfoframe(
+	struct drm_encoder *encoder,
+	enum r600_hdmi_color_format color_format,
+	int active_information_present,
+	uint8_t active_format_aspect_ratio,
+	uint8_t scan_information,
+	uint8_t colorimetry,
+	uint8_t ex_colorimetry,
+	uint8_t quantization,
+	int ITC,
+	uint8_t picture_aspect_ratio,
+	uint8_t video_format_identification,
+	uint8_t pixel_repetition,
+	uint8_t non_uniform_picture_scaling,
+	uint8_t bar_info_data_valid,
+	uint16_t top_bar,
+	uint16_t bottom_bar,
+	uint16_t left_bar,
+	uint16_t right_bar
+)
+{
+	struct drm_device *dev = encoder->dev;
+	struct radeon_device *rdev = dev->dev_private;
+	uint32_t offset = to_radeon_encoder(encoder)->hdmi_offset;
+
+	uint8_t frame[14];
+
+	frame[0x0] = 0;
+	frame[0x1] =
+		(scan_information & 0x3) |
+		((bar_info_data_valid & 0x3) << 2) |
+		((active_information_present & 0x1) << 4) |
+		((color_format & 0x3) << 5);
+	frame[0x2] =
+		(active_format_aspect_ratio & 0xF) |
+		((picture_aspect_ratio & 0x3) << 4) |
+		((colorimetry & 0x3) << 6);
+	frame[0x3] =
+		(non_uniform_picture_scaling & 0x3) |
+		((quantization & 0x3) << 2) |
+		((ex_colorimetry & 0x7) << 4) |
+		((ITC & 0x1) << 7);
+	frame[0x4] = (video_format_identification & 0x7F);
+	frame[0x5] = (pixel_repetition & 0xF);
+	frame[0x6] = (top_bar & 0xFF);
+	frame[0x7] = (top_bar >> 8);
+	frame[0x8] = (bottom_bar & 0xFF);
+	frame[0x9] = (bottom_bar >> 8);
+	frame[0xA] = (left_bar & 0xFF);
+	frame[0xB] = (left_bar >> 8);
+	frame[0xC] = (right_bar & 0xFF);
+	frame[0xD] = (right_bar >> 8);
+
+	r600_hdmi_infoframe_checksum(0x82, 0x02, 0x0D, frame);
+
+	WREG32(offset+R600_HDMI_VIDEOINFOFRAME_0,
+		frame[0x0] | (frame[0x1] << 8) | (frame[0x2] << 16) | (frame[0x3] << 24));
+	WREG32(offset+R600_HDMI_VIDEOINFOFRAME_1,
+		frame[0x4] | (frame[0x5] << 8) | (frame[0x6] << 16) | (frame[0x7] << 24));
+	WREG32(offset+R600_HDMI_VIDEOINFOFRAME_2,
+		frame[0x8] | (frame[0x9] << 8) | (frame[0xA] << 16) | (frame[0xB] << 24));
+	WREG32(offset+R600_HDMI_VIDEOINFOFRAME_3,
+		frame[0xC] | (frame[0xD] << 8));
+}
+
+/*
+ * build a Audio Info Frame
+ */
+static void r600_hdmi_audioinfoframe(
+	struct drm_encoder *encoder,
+	uint8_t channel_count,
+	uint8_t coding_type,
+	uint8_t sample_size,
+	uint8_t sample_frequency,
+	uint8_t format,
+	uint8_t channel_allocation,
+	uint8_t level_shift,
+	int downmix_inhibit
+)
+{
+	struct drm_device *dev = encoder->dev;
+	struct radeon_device *rdev = dev->dev_private;
+	uint32_t offset = to_radeon_encoder(encoder)->hdmi_offset;
+
+	uint8_t frame[11];
+
+	frame[0x0] = 0;
+	frame[0x1] = (channel_count & 0x7) | ((coding_type & 0xF) << 4);
+	frame[0x2] = (sample_size & 0x3) | ((sample_frequency & 0x7) << 2);
+	frame[0x3] = format;
+	frame[0x4] = channel_allocation;
+	frame[0x5] = ((level_shift & 0xF) << 3) | ((downmix_inhibit & 0x1) << 7);
+	frame[0x6] = 0;
+	frame[0x7] = 0;
+	frame[0x8] = 0;
+	frame[0x9] = 0;
+	frame[0xA] = 0;
+
+	r600_hdmi_infoframe_checksum(0x84, 0x01, 0x0A, frame);
+
+	WREG32(offset+R600_HDMI_AUDIOINFOFRAME_0,
+		frame[0x0] | (frame[0x1] << 8) | (frame[0x2] << 16) | (frame[0x3] << 24));
+	WREG32(offset+R600_HDMI_AUDIOINFOFRAME_1,
+		frame[0x4] | (frame[0x5] << 8) | (frame[0x6] << 16) | (frame[0x8] << 24));
+}
+
+/*
+ * test if audio buffer is filled enough to start playing
+ */
+static int r600_hdmi_is_audio_buffer_filled(struct drm_encoder *encoder)
+{
+	struct drm_device *dev = encoder->dev;
+	struct radeon_device *rdev = dev->dev_private;
+	uint32_t offset = to_radeon_encoder(encoder)->hdmi_offset;
+
+	return (RREG32(offset+R600_HDMI_STATUS) & 0x10) != 0;
+}
+
+/*
+ * have buffer status changed since last call?
+ */
+int r600_hdmi_buffer_status_changed(struct drm_encoder *encoder)
+{
+	struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
+	int status, result;
+
+	if (!radeon_encoder->hdmi_offset)
+		return 0;
+
+	status = r600_hdmi_is_audio_buffer_filled(encoder);
+	result = radeon_encoder->hdmi_buffer_status != status;
+	radeon_encoder->hdmi_buffer_status = status;
+
+	return result;
+}
+
+/*
+ * write the audio workaround status to the hardware
+ */
+void r600_hdmi_audio_workaround(struct drm_encoder *encoder)
+{
+	struct drm_device *dev = encoder->dev;
+	struct radeon_device *rdev = dev->dev_private;
+	struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
+	uint32_t offset = radeon_encoder->hdmi_offset;
+
+	if (!offset)
+		return;
+
+	if (r600_hdmi_is_audio_buffer_filled(encoder)) {
+		/* disable audio workaround and start delivering of audio frames */
+		WREG32_P(offset+R600_HDMI_CNTL, 0x00000001, ~0x00001001);
+
+	} else if (radeon_encoder->hdmi_audio_workaround) {
+		/* enable audio workaround and start delivering of audio frames */
+		WREG32_P(offset+R600_HDMI_CNTL, 0x00001001, ~0x00001001);
+
+	} else {
+		/* disable audio workaround and stop delivering of audio frames */
+		WREG32_P(offset+R600_HDMI_CNTL, 0x00000000, ~0x00001001);
+	}
+}
+
+
+/*
+ * update the info frames with the data from the current display mode
+ */
+void r600_hdmi_setmode(struct drm_encoder *encoder, struct drm_display_mode *mode)
+{
+	struct drm_device *dev = encoder->dev;
+	struct radeon_device *rdev = dev->dev_private;
+	uint32_t offset = to_radeon_encoder(encoder)->hdmi_offset;
+
+	if (!offset)
+		return;
+
+	r600_audio_set_clock(encoder, mode->clock);
+
+	WREG32(offset+R600_HDMI_UNKNOWN_0, 0x1000);
+	WREG32(offset+R600_HDMI_UNKNOWN_1, 0x0);
+	WREG32(offset+R600_HDMI_UNKNOWN_2, 0x1000);
+
+	r600_hdmi_update_ACR(encoder, mode->clock);
+
+	WREG32(offset+R600_HDMI_VIDEOCNTL, 0x13);
+
+	WREG32(offset+R600_HDMI_VERSION, 0x202);
+
+	r600_hdmi_videoinfoframe(encoder, RGB, 0, 0, 0, 0,
+		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0);
+
+	/* it's unknown what these bits do excatly, but it's indeed quite usefull for debugging */
+	WREG32(offset+R600_HDMI_AUDIO_DEBUG_0, 0x00FFFFFF);
+	WREG32(offset+R600_HDMI_AUDIO_DEBUG_1, 0x007FFFFF);
+	WREG32(offset+R600_HDMI_AUDIO_DEBUG_2, 0x00000001);
+	WREG32(offset+R600_HDMI_AUDIO_DEBUG_3, 0x00000001);
+
+	r600_hdmi_audio_workaround(encoder);
+
+	/* audio packets per line, does anyone know how to calc this ? */
+	WREG32_P(offset+R600_HDMI_CNTL, 0x00040000, ~0x001F0000);
+
+	/* update? reset? don't realy know */
+	WREG32_P(offset+R600_HDMI_CNTL, 0x14000000, ~0x14000000);
+}
+
+/*
+ * update settings with current parameters from audio engine
+ */
+void r600_hdmi_update_audio_settings(struct drm_encoder *encoder,
+				     int channels,
+				     int rate,
+				     int bps,
+				     uint8_t status_bits,
+				     uint8_t category_code)
+{
+	struct drm_device *dev = encoder->dev;
+	struct radeon_device *rdev = dev->dev_private;
+	uint32_t offset = to_radeon_encoder(encoder)->hdmi_offset;
+
+	uint32_t iec;
+
+	if (!offset)
+		return;
+
+	DRM_DEBUG("%s with %d channels, %d Hz sampling rate, %d bits per sample,\n",
+		 r600_hdmi_is_audio_buffer_filled(encoder) ? "playing" : "stopped",
+		channels, rate, bps);
+	DRM_DEBUG("0x%02X IEC60958 status bits and 0x%02X category code\n",
+		  (int)status_bits, (int)category_code);
+
+	iec = 0;
+	if (status_bits & AUDIO_STATUS_PROFESSIONAL)
+		iec |= 1 << 0;
+	if (status_bits & AUDIO_STATUS_NONAUDIO)
+		iec |= 1 << 1;
+	if (status_bits & AUDIO_STATUS_COPYRIGHT)
+		iec |= 1 << 2;
+	if (status_bits & AUDIO_STATUS_EMPHASIS)
+		iec |= 1 << 3;
+
+	iec |= category_code << 8;
+
+	switch (rate) {
+	case  32000: iec |= 0x3 << 24; break;
+	case  44100: iec |= 0x0 << 24; break;
+	case  88200: iec |= 0x8 << 24; break;
+	case 176400: iec |= 0xc << 24; break;
+	case  48000: iec |= 0x2 << 24; break;
+	case  96000: iec |= 0xa << 24; break;
+	case 192000: iec |= 0xe << 24; break;
+	}
+
+	WREG32(offset+R600_HDMI_IEC60958_1, iec);
+
+	iec = 0;
+	switch (bps) {
+	case 16: iec |= 0x2; break;
+	case 20: iec |= 0x3; break;
+	case 24: iec |= 0xb; break;
+	}
+	if (status_bits & AUDIO_STATUS_V)
+		iec |= 0x5 << 16;
+
+	WREG32_P(offset+R600_HDMI_IEC60958_2, iec, ~0x5000f);
+
+	/* 0x021 or 0x031 sets the audio frame length */
+	WREG32(offset+R600_HDMI_AUDIOCNTL, 0x31);
+	r600_hdmi_audioinfoframe(encoder, channels-1, 0, 0, 0, 0, 0, 0, 0);
+
+	r600_hdmi_audio_workaround(encoder);
+
+	/* update? reset? don't realy know */
+	WREG32_P(offset+R600_HDMI_CNTL, 0x04000000, ~0x04000000);
+}
+
+/*
+ * enable/disable the HDMI engine
+ */
+void r600_hdmi_enable(struct drm_encoder *encoder, int enable)
+{
+	struct drm_device *dev = encoder->dev;
+	struct radeon_device *rdev = dev->dev_private;
+	struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
+	uint32_t offset = to_radeon_encoder(encoder)->hdmi_offset;
+
+	if (!offset)
+		return;
+
+	DRM_DEBUG("%s HDMI interface @ 0x%04X\n", enable ? "Enabling" : "Disabling", offset);
+
+	/* some version of atombios ignore the enable HDMI flag
+	 * so enabling/disabling HDMI was moved here for TMDS1+2 */
+	switch (radeon_encoder->encoder_id) {
+	case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_TMDS1:
+		WREG32_P(AVIVO_TMDSA_CNTL, enable ? 0x4 : 0x0, ~0x4);
+		WREG32(offset+R600_HDMI_ENABLE, enable ? 0x101 : 0x0);
+		break;
+
+	case ENCODER_OBJECT_ID_INTERNAL_LVTM1:
+		WREG32_P(AVIVO_LVTMA_CNTL, enable ? 0x4 : 0x0, ~0x4);
+		WREG32(offset+R600_HDMI_ENABLE, enable ? 0x105 : 0x0);
+		break;
+
+	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY:
+	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1:
+	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2:
+	case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_LVTMA:
+		/* This part is doubtfull in my opinion */
+		WREG32(offset+R600_HDMI_ENABLE, enable ? 0x110 : 0x0);
+		break;
+
+	default:
+		DRM_ERROR("unknown HDMI output type\n");
+		break;
+	}
+}
+
+/*
+ * determin at which register offset the HDMI encoder is
+ */
+void r600_hdmi_init(struct drm_encoder *encoder)
+{
+	struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
+
+	switch (radeon_encoder->encoder_id) {
+	case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_TMDS1:
+	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY:
+	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1:
+		radeon_encoder->hdmi_offset = R600_HDMI_TMDS1;
+		break;
+
+	case ENCODER_OBJECT_ID_INTERNAL_LVTM1:
+		switch (r600_audio_tmds_index(encoder)) {
+		case 0:
+			radeon_encoder->hdmi_offset = R600_HDMI_TMDS1;
+			break;
+		case 1:
+			radeon_encoder->hdmi_offset = R600_HDMI_TMDS2;
+			break;
+		default:
+			radeon_encoder->hdmi_offset = 0;
+			break;
+		}
+	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2:
+		radeon_encoder->hdmi_offset = R600_HDMI_TMDS2;
+		break;
+
+	case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_LVTMA:
+		radeon_encoder->hdmi_offset = R600_HDMI_DIG;
+		break;
+
+	default:
+		radeon_encoder->hdmi_offset = 0;
+		break;
+	}
+
+	DRM_DEBUG("using HDMI engine at offset 0x%04X for encoder 0x%x\n",
+		  radeon_encoder->hdmi_offset, radeon_encoder->encoder_id);
+
+	/* TODO: make this configureable */
+	radeon_encoder->hdmi_audio_workaround = 0;
+}
diff --git a/drivers/gpu/drm/radeon/r600_reg.h b/drivers/gpu/drm/radeon/r600_reg.h
index e2d1f5f33f7e..d0e28ffdeda9 100644
--- a/drivers/gpu/drm/radeon/r600_reg.h
+++ b/drivers/gpu/drm/radeon/r600_reg.h
@@ -110,5 +110,79 @@
 #define R600_BIOS_6_SCRATCH               0x173c
 #define R600_BIOS_7_SCRATCH               0x1740
 
+/* Audio, these regs were reverse enginered,
+ * so the chance is high that the naming is wrong
+ * R6xx+ ??? */
+
+/* Audio clocks */
+#define R600_AUDIO_PLL1_MUL               0x0514
+#define R600_AUDIO_PLL1_DIV               0x0518
+#define R600_AUDIO_PLL2_MUL               0x0524
+#define R600_AUDIO_PLL2_DIV               0x0528
+#define R600_AUDIO_CLK_SRCSEL             0x0534
+
+/* Audio general */
+#define R600_AUDIO_ENABLE                 0x7300
+#define R600_AUDIO_TIMING                 0x7344
+
+/* Audio params */
+#define R600_AUDIO_VENDOR_ID              0x7380
+#define R600_AUDIO_REVISION_ID            0x7384
+#define R600_AUDIO_ROOT_NODE_COUNT        0x7388
+#define R600_AUDIO_NID1_NODE_COUNT        0x738c
+#define R600_AUDIO_NID1_TYPE              0x7390
+#define R600_AUDIO_SUPPORTED_SIZE_RATE    0x7394
+#define R600_AUDIO_SUPPORTED_CODEC        0x7398
+#define R600_AUDIO_SUPPORTED_POWER_STATES 0x739c
+#define R600_AUDIO_NID2_CAPS              0x73a0
+#define R600_AUDIO_NID3_CAPS              0x73a4
+#define R600_AUDIO_NID3_PIN_CAPS          0x73a8
+
+/* Audio conn list */
+#define R600_AUDIO_CONN_LIST_LEN          0x73ac
+#define R600_AUDIO_CONN_LIST              0x73b0
+
+/* Audio verbs */
+#define R600_AUDIO_RATE_BPS_CHANNEL       0x73c0
+#define R600_AUDIO_PLAYING                0x73c4
+#define R600_AUDIO_IMPLEMENTATION_ID      0x73c8
+#define R600_AUDIO_CONFIG_DEFAULT         0x73cc
+#define R600_AUDIO_PIN_SENSE              0x73d0
+#define R600_AUDIO_PIN_WIDGET_CNTL        0x73d4
+#define R600_AUDIO_STATUS_BITS            0x73d8
+
+/* HDMI base register addresses */
+#define R600_HDMI_TMDS1                   0x7400
+#define R600_HDMI_TMDS2                   0x7700
+#define R600_HDMI_DIG                     0x7800
+
+/* HDMI registers */
+#define R600_HDMI_ENABLE           0x00
+#define R600_HDMI_STATUS           0x04
+#define R600_HDMI_CNTL             0x08
+#define R600_HDMI_UNKNOWN_0        0x0C
+#define R600_HDMI_AUDIOCNTL        0x10
+#define R600_HDMI_VIDEOCNTL        0x14
+#define R600_HDMI_VERSION          0x18
+#define R600_HDMI_UNKNOWN_1        0x28
+#define R600_HDMI_VIDEOINFOFRAME_0 0x54
+#define R600_HDMI_VIDEOINFOFRAME_1 0x58
+#define R600_HDMI_VIDEOINFOFRAME_2 0x5c
+#define R600_HDMI_VIDEOINFOFRAME_3 0x60
+#define R600_HDMI_32kHz_CTS        0xac
+#define R600_HDMI_32kHz_N          0xb0
+#define R600_HDMI_44_1kHz_CTS      0xb4
+#define R600_HDMI_44_1kHz_N        0xb8
+#define R600_HDMI_48kHz_CTS        0xbc
+#define R600_HDMI_48kHz_N          0xc0
+#define R600_HDMI_AUDIOINFOFRAME_0 0xcc
+#define R600_HDMI_AUDIOINFOFRAME_1 0xd0
+#define R600_HDMI_IEC60958_1       0xd4
+#define R600_HDMI_IEC60958_2       0xd8
+#define R600_HDMI_UNKNOWN_2        0xdc
+#define R600_HDMI_AUDIO_DEBUG_0    0xe0
+#define R600_HDMI_AUDIO_DEBUG_1    0xe4
+#define R600_HDMI_AUDIO_DEBUG_2    0xe8
+#define R600_HDMI_AUDIO_DEBUG_3    0xec
 
 #endif
diff --git a/drivers/gpu/drm/radeon/r600d.h b/drivers/gpu/drm/radeon/r600d.h
index 4a9028a85c9b..5b2e4d442823 100644
--- a/drivers/gpu/drm/radeon/r600d.h
+++ b/drivers/gpu/drm/radeon/r600d.h
@@ -119,6 +119,7 @@
 #define	DB_DEBUG					0x9830
 #define		PREZ_MUST_WAIT_FOR_POSTZ_DONE			(1 << 31)
 #define	DB_DEPTH_BASE					0x2800C
+#define	DB_HTILE_DATA_BASE				0x28014
 #define	DB_WATERMARKS					0x9838
 #define		DEPTH_FREE(x)					((x) << 0)
 #define		DEPTH_FLUSH(x)					((x) << 5)
@@ -171,6 +172,14 @@
 #define SQ_STACK_RESOURCE_MGMT_2                          0x8c14
 #       define NUM_GS_STACK_ENTRIES(x)                    ((x) << 0)
 #       define NUM_ES_STACK_ENTRIES(x)                    ((x) << 16)
+#define SQ_ESGS_RING_BASE                               0x8c40
+#define SQ_GSVS_RING_BASE                               0x8c48
+#define SQ_ESTMP_RING_BASE                              0x8c50
+#define SQ_GSTMP_RING_BASE                              0x8c58
+#define SQ_VSTMP_RING_BASE                              0x8c60
+#define SQ_PSTMP_RING_BASE                              0x8c68
+#define SQ_FBUF_RING_BASE                               0x8c70
+#define SQ_REDUC_RING_BASE                              0x8c78
 
 #define GRBM_CNTL                                       0x8000
 #       define GRBM_READ_TIMEOUT(x)                     ((x) << 0)
@@ -271,6 +280,10 @@
 #define	PCIE_PORT_INDEX					0x0038
 #define	PCIE_PORT_DATA					0x003C
 
+#define CHMAP						0x2004
+#define		NOOFCHAN_SHIFT					12
+#define		NOOFCHAN_MASK					0x00003000
+
 #define RAMCFG						0x2408
 #define		NOOFBANK_SHIFT					0
 #define		NOOFBANK_MASK					0x00000001
@@ -352,6 +365,7 @@
 
 
 #define	SX_MISC						0x28350
+#define	SX_MEMORY_EXPORT_BASE				0x9010
 #define	SX_DEBUG_1					0x9054
 #define		SMX_EVENT_RELEASE				(1 << 0)
 #define		ENABLE_NEW_SMX_ADDRESS				(1 << 16)
@@ -442,7 +456,215 @@
 #define         WAIT_2D_IDLECLEAN_bit                           (1 << 16)
 #define         WAIT_3D_IDLECLEAN_bit                           (1 << 17)
 
+#define IH_RB_CNTL                                        0x3e00
+#       define IH_RB_ENABLE                               (1 << 0)
+#       define IH_IB_SIZE(x)                              ((x) << 1) /* log2 */
+#       define IH_RB_FULL_DRAIN_ENABLE                    (1 << 6)
+#       define IH_WPTR_WRITEBACK_ENABLE                   (1 << 8)
+#       define IH_WPTR_WRITEBACK_TIMER(x)                 ((x) << 9) /* log2 */
+#       define IH_WPTR_OVERFLOW_ENABLE                    (1 << 16)
+#       define IH_WPTR_OVERFLOW_CLEAR                     (1 << 31)
+#define IH_RB_BASE                                        0x3e04
+#define IH_RB_RPTR                                        0x3e08
+#define IH_RB_WPTR                                        0x3e0c
+#       define RB_OVERFLOW                                (1 << 0)
+#       define WPTR_OFFSET_MASK                           0x3fffc
+#define IH_RB_WPTR_ADDR_HI                                0x3e10
+#define IH_RB_WPTR_ADDR_LO                                0x3e14
+#define IH_CNTL                                           0x3e18
+#       define ENABLE_INTR                                (1 << 0)
+#       define IH_MC_SWAP(x)                              ((x) << 2)
+#       define IH_MC_SWAP_NONE                            0
+#       define IH_MC_SWAP_16BIT                           1
+#       define IH_MC_SWAP_32BIT                           2
+#       define IH_MC_SWAP_64BIT                           3
+#       define RPTR_REARM                                 (1 << 4)
+#       define MC_WRREQ_CREDIT(x)                         ((x) << 15)
+#       define MC_WR_CLEAN_CNT(x)                         ((x) << 20)
+
+#define RLC_CNTL                                          0x3f00
+#       define RLC_ENABLE                                 (1 << 0)
+#define RLC_HB_BASE                                       0x3f10
+#define RLC_HB_CNTL                                       0x3f0c
+#define RLC_HB_RPTR                                       0x3f20
+#define RLC_HB_WPTR                                       0x3f1c
+#define RLC_HB_WPTR_LSB_ADDR                              0x3f14
+#define RLC_HB_WPTR_MSB_ADDR                              0x3f18
+#define RLC_MC_CNTL                                       0x3f44
+#define RLC_UCODE_CNTL                                    0x3f48
+#define RLC_UCODE_ADDR                                    0x3f2c
+#define RLC_UCODE_DATA                                    0x3f30
+
+#define SRBM_SOFT_RESET                                   0xe60
+#       define SOFT_RESET_RLC                             (1 << 13)
+
+#define CP_INT_CNTL                                       0xc124
+#       define CNTX_BUSY_INT_ENABLE                       (1 << 19)
+#       define CNTX_EMPTY_INT_ENABLE                      (1 << 20)
+#       define SCRATCH_INT_ENABLE                         (1 << 25)
+#       define TIME_STAMP_INT_ENABLE                      (1 << 26)
+#       define IB2_INT_ENABLE                             (1 << 29)
+#       define IB1_INT_ENABLE                             (1 << 30)
+#       define RB_INT_ENABLE                              (1 << 31)
+#define CP_INT_STATUS                                     0xc128
+#       define SCRATCH_INT_STAT                           (1 << 25)
+#       define TIME_STAMP_INT_STAT                        (1 << 26)
+#       define IB2_INT_STAT                               (1 << 29)
+#       define IB1_INT_STAT                               (1 << 30)
+#       define RB_INT_STAT                                (1 << 31)
+
+#define GRBM_INT_CNTL                                     0x8060
+#       define RDERR_INT_ENABLE                           (1 << 0)
+#       define WAIT_COUNT_TIMEOUT_INT_ENABLE              (1 << 1)
+#       define GUI_IDLE_INT_ENABLE                        (1 << 19)
+
+#define INTERRUPT_CNTL                                    0x5468
+#       define IH_DUMMY_RD_OVERRIDE                       (1 << 0)
+#       define IH_DUMMY_RD_EN                             (1 << 1)
+#       define IH_REQ_NONSNOOP_EN                         (1 << 3)
+#       define GEN_IH_INT_EN                              (1 << 8)
+#define INTERRUPT_CNTL2                                   0x546c
+
+#define D1MODE_VBLANK_STATUS                              0x6534
+#define D2MODE_VBLANK_STATUS                              0x6d34
+#       define DxMODE_VBLANK_OCCURRED                     (1 << 0)
+#       define DxMODE_VBLANK_ACK                          (1 << 4)
+#       define DxMODE_VBLANK_STAT                         (1 << 12)
+#       define DxMODE_VBLANK_INTERRUPT                    (1 << 16)
+#       define DxMODE_VBLANK_INTERRUPT_TYPE               (1 << 17)
+#define D1MODE_VLINE_STATUS                               0x653c
+#define D2MODE_VLINE_STATUS                               0x6d3c
+#       define DxMODE_VLINE_OCCURRED                      (1 << 0)
+#       define DxMODE_VLINE_ACK                           (1 << 4)
+#       define DxMODE_VLINE_STAT                          (1 << 12)
+#       define DxMODE_VLINE_INTERRUPT                     (1 << 16)
+#       define DxMODE_VLINE_INTERRUPT_TYPE                (1 << 17)
+#define DxMODE_INT_MASK                                   0x6540
+#       define D1MODE_VBLANK_INT_MASK                     (1 << 0)
+#       define D1MODE_VLINE_INT_MASK                      (1 << 4)
+#       define D2MODE_VBLANK_INT_MASK                     (1 << 8)
+#       define D2MODE_VLINE_INT_MASK                      (1 << 12)
+#define DCE3_DISP_INTERRUPT_STATUS                        0x7ddc
+#       define DC_HPD1_INTERRUPT                          (1 << 18)
+#       define DC_HPD2_INTERRUPT                          (1 << 19)
+#define DISP_INTERRUPT_STATUS                             0x7edc
+#       define LB_D1_VLINE_INTERRUPT                      (1 << 2)
+#       define LB_D2_VLINE_INTERRUPT                      (1 << 3)
+#       define LB_D1_VBLANK_INTERRUPT                     (1 << 4)
+#       define LB_D2_VBLANK_INTERRUPT                     (1 << 5)
+#       define DACA_AUTODETECT_INTERRUPT                  (1 << 16)
+#       define DACB_AUTODETECT_INTERRUPT                  (1 << 17)
+#       define DC_HOT_PLUG_DETECT1_INTERRUPT              (1 << 18)
+#       define DC_HOT_PLUG_DETECT2_INTERRUPT              (1 << 19)
+#       define DC_I2C_SW_DONE_INTERRUPT                   (1 << 20)
+#       define DC_I2C_HW_DONE_INTERRUPT                   (1 << 21)
+#define DISP_INTERRUPT_STATUS_CONTINUE                    0x7ee8
+#define DCE3_DISP_INTERRUPT_STATUS_CONTINUE               0x7de8
+#       define DC_HPD4_INTERRUPT                          (1 << 14)
+#       define DC_HPD4_RX_INTERRUPT                       (1 << 15)
+#       define DC_HPD3_INTERRUPT                          (1 << 28)
+#       define DC_HPD1_RX_INTERRUPT                       (1 << 29)
+#       define DC_HPD2_RX_INTERRUPT                       (1 << 30)
+#define DCE3_DISP_INTERRUPT_STATUS_CONTINUE2              0x7dec
+#       define DC_HPD3_RX_INTERRUPT                       (1 << 0)
+#       define DIGA_DP_VID_STREAM_DISABLE_INTERRUPT       (1 << 1)
+#       define DIGA_DP_STEER_FIFO_OVERFLOW_INTERRUPT      (1 << 2)
+#       define DIGB_DP_VID_STREAM_DISABLE_INTERRUPT       (1 << 3)
+#       define DIGB_DP_STEER_FIFO_OVERFLOW_INTERRUPT      (1 << 4)
+#       define AUX1_SW_DONE_INTERRUPT                     (1 << 5)
+#       define AUX1_LS_DONE_INTERRUPT                     (1 << 6)
+#       define AUX2_SW_DONE_INTERRUPT                     (1 << 7)
+#       define AUX2_LS_DONE_INTERRUPT                     (1 << 8)
+#       define AUX3_SW_DONE_INTERRUPT                     (1 << 9)
+#       define AUX3_LS_DONE_INTERRUPT                     (1 << 10)
+#       define AUX4_SW_DONE_INTERRUPT                     (1 << 11)
+#       define AUX4_LS_DONE_INTERRUPT                     (1 << 12)
+#       define DIGA_DP_FAST_TRAINING_COMPLETE_INTERRUPT   (1 << 13)
+#       define DIGB_DP_FAST_TRAINING_COMPLETE_INTERRUPT   (1 << 14)
+/* DCE 3.2 */
+#       define AUX5_SW_DONE_INTERRUPT                     (1 << 15)
+#       define AUX5_LS_DONE_INTERRUPT                     (1 << 16)
+#       define AUX6_SW_DONE_INTERRUPT                     (1 << 17)
+#       define AUX6_LS_DONE_INTERRUPT                     (1 << 18)
+#       define DC_HPD5_INTERRUPT                          (1 << 19)
+#       define DC_HPD5_RX_INTERRUPT                       (1 << 20)
+#       define DC_HPD6_INTERRUPT                          (1 << 21)
+#       define DC_HPD6_RX_INTERRUPT                       (1 << 22)
+
+#define DACA_AUTO_DETECT_CONTROL                          0x7828
+#define DACB_AUTO_DETECT_CONTROL                          0x7a28
+#define DCE3_DACA_AUTO_DETECT_CONTROL                     0x7028
+#define DCE3_DACB_AUTO_DETECT_CONTROL                     0x7128
+#       define DACx_AUTODETECT_MODE(x)                    ((x) << 0)
+#       define DACx_AUTODETECT_MODE_NONE                  0
+#       define DACx_AUTODETECT_MODE_CONNECT               1
+#       define DACx_AUTODETECT_MODE_DISCONNECT            2
+#       define DACx_AUTODETECT_FRAME_TIME_COUNTER(x)      ((x) << 8)
+/* bit 18 = R/C, 17 = G/Y, 16 = B/Comp */
+#       define DACx_AUTODETECT_CHECK_MASK(x)              ((x) << 16)
 
+#define DCE3_DACA_AUTODETECT_INT_CONTROL                  0x7038
+#define DCE3_DACB_AUTODETECT_INT_CONTROL                  0x7138
+#define DACA_AUTODETECT_INT_CONTROL                       0x7838
+#define DACB_AUTODETECT_INT_CONTROL                       0x7a38
+#       define DACx_AUTODETECT_ACK                        (1 << 0)
+#       define DACx_AUTODETECT_INT_ENABLE                 (1 << 16)
+
+#define DC_HOT_PLUG_DETECT1_CONTROL                       0x7d00
+#define DC_HOT_PLUG_DETECT2_CONTROL                       0x7d10
+#define DC_HOT_PLUG_DETECT3_CONTROL                       0x7d24
+#       define DC_HOT_PLUG_DETECTx_EN                     (1 << 0)
+
+#define DC_HOT_PLUG_DETECT1_INT_STATUS                    0x7d04
+#define DC_HOT_PLUG_DETECT2_INT_STATUS                    0x7d14
+#define DC_HOT_PLUG_DETECT3_INT_STATUS                    0x7d28
+#       define DC_HOT_PLUG_DETECTx_INT_STATUS             (1 << 0)
+#       define DC_HOT_PLUG_DETECTx_SENSE                  (1 << 1)
+
+/* DCE 3.0 */
+#define DC_HPD1_INT_STATUS                                0x7d00
+#define DC_HPD2_INT_STATUS                                0x7d0c
+#define DC_HPD3_INT_STATUS                                0x7d18
+#define DC_HPD4_INT_STATUS                                0x7d24
+/* DCE 3.2 */
+#define DC_HPD5_INT_STATUS                                0x7dc0
+#define DC_HPD6_INT_STATUS                                0x7df4
+#       define DC_HPDx_INT_STATUS                         (1 << 0)
+#       define DC_HPDx_SENSE                              (1 << 1)
+#       define DC_HPDx_RX_INT_STATUS                      (1 << 8)
+
+#define DC_HOT_PLUG_DETECT1_INT_CONTROL                   0x7d08
+#define DC_HOT_PLUG_DETECT2_INT_CONTROL                   0x7d18
+#define DC_HOT_PLUG_DETECT3_INT_CONTROL                   0x7d2c
+#       define DC_HOT_PLUG_DETECTx_INT_ACK                (1 << 0)
+#       define DC_HOT_PLUG_DETECTx_INT_POLARITY           (1 << 8)
+#       define DC_HOT_PLUG_DETECTx_INT_EN                 (1 << 16)
+/* DCE 3.0 */
+#define DC_HPD1_INT_CONTROL                               0x7d04
+#define DC_HPD2_INT_CONTROL                               0x7d10
+#define DC_HPD3_INT_CONTROL                               0x7d1c
+#define DC_HPD4_INT_CONTROL                               0x7d28
+/* DCE 3.2 */
+#define DC_HPD5_INT_CONTROL                               0x7dc4
+#define DC_HPD6_INT_CONTROL                               0x7df8
+#       define DC_HPDx_INT_ACK                            (1 << 0)
+#       define DC_HPDx_INT_POLARITY                       (1 << 8)
+#       define DC_HPDx_INT_EN                             (1 << 16)
+#       define DC_HPDx_RX_INT_ACK                         (1 << 20)
+#       define DC_HPDx_RX_INT_EN                          (1 << 24)
+
+/* DCE 3.0 */
+#define DC_HPD1_CONTROL                                   0x7d08
+#define DC_HPD2_CONTROL                                   0x7d14
+#define DC_HPD3_CONTROL                                   0x7d20
+#define DC_HPD4_CONTROL                                   0x7d2c
+/* DCE 3.2 */
+#define DC_HPD5_CONTROL                                   0x7dc8
+#define DC_HPD6_CONTROL                                   0x7dfc
+#       define DC_HPDx_CONNECTION_TIMER(x)                ((x) << 0)
+#       define DC_HPDx_RX_INT_TIMER(x)                    ((x) << 16)
+/* DCE 3.2 */
+#       define DC_HPDx_EN                                 (1 << 28)
 
 /*
  * PM4
@@ -486,7 +708,6 @@
 #define	PACKET3_WAIT_REG_MEM				0x3C
 #define	PACKET3_MEM_WRITE				0x3D
 #define	PACKET3_INDIRECT_BUFFER				0x32
-#define	PACKET3_CP_INTERRUPT				0x40
 #define	PACKET3_SURFACE_SYNC				0x43
 #              define PACKET3_CB0_DEST_BASE_ENA    (1 << 6)
 #              define PACKET3_TC_ACTION_ENA        (1 << 23)
@@ -643,6 +864,7 @@
 #define		G_000E50_MCDW_BUSY(x)			(((x) >> 13) & 1)
 #define		G_000E50_SEM_BUSY(x)			(((x) >> 14) & 1)
 #define		G_000E50_RLC_BUSY(x)			(((x) >> 15) & 1)
+#define		G_000E50_BIF_BUSY(x)			(((x) >> 29) & 1)
 #define	R_000E60_SRBM_SOFT_RESET			0x0E60
 #define		S_000E60_SOFT_RESET_BIF(x)		(((x) & 1) << 1)
 #define		S_000E60_SOFT_RESET_CG(x)		(((x) & 1) << 2)
@@ -659,4 +881,495 @@
 #define		S_000E60_SOFT_RESET_TSC(x)		(((x) & 1) << 16)
 #define		S_000E60_SOFT_RESET_VMC(x)		(((x) & 1) << 17)
 
+#define R_005480_HDP_MEM_COHERENCY_FLUSH_CNTL		0x5480
+
+#define R_028C04_PA_SC_AA_CONFIG                     0x028C04
+#define   S_028C04_MSAA_NUM_SAMPLES(x)                 (((x) & 0x3) << 0)
+#define   G_028C04_MSAA_NUM_SAMPLES(x)                 (((x) >> 0) & 0x3)
+#define   C_028C04_MSAA_NUM_SAMPLES                    0xFFFFFFFC
+#define   S_028C04_AA_MASK_CENTROID_DTMN(x)            (((x) & 0x1) << 4)
+#define   G_028C04_AA_MASK_CENTROID_DTMN(x)            (((x) >> 4) & 0x1)
+#define   C_028C04_AA_MASK_CENTROID_DTMN               0xFFFFFFEF
+#define   S_028C04_MAX_SAMPLE_DIST(x)                  (((x) & 0xF) << 13)
+#define   G_028C04_MAX_SAMPLE_DIST(x)                  (((x) >> 13) & 0xF)
+#define   C_028C04_MAX_SAMPLE_DIST                     0xFFFE1FFF
+#define R_0280E0_CB_COLOR0_FRAG                      0x0280E0
+#define   S_0280E0_BASE_256B(x)                        (((x) & 0xFFFFFFFF) << 0)
+#define   G_0280E0_BASE_256B(x)                        (((x) >> 0) & 0xFFFFFFFF)
+#define   C_0280E0_BASE_256B                           0x00000000
+#define R_0280E4_CB_COLOR1_FRAG                      0x0280E4
+#define R_0280E8_CB_COLOR2_FRAG                      0x0280E8
+#define R_0280EC_CB_COLOR3_FRAG                      0x0280EC
+#define R_0280F0_CB_COLOR4_FRAG                      0x0280F0
+#define R_0280F4_CB_COLOR5_FRAG                      0x0280F4
+#define R_0280F8_CB_COLOR6_FRAG                      0x0280F8
+#define R_0280FC_CB_COLOR7_FRAG                      0x0280FC
+#define R_0280C0_CB_COLOR0_TILE                      0x0280C0
+#define   S_0280C0_BASE_256B(x)                        (((x) & 0xFFFFFFFF) << 0)
+#define   G_0280C0_BASE_256B(x)                        (((x) >> 0) & 0xFFFFFFFF)
+#define   C_0280C0_BASE_256B                           0x00000000
+#define R_0280C4_CB_COLOR1_TILE                      0x0280C4
+#define R_0280C8_CB_COLOR2_TILE                      0x0280C8
+#define R_0280CC_CB_COLOR3_TILE                      0x0280CC
+#define R_0280D0_CB_COLOR4_TILE                      0x0280D0
+#define R_0280D4_CB_COLOR5_TILE                      0x0280D4
+#define R_0280D8_CB_COLOR6_TILE                      0x0280D8
+#define R_0280DC_CB_COLOR7_TILE                      0x0280DC
+#define R_0280A0_CB_COLOR0_INFO                      0x0280A0
+#define   S_0280A0_ENDIAN(x)                           (((x) & 0x3) << 0)
+#define   G_0280A0_ENDIAN(x)                           (((x) >> 0) & 0x3)
+#define   C_0280A0_ENDIAN                              0xFFFFFFFC
+#define   S_0280A0_FORMAT(x)                           (((x) & 0x3F) << 2)
+#define   G_0280A0_FORMAT(x)                           (((x) >> 2) & 0x3F)
+#define   C_0280A0_FORMAT                              0xFFFFFF03
+#define     V_0280A0_COLOR_INVALID                     0x00000000
+#define     V_0280A0_COLOR_8                           0x00000001
+#define     V_0280A0_COLOR_4_4                         0x00000002
+#define     V_0280A0_COLOR_3_3_2                       0x00000003
+#define     V_0280A0_COLOR_16                          0x00000005
+#define     V_0280A0_COLOR_16_FLOAT                    0x00000006
+#define     V_0280A0_COLOR_8_8                         0x00000007
+#define     V_0280A0_COLOR_5_6_5                       0x00000008
+#define     V_0280A0_COLOR_6_5_5                       0x00000009
+#define     V_0280A0_COLOR_1_5_5_5                     0x0000000A
+#define     V_0280A0_COLOR_4_4_4_4                     0x0000000B
+#define     V_0280A0_COLOR_5_5_5_1                     0x0000000C
+#define     V_0280A0_COLOR_32                          0x0000000D
+#define     V_0280A0_COLOR_32_FLOAT                    0x0000000E
+#define     V_0280A0_COLOR_16_16                       0x0000000F
+#define     V_0280A0_COLOR_16_16_FLOAT                 0x00000010
+#define     V_0280A0_COLOR_8_24                        0x00000011
+#define     V_0280A0_COLOR_8_24_FLOAT                  0x00000012
+#define     V_0280A0_COLOR_24_8                        0x00000013
+#define     V_0280A0_COLOR_24_8_FLOAT                  0x00000014
+#define     V_0280A0_COLOR_10_11_11                    0x00000015
+#define     V_0280A0_COLOR_10_11_11_FLOAT              0x00000016
+#define     V_0280A0_COLOR_11_11_10                    0x00000017
+#define     V_0280A0_COLOR_11_11_10_FLOAT              0x00000018
+#define     V_0280A0_COLOR_2_10_10_10                  0x00000019
+#define     V_0280A0_COLOR_8_8_8_8                     0x0000001A
+#define     V_0280A0_COLOR_10_10_10_2                  0x0000001B
+#define     V_0280A0_COLOR_X24_8_32_FLOAT              0x0000001C
+#define     V_0280A0_COLOR_32_32                       0x0000001D
+#define     V_0280A0_COLOR_32_32_FLOAT                 0x0000001E
+#define     V_0280A0_COLOR_16_16_16_16                 0x0000001F
+#define     V_0280A0_COLOR_16_16_16_16_FLOAT           0x00000020
+#define     V_0280A0_COLOR_32_32_32_32                 0x00000022
+#define     V_0280A0_COLOR_32_32_32_32_FLOAT           0x00000023
+#define   S_0280A0_ARRAY_MODE(x)                       (((x) & 0xF) << 8)
+#define   G_0280A0_ARRAY_MODE(x)                       (((x) >> 8) & 0xF)
+#define   C_0280A0_ARRAY_MODE                          0xFFFFF0FF
+#define     V_0280A0_ARRAY_LINEAR_GENERAL              0x00000000
+#define     V_0280A0_ARRAY_LINEAR_ALIGNED              0x00000001
+#define     V_0280A0_ARRAY_1D_TILED_THIN1              0x00000002
+#define     V_0280A0_ARRAY_2D_TILED_THIN1              0x00000004
+#define   S_0280A0_NUMBER_TYPE(x)                      (((x) & 0x7) << 12)
+#define   G_0280A0_NUMBER_TYPE(x)                      (((x) >> 12) & 0x7)
+#define   C_0280A0_NUMBER_TYPE                         0xFFFF8FFF
+#define   S_0280A0_READ_SIZE(x)                        (((x) & 0x1) << 15)
+#define   G_0280A0_READ_SIZE(x)                        (((x) >> 15) & 0x1)
+#define   C_0280A0_READ_SIZE                           0xFFFF7FFF
+#define   S_0280A0_COMP_SWAP(x)                        (((x) & 0x3) << 16)
+#define   G_0280A0_COMP_SWAP(x)                        (((x) >> 16) & 0x3)
+#define   C_0280A0_COMP_SWAP                           0xFFFCFFFF
+#define   S_0280A0_TILE_MODE(x)                        (((x) & 0x3) << 18)
+#define   G_0280A0_TILE_MODE(x)                        (((x) >> 18) & 0x3)
+#define   C_0280A0_TILE_MODE                           0xFFF3FFFF
+#define   S_0280A0_BLEND_CLAMP(x)                      (((x) & 0x1) << 20)
+#define   G_0280A0_BLEND_CLAMP(x)                      (((x) >> 20) & 0x1)
+#define   C_0280A0_BLEND_CLAMP                         0xFFEFFFFF
+#define   S_0280A0_CLEAR_COLOR(x)                      (((x) & 0x1) << 21)
+#define   G_0280A0_CLEAR_COLOR(x)                      (((x) >> 21) & 0x1)
+#define   C_0280A0_CLEAR_COLOR                         0xFFDFFFFF
+#define   S_0280A0_BLEND_BYPASS(x)                     (((x) & 0x1) << 22)
+#define   G_0280A0_BLEND_BYPASS(x)                     (((x) >> 22) & 0x1)
+#define   C_0280A0_BLEND_BYPASS                        0xFFBFFFFF
+#define   S_0280A0_BLEND_FLOAT32(x)                    (((x) & 0x1) << 23)
+#define   G_0280A0_BLEND_FLOAT32(x)                    (((x) >> 23) & 0x1)
+#define   C_0280A0_BLEND_FLOAT32                       0xFF7FFFFF
+#define   S_0280A0_SIMPLE_FLOAT(x)                     (((x) & 0x1) << 24)
+#define   G_0280A0_SIMPLE_FLOAT(x)                     (((x) >> 24) & 0x1)
+#define   C_0280A0_SIMPLE_FLOAT                        0xFEFFFFFF
+#define   S_0280A0_ROUND_MODE(x)                       (((x) & 0x1) << 25)
+#define   G_0280A0_ROUND_MODE(x)                       (((x) >> 25) & 0x1)
+#define   C_0280A0_ROUND_MODE                          0xFDFFFFFF
+#define   S_0280A0_TILE_COMPACT(x)                     (((x) & 0x1) << 26)
+#define   G_0280A0_TILE_COMPACT(x)                     (((x) >> 26) & 0x1)
+#define   C_0280A0_TILE_COMPACT                        0xFBFFFFFF
+#define   S_0280A0_SOURCE_FORMAT(x)                    (((x) & 0x1) << 27)
+#define   G_0280A0_SOURCE_FORMAT(x)                    (((x) >> 27) & 0x1)
+#define   C_0280A0_SOURCE_FORMAT                       0xF7FFFFFF
+#define R_0280A4_CB_COLOR1_INFO                      0x0280A4
+#define R_0280A8_CB_COLOR2_INFO                      0x0280A8
+#define R_0280AC_CB_COLOR3_INFO                      0x0280AC
+#define R_0280B0_CB_COLOR4_INFO                      0x0280B0
+#define R_0280B4_CB_COLOR5_INFO                      0x0280B4
+#define R_0280B8_CB_COLOR6_INFO                      0x0280B8
+#define R_0280BC_CB_COLOR7_INFO                      0x0280BC
+#define R_028060_CB_COLOR0_SIZE                      0x028060
+#define   S_028060_PITCH_TILE_MAX(x)                   (((x) & 0x3FF) << 0)
+#define   G_028060_PITCH_TILE_MAX(x)                   (((x) >> 0) & 0x3FF)
+#define   C_028060_PITCH_TILE_MAX                      0xFFFFFC00
+#define   S_028060_SLICE_TILE_MAX(x)                   (((x) & 0xFFFFF) << 10)
+#define   G_028060_SLICE_TILE_MAX(x)                   (((x) >> 10) & 0xFFFFF)
+#define   C_028060_SLICE_TILE_MAX                      0xC00003FF
+#define R_028064_CB_COLOR1_SIZE                      0x028064
+#define R_028068_CB_COLOR2_SIZE                      0x028068
+#define R_02806C_CB_COLOR3_SIZE                      0x02806C
+#define R_028070_CB_COLOR4_SIZE                      0x028070
+#define R_028074_CB_COLOR5_SIZE                      0x028074
+#define R_028078_CB_COLOR6_SIZE                      0x028078
+#define R_02807C_CB_COLOR7_SIZE                      0x02807C
+#define R_028238_CB_TARGET_MASK                      0x028238
+#define   S_028238_TARGET0_ENABLE(x)                   (((x) & 0xF) << 0)
+#define   G_028238_TARGET0_ENABLE(x)                   (((x) >> 0) & 0xF)
+#define   C_028238_TARGET0_ENABLE                      0xFFFFFFF0
+#define   S_028238_TARGET1_ENABLE(x)                   (((x) & 0xF) << 4)
+#define   G_028238_TARGET1_ENABLE(x)                   (((x) >> 4) & 0xF)
+#define   C_028238_TARGET1_ENABLE                      0xFFFFFF0F
+#define   S_028238_TARGET2_ENABLE(x)                   (((x) & 0xF) << 8)
+#define   G_028238_TARGET2_ENABLE(x)                   (((x) >> 8) & 0xF)
+#define   C_028238_TARGET2_ENABLE                      0xFFFFF0FF
+#define   S_028238_TARGET3_ENABLE(x)                   (((x) & 0xF) << 12)
+#define   G_028238_TARGET3_ENABLE(x)                   (((x) >> 12) & 0xF)
+#define   C_028238_TARGET3_ENABLE                      0xFFFF0FFF
+#define   S_028238_TARGET4_ENABLE(x)                   (((x) & 0xF) << 16)
+#define   G_028238_TARGET4_ENABLE(x)                   (((x) >> 16) & 0xF)
+#define   C_028238_TARGET4_ENABLE                      0xFFF0FFFF
+#define   S_028238_TARGET5_ENABLE(x)                   (((x) & 0xF) << 20)
+#define   G_028238_TARGET5_ENABLE(x)                   (((x) >> 20) & 0xF)
+#define   C_028238_TARGET5_ENABLE                      0xFF0FFFFF
+#define   S_028238_TARGET6_ENABLE(x)                   (((x) & 0xF) << 24)
+#define   G_028238_TARGET6_ENABLE(x)                   (((x) >> 24) & 0xF)
+#define   C_028238_TARGET6_ENABLE                      0xF0FFFFFF
+#define   S_028238_TARGET7_ENABLE(x)                   (((x) & 0xF) << 28)
+#define   G_028238_TARGET7_ENABLE(x)                   (((x) >> 28) & 0xF)
+#define   C_028238_TARGET7_ENABLE                      0x0FFFFFFF
+#define R_02823C_CB_SHADER_MASK                      0x02823C
+#define   S_02823C_OUTPUT0_ENABLE(x)                   (((x) & 0xF) << 0)
+#define   G_02823C_OUTPUT0_ENABLE(x)                   (((x) >> 0) & 0xF)
+#define   C_02823C_OUTPUT0_ENABLE                      0xFFFFFFF0
+#define   S_02823C_OUTPUT1_ENABLE(x)                   (((x) & 0xF) << 4)
+#define   G_02823C_OUTPUT1_ENABLE(x)                   (((x) >> 4) & 0xF)
+#define   C_02823C_OUTPUT1_ENABLE                      0xFFFFFF0F
+#define   S_02823C_OUTPUT2_ENABLE(x)                   (((x) & 0xF) << 8)
+#define   G_02823C_OUTPUT2_ENABLE(x)                   (((x) >> 8) & 0xF)
+#define   C_02823C_OUTPUT2_ENABLE                      0xFFFFF0FF
+#define   S_02823C_OUTPUT3_ENABLE(x)                   (((x) & 0xF) << 12)
+#define   G_02823C_OUTPUT3_ENABLE(x)                   (((x) >> 12) & 0xF)
+#define   C_02823C_OUTPUT3_ENABLE                      0xFFFF0FFF
+#define   S_02823C_OUTPUT4_ENABLE(x)                   (((x) & 0xF) << 16)
+#define   G_02823C_OUTPUT4_ENABLE(x)                   (((x) >> 16) & 0xF)
+#define   C_02823C_OUTPUT4_ENABLE                      0xFFF0FFFF
+#define   S_02823C_OUTPUT5_ENABLE(x)                   (((x) & 0xF) << 20)
+#define   G_02823C_OUTPUT5_ENABLE(x)                   (((x) >> 20) & 0xF)
+#define   C_02823C_OUTPUT5_ENABLE                      0xFF0FFFFF
+#define   S_02823C_OUTPUT6_ENABLE(x)                   (((x) & 0xF) << 24)
+#define   G_02823C_OUTPUT6_ENABLE(x)                   (((x) >> 24) & 0xF)
+#define   C_02823C_OUTPUT6_ENABLE                      0xF0FFFFFF
+#define   S_02823C_OUTPUT7_ENABLE(x)                   (((x) & 0xF) << 28)
+#define   G_02823C_OUTPUT7_ENABLE(x)                   (((x) >> 28) & 0xF)
+#define   C_02823C_OUTPUT7_ENABLE                      0x0FFFFFFF
+#define R_028AB0_VGT_STRMOUT_EN                      0x028AB0
+#define   S_028AB0_STREAMOUT(x)                        (((x) & 0x1) << 0)
+#define   G_028AB0_STREAMOUT(x)                        (((x) >> 0) & 0x1)
+#define   C_028AB0_STREAMOUT                           0xFFFFFFFE
+#define R_028B20_VGT_STRMOUT_BUFFER_EN               0x028B20
+#define   S_028B20_BUFFER_0_EN(x)                      (((x) & 0x1) << 0)
+#define   G_028B20_BUFFER_0_EN(x)                      (((x) >> 0) & 0x1)
+#define   C_028B20_BUFFER_0_EN                         0xFFFFFFFE
+#define   S_028B20_BUFFER_1_EN(x)                      (((x) & 0x1) << 1)
+#define   G_028B20_BUFFER_1_EN(x)                      (((x) >> 1) & 0x1)
+#define   C_028B20_BUFFER_1_EN                         0xFFFFFFFD
+#define   S_028B20_BUFFER_2_EN(x)                      (((x) & 0x1) << 2)
+#define   G_028B20_BUFFER_2_EN(x)                      (((x) >> 2) & 0x1)
+#define   C_028B20_BUFFER_2_EN                         0xFFFFFFFB
+#define   S_028B20_BUFFER_3_EN(x)                      (((x) & 0x1) << 3)
+#define   G_028B20_BUFFER_3_EN(x)                      (((x) >> 3) & 0x1)
+#define   C_028B20_BUFFER_3_EN                         0xFFFFFFF7
+#define   S_028B20_SIZE(x)                             (((x) & 0xFFFFFFFF) << 0)
+#define   G_028B20_SIZE(x)                             (((x) >> 0) & 0xFFFFFFFF)
+#define   C_028B20_SIZE                                0x00000000
+#define R_038000_SQ_TEX_RESOURCE_WORD0_0             0x038000
+#define   S_038000_DIM(x)                              (((x) & 0x7) << 0)
+#define   G_038000_DIM(x)                              (((x) >> 0) & 0x7)
+#define   C_038000_DIM                                 0xFFFFFFF8
+#define     V_038000_SQ_TEX_DIM_1D                     0x00000000
+#define     V_038000_SQ_TEX_DIM_2D                     0x00000001
+#define     V_038000_SQ_TEX_DIM_3D                     0x00000002
+#define     V_038000_SQ_TEX_DIM_CUBEMAP                0x00000003
+#define     V_038000_SQ_TEX_DIM_1D_ARRAY               0x00000004
+#define     V_038000_SQ_TEX_DIM_2D_ARRAY               0x00000005
+#define     V_038000_SQ_TEX_DIM_2D_MSAA                0x00000006
+#define     V_038000_SQ_TEX_DIM_2D_ARRAY_MSAA          0x00000007
+#define   S_038000_TILE_MODE(x)                        (((x) & 0xF) << 3)
+#define   G_038000_TILE_MODE(x)                        (((x) >> 3) & 0xF)
+#define   C_038000_TILE_MODE                           0xFFFFFF87
+#define   S_038000_TILE_TYPE(x)                        (((x) & 0x1) << 7)
+#define   G_038000_TILE_TYPE(x)                        (((x) >> 7) & 0x1)
+#define   C_038000_TILE_TYPE                           0xFFFFFF7F
+#define   S_038000_PITCH(x)                            (((x) & 0x7FF) << 8)
+#define   G_038000_PITCH(x)                            (((x) >> 8) & 0x7FF)
+#define   C_038000_PITCH                               0xFFF800FF
+#define   S_038000_TEX_WIDTH(x)                        (((x) & 0x1FFF) << 19)
+#define   G_038000_TEX_WIDTH(x)                        (((x) >> 19) & 0x1FFF)
+#define   C_038000_TEX_WIDTH                           0x0007FFFF
+#define R_038004_SQ_TEX_RESOURCE_WORD1_0             0x038004
+#define   S_038004_TEX_HEIGHT(x)                       (((x) & 0x1FFF) << 0)
+#define   G_038004_TEX_HEIGHT(x)                       (((x) >> 0) & 0x1FFF)
+#define   C_038004_TEX_HEIGHT                          0xFFFFE000
+#define   S_038004_TEX_DEPTH(x)                        (((x) & 0x1FFF) << 13)
+#define   G_038004_TEX_DEPTH(x)                        (((x) >> 13) & 0x1FFF)
+#define   C_038004_TEX_DEPTH                           0xFC001FFF
+#define   S_038004_DATA_FORMAT(x)                      (((x) & 0x3F) << 26)
+#define   G_038004_DATA_FORMAT(x)                      (((x) >> 26) & 0x3F)
+#define   C_038004_DATA_FORMAT                         0x03FFFFFF
+#define     V_038004_COLOR_INVALID                     0x00000000
+#define     V_038004_COLOR_8                           0x00000001
+#define     V_038004_COLOR_4_4                         0x00000002
+#define     V_038004_COLOR_3_3_2                       0x00000003
+#define     V_038004_COLOR_16                          0x00000005
+#define     V_038004_COLOR_16_FLOAT                    0x00000006
+#define     V_038004_COLOR_8_8                         0x00000007
+#define     V_038004_COLOR_5_6_5                       0x00000008
+#define     V_038004_COLOR_6_5_5                       0x00000009
+#define     V_038004_COLOR_1_5_5_5                     0x0000000A
+#define     V_038004_COLOR_4_4_4_4                     0x0000000B
+#define     V_038004_COLOR_5_5_5_1                     0x0000000C
+#define     V_038004_COLOR_32                          0x0000000D
+#define     V_038004_COLOR_32_FLOAT                    0x0000000E
+#define     V_038004_COLOR_16_16                       0x0000000F
+#define     V_038004_COLOR_16_16_FLOAT                 0x00000010
+#define     V_038004_COLOR_8_24                        0x00000011
+#define     V_038004_COLOR_8_24_FLOAT                  0x00000012
+#define     V_038004_COLOR_24_8                        0x00000013
+#define     V_038004_COLOR_24_8_FLOAT                  0x00000014
+#define     V_038004_COLOR_10_11_11                    0x00000015
+#define     V_038004_COLOR_10_11_11_FLOAT              0x00000016
+#define     V_038004_COLOR_11_11_10                    0x00000017
+#define     V_038004_COLOR_11_11_10_FLOAT              0x00000018
+#define     V_038004_COLOR_2_10_10_10                  0x00000019
+#define     V_038004_COLOR_8_8_8_8                     0x0000001A
+#define     V_038004_COLOR_10_10_10_2                  0x0000001B
+#define     V_038004_COLOR_X24_8_32_FLOAT              0x0000001C
+#define     V_038004_COLOR_32_32                       0x0000001D
+#define     V_038004_COLOR_32_32_FLOAT                 0x0000001E
+#define     V_038004_COLOR_16_16_16_16                 0x0000001F
+#define     V_038004_COLOR_16_16_16_16_FLOAT           0x00000020
+#define     V_038004_COLOR_32_32_32_32                 0x00000022
+#define     V_038004_COLOR_32_32_32_32_FLOAT           0x00000023
+#define     V_038004_FMT_1                             0x00000025
+#define     V_038004_FMT_GB_GR                         0x00000027
+#define     V_038004_FMT_BG_RG                         0x00000028
+#define     V_038004_FMT_32_AS_8                       0x00000029
+#define     V_038004_FMT_32_AS_8_8                     0x0000002A
+#define     V_038004_FMT_5_9_9_9_SHAREDEXP             0x0000002B
+#define     V_038004_FMT_8_8_8                         0x0000002C
+#define     V_038004_FMT_16_16_16                      0x0000002D
+#define     V_038004_FMT_16_16_16_FLOAT                0x0000002E
+#define     V_038004_FMT_32_32_32                      0x0000002F
+#define     V_038004_FMT_32_32_32_FLOAT                0x00000030
+#define R_038010_SQ_TEX_RESOURCE_WORD4_0             0x038010
+#define   S_038010_FORMAT_COMP_X(x)                    (((x) & 0x3) << 0)
+#define   G_038010_FORMAT_COMP_X(x)                    (((x) >> 0) & 0x3)
+#define   C_038010_FORMAT_COMP_X                       0xFFFFFFFC
+#define   S_038010_FORMAT_COMP_Y(x)                    (((x) & 0x3) << 2)
+#define   G_038010_FORMAT_COMP_Y(x)                    (((x) >> 2) & 0x3)
+#define   C_038010_FORMAT_COMP_Y                       0xFFFFFFF3
+#define   S_038010_FORMAT_COMP_Z(x)                    (((x) & 0x3) << 4)
+#define   G_038010_FORMAT_COMP_Z(x)                    (((x) >> 4) & 0x3)
+#define   C_038010_FORMAT_COMP_Z                       0xFFFFFFCF
+#define   S_038010_FORMAT_COMP_W(x)                    (((x) & 0x3) << 6)
+#define   G_038010_FORMAT_COMP_W(x)                    (((x) >> 6) & 0x3)
+#define   C_038010_FORMAT_COMP_W                       0xFFFFFF3F
+#define   S_038010_NUM_FORMAT_ALL(x)                   (((x) & 0x3) << 8)
+#define   G_038010_NUM_FORMAT_ALL(x)                   (((x) >> 8) & 0x3)
+#define   C_038010_NUM_FORMAT_ALL                      0xFFFFFCFF
+#define   S_038010_SRF_MODE_ALL(x)                     (((x) & 0x1) << 10)
+#define   G_038010_SRF_MODE_ALL(x)                     (((x) >> 10) & 0x1)
+#define   C_038010_SRF_MODE_ALL                        0xFFFFFBFF
+#define   S_038010_FORCE_DEGAMMA(x)                    (((x) & 0x1) << 11)
+#define   G_038010_FORCE_DEGAMMA(x)                    (((x) >> 11) & 0x1)
+#define   C_038010_FORCE_DEGAMMA                       0xFFFFF7FF
+#define   S_038010_ENDIAN_SWAP(x)                      (((x) & 0x3) << 12)
+#define   G_038010_ENDIAN_SWAP(x)                      (((x) >> 12) & 0x3)
+#define   C_038010_ENDIAN_SWAP                         0xFFFFCFFF
+#define   S_038010_REQUEST_SIZE(x)                     (((x) & 0x3) << 14)
+#define   G_038010_REQUEST_SIZE(x)                     (((x) >> 14) & 0x3)
+#define   C_038010_REQUEST_SIZE                        0xFFFF3FFF
+#define   S_038010_DST_SEL_X(x)                        (((x) & 0x7) << 16)
+#define   G_038010_DST_SEL_X(x)                        (((x) >> 16) & 0x7)
+#define   C_038010_DST_SEL_X                           0xFFF8FFFF
+#define   S_038010_DST_SEL_Y(x)                        (((x) & 0x7) << 19)
+#define   G_038010_DST_SEL_Y(x)                        (((x) >> 19) & 0x7)
+#define   C_038010_DST_SEL_Y                           0xFFC7FFFF
+#define   S_038010_DST_SEL_Z(x)                        (((x) & 0x7) << 22)
+#define   G_038010_DST_SEL_Z(x)                        (((x) >> 22) & 0x7)
+#define   C_038010_DST_SEL_Z                           0xFE3FFFFF
+#define   S_038010_DST_SEL_W(x)                        (((x) & 0x7) << 25)
+#define   G_038010_DST_SEL_W(x)                        (((x) >> 25) & 0x7)
+#define   C_038010_DST_SEL_W                           0xF1FFFFFF
+#define   S_038010_BASE_LEVEL(x)                       (((x) & 0xF) << 28)
+#define   G_038010_BASE_LEVEL(x)                       (((x) >> 28) & 0xF)
+#define   C_038010_BASE_LEVEL                          0x0FFFFFFF
+#define R_038014_SQ_TEX_RESOURCE_WORD5_0             0x038014
+#define   S_038014_LAST_LEVEL(x)                       (((x) & 0xF) << 0)
+#define   G_038014_LAST_LEVEL(x)                       (((x) >> 0) & 0xF)
+#define   C_038014_LAST_LEVEL                          0xFFFFFFF0
+#define   S_038014_BASE_ARRAY(x)                       (((x) & 0x1FFF) << 4)
+#define   G_038014_BASE_ARRAY(x)                       (((x) >> 4) & 0x1FFF)
+#define   C_038014_BASE_ARRAY                          0xFFFE000F
+#define   S_038014_LAST_ARRAY(x)                       (((x) & 0x1FFF) << 17)
+#define   G_038014_LAST_ARRAY(x)                       (((x) >> 17) & 0x1FFF)
+#define   C_038014_LAST_ARRAY                          0xC001FFFF
+#define R_0288A8_SQ_ESGS_RING_ITEMSIZE               0x0288A8
+#define   S_0288A8_ITEMSIZE(x)                         (((x) & 0x7FFF) << 0)
+#define   G_0288A8_ITEMSIZE(x)                         (((x) >> 0) & 0x7FFF)
+#define   C_0288A8_ITEMSIZE                            0xFFFF8000
+#define R_008C44_SQ_ESGS_RING_SIZE                   0x008C44
+#define   S_008C44_MEM_SIZE(x)                         (((x) & 0xFFFFFFFF) << 0)
+#define   G_008C44_MEM_SIZE(x)                         (((x) >> 0) & 0xFFFFFFFF)
+#define   C_008C44_MEM_SIZE                            0x00000000
+#define R_0288B0_SQ_ESTMP_RING_ITEMSIZE              0x0288B0
+#define   S_0288B0_ITEMSIZE(x)                         (((x) & 0x7FFF) << 0)
+#define   G_0288B0_ITEMSIZE(x)                         (((x) >> 0) & 0x7FFF)
+#define   C_0288B0_ITEMSIZE                            0xFFFF8000
+#define R_008C54_SQ_ESTMP_RING_SIZE                  0x008C54
+#define   S_008C54_MEM_SIZE(x)                         (((x) & 0xFFFFFFFF) << 0)
+#define   G_008C54_MEM_SIZE(x)                         (((x) >> 0) & 0xFFFFFFFF)
+#define   C_008C54_MEM_SIZE                            0x00000000
+#define R_0288C0_SQ_FBUF_RING_ITEMSIZE               0x0288C0
+#define   S_0288C0_ITEMSIZE(x)                         (((x) & 0x7FFF) << 0)
+#define   G_0288C0_ITEMSIZE(x)                         (((x) >> 0) & 0x7FFF)
+#define   C_0288C0_ITEMSIZE                            0xFFFF8000
+#define R_008C74_SQ_FBUF_RING_SIZE                   0x008C74
+#define   S_008C74_MEM_SIZE(x)                         (((x) & 0xFFFFFFFF) << 0)
+#define   G_008C74_MEM_SIZE(x)                         (((x) >> 0) & 0xFFFFFFFF)
+#define   C_008C74_MEM_SIZE                            0x00000000
+#define R_0288B4_SQ_GSTMP_RING_ITEMSIZE              0x0288B4
+#define   S_0288B4_ITEMSIZE(x)                         (((x) & 0x7FFF) << 0)
+#define   G_0288B4_ITEMSIZE(x)                         (((x) >> 0) & 0x7FFF)
+#define   C_0288B4_ITEMSIZE                            0xFFFF8000
+#define R_008C5C_SQ_GSTMP_RING_SIZE                  0x008C5C
+#define   S_008C5C_MEM_SIZE(x)                         (((x) & 0xFFFFFFFF) << 0)
+#define   G_008C5C_MEM_SIZE(x)                         (((x) >> 0) & 0xFFFFFFFF)
+#define   C_008C5C_MEM_SIZE                            0x00000000
+#define R_0288AC_SQ_GSVS_RING_ITEMSIZE               0x0288AC
+#define   S_0288AC_ITEMSIZE(x)                         (((x) & 0x7FFF) << 0)
+#define   G_0288AC_ITEMSIZE(x)                         (((x) >> 0) & 0x7FFF)
+#define   C_0288AC_ITEMSIZE                            0xFFFF8000
+#define R_008C4C_SQ_GSVS_RING_SIZE                   0x008C4C
+#define   S_008C4C_MEM_SIZE(x)                         (((x) & 0xFFFFFFFF) << 0)
+#define   G_008C4C_MEM_SIZE(x)                         (((x) >> 0) & 0xFFFFFFFF)
+#define   C_008C4C_MEM_SIZE                            0x00000000
+#define R_0288BC_SQ_PSTMP_RING_ITEMSIZE              0x0288BC
+#define   S_0288BC_ITEMSIZE(x)                         (((x) & 0x7FFF) << 0)
+#define   G_0288BC_ITEMSIZE(x)                         (((x) >> 0) & 0x7FFF)
+#define   C_0288BC_ITEMSIZE                            0xFFFF8000
+#define R_008C6C_SQ_PSTMP_RING_SIZE                  0x008C6C
+#define   S_008C6C_MEM_SIZE(x)                         (((x) & 0xFFFFFFFF) << 0)
+#define   G_008C6C_MEM_SIZE(x)                         (((x) >> 0) & 0xFFFFFFFF)
+#define   C_008C6C_MEM_SIZE                            0x00000000
+#define R_0288C4_SQ_REDUC_RING_ITEMSIZE              0x0288C4
+#define   S_0288C4_ITEMSIZE(x)                         (((x) & 0x7FFF) << 0)
+#define   G_0288C4_ITEMSIZE(x)                         (((x) >> 0) & 0x7FFF)
+#define   C_0288C4_ITEMSIZE                            0xFFFF8000
+#define R_008C7C_SQ_REDUC_RING_SIZE                  0x008C7C
+#define   S_008C7C_MEM_SIZE(x)                         (((x) & 0xFFFFFFFF) << 0)
+#define   G_008C7C_MEM_SIZE(x)                         (((x) >> 0) & 0xFFFFFFFF)
+#define   C_008C7C_MEM_SIZE                            0x00000000
+#define R_0288B8_SQ_VSTMP_RING_ITEMSIZE              0x0288B8
+#define   S_0288B8_ITEMSIZE(x)                         (((x) & 0x7FFF) << 0)
+#define   G_0288B8_ITEMSIZE(x)                         (((x) >> 0) & 0x7FFF)
+#define   C_0288B8_ITEMSIZE                            0xFFFF8000
+#define R_008C64_SQ_VSTMP_RING_SIZE                  0x008C64
+#define   S_008C64_MEM_SIZE(x)                         (((x) & 0xFFFFFFFF) << 0)
+#define   G_008C64_MEM_SIZE(x)                         (((x) >> 0) & 0xFFFFFFFF)
+#define   C_008C64_MEM_SIZE                            0x00000000
+#define R_0288C8_SQ_GS_VERT_ITEMSIZE                 0x0288C8
+#define   S_0288C8_ITEMSIZE(x)                         (((x) & 0x7FFF) << 0)
+#define   G_0288C8_ITEMSIZE(x)                         (((x) >> 0) & 0x7FFF)
+#define   C_0288C8_ITEMSIZE                            0xFFFF8000
+#define R_028010_DB_DEPTH_INFO                       0x028010
+#define   S_028010_FORMAT(x)                           (((x) & 0x7) << 0)
+#define   G_028010_FORMAT(x)                           (((x) >> 0) & 0x7)
+#define   C_028010_FORMAT                              0xFFFFFFF8
+#define     V_028010_DEPTH_INVALID                     0x00000000
+#define     V_028010_DEPTH_16                          0x00000001
+#define     V_028010_DEPTH_X8_24                       0x00000002
+#define     V_028010_DEPTH_8_24                        0x00000003
+#define     V_028010_DEPTH_X8_24_FLOAT                 0x00000004
+#define     V_028010_DEPTH_8_24_FLOAT                  0x00000005
+#define     V_028010_DEPTH_32_FLOAT                    0x00000006
+#define     V_028010_DEPTH_X24_8_32_FLOAT              0x00000007
+#define   S_028010_READ_SIZE(x)                        (((x) & 0x1) << 3)
+#define   G_028010_READ_SIZE(x)                        (((x) >> 3) & 0x1)
+#define   C_028010_READ_SIZE                           0xFFFFFFF7
+#define   S_028010_ARRAY_MODE(x)                       (((x) & 0xF) << 15)
+#define   G_028010_ARRAY_MODE(x)                       (((x) >> 15) & 0xF)
+#define   C_028010_ARRAY_MODE                          0xFFF87FFF
+#define   S_028010_TILE_SURFACE_ENABLE(x)              (((x) & 0x1) << 25)
+#define   G_028010_TILE_SURFACE_ENABLE(x)              (((x) >> 25) & 0x1)
+#define   C_028010_TILE_SURFACE_ENABLE                 0xFDFFFFFF
+#define   S_028010_TILE_COMPACT(x)                     (((x) & 0x1) << 26)
+#define   G_028010_TILE_COMPACT(x)                     (((x) >> 26) & 0x1)
+#define   C_028010_TILE_COMPACT                        0xFBFFFFFF
+#define   S_028010_ZRANGE_PRECISION(x)                 (((x) & 0x1) << 31)
+#define   G_028010_ZRANGE_PRECISION(x)                 (((x) >> 31) & 0x1)
+#define   C_028010_ZRANGE_PRECISION                    0x7FFFFFFF
+#define R_028000_DB_DEPTH_SIZE                       0x028000
+#define   S_028000_PITCH_TILE_MAX(x)                   (((x) & 0x3FF) << 0)
+#define   G_028000_PITCH_TILE_MAX(x)                   (((x) >> 0) & 0x3FF)
+#define   C_028000_PITCH_TILE_MAX                      0xFFFFFC00
+#define   S_028000_SLICE_TILE_MAX(x)                   (((x) & 0xFFFFF) << 10)
+#define   G_028000_SLICE_TILE_MAX(x)                   (((x) >> 10) & 0xFFFFF)
+#define   C_028000_SLICE_TILE_MAX                      0xC00003FF
+#define R_028004_DB_DEPTH_VIEW                       0x028004
+#define   S_028004_SLICE_START(x)                      (((x) & 0x7FF) << 0)
+#define   G_028004_SLICE_START(x)                      (((x) >> 0) & 0x7FF)
+#define   C_028004_SLICE_START                         0xFFFFF800
+#define   S_028004_SLICE_MAX(x)                        (((x) & 0x7FF) << 13)
+#define   G_028004_SLICE_MAX(x)                        (((x) >> 13) & 0x7FF)
+#define   C_028004_SLICE_MAX                           0xFF001FFF
+#define R_028800_DB_DEPTH_CONTROL                    0x028800
+#define   S_028800_STENCIL_ENABLE(x)                   (((x) & 0x1) << 0)
+#define   G_028800_STENCIL_ENABLE(x)                   (((x) >> 0) & 0x1)
+#define   C_028800_STENCIL_ENABLE                      0xFFFFFFFE
+#define   S_028800_Z_ENABLE(x)                         (((x) & 0x1) << 1)
+#define   G_028800_Z_ENABLE(x)                         (((x) >> 1) & 0x1)
+#define   C_028800_Z_ENABLE                            0xFFFFFFFD
+#define   S_028800_Z_WRITE_ENABLE(x)                   (((x) & 0x1) << 2)
+#define   G_028800_Z_WRITE_ENABLE(x)                   (((x) >> 2) & 0x1)
+#define   C_028800_Z_WRITE_ENABLE                      0xFFFFFFFB
+#define   S_028800_ZFUNC(x)                            (((x) & 0x7) << 4)
+#define   G_028800_ZFUNC(x)                            (((x) >> 4) & 0x7)
+#define   C_028800_ZFUNC                               0xFFFFFF8F
+#define   S_028800_BACKFACE_ENABLE(x)                  (((x) & 0x1) << 7)
+#define   G_028800_BACKFACE_ENABLE(x)                  (((x) >> 7) & 0x1)
+#define   C_028800_BACKFACE_ENABLE                     0xFFFFFF7F
+#define   S_028800_STENCILFUNC(x)                      (((x) & 0x7) << 8)
+#define   G_028800_STENCILFUNC(x)                      (((x) >> 8) & 0x7)
+#define   C_028800_STENCILFUNC                         0xFFFFF8FF
+#define   S_028800_STENCILFAIL(x)                      (((x) & 0x7) << 11)
+#define   G_028800_STENCILFAIL(x)                      (((x) >> 11) & 0x7)
+#define   C_028800_STENCILFAIL                         0xFFFFC7FF
+#define   S_028800_STENCILZPASS(x)                     (((x) & 0x7) << 14)
+#define   G_028800_STENCILZPASS(x)                     (((x) >> 14) & 0x7)
+#define   C_028800_STENCILZPASS                        0xFFFE3FFF
+#define   S_028800_STENCILZFAIL(x)                     (((x) & 0x7) << 17)
+#define   G_028800_STENCILZFAIL(x)                     (((x) >> 17) & 0x7)
+#define   C_028800_STENCILZFAIL                        0xFFF1FFFF
+#define   S_028800_STENCILFUNC_BF(x)                   (((x) & 0x7) << 20)
+#define   G_028800_STENCILFUNC_BF(x)                   (((x) >> 20) & 0x7)
+#define   C_028800_STENCILFUNC_BF                      0xFF8FFFFF
+#define   S_028800_STENCILFAIL_BF(x)                   (((x) & 0x7) << 23)
+#define   G_028800_STENCILFAIL_BF(x)                   (((x) >> 23) & 0x7)
+#define   C_028800_STENCILFAIL_BF                      0xFC7FFFFF
+#define   S_028800_STENCILZPASS_BF(x)                  (((x) & 0x7) << 26)
+#define   G_028800_STENCILZPASS_BF(x)                  (((x) >> 26) & 0x7)
+#define   C_028800_STENCILZPASS_BF                     0xE3FFFFFF
+#define   S_028800_STENCILZFAIL_BF(x)                  (((x) & 0x7) << 29)
+#define   G_028800_STENCILZFAIL_BF(x)                  (((x) >> 29) & 0x7)
+#define   C_028800_STENCILZFAIL_BF                     0x1FFFFFFF
+
 #endif
diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h
index 6311b1362594..829e26e8a4bb 100644
--- a/drivers/gpu/drm/radeon/radeon.h
+++ b/drivers/gpu/drm/radeon/radeon.h
@@ -28,8 +28,6 @@
 #ifndef __RADEON_H__
 #define __RADEON_H__
 
-#include "radeon_object.h"
-
 /* TODO: Here are things that needs to be done :
  *	- surface allocator & initializer : (bit like scratch reg) should
  *	  initialize HDP_ stuff on RS600, R600, R700 hw, well anythings
@@ -44,11 +42,34 @@
  *	- TESTING, TESTING, TESTING
  */
 
+/* Initialization path:
+ *  We expect that acceleration initialization might fail for various
+ *  reasons even thought we work hard to make it works on most
+ *  configurations. In order to still have a working userspace in such
+ *  situation the init path must succeed up to the memory controller
+ *  initialization point. Failure before this point are considered as
+ *  fatal error. Here is the init callchain :
+ *      radeon_device_init  perform common structure, mutex initialization
+ *      asic_init           setup the GPU memory layout and perform all
+ *                          one time initialization (failure in this
+ *                          function are considered fatal)
+ *      asic_startup        setup the GPU acceleration, in order to
+ *                          follow guideline the first thing this
+ *                          function should do is setting the GPU
+ *                          memory controller (only MC setup failure
+ *                          are considered as fatal)
+ */
+
 #include <asm/atomic.h>
 #include <linux/wait.h>
 #include <linux/list.h>
 #include <linux/kref.h>
 
+#include <ttm/ttm_bo_api.h>
+#include <ttm/ttm_bo_driver.h>
+#include <ttm/ttm_placement.h>
+#include <ttm/ttm_module.h>
+
 #include "radeon_family.h"
 #include "radeon_mode.h"
 #include "radeon_reg.h"
@@ -67,12 +88,16 @@ extern int radeon_benchmarking;
 extern int radeon_testing;
 extern int radeon_connector_table;
 extern int radeon_tv;
+extern int radeon_new_pll;
+extern int radeon_dynpm;
+extern int radeon_audio;
 
 /*
  * Copy from radeon_drv.h so we don't have to include both and have conflicting
  * symbol;
  */
 #define RADEON_MAX_USEC_TIMEOUT		100000	/* 100 ms */
+/* RADEON_IB_POOL_SIZE must be a power of 2 */
 #define RADEON_IB_POOL_SIZE		16
 #define RADEON_DEBUGFS_MAX_NUM_FILES	32
 #define RADEONFB_CONN_LIMIT		4
@@ -94,6 +119,21 @@ struct radeon_device;
 /*
  * BIOS.
  */
+#define ATRM_BIOS_PAGE 4096
+
+#if defined(CONFIG_VGA_SWITCHEROO)
+bool radeon_atrm_supported(struct pci_dev *pdev);
+int radeon_atrm_get_bios_chunk(uint8_t *bios, int offset, int len);
+#else
+static inline bool radeon_atrm_supported(struct pci_dev *pdev)
+{
+	return false;
+}
+
+static inline int radeon_atrm_get_bios_chunk(uint8_t *bios, int offset, int len){
+	return -EINVAL;
+}
+#endif
 bool radeon_get_bios(struct radeon_device *rdev);
 
 
@@ -114,13 +154,23 @@ void radeon_dummy_page_fini(struct radeon_device *rdev);
 struct radeon_clock {
 	struct radeon_pll p1pll;
 	struct radeon_pll p2pll;
+	struct radeon_pll dcpll;
 	struct radeon_pll spll;
 	struct radeon_pll mpll;
 	/* 10 Khz units */
 	uint32_t default_mclk;
 	uint32_t default_sclk;
+	uint32_t default_dispclk;
+	uint32_t dp_extclk;
 };
 
+/*
+ * Power management
+ */
+int radeon_pm_init(struct radeon_device *rdev);
+void radeon_pm_compute_clocks(struct radeon_device *rdev);
+void radeon_combios_get_power_modes(struct radeon_device *rdev);
+void radeon_atombios_get_power_modes(struct radeon_device *rdev);
 
 /*
  * Fences.
@@ -135,6 +185,7 @@ struct radeon_fence_driver {
 	struct list_head		created;
 	struct list_head		emited;
 	struct list_head		signaled;
+	bool				initialized;
 };
 
 struct radeon_fence {
@@ -164,76 +215,63 @@ void radeon_fence_unref(struct radeon_fence **fence);
  * Tiling registers
  */
 struct radeon_surface_reg {
-	struct radeon_object *robj;
+	struct radeon_bo *bo;
 };
 
 #define RADEON_GEM_MAX_SURFACES 8
 
 /*
- * Radeon buffer.
+ * TTM.
  */
-struct radeon_object;
+struct radeon_mman {
+	struct ttm_bo_global_ref        bo_global_ref;
+	struct ttm_global_reference	mem_global_ref;
+	struct ttm_bo_device		bdev;
+	bool				mem_global_referenced;
+	bool				initialized;
+};
+
+struct radeon_bo {
+	/* Protected by gem.mutex */
+	struct list_head		list;
+	/* Protected by tbo.reserved */
+	u32				placements[3];
+	struct ttm_placement		placement;
+	struct ttm_buffer_object	tbo;
+	struct ttm_bo_kmap_obj		kmap;
+	unsigned			pin_count;
+	void				*kptr;
+	u32				tiling_flags;
+	u32				pitch;
+	int				surface_reg;
+	/* Constant after initialization */
+	struct radeon_device		*rdev;
+	struct drm_gem_object		*gobj;
+};
 
-struct radeon_object_list {
+struct radeon_bo_list {
 	struct list_head	list;
-	struct radeon_object	*robj;
+	struct radeon_bo	*bo;
 	uint64_t		gpu_offset;
 	unsigned		rdomain;
 	unsigned		wdomain;
-	uint32_t                tiling_flags;
-};
-
-int radeon_object_init(struct radeon_device *rdev);
-void radeon_object_fini(struct radeon_device *rdev);
-int radeon_object_create(struct radeon_device *rdev,
-			 struct drm_gem_object *gobj,
-			 unsigned long size,
-			 bool kernel,
-			 uint32_t domain,
-			 bool interruptible,
-			 struct radeon_object **robj_ptr);
-int radeon_object_kmap(struct radeon_object *robj, void **ptr);
-void radeon_object_kunmap(struct radeon_object *robj);
-void radeon_object_unref(struct radeon_object **robj);
-int radeon_object_pin(struct radeon_object *robj, uint32_t domain,
-		      uint64_t *gpu_addr);
-void radeon_object_unpin(struct radeon_object *robj);
-int radeon_object_wait(struct radeon_object *robj);
-int radeon_object_busy_domain(struct radeon_object *robj, uint32_t *cur_placement);
-int radeon_object_evict_vram(struct radeon_device *rdev);
-int radeon_object_mmap(struct radeon_object *robj, uint64_t *offset);
-void radeon_object_force_delete(struct radeon_device *rdev);
-void radeon_object_list_add_object(struct radeon_object_list *lobj,
-				   struct list_head *head);
-int radeon_object_list_validate(struct list_head *head, void *fence);
-void radeon_object_list_unvalidate(struct list_head *head);
-void radeon_object_list_clean(struct list_head *head);
-int radeon_object_fbdev_mmap(struct radeon_object *robj,
-			     struct vm_area_struct *vma);
-unsigned long radeon_object_size(struct radeon_object *robj);
-void radeon_object_clear_surface_reg(struct radeon_object *robj);
-int radeon_object_check_tiling(struct radeon_object *robj, bool has_moved,
-			       bool force_drop);
-void radeon_object_set_tiling_flags(struct radeon_object *robj,
-				    uint32_t tiling_flags, uint32_t pitch);
-void radeon_object_get_tiling_flags(struct radeon_object *robj, uint32_t *tiling_flags, uint32_t *pitch);
-void radeon_bo_move_notify(struct ttm_buffer_object *bo,
-			   struct ttm_mem_reg *mem);
-void radeon_bo_fault_reserve_notify(struct ttm_buffer_object *bo);
+	u32			tiling_flags;
+};
+
 /*
  * GEM objects.
  */
 struct radeon_gem {
+	struct mutex		mutex;
 	struct list_head	objects;
 };
 
 int radeon_gem_init(struct radeon_device *rdev);
 void radeon_gem_fini(struct radeon_device *rdev);
 int radeon_gem_object_create(struct radeon_device *rdev, int size,
-			     int alignment, int initial_domain,
-			     bool discardable, bool kernel,
-			     bool interruptible,
-			     struct drm_gem_object **obj);
+				int alignment, int initial_domain,
+				bool discardable, bool kernel,
+				struct drm_gem_object **obj);
 int radeon_gem_object_pin(struct drm_gem_object *obj, uint32_t pin_domain,
 			  uint64_t *gpu_addr);
 void radeon_gem_object_unpin(struct drm_gem_object *obj);
@@ -249,7 +287,7 @@ struct radeon_gart_table_ram {
 };
 
 struct radeon_gart_table_vram {
-	struct radeon_object		*robj;
+	struct radeon_bo		*robj;
 	volatile uint32_t		*ptr;
 };
 
@@ -258,6 +296,9 @@ union radeon_gart_table {
 	struct radeon_gart_table_vram	vram;
 };
 
+#define RADEON_GPU_PAGE_SIZE 4096
+#define RADEON_GPU_PAGE_MASK (RADEON_GPU_PAGE_SIZE - 1)
+
 struct radeon_gart {
 	dma_addr_t			table_addr;
 	unsigned			num_gpu_pages;
@@ -291,21 +332,21 @@ struct radeon_mc {
 	/* for some chips with <= 32MB we need to lie
 	 * about vram size near mc fb location */
 	u64			mc_vram_size;
-	u64			gtt_location;
+	u64			visible_vram_size;
 	u64			gtt_size;
 	u64			gtt_start;
 	u64			gtt_end;
-	u64			vram_location;
 	u64			vram_start;
 	u64			vram_end;
 	unsigned		vram_width;
 	u64			real_vram_size;
 	int			vram_mtrr;
 	bool			vram_is_ddr;
+	bool			igp_sideport_enabled;
 };
 
-int radeon_mc_setup(struct radeon_device *rdev);
-
+bool radeon_combios_sideport_present(struct radeon_device *rdev);
+bool radeon_atombios_sideport_present(struct radeon_device *rdev);
 
 /*
  * GPU scratch registers structures, functions & helpers
@@ -328,22 +369,29 @@ struct radeon_irq {
 	bool		sw_int;
 	/* FIXME: use a define max crtc rather than hardcode it */
 	bool		crtc_vblank_int[2];
+	wait_queue_head_t	vblank_queue;
+	/* FIXME: use defines for max hpd/dacs */
+	bool            hpd[6];
+	spinlock_t sw_lock;
+	int sw_refcount;
 };
 
 int radeon_irq_kms_init(struct radeon_device *rdev);
 void radeon_irq_kms_fini(struct radeon_device *rdev);
-
+void radeon_irq_kms_sw_irq_get(struct radeon_device *rdev);
+void radeon_irq_kms_sw_irq_put(struct radeon_device *rdev);
 
 /*
  * CP & ring.
  */
 struct radeon_ib {
 	struct list_head	list;
-	unsigned long		idx;
+	unsigned		idx;
 	uint64_t		gpu_addr;
 	struct radeon_fence	*fence;
-	volatile uint32_t	*ptr;
+	uint32_t		*ptr;
 	uint32_t		length_dw;
+	bool			free;
 };
 
 /*
@@ -352,15 +400,15 @@ struct radeon_ib {
  */
 struct radeon_ib_pool {
 	struct mutex		mutex;
-	struct radeon_object	*robj;
-	struct list_head	scheduled_ibs;
+	struct radeon_bo	*robj;
+	struct list_head	bogus_ib;
 	struct radeon_ib	ibs[RADEON_IB_POOL_SIZE];
 	bool			ready;
-	DECLARE_BITMAP(alloc_bm, RADEON_IB_POOL_SIZE);
+	unsigned		head_id;
 };
 
 struct radeon_cp {
-	struct radeon_object	*ring_obj;
+	struct radeon_bo	*ring_obj;
 	volatile uint32_t	*ring;
 	unsigned		rptr;
 	unsigned		wptr;
@@ -375,8 +423,25 @@ struct radeon_cp {
 	bool			ready;
 };
 
+/*
+ * R6xx+ IH ring
+ */
+struct r600_ih {
+	struct radeon_bo	*ring_obj;
+	volatile uint32_t	*ring;
+	unsigned		rptr;
+	unsigned		wptr;
+	unsigned		wptr_old;
+	unsigned		ring_size;
+	uint64_t		gpu_addr;
+	uint32_t		ptr_mask;
+	spinlock_t              lock;
+	bool                    enabled;
+};
+
 struct r600_blit {
-	struct radeon_object	*shader_obj;
+	struct mutex		mutex;
+	struct radeon_bo	*shader_obj;
 	u64 shader_gpu_addr;
 	u32 vs_offset, ps_offset;
 	u32 state_offset;
@@ -391,6 +456,7 @@ int radeon_ib_schedule(struct radeon_device *rdev, struct radeon_ib *ib);
 int radeon_ib_pool_init(struct radeon_device *rdev);
 void radeon_ib_pool_fini(struct radeon_device *rdev);
 int radeon_ib_test(struct radeon_device *rdev);
+extern void radeon_ib_bogus_add(struct radeon_device *rdev, struct radeon_ib *ib);
 /* Ring access between begin & end cannot sleep */
 void radeon_ring_free_size(struct radeon_device *rdev);
 int radeon_ring_lock(struct radeon_device *rdev, unsigned ndw);
@@ -406,8 +472,8 @@ void radeon_ring_fini(struct radeon_device *rdev);
  */
 struct radeon_cs_reloc {
 	struct drm_gem_object		*gobj;
-	struct radeon_object		*robj;
-	struct radeon_object_list	lobj;
+	struct radeon_bo		*robj;
+	struct radeon_bo_list		lobj;
 	uint32_t			handle;
 	uint32_t			flags;
 };
@@ -415,10 +481,16 @@ struct radeon_cs_reloc {
 struct radeon_cs_chunk {
 	uint32_t		chunk_id;
 	uint32_t		length_dw;
+	int kpage_idx[2];
+	uint32_t                *kpage[2];
 	uint32_t		*kdata;
+	void __user *user_ptr;
+	int last_copied_page;
+	int last_page_index;
 };
 
 struct radeon_cs_parser {
+	struct device		*dev;
 	struct radeon_device	*rdev;
 	struct drm_file		*filp;
 	/* chunks */
@@ -438,8 +510,38 @@ struct radeon_cs_parser {
 	struct radeon_ib	*ib;
 	void			*track;
 	unsigned		family;
+	int parser_error;
 };
 
+extern int radeon_cs_update_pages(struct radeon_cs_parser *p, int pg_idx);
+extern int radeon_cs_finish_pages(struct radeon_cs_parser *p);
+
+
+static inline u32 radeon_get_ib_value(struct radeon_cs_parser *p, int idx)
+{
+	struct radeon_cs_chunk *ibc = &p->chunks[p->chunk_ib_idx];
+	u32 pg_idx, pg_offset;
+	u32 idx_value = 0;
+	int new_page;
+
+	pg_idx = (idx * 4) / PAGE_SIZE;
+	pg_offset = (idx * 4) % PAGE_SIZE;
+
+	if (ibc->kpage_idx[0] == pg_idx)
+		return ibc->kpage[0][pg_offset/4];
+	if (ibc->kpage_idx[1] == pg_idx)
+		return ibc->kpage[1][pg_offset/4];
+
+	new_page = radeon_cs_update_pages(p, pg_idx);
+	if (new_page < 0) {
+		p->parser_error = new_page;
+		return 0;
+	}
+
+	idx_value = ibc->kpage[new_page][pg_offset/4];
+	return idx_value;
+}
+
 struct radeon_cs_packet {
 	unsigned	idx;
 	unsigned	type;
@@ -460,6 +562,7 @@ typedef int (*radeon_packet3_check_t)(struct radeon_cs_parser *p,
  * AGP
  */
 int radeon_agp_init(struct radeon_device *rdev);
+void radeon_agp_resume(struct radeon_device *rdev);
 void radeon_agp_fini(struct radeon_device *rdev);
 
 
@@ -467,7 +570,7 @@ void radeon_agp_fini(struct radeon_device *rdev);
  * Writeback
  */
 struct radeon_wb {
-	struct radeon_object	*wb_obj;
+	struct radeon_bo	*wb_obj;
 	volatile uint32_t	*wb;
 	uint64_t		gpu_addr;
 };
@@ -491,7 +594,99 @@ struct radeon_wb {
  * Equation between gpu/memory clock and available bandwidth is hw dependent
  * (type of memory, bus size, efficiency, ...)
  */
+enum radeon_pm_state {
+	PM_STATE_DISABLED,
+	PM_STATE_MINIMUM,
+	PM_STATE_PAUSED,
+	PM_STATE_ACTIVE
+};
+enum radeon_pm_action {
+	PM_ACTION_NONE,
+	PM_ACTION_MINIMUM,
+	PM_ACTION_DOWNCLOCK,
+	PM_ACTION_UPCLOCK
+};
+
+enum radeon_voltage_type {
+	VOLTAGE_NONE = 0,
+	VOLTAGE_GPIO,
+	VOLTAGE_VDDC,
+	VOLTAGE_SW
+};
+
+enum radeon_pm_state_type {
+	POWER_STATE_TYPE_DEFAULT,
+	POWER_STATE_TYPE_POWERSAVE,
+	POWER_STATE_TYPE_BATTERY,
+	POWER_STATE_TYPE_BALANCED,
+	POWER_STATE_TYPE_PERFORMANCE,
+};
+
+enum radeon_pm_clock_mode_type {
+	POWER_MODE_TYPE_DEFAULT,
+	POWER_MODE_TYPE_LOW,
+	POWER_MODE_TYPE_MID,
+	POWER_MODE_TYPE_HIGH,
+};
+
+struct radeon_voltage {
+	enum radeon_voltage_type type;
+	/* gpio voltage */
+	struct radeon_gpio_rec gpio;
+	u32 delay; /* delay in usec from voltage drop to sclk change */
+	bool active_high; /* voltage drop is active when bit is high */
+	/* VDDC voltage */
+	u8 vddc_id; /* index into vddc voltage table */
+	u8 vddci_id; /* index into vddci voltage table */
+	bool vddci_enabled;
+	/* r6xx+ sw */
+	u32 voltage;
+};
+
+struct radeon_pm_non_clock_info {
+	/* pcie lanes */
+	int pcie_lanes;
+	/* standardized non-clock flags */
+	u32 flags;
+};
+
+struct radeon_pm_clock_info {
+	/* memory clock */
+	u32 mclk;
+	/* engine clock */
+	u32 sclk;
+	/* voltage info */
+	struct radeon_voltage voltage;
+	/* standardized clock flags - not sure we'll need these */
+	u32 flags;
+};
+
+struct radeon_power_state {
+	enum radeon_pm_state_type type;
+	/* XXX: use a define for num clock modes */
+	struct radeon_pm_clock_info clock_info[8];
+	/* number of valid clock modes in this power state */
+	int num_clock_modes;
+	struct radeon_pm_clock_info *default_clock_mode;
+	/* non clock info about this state */
+	struct radeon_pm_non_clock_info non_clock_info;
+	bool voltage_drop_active;
+};
+
+/*
+ * Some modes are overclocked by very low value, accept them
+ */
+#define RADEON_MODE_OVERCLOCK_MARGIN 500 /* 5 MHz */
+
 struct radeon_pm {
+	struct mutex		mutex;
+	struct delayed_work	idle_work;
+	enum radeon_pm_state	state;
+	enum radeon_pm_action	planned_action;
+	unsigned long		action_timeout;
+	bool 			downclocked;
+	int			active_crtcs;
+	int			req_vblank;
 	fixed20_12		max_bandwidth;
 	fixed20_12		igp_sideport_mclk;
 	fixed20_12		igp_system_mclk;
@@ -503,6 +698,15 @@ struct radeon_pm {
 	fixed20_12		core_bandwidth;
 	fixed20_12		sclk;
 	fixed20_12		needed_bandwidth;
+	/* XXX: use a define for num power modes */
+	struct radeon_power_state power_state[8];
+	/* number of valid power states */
+	int                     num_power_states;
+	struct radeon_power_state *current_power_state;
+	struct radeon_pm_clock_info *current_clock_mode;
+	struct radeon_power_state *requested_power_state;
+	struct radeon_pm_clock_info *requested_clock_mode;
+	struct radeon_power_state *default_power_state;
 };
 
 
@@ -537,18 +741,8 @@ struct radeon_asic {
 	void (*fini)(struct radeon_device *rdev);
 	int (*resume)(struct radeon_device *rdev);
 	int (*suspend)(struct radeon_device *rdev);
-	void (*errata)(struct radeon_device *rdev);
-	void (*vram_info)(struct radeon_device *rdev);
 	void (*vga_set_state)(struct radeon_device *rdev, bool state);
 	int (*gpu_reset)(struct radeon_device *rdev);
-	int (*mc_init)(struct radeon_device *rdev);
-	void (*mc_fini)(struct radeon_device *rdev);
-	int (*wb_init)(struct radeon_device *rdev);
-	void (*wb_fini)(struct radeon_device *rdev);
-	int (*gart_init)(struct radeon_device *rdev);
-	void (*gart_fini)(struct radeon_device *rdev);
-	int (*gart_enable)(struct radeon_device *rdev);
-	void (*gart_disable)(struct radeon_device *rdev);
 	void (*gart_tlb_flush)(struct radeon_device *rdev);
 	int (*gart_set_page)(struct radeon_device *rdev, int i, uint64_t addr);
 	int (*cp_init)(struct radeon_device *rdev, unsigned ring_size);
@@ -558,7 +752,6 @@ struct radeon_asic {
 	void (*ring_start)(struct radeon_device *rdev);
 	int (*ring_test)(struct radeon_device *rdev);
 	void (*ring_ib_execute)(struct radeon_device *rdev, struct radeon_ib *ib);
-	int (*ib_test)(struct radeon_device *rdev);
 	int (*irq_set)(struct radeon_device *rdev);
 	int (*irq_process)(struct radeon_device *rdev);
 	u32 (*get_vblank_counter)(struct radeon_device *rdev, int crtc);
@@ -579,8 +772,11 @@ struct radeon_asic {
 		    uint64_t dst_offset,
 		    unsigned num_pages,
 		    struct radeon_fence *fence);
+	uint32_t (*get_engine_clock)(struct radeon_device *rdev);
 	void (*set_engine_clock)(struct radeon_device *rdev, uint32_t eng_clock);
+	uint32_t (*get_memory_clock)(struct radeon_device *rdev);
 	void (*set_memory_clock)(struct radeon_device *rdev, uint32_t mem_clock);
+	int (*get_pcie_lanes)(struct radeon_device *rdev);
 	void (*set_pcie_lanes)(struct radeon_device *rdev, int lanes);
 	void (*set_clock_gating)(struct radeon_device *rdev, int enable);
 	int (*set_surface_reg)(struct radeon_device *rdev, int reg,
@@ -588,6 +784,17 @@ struct radeon_asic {
 			       uint32_t offset, uint32_t obj_size);
 	int (*clear_surface_reg)(struct radeon_device *rdev, int reg);
 	void (*bandwidth_update)(struct radeon_device *rdev);
+	void (*hpd_init)(struct radeon_device *rdev);
+	void (*hpd_fini)(struct radeon_device *rdev);
+	bool (*hpd_sense)(struct radeon_device *rdev, enum radeon_hpd_id hpd);
+	void (*hpd_set_polarity)(struct radeon_device *rdev, enum radeon_hpd_id hpd);
+	/* ioctl hw specific callback. Some hw might want to perform special
+	 * operation on specific ioctl. For instance on wait idle some hw
+	 * might want to perform and HDP flush through MMIO as it seems that
+	 * some R6XX/R7XX hw doesn't take HDP flush into account if programmed
+	 * through ring.
+	 */
+	void (*ioctl_wait_idle)(struct radeon_device *rdev, struct radeon_bo *bo);
 };
 
 /*
@@ -596,11 +803,14 @@ struct radeon_asic {
 struct r100_asic {
 	const unsigned	*reg_safe_bm;
 	unsigned	reg_safe_bm_size;
+	u32		hdp_cntl;
 };
 
 struct r300_asic {
 	const unsigned	*reg_safe_bm;
 	unsigned	reg_safe_bm_size;
+	u32		resync_scratch;
+	u32		hdp_cntl;
 };
 
 struct r600_asic {
@@ -617,6 +827,9 @@ struct r600_asic {
 	unsigned sx_max_export_pos_size;
 	unsigned sx_max_export_smx_size;
 	unsigned sq_num_cf_insts;
+	unsigned tiling_nbanks;
+	unsigned tiling_npipes;
+	unsigned tiling_group_size;
 };
 
 struct rv770_asic {
@@ -637,6 +850,9 @@ struct rv770_asic {
 	unsigned sc_prim_fifo_size;
 	unsigned sc_hiz_tile_fifo_size;
 	unsigned sc_earlyz_tile_fifo_fize;
+	unsigned tiling_nbanks;
+	unsigned tiling_npipes;
+	unsigned tiling_group_size;
 };
 
 union radeon_asic_config {
@@ -700,9 +916,9 @@ struct radeon_device {
 	uint8_t				*bios;
 	bool				is_atom_bios;
 	uint16_t			bios_header_start;
-	struct radeon_object		*stollen_vga_memory;
+	struct radeon_bo		*stollen_vga_memory;
 	struct fb_info			*fbdev_info;
-	struct radeon_object		*fbdev_robj;
+	struct radeon_bo		*fbdev_rbo;
 	struct radeon_framebuffer	*fbdev_rfb;
 	/* Register mmio */
 	resource_size_t			rmmio_base;
@@ -736,12 +952,28 @@ struct radeon_device {
 	bool				shutdown;
 	bool				suspend;
 	bool				need_dma32;
-	bool				new_init_path;
 	bool				accel_working;
 	struct radeon_surface_reg surface_regs[RADEON_GEM_MAX_SURFACES];
 	const struct firmware *me_fw;	/* all family ME firmware */
 	const struct firmware *pfp_fw;	/* r6/700 PFP firmware */
+	const struct firmware *rlc_fw;	/* r6/700 RLC firmware */
 	struct r600_blit r600_blit;
+	int msi_enabled; /* msi enabled */
+	struct r600_ih ih; /* r6/700 interrupt ring */
+	struct workqueue_struct *wq;
+	struct work_struct hotplug_work;
+	int num_crtc; /* number of crtcs */
+	struct mutex dc_hw_i2c_mutex; /* display controller hw i2c mutex */
+
+	/* audio stuff */
+	struct timer_list	audio_timer;
+	int			audio_channels;
+	int			audio_rate;
+	int			audio_bits_per_sample;
+	uint8_t			audio_status_bits;
+	uint8_t			audio_category_code;
+
+	bool powered_down;
 };
 
 int radeon_device_init(struct radeon_device *rdev,
@@ -760,7 +992,7 @@ void r600_kms_blit_copy(struct radeon_device *rdev,
 
 static inline uint32_t r100_mm_rreg(struct radeon_device *rdev, uint32_t reg)
 {
-	if (reg < 0x10000)
+	if (reg < rdev->rmmio_size)
 		return readl(((void __iomem *)rdev->rmmio) + reg);
 	else {
 		writel(reg, ((void __iomem *)rdev->rmmio) + RADEON_MM_INDEX);
@@ -770,7 +1002,7 @@ static inline uint32_t r100_mm_rreg(struct radeon_device *rdev, uint32_t reg)
 
 static inline void r100_mm_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v)
 {
-	if (reg < 0x10000)
+	if (reg < rdev->rmmio_size)
 		writel(v, ((void __iomem *)rdev->rmmio) + reg);
 	else {
 		writel(reg, ((void __iomem *)rdev->rmmio) + RADEON_MM_INDEX);
@@ -778,6 +1010,10 @@ static inline void r100_mm_wreg(struct radeon_device *rdev, uint32_t reg, uint32
 	}
 }
 
+/*
+ * Cast helper
+ */
+#define to_radeon_fence(p) ((struct radeon_fence *)(p))
 
 /*
  * Registers read & write functions.
@@ -795,6 +1031,8 @@ static inline void r100_mm_wreg(struct radeon_device *rdev, uint32_t reg, uint32
 #define WREG32_MC(reg, v) rdev->mc_wreg(rdev, (reg), (v))
 #define RREG32_PCIE(reg) rv370_pcie_rreg(rdev, (reg))
 #define WREG32_PCIE(reg, v) rv370_pcie_wreg(rdev, (reg), (v))
+#define RREG32_PCIE_P(reg) rdev->pciep_rreg(rdev, (reg))
+#define WREG32_PCIE_P(reg, v) rdev->pciep_wreg(rdev, (reg), (v))
 #define WREG32_P(reg, val, mask)				\
 	do {							\
 		uint32_t tmp_ = RREG32(reg);			\
@@ -856,7 +1094,7 @@ void r100_pll_errata_after_index(struct radeon_device *rdev);
 #define ASIC_IS_AVIVO(rdev) ((rdev->family >= CHIP_RS600))
 #define ASIC_IS_DCE3(rdev) ((rdev->family >= CHIP_RV620))
 #define ASIC_IS_DCE32(rdev) ((rdev->family >= CHIP_RV730))
-
+#define ASIC_IS_DCE4(rdev) ((rdev->family >= CHIP_CEDAR))
 
 /*
  * BIOS helpers.
@@ -896,28 +1134,14 @@ static inline void radeon_ring_write(struct radeon_device *rdev, uint32_t v)
 #define radeon_resume(rdev) (rdev)->asic->resume((rdev))
 #define radeon_suspend(rdev) (rdev)->asic->suspend((rdev))
 #define radeon_cs_parse(p) rdev->asic->cs_parse((p))
-#define radeon_errata(rdev) (rdev)->asic->errata((rdev))
-#define radeon_vram_info(rdev) (rdev)->asic->vram_info((rdev))
 #define radeon_vga_set_state(rdev, state) (rdev)->asic->vga_set_state((rdev), (state))
 #define radeon_gpu_reset(rdev) (rdev)->asic->gpu_reset((rdev))
-#define radeon_mc_init(rdev) (rdev)->asic->mc_init((rdev))
-#define radeon_mc_fini(rdev) (rdev)->asic->mc_fini((rdev))
-#define radeon_wb_init(rdev) (rdev)->asic->wb_init((rdev))
-#define radeon_wb_fini(rdev) (rdev)->asic->wb_fini((rdev))
-#define radeon_gpu_gart_init(rdev) (rdev)->asic->gart_init((rdev))
-#define radeon_gpu_gart_fini(rdev) (rdev)->asic->gart_fini((rdev))
-#define radeon_gart_enable(rdev) (rdev)->asic->gart_enable((rdev))
-#define radeon_gart_disable(rdev) (rdev)->asic->gart_disable((rdev))
 #define radeon_gart_tlb_flush(rdev) (rdev)->asic->gart_tlb_flush((rdev))
 #define radeon_gart_set_page(rdev, i, p) (rdev)->asic->gart_set_page((rdev), (i), (p))
-#define radeon_cp_init(rdev,rsize) (rdev)->asic->cp_init((rdev), (rsize))
-#define radeon_cp_fini(rdev) (rdev)->asic->cp_fini((rdev))
-#define radeon_cp_disable(rdev) (rdev)->asic->cp_disable((rdev))
 #define radeon_cp_commit(rdev) (rdev)->asic->cp_commit((rdev))
 #define radeon_ring_start(rdev) (rdev)->asic->ring_start((rdev))
 #define radeon_ring_test(rdev) (rdev)->asic->ring_test((rdev))
 #define radeon_ring_ib_execute(rdev, ib) (rdev)->asic->ring_ib_execute((rdev), (ib))
-#define radeon_ib_test(rdev) (rdev)->asic->ib_test((rdev))
 #define radeon_irq_set(rdev) (rdev)->asic->irq_set((rdev))
 #define radeon_irq_process(rdev) (rdev)->asic->irq_process((rdev))
 #define radeon_get_vblank_counter(rdev, crtc) (rdev)->asic->get_vblank_counter((rdev), (crtc))
@@ -925,24 +1149,43 @@ static inline void radeon_ring_write(struct radeon_device *rdev, uint32_t v)
 #define radeon_copy_blit(rdev, s, d, np, f) (rdev)->asic->copy_blit((rdev), (s), (d), (np), (f))
 #define radeon_copy_dma(rdev, s, d, np, f) (rdev)->asic->copy_dma((rdev), (s), (d), (np), (f))
 #define radeon_copy(rdev, s, d, np, f) (rdev)->asic->copy((rdev), (s), (d), (np), (f))
+#define radeon_get_engine_clock(rdev) (rdev)->asic->get_engine_clock((rdev))
 #define radeon_set_engine_clock(rdev, e) (rdev)->asic->set_engine_clock((rdev), (e))
-#define radeon_set_memory_clock(rdev, e) (rdev)->asic->set_engine_clock((rdev), (e))
+#define radeon_get_memory_clock(rdev) (rdev)->asic->get_memory_clock((rdev))
+#define radeon_set_memory_clock(rdev, e) (rdev)->asic->set_memory_clock((rdev), (e))
+#define radeon_get_pcie_lanes(rdev) (rdev)->asic->get_pcie_lanes((rdev))
 #define radeon_set_pcie_lanes(rdev, l) (rdev)->asic->set_pcie_lanes((rdev), (l))
 #define radeon_set_clock_gating(rdev, e) (rdev)->asic->set_clock_gating((rdev), (e))
 #define radeon_set_surface_reg(rdev, r, f, p, o, s) ((rdev)->asic->set_surface_reg((rdev), (r), (f), (p), (o), (s)))
 #define radeon_clear_surface_reg(rdev, r) ((rdev)->asic->clear_surface_reg((rdev), (r)))
 #define radeon_bandwidth_update(rdev) (rdev)->asic->bandwidth_update((rdev))
+#define radeon_hpd_init(rdev) (rdev)->asic->hpd_init((rdev))
+#define radeon_hpd_fini(rdev) (rdev)->asic->hpd_fini((rdev))
+#define radeon_hpd_sense(rdev, hpd) (rdev)->asic->hpd_sense((rdev), (hpd))
+#define radeon_hpd_set_polarity(rdev, hpd) (rdev)->asic->hpd_set_polarity((rdev), (hpd))
 
 /* Common functions */
+/* AGP */
+extern void radeon_agp_disable(struct radeon_device *rdev);
 extern int radeon_gart_table_vram_pin(struct radeon_device *rdev);
+extern void radeon_gart_restore(struct radeon_device *rdev);
 extern int radeon_modeset_init(struct radeon_device *rdev);
 extern void radeon_modeset_fini(struct radeon_device *rdev);
 extern bool radeon_card_posted(struct radeon_device *rdev);
+extern bool radeon_boot_test_post_card(struct radeon_device *rdev);
 extern int radeon_clocks_init(struct radeon_device *rdev);
 extern void radeon_clocks_fini(struct radeon_device *rdev);
 extern void radeon_scratch_init(struct radeon_device *rdev);
 extern void radeon_surface_init(struct radeon_device *rdev);
 extern int radeon_cs_parser_init(struct radeon_cs_parser *p, void *data);
+extern void radeon_legacy_set_clock_gating(struct radeon_device *rdev, int enable);
+extern void radeon_atom_set_clock_gating(struct radeon_device *rdev, int enable);
+extern void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain);
+extern bool radeon_ttm_bo_is_radeon_bo(struct ttm_buffer_object *bo);
+extern void radeon_vram_location(struct radeon_device *rdev, struct radeon_mc *mc, u64 base);
+extern void radeon_gtt_location(struct radeon_device *rdev, struct radeon_mc *mc);
+extern int radeon_resume_kms(struct drm_device *dev);
+extern int radeon_suspend_kms(struct drm_device *dev, pm_message_t state);
 
 /* r100,rv100,rs100,rv200,rs200,r200,rv250,rs300,rv280 */
 struct r100_mc_save {
@@ -974,11 +1217,32 @@ extern void r100_vram_init_sizes(struct radeon_device *rdev);
 extern void r100_wb_disable(struct radeon_device *rdev);
 extern void r100_wb_fini(struct radeon_device *rdev);
 extern int r100_wb_init(struct radeon_device *rdev);
+extern void r100_hdp_reset(struct radeon_device *rdev);
+extern int r100_rb2d_reset(struct radeon_device *rdev);
+extern int r100_cp_reset(struct radeon_device *rdev);
+extern void r100_vga_render_disable(struct radeon_device *rdev);
+extern int r100_cs_track_check_pkt3_indx_buffer(struct radeon_cs_parser *p,
+						struct radeon_cs_packet *pkt,
+						struct radeon_bo *robj);
+extern int r100_cs_parse_packet0(struct radeon_cs_parser *p,
+				struct radeon_cs_packet *pkt,
+				const unsigned *auth, unsigned n,
+				radeon_packet0_check_t check);
+extern int r100_cs_packet_parse(struct radeon_cs_parser *p,
+				struct radeon_cs_packet *pkt,
+				unsigned idx);
+extern void r100_enable_bm(struct radeon_device *rdev);
+extern void r100_set_common_regs(struct radeon_device *rdev);
+
+/* rv200,rv250,rv280 */
+extern void r200_set_safe_registers(struct radeon_device *rdev);
 
 /* r300,r350,rv350,rv370,rv380 */
 extern void r300_set_reg_safe(struct radeon_device *rdev);
 extern void r300_mc_program(struct radeon_device *rdev);
-extern void r300_vram_info(struct radeon_device *rdev);
+extern void r300_mc_init(struct radeon_device *rdev);
+extern void r300_clock_startup(struct radeon_device *rdev);
+extern int r300_mc_wait_for_idle(struct radeon_device *rdev);
 extern int rv370_pcie_gart_init(struct radeon_device *rdev);
 extern void rv370_pcie_gart_fini(struct radeon_device *rdev);
 extern int rv370_pcie_gart_enable(struct radeon_device *rdev);
@@ -988,9 +1252,37 @@ extern void rv370_pcie_gart_disable(struct radeon_device *rdev);
 extern u32 r420_mc_rreg(struct radeon_device *rdev, u32 reg);
 extern void r420_mc_wreg(struct radeon_device *rdev, u32 reg, u32 v);
 extern int r420_debugfs_pipes_info_init(struct radeon_device *rdev);
+extern void r420_pipes_init(struct radeon_device *rdev);
 
 /* rv515 */
+struct rv515_mc_save {
+	u32 d1vga_control;
+	u32 d2vga_control;
+	u32 vga_render_control;
+	u32 vga_hdp_control;
+	u32 d1crtc_control;
+	u32 d2crtc_control;
+};
 extern void rv515_bandwidth_avivo_update(struct radeon_device *rdev);
+extern void rv515_vga_render_disable(struct radeon_device *rdev);
+extern void rv515_set_safe_registers(struct radeon_device *rdev);
+extern void rv515_mc_stop(struct radeon_device *rdev, struct rv515_mc_save *save);
+extern void rv515_mc_resume(struct radeon_device *rdev, struct rv515_mc_save *save);
+extern void rv515_clock_startup(struct radeon_device *rdev);
+extern void rv515_debugfs(struct radeon_device *rdev);
+extern int rv515_suspend(struct radeon_device *rdev);
+
+/* rs400 */
+extern int rs400_gart_init(struct radeon_device *rdev);
+extern int rs400_gart_enable(struct radeon_device *rdev);
+extern void rs400_gart_adjust_size(struct radeon_device *rdev);
+extern void rs400_gart_disable(struct radeon_device *rdev);
+extern void rs400_gart_fini(struct radeon_device *rdev);
+
+/* rs600 */
+extern void rs600_set_safe_registers(struct radeon_device *rdev);
+extern int rs600_irq_set(struct radeon_device *rdev);
+extern void rs600_irq_disable(struct radeon_device *rdev);
 
 /* rs690, rs740 */
 extern void rs690_line_buffer_adjust(struct radeon_device *rdev,
@@ -998,23 +1290,56 @@ extern void rs690_line_buffer_adjust(struct radeon_device *rdev,
 					struct drm_display_mode *mode2);
 
 /* r600, rv610, rv630, rv620, rv635, rv670, rs780, rs880 */
+extern void r600_vram_gtt_location(struct radeon_device *rdev, struct radeon_mc *mc);
 extern bool r600_card_posted(struct radeon_device *rdev);
 extern void r600_cp_stop(struct radeon_device *rdev);
 extern void r600_ring_init(struct radeon_device *rdev, unsigned ring_size);
 extern int r600_cp_resume(struct radeon_device *rdev);
+extern void r600_cp_fini(struct radeon_device *rdev);
 extern int r600_count_pipe_bits(uint32_t val);
-extern int r600_gart_clear_page(struct radeon_device *rdev, int i);
 extern int r600_mc_wait_for_idle(struct radeon_device *rdev);
 extern int r600_pcie_gart_init(struct radeon_device *rdev);
 extern void r600_pcie_gart_tlb_flush(struct radeon_device *rdev);
 extern int r600_ib_test(struct radeon_device *rdev);
 extern int r600_ring_test(struct radeon_device *rdev);
-extern int r600_wb_init(struct radeon_device *rdev);
 extern void r600_wb_fini(struct radeon_device *rdev);
+extern int r600_wb_enable(struct radeon_device *rdev);
+extern void r600_wb_disable(struct radeon_device *rdev);
 extern void r600_scratch_init(struct radeon_device *rdev);
 extern int r600_blit_init(struct radeon_device *rdev);
 extern void r600_blit_fini(struct radeon_device *rdev);
-extern int r600_cp_init_microcode(struct radeon_device *rdev);
+extern int r600_init_microcode(struct radeon_device *rdev);
 extern int r600_gpu_reset(struct radeon_device *rdev);
+/* r600 irq */
+extern int r600_irq_init(struct radeon_device *rdev);
+extern void r600_irq_fini(struct radeon_device *rdev);
+extern void r600_ih_ring_init(struct radeon_device *rdev, unsigned ring_size);
+extern int r600_irq_set(struct radeon_device *rdev);
+extern void r600_irq_suspend(struct radeon_device *rdev);
+/* r600 audio */
+extern int r600_audio_init(struct radeon_device *rdev);
+extern int r600_audio_tmds_index(struct drm_encoder *encoder);
+extern void r600_audio_set_clock(struct drm_encoder *encoder, int clock);
+extern void r600_audio_fini(struct radeon_device *rdev);
+extern void r600_hdmi_init(struct drm_encoder *encoder);
+extern void r600_hdmi_enable(struct drm_encoder *encoder, int enable);
+extern void r600_hdmi_setmode(struct drm_encoder *encoder, struct drm_display_mode *mode);
+extern int r600_hdmi_buffer_status_changed(struct drm_encoder *encoder);
+extern void r600_hdmi_update_audio_settings(struct drm_encoder *encoder,
+					    int channels,
+					    int rate,
+					    int bps,
+					    uint8_t status_bits,
+					    uint8_t category_code);
+
+/* evergreen */
+struct evergreen_mc_save {
+	u32 vga_control[6];
+	u32 vga_render_control;
+	u32 vga_hdp_control;
+	u32 crtc_control[6];
+};
+
+#include "radeon_object.h"
 
 #endif
diff --git a/drivers/gpu/drm/radeon/radeon_agp.c b/drivers/gpu/drm/radeon/radeon_agp.c
index 23ea9955ac59..c4457791dff1 100644
--- a/drivers/gpu/drm/radeon/radeon_agp.c
+++ b/drivers/gpu/drm/radeon/radeon_agp.c
@@ -144,9 +144,19 @@ int radeon_agp_init(struct radeon_device *rdev)
 
 	ret = drm_agp_info(rdev->ddev, &info);
 	if (ret) {
+		drm_agp_release(rdev->ddev);
 		DRM_ERROR("Unable to get AGP info: %d\n", ret);
 		return ret;
 	}
+
+	if (rdev->ddev->agp->agp_info.aper_size < 32) {
+		drm_agp_release(rdev->ddev);
+		dev_warn(rdev->dev, "AGP aperture too small (%zuM) "
+			"need at least 32M, disabling AGP\n",
+			rdev->ddev->agp->agp_info.aper_size);
+		return -EINVAL;
+	}
+
 	mode.mode = info.mode;
 	agp_status = (RREG32(RADEON_AGP_STATUS) | RADEON_AGPv3_MODE) & mode.mode;
 	is_v3 = !!(agp_status & RADEON_AGPv3_MODE);
@@ -221,11 +231,16 @@ int radeon_agp_init(struct radeon_device *rdev)
 	ret = drm_agp_enable(rdev->ddev, mode);
 	if (ret) {
 		DRM_ERROR("Unable to enable AGP (mode = 0x%lx)\n", mode.mode);
+		drm_agp_release(rdev->ddev);
 		return ret;
 	}
 
 	rdev->mc.agp_base = rdev->ddev->agp->agp_info.aper_base;
 	rdev->mc.gtt_size = rdev->ddev->agp->agp_info.aper_size << 20;
+	rdev->mc.gtt_start = rdev->mc.agp_base;
+	rdev->mc.gtt_end = rdev->mc.gtt_start + rdev->mc.gtt_size - 1;
+	dev_info(rdev->dev, "GTT: %lluM 0x%08llX - 0x%08llX\n",
+		rdev->mc.gtt_size >> 20, rdev->mc.gtt_start, rdev->mc.gtt_end);
 
 	/* workaround some hw issues */
 	if (rdev->family < CHIP_R200) {
@@ -237,13 +252,23 @@ int radeon_agp_init(struct radeon_device *rdev)
 #endif
 }
 
-void radeon_agp_fini(struct radeon_device *rdev)
+void radeon_agp_resume(struct radeon_device *rdev)
 {
 #if __OS_HAS_AGP
+	int r;
 	if (rdev->flags & RADEON_IS_AGP) {
-		if (rdev->ddev->agp && rdev->ddev->agp->acquired) {
-			drm_agp_release(rdev->ddev);
-		}
+		r = radeon_agp_init(rdev);
+		if (r)
+			dev_warn(rdev->dev, "radeon AGP reinit failed\n");
+	}
+#endif
+}
+
+void radeon_agp_fini(struct radeon_device *rdev)
+{
+#if __OS_HAS_AGP
+	if (rdev->ddev->agp && rdev->ddev->agp->acquired) {
+		drm_agp_release(rdev->ddev);
 	}
 #endif
 }
diff --git a/drivers/gpu/drm/radeon/radeon_asic.h b/drivers/gpu/drm/radeon/radeon_asic.h
index 8968f78fa1e3..d3a157b2bcb7 100644
--- a/drivers/gpu/drm/radeon/radeon_asic.h
+++ b/drivers/gpu/drm/radeon/radeon_asic.h
@@ -31,38 +31,31 @@
 /*
  * common functions
  */
+uint32_t radeon_legacy_get_engine_clock(struct radeon_device *rdev);
 void radeon_legacy_set_engine_clock(struct radeon_device *rdev, uint32_t eng_clock);
+uint32_t radeon_legacy_get_memory_clock(struct radeon_device *rdev);
 void radeon_legacy_set_clock_gating(struct radeon_device *rdev, int enable);
 
+uint32_t radeon_atom_get_engine_clock(struct radeon_device *rdev);
 void radeon_atom_set_engine_clock(struct radeon_device *rdev, uint32_t eng_clock);
+uint32_t radeon_atom_get_memory_clock(struct radeon_device *rdev);
 void radeon_atom_set_memory_clock(struct radeon_device *rdev, uint32_t mem_clock);
 void radeon_atom_set_clock_gating(struct radeon_device *rdev, int enable);
 
 /*
- * r100,rv100,rs100,rv200,rs200,r200,rv250,rs300,rv280
+ * r100,rv100,rs100,rv200,rs200
  */
-int r100_init(struct radeon_device *rdev);
-int r200_init(struct radeon_device *rdev);
+extern int r100_init(struct radeon_device *rdev);
+extern void r100_fini(struct radeon_device *rdev);
+extern int r100_suspend(struct radeon_device *rdev);
+extern int r100_resume(struct radeon_device *rdev);
 uint32_t r100_mm_rreg(struct radeon_device *rdev, uint32_t reg);
 void r100_mm_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v);
-void r100_errata(struct radeon_device *rdev);
-void r100_vram_info(struct radeon_device *rdev);
 void r100_vga_set_state(struct radeon_device *rdev, bool state);
 int r100_gpu_reset(struct radeon_device *rdev);
-int r100_mc_init(struct radeon_device *rdev);
-void r100_mc_fini(struct radeon_device *rdev);
 u32 r100_get_vblank_counter(struct radeon_device *rdev, int crtc);
-int r100_wb_init(struct radeon_device *rdev);
-void r100_wb_fini(struct radeon_device *rdev);
-int r100_pci_gart_init(struct radeon_device *rdev);
-void r100_pci_gart_fini(struct radeon_device *rdev);
-int r100_pci_gart_enable(struct radeon_device *rdev);
-void r100_pci_gart_disable(struct radeon_device *rdev);
 void r100_pci_gart_tlb_flush(struct radeon_device *rdev);
 int r100_pci_gart_set_page(struct radeon_device *rdev, int i, uint64_t addr);
-int r100_cp_init(struct radeon_device *rdev, unsigned ring_size);
-void r100_cp_fini(struct radeon_device *rdev);
-void r100_cp_disable(struct radeon_device *rdev);
 void r100_cp_commit(struct radeon_device *rdev);
 void r100_ring_start(struct radeon_device *rdev);
 int r100_irq_set(struct radeon_device *rdev);
@@ -83,33 +76,26 @@ int r100_set_surface_reg(struct radeon_device *rdev, int reg,
 int r100_clear_surface_reg(struct radeon_device *rdev, int reg);
 void r100_bandwidth_update(struct radeon_device *rdev);
 void r100_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib);
-int r100_ib_test(struct radeon_device *rdev);
 int r100_ring_test(struct radeon_device *rdev);
+void r100_hpd_init(struct radeon_device *rdev);
+void r100_hpd_fini(struct radeon_device *rdev);
+bool r100_hpd_sense(struct radeon_device *rdev, enum radeon_hpd_id hpd);
+void r100_hpd_set_polarity(struct radeon_device *rdev,
+			   enum radeon_hpd_id hpd);
 
 static struct radeon_asic r100_asic = {
 	.init = &r100_init,
-	.errata = &r100_errata,
-	.vram_info = &r100_vram_info,
+	.fini = &r100_fini,
+	.suspend = &r100_suspend,
+	.resume = &r100_resume,
 	.vga_set_state = &r100_vga_set_state,
 	.gpu_reset = &r100_gpu_reset,
-	.mc_init = &r100_mc_init,
-	.mc_fini = &r100_mc_fini,
-	.wb_init = &r100_wb_init,
-	.wb_fini = &r100_wb_fini,
-	.gart_init = &r100_pci_gart_init,
-	.gart_fini = &r100_pci_gart_fini,
-	.gart_enable = &r100_pci_gart_enable,
-	.gart_disable = &r100_pci_gart_disable,
 	.gart_tlb_flush = &r100_pci_gart_tlb_flush,
 	.gart_set_page = &r100_pci_gart_set_page,
-	.cp_init = &r100_cp_init,
-	.cp_fini = &r100_cp_fini,
-	.cp_disable = &r100_cp_disable,
 	.cp_commit = &r100_cp_commit,
 	.ring_start = &r100_ring_start,
 	.ring_test = &r100_ring_test,
 	.ring_ib_execute = &r100_ring_ib_execute,
-	.ib_test = &r100_ib_test,
 	.irq_set = &r100_irq_set,
 	.irq_process = &r100_irq_process,
 	.get_vblank_counter = &r100_get_vblank_counter,
@@ -118,83 +104,162 @@ static struct radeon_asic r100_asic = {
 	.copy_blit = &r100_copy_blit,
 	.copy_dma = NULL,
 	.copy = &r100_copy_blit,
+	.get_engine_clock = &radeon_legacy_get_engine_clock,
 	.set_engine_clock = &radeon_legacy_set_engine_clock,
+	.get_memory_clock = &radeon_legacy_get_memory_clock,
 	.set_memory_clock = NULL,
+	.get_pcie_lanes = NULL,
 	.set_pcie_lanes = NULL,
 	.set_clock_gating = &radeon_legacy_set_clock_gating,
 	.set_surface_reg = r100_set_surface_reg,
 	.clear_surface_reg = r100_clear_surface_reg,
 	.bandwidth_update = &r100_bandwidth_update,
+	.hpd_init = &r100_hpd_init,
+	.hpd_fini = &r100_hpd_fini,
+	.hpd_sense = &r100_hpd_sense,
+	.hpd_set_polarity = &r100_hpd_set_polarity,
+	.ioctl_wait_idle = NULL,
+};
+
+/*
+ * r200,rv250,rs300,rv280
+ */
+extern int r200_copy_dma(struct radeon_device *rdev,
+			uint64_t src_offset,
+			uint64_t dst_offset,
+			unsigned num_pages,
+			struct radeon_fence *fence);
+static struct radeon_asic r200_asic = {
+	.init = &r100_init,
+	.fini = &r100_fini,
+	.suspend = &r100_suspend,
+	.resume = &r100_resume,
+	.vga_set_state = &r100_vga_set_state,
+	.gpu_reset = &r100_gpu_reset,
+	.gart_tlb_flush = &r100_pci_gart_tlb_flush,
+	.gart_set_page = &r100_pci_gart_set_page,
+	.cp_commit = &r100_cp_commit,
+	.ring_start = &r100_ring_start,
+	.ring_test = &r100_ring_test,
+	.ring_ib_execute = &r100_ring_ib_execute,
+	.irq_set = &r100_irq_set,
+	.irq_process = &r100_irq_process,
+	.get_vblank_counter = &r100_get_vblank_counter,
+	.fence_ring_emit = &r100_fence_ring_emit,
+	.cs_parse = &r100_cs_parse,
+	.copy_blit = &r100_copy_blit,
+	.copy_dma = &r200_copy_dma,
+	.copy = &r100_copy_blit,
+	.get_engine_clock = &radeon_legacy_get_engine_clock,
+	.set_engine_clock = &radeon_legacy_set_engine_clock,
+	.get_memory_clock = &radeon_legacy_get_memory_clock,
+	.set_memory_clock = NULL,
+	.set_pcie_lanes = NULL,
+	.set_clock_gating = &radeon_legacy_set_clock_gating,
+	.set_surface_reg = r100_set_surface_reg,
+	.clear_surface_reg = r100_clear_surface_reg,
+	.bandwidth_update = &r100_bandwidth_update,
+	.hpd_init = &r100_hpd_init,
+	.hpd_fini = &r100_hpd_fini,
+	.hpd_sense = &r100_hpd_sense,
+	.hpd_set_polarity = &r100_hpd_set_polarity,
+	.ioctl_wait_idle = NULL,
 };
 
 
 /*
  * r300,r350,rv350,rv380
  */
-int r300_init(struct radeon_device *rdev);
-void r300_errata(struct radeon_device *rdev);
-void r300_vram_info(struct radeon_device *rdev);
-int r300_gpu_reset(struct radeon_device *rdev);
-int r300_mc_init(struct radeon_device *rdev);
-void r300_mc_fini(struct radeon_device *rdev);
-void r300_ring_start(struct radeon_device *rdev);
-void r300_fence_ring_emit(struct radeon_device *rdev,
-			  struct radeon_fence *fence);
-int r300_cs_parse(struct radeon_cs_parser *p);
-int rv370_pcie_gart_init(struct radeon_device *rdev);
-void rv370_pcie_gart_fini(struct radeon_device *rdev);
-int rv370_pcie_gart_enable(struct radeon_device *rdev);
-void rv370_pcie_gart_disable(struct radeon_device *rdev);
-void rv370_pcie_gart_tlb_flush(struct radeon_device *rdev);
-int rv370_pcie_gart_set_page(struct radeon_device *rdev, int i, uint64_t addr);
-uint32_t rv370_pcie_rreg(struct radeon_device *rdev, uint32_t reg);
-void rv370_pcie_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v);
-void rv370_set_pcie_lanes(struct radeon_device *rdev, int lanes);
-int r300_copy_dma(struct radeon_device *rdev,
-		  uint64_t src_offset,
-		  uint64_t dst_offset,
-		  unsigned num_pages,
-		  struct radeon_fence *fence);
+extern int r300_init(struct radeon_device *rdev);
+extern void r300_fini(struct radeon_device *rdev);
+extern int r300_suspend(struct radeon_device *rdev);
+extern int r300_resume(struct radeon_device *rdev);
+extern int r300_gpu_reset(struct radeon_device *rdev);
+extern void r300_ring_start(struct radeon_device *rdev);
+extern void r300_fence_ring_emit(struct radeon_device *rdev,
+				struct radeon_fence *fence);
+extern int r300_cs_parse(struct radeon_cs_parser *p);
+extern void rv370_pcie_gart_tlb_flush(struct radeon_device *rdev);
+extern int rv370_pcie_gart_set_page(struct radeon_device *rdev, int i, uint64_t addr);
+extern uint32_t rv370_pcie_rreg(struct radeon_device *rdev, uint32_t reg);
+extern void rv370_pcie_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v);
+extern void rv370_set_pcie_lanes(struct radeon_device *rdev, int lanes);
+extern int rv370_get_pcie_lanes(struct radeon_device *rdev);
 
 static struct radeon_asic r300_asic = {
 	.init = &r300_init,
-	.errata = &r300_errata,
-	.vram_info = &r300_vram_info,
+	.fini = &r300_fini,
+	.suspend = &r300_suspend,
+	.resume = &r300_resume,
 	.vga_set_state = &r100_vga_set_state,
 	.gpu_reset = &r300_gpu_reset,
-	.mc_init = &r300_mc_init,
-	.mc_fini = &r300_mc_fini,
-	.wb_init = &r100_wb_init,
-	.wb_fini = &r100_wb_fini,
-	.gart_init = &r100_pci_gart_init,
-	.gart_fini = &r100_pci_gart_fini,
-	.gart_enable = &r100_pci_gart_enable,
-	.gart_disable = &r100_pci_gart_disable,
 	.gart_tlb_flush = &r100_pci_gart_tlb_flush,
 	.gart_set_page = &r100_pci_gart_set_page,
-	.cp_init = &r100_cp_init,
-	.cp_fini = &r100_cp_fini,
-	.cp_disable = &r100_cp_disable,
 	.cp_commit = &r100_cp_commit,
 	.ring_start = &r300_ring_start,
 	.ring_test = &r100_ring_test,
 	.ring_ib_execute = &r100_ring_ib_execute,
-	.ib_test = &r100_ib_test,
 	.irq_set = &r100_irq_set,
 	.irq_process = &r100_irq_process,
 	.get_vblank_counter = &r100_get_vblank_counter,
 	.fence_ring_emit = &r300_fence_ring_emit,
 	.cs_parse = &r300_cs_parse,
 	.copy_blit = &r100_copy_blit,
-	.copy_dma = &r300_copy_dma,
+	.copy_dma = &r200_copy_dma,
+	.copy = &r100_copy_blit,
+	.get_engine_clock = &radeon_legacy_get_engine_clock,
+	.set_engine_clock = &radeon_legacy_set_engine_clock,
+	.get_memory_clock = &radeon_legacy_get_memory_clock,
+	.set_memory_clock = NULL,
+	.get_pcie_lanes = &rv370_get_pcie_lanes,
+	.set_pcie_lanes = &rv370_set_pcie_lanes,
+	.set_clock_gating = &radeon_legacy_set_clock_gating,
+	.set_surface_reg = r100_set_surface_reg,
+	.clear_surface_reg = r100_clear_surface_reg,
+	.bandwidth_update = &r100_bandwidth_update,
+	.hpd_init = &r100_hpd_init,
+	.hpd_fini = &r100_hpd_fini,
+	.hpd_sense = &r100_hpd_sense,
+	.hpd_set_polarity = &r100_hpd_set_polarity,
+	.ioctl_wait_idle = NULL,
+};
+
+
+static struct radeon_asic r300_asic_pcie = {
+	.init = &r300_init,
+	.fini = &r300_fini,
+	.suspend = &r300_suspend,
+	.resume = &r300_resume,
+	.vga_set_state = &r100_vga_set_state,
+	.gpu_reset = &r300_gpu_reset,
+	.gart_tlb_flush = &rv370_pcie_gart_tlb_flush,
+	.gart_set_page = &rv370_pcie_gart_set_page,
+	.cp_commit = &r100_cp_commit,
+	.ring_start = &r300_ring_start,
+	.ring_test = &r100_ring_test,
+	.ring_ib_execute = &r100_ring_ib_execute,
+	.irq_set = &r100_irq_set,
+	.irq_process = &r100_irq_process,
+	.get_vblank_counter = &r100_get_vblank_counter,
+	.fence_ring_emit = &r300_fence_ring_emit,
+	.cs_parse = &r300_cs_parse,
+	.copy_blit = &r100_copy_blit,
+	.copy_dma = &r200_copy_dma,
 	.copy = &r100_copy_blit,
+	.get_engine_clock = &radeon_legacy_get_engine_clock,
 	.set_engine_clock = &radeon_legacy_set_engine_clock,
+	.get_memory_clock = &radeon_legacy_get_memory_clock,
 	.set_memory_clock = NULL,
 	.set_pcie_lanes = &rv370_set_pcie_lanes,
 	.set_clock_gating = &radeon_legacy_set_clock_gating,
 	.set_surface_reg = r100_set_surface_reg,
 	.clear_surface_reg = r100_clear_surface_reg,
 	.bandwidth_update = &r100_bandwidth_update,
+	.hpd_init = &r100_hpd_init,
+	.hpd_fini = &r100_hpd_fini,
+	.hpd_sense = &r100_hpd_sense,
+	.hpd_set_polarity = &r100_hpd_set_polarity,
+	.ioctl_wait_idle = NULL,
 };
 
 /*
@@ -209,210 +274,196 @@ static struct radeon_asic r420_asic = {
 	.fini = &r420_fini,
 	.suspend = &r420_suspend,
 	.resume = &r420_resume,
-	.errata = NULL,
-	.vram_info = NULL,
 	.vga_set_state = &r100_vga_set_state,
 	.gpu_reset = &r300_gpu_reset,
-	.mc_init = NULL,
-	.mc_fini = NULL,
-	.wb_init = NULL,
-	.wb_fini = NULL,
-	.gart_enable = NULL,
-	.gart_disable = NULL,
 	.gart_tlb_flush = &rv370_pcie_gart_tlb_flush,
 	.gart_set_page = &rv370_pcie_gart_set_page,
-	.cp_init = NULL,
-	.cp_fini = NULL,
-	.cp_disable = NULL,
 	.cp_commit = &r100_cp_commit,
 	.ring_start = &r300_ring_start,
 	.ring_test = &r100_ring_test,
 	.ring_ib_execute = &r100_ring_ib_execute,
-	.ib_test = NULL,
 	.irq_set = &r100_irq_set,
 	.irq_process = &r100_irq_process,
 	.get_vblank_counter = &r100_get_vblank_counter,
 	.fence_ring_emit = &r300_fence_ring_emit,
 	.cs_parse = &r300_cs_parse,
 	.copy_blit = &r100_copy_blit,
-	.copy_dma = &r300_copy_dma,
+	.copy_dma = &r200_copy_dma,
 	.copy = &r100_copy_blit,
+	.get_engine_clock = &radeon_atom_get_engine_clock,
 	.set_engine_clock = &radeon_atom_set_engine_clock,
+	.get_memory_clock = &radeon_atom_get_memory_clock,
 	.set_memory_clock = &radeon_atom_set_memory_clock,
+	.get_pcie_lanes = &rv370_get_pcie_lanes,
 	.set_pcie_lanes = &rv370_set_pcie_lanes,
 	.set_clock_gating = &radeon_atom_set_clock_gating,
 	.set_surface_reg = r100_set_surface_reg,
 	.clear_surface_reg = r100_clear_surface_reg,
 	.bandwidth_update = &r100_bandwidth_update,
+	.hpd_init = &r100_hpd_init,
+	.hpd_fini = &r100_hpd_fini,
+	.hpd_sense = &r100_hpd_sense,
+	.hpd_set_polarity = &r100_hpd_set_polarity,
+	.ioctl_wait_idle = NULL,
 };
 
 
 /*
  * rs400,rs480
  */
-void rs400_errata(struct radeon_device *rdev);
-void rs400_vram_info(struct radeon_device *rdev);
-int rs400_mc_init(struct radeon_device *rdev);
-void rs400_mc_fini(struct radeon_device *rdev);
-int rs400_gart_init(struct radeon_device *rdev);
-void rs400_gart_fini(struct radeon_device *rdev);
-int rs400_gart_enable(struct radeon_device *rdev);
-void rs400_gart_disable(struct radeon_device *rdev);
+extern int rs400_init(struct radeon_device *rdev);
+extern void rs400_fini(struct radeon_device *rdev);
+extern int rs400_suspend(struct radeon_device *rdev);
+extern int rs400_resume(struct radeon_device *rdev);
 void rs400_gart_tlb_flush(struct radeon_device *rdev);
 int rs400_gart_set_page(struct radeon_device *rdev, int i, uint64_t addr);
 uint32_t rs400_mc_rreg(struct radeon_device *rdev, uint32_t reg);
 void rs400_mc_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v);
 static struct radeon_asic rs400_asic = {
-	.init = &r300_init,
-	.errata = &rs400_errata,
-	.vram_info = &rs400_vram_info,
+	.init = &rs400_init,
+	.fini = &rs400_fini,
+	.suspend = &rs400_suspend,
+	.resume = &rs400_resume,
 	.vga_set_state = &r100_vga_set_state,
 	.gpu_reset = &r300_gpu_reset,
-	.mc_init = &rs400_mc_init,
-	.mc_fini = &rs400_mc_fini,
-	.wb_init = &r100_wb_init,
-	.wb_fini = &r100_wb_fini,
-	.gart_init = &rs400_gart_init,
-	.gart_fini = &rs400_gart_fini,
-	.gart_enable = &rs400_gart_enable,
-	.gart_disable = &rs400_gart_disable,
 	.gart_tlb_flush = &rs400_gart_tlb_flush,
 	.gart_set_page = &rs400_gart_set_page,
-	.cp_init = &r100_cp_init,
-	.cp_fini = &r100_cp_fini,
-	.cp_disable = &r100_cp_disable,
 	.cp_commit = &r100_cp_commit,
 	.ring_start = &r300_ring_start,
 	.ring_test = &r100_ring_test,
 	.ring_ib_execute = &r100_ring_ib_execute,
-	.ib_test = &r100_ib_test,
 	.irq_set = &r100_irq_set,
 	.irq_process = &r100_irq_process,
 	.get_vblank_counter = &r100_get_vblank_counter,
 	.fence_ring_emit = &r300_fence_ring_emit,
 	.cs_parse = &r300_cs_parse,
 	.copy_blit = &r100_copy_blit,
-	.copy_dma = &r300_copy_dma,
+	.copy_dma = &r200_copy_dma,
 	.copy = &r100_copy_blit,
+	.get_engine_clock = &radeon_legacy_get_engine_clock,
 	.set_engine_clock = &radeon_legacy_set_engine_clock,
+	.get_memory_clock = &radeon_legacy_get_memory_clock,
 	.set_memory_clock = NULL,
+	.get_pcie_lanes = NULL,
 	.set_pcie_lanes = NULL,
 	.set_clock_gating = &radeon_legacy_set_clock_gating,
 	.set_surface_reg = r100_set_surface_reg,
 	.clear_surface_reg = r100_clear_surface_reg,
 	.bandwidth_update = &r100_bandwidth_update,
+	.hpd_init = &r100_hpd_init,
+	.hpd_fini = &r100_hpd_fini,
+	.hpd_sense = &r100_hpd_sense,
+	.hpd_set_polarity = &r100_hpd_set_polarity,
+	.ioctl_wait_idle = NULL,
 };
 
 
 /*
  * rs600.
  */
-int rs600_init(struct radeon_device *rdev);
-void rs600_errata(struct radeon_device *rdev);
-void rs600_vram_info(struct radeon_device *rdev);
-int rs600_mc_init(struct radeon_device *rdev);
-void rs600_mc_fini(struct radeon_device *rdev);
+extern int rs600_init(struct radeon_device *rdev);
+extern void rs600_fini(struct radeon_device *rdev);
+extern int rs600_suspend(struct radeon_device *rdev);
+extern int rs600_resume(struct radeon_device *rdev);
 int rs600_irq_set(struct radeon_device *rdev);
 int rs600_irq_process(struct radeon_device *rdev);
 u32 rs600_get_vblank_counter(struct radeon_device *rdev, int crtc);
-int rs600_gart_init(struct radeon_device *rdev);
-void rs600_gart_fini(struct radeon_device *rdev);
-int rs600_gart_enable(struct radeon_device *rdev);
-void rs600_gart_disable(struct radeon_device *rdev);
 void rs600_gart_tlb_flush(struct radeon_device *rdev);
 int rs600_gart_set_page(struct radeon_device *rdev, int i, uint64_t addr);
 uint32_t rs600_mc_rreg(struct radeon_device *rdev, uint32_t reg);
 void rs600_mc_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v);
 void rs600_bandwidth_update(struct radeon_device *rdev);
+void rs600_hpd_init(struct radeon_device *rdev);
+void rs600_hpd_fini(struct radeon_device *rdev);
+bool rs600_hpd_sense(struct radeon_device *rdev, enum radeon_hpd_id hpd);
+void rs600_hpd_set_polarity(struct radeon_device *rdev,
+			    enum radeon_hpd_id hpd);
+
 static struct radeon_asic rs600_asic = {
 	.init = &rs600_init,
-	.errata = &rs600_errata,
-	.vram_info = &rs600_vram_info,
+	.fini = &rs600_fini,
+	.suspend = &rs600_suspend,
+	.resume = &rs600_resume,
 	.vga_set_state = &r100_vga_set_state,
 	.gpu_reset = &r300_gpu_reset,
-	.mc_init = &rs600_mc_init,
-	.mc_fini = &rs600_mc_fini,
-	.wb_init = &r100_wb_init,
-	.wb_fini = &r100_wb_fini,
-	.gart_init = &rs600_gart_init,
-	.gart_fini = &rs600_gart_fini,
-	.gart_enable = &rs600_gart_enable,
-	.gart_disable = &rs600_gart_disable,
 	.gart_tlb_flush = &rs600_gart_tlb_flush,
 	.gart_set_page = &rs600_gart_set_page,
-	.cp_init = &r100_cp_init,
-	.cp_fini = &r100_cp_fini,
-	.cp_disable = &r100_cp_disable,
 	.cp_commit = &r100_cp_commit,
 	.ring_start = &r300_ring_start,
 	.ring_test = &r100_ring_test,
 	.ring_ib_execute = &r100_ring_ib_execute,
-	.ib_test = &r100_ib_test,
 	.irq_set = &rs600_irq_set,
 	.irq_process = &rs600_irq_process,
 	.get_vblank_counter = &rs600_get_vblank_counter,
 	.fence_ring_emit = &r300_fence_ring_emit,
 	.cs_parse = &r300_cs_parse,
 	.copy_blit = &r100_copy_blit,
-	.copy_dma = &r300_copy_dma,
+	.copy_dma = &r200_copy_dma,
 	.copy = &r100_copy_blit,
+	.get_engine_clock = &radeon_atom_get_engine_clock,
 	.set_engine_clock = &radeon_atom_set_engine_clock,
+	.get_memory_clock = &radeon_atom_get_memory_clock,
 	.set_memory_clock = &radeon_atom_set_memory_clock,
+	.get_pcie_lanes = NULL,
 	.set_pcie_lanes = NULL,
 	.set_clock_gating = &radeon_atom_set_clock_gating,
+	.set_surface_reg = r100_set_surface_reg,
+	.clear_surface_reg = r100_clear_surface_reg,
 	.bandwidth_update = &rs600_bandwidth_update,
+	.hpd_init = &rs600_hpd_init,
+	.hpd_fini = &rs600_hpd_fini,
+	.hpd_sense = &rs600_hpd_sense,
+	.hpd_set_polarity = &rs600_hpd_set_polarity,
+	.ioctl_wait_idle = NULL,
 };
 
 
 /*
  * rs690,rs740
  */
-void rs690_errata(struct radeon_device *rdev);
-void rs690_vram_info(struct radeon_device *rdev);
-int rs690_mc_init(struct radeon_device *rdev);
-void rs690_mc_fini(struct radeon_device *rdev);
+int rs690_init(struct radeon_device *rdev);
+void rs690_fini(struct radeon_device *rdev);
+int rs690_resume(struct radeon_device *rdev);
+int rs690_suspend(struct radeon_device *rdev);
 uint32_t rs690_mc_rreg(struct radeon_device *rdev, uint32_t reg);
 void rs690_mc_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v);
 void rs690_bandwidth_update(struct radeon_device *rdev);
 static struct radeon_asic rs690_asic = {
-	.init = &rs600_init,
-	.errata = &rs690_errata,
-	.vram_info = &rs690_vram_info,
+	.init = &rs690_init,
+	.fini = &rs690_fini,
+	.suspend = &rs690_suspend,
+	.resume = &rs690_resume,
 	.vga_set_state = &r100_vga_set_state,
 	.gpu_reset = &r300_gpu_reset,
-	.mc_init = &rs690_mc_init,
-	.mc_fini = &rs690_mc_fini,
-	.wb_init = &r100_wb_init,
-	.wb_fini = &r100_wb_fini,
-	.gart_init = &rs400_gart_init,
-	.gart_fini = &rs400_gart_fini,
-	.gart_enable = &rs400_gart_enable,
-	.gart_disable = &rs400_gart_disable,
 	.gart_tlb_flush = &rs400_gart_tlb_flush,
 	.gart_set_page = &rs400_gart_set_page,
-	.cp_init = &r100_cp_init,
-	.cp_fini = &r100_cp_fini,
-	.cp_disable = &r100_cp_disable,
 	.cp_commit = &r100_cp_commit,
 	.ring_start = &r300_ring_start,
 	.ring_test = &r100_ring_test,
 	.ring_ib_execute = &r100_ring_ib_execute,
-	.ib_test = &r100_ib_test,
 	.irq_set = &rs600_irq_set,
 	.irq_process = &rs600_irq_process,
 	.get_vblank_counter = &rs600_get_vblank_counter,
 	.fence_ring_emit = &r300_fence_ring_emit,
 	.cs_parse = &r300_cs_parse,
 	.copy_blit = &r100_copy_blit,
-	.copy_dma = &r300_copy_dma,
-	.copy = &r300_copy_dma,
+	.copy_dma = &r200_copy_dma,
+	.copy = &r200_copy_dma,
+	.get_engine_clock = &radeon_atom_get_engine_clock,
 	.set_engine_clock = &radeon_atom_set_engine_clock,
+	.get_memory_clock = &radeon_atom_get_memory_clock,
 	.set_memory_clock = &radeon_atom_set_memory_clock,
+	.get_pcie_lanes = NULL,
 	.set_pcie_lanes = NULL,
 	.set_clock_gating = &radeon_atom_set_clock_gating,
 	.set_surface_reg = r100_set_surface_reg,
 	.clear_surface_reg = r100_clear_surface_reg,
 	.bandwidth_update = &rs690_bandwidth_update,
+	.hpd_init = &rs600_hpd_init,
+	.hpd_fini = &rs600_hpd_fini,
+	.hpd_sense = &rs600_hpd_sense,
+	.hpd_set_polarity = &rs600_hpd_set_polarity,
+	.ioctl_wait_idle = NULL,
 };
 
 
@@ -420,106 +471,96 @@ static struct radeon_asic rs690_asic = {
  * rv515
  */
 int rv515_init(struct radeon_device *rdev);
-void rv515_errata(struct radeon_device *rdev);
-void rv515_vram_info(struct radeon_device *rdev);
+void rv515_fini(struct radeon_device *rdev);
 int rv515_gpu_reset(struct radeon_device *rdev);
-int rv515_mc_init(struct radeon_device *rdev);
-void rv515_mc_fini(struct radeon_device *rdev);
 uint32_t rv515_mc_rreg(struct radeon_device *rdev, uint32_t reg);
 void rv515_mc_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v);
 void rv515_ring_start(struct radeon_device *rdev);
 uint32_t rv515_pcie_rreg(struct radeon_device *rdev, uint32_t reg);
 void rv515_pcie_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v);
 void rv515_bandwidth_update(struct radeon_device *rdev);
+int rv515_resume(struct radeon_device *rdev);
+int rv515_suspend(struct radeon_device *rdev);
 static struct radeon_asic rv515_asic = {
 	.init = &rv515_init,
-	.errata = &rv515_errata,
-	.vram_info = &rv515_vram_info,
+	.fini = &rv515_fini,
+	.suspend = &rv515_suspend,
+	.resume = &rv515_resume,
 	.vga_set_state = &r100_vga_set_state,
 	.gpu_reset = &rv515_gpu_reset,
-	.mc_init = &rv515_mc_init,
-	.mc_fini = &rv515_mc_fini,
-	.wb_init = &r100_wb_init,
-	.wb_fini = &r100_wb_fini,
-	.gart_init = &rv370_pcie_gart_init,
-	.gart_fini = &rv370_pcie_gart_fini,
-	.gart_enable = &rv370_pcie_gart_enable,
-	.gart_disable = &rv370_pcie_gart_disable,
 	.gart_tlb_flush = &rv370_pcie_gart_tlb_flush,
 	.gart_set_page = &rv370_pcie_gart_set_page,
-	.cp_init = &r100_cp_init,
-	.cp_fini = &r100_cp_fini,
-	.cp_disable = &r100_cp_disable,
 	.cp_commit = &r100_cp_commit,
 	.ring_start = &rv515_ring_start,
 	.ring_test = &r100_ring_test,
 	.ring_ib_execute = &r100_ring_ib_execute,
-	.ib_test = &r100_ib_test,
 	.irq_set = &rs600_irq_set,
 	.irq_process = &rs600_irq_process,
 	.get_vblank_counter = &rs600_get_vblank_counter,
 	.fence_ring_emit = &r300_fence_ring_emit,
 	.cs_parse = &r300_cs_parse,
 	.copy_blit = &r100_copy_blit,
-	.copy_dma = &r300_copy_dma,
+	.copy_dma = &r200_copy_dma,
 	.copy = &r100_copy_blit,
+	.get_engine_clock = &radeon_atom_get_engine_clock,
 	.set_engine_clock = &radeon_atom_set_engine_clock,
+	.get_memory_clock = &radeon_atom_get_memory_clock,
 	.set_memory_clock = &radeon_atom_set_memory_clock,
+	.get_pcie_lanes = &rv370_get_pcie_lanes,
 	.set_pcie_lanes = &rv370_set_pcie_lanes,
 	.set_clock_gating = &radeon_atom_set_clock_gating,
 	.set_surface_reg = r100_set_surface_reg,
 	.clear_surface_reg = r100_clear_surface_reg,
 	.bandwidth_update = &rv515_bandwidth_update,
+	.hpd_init = &rs600_hpd_init,
+	.hpd_fini = &rs600_hpd_fini,
+	.hpd_sense = &rs600_hpd_sense,
+	.hpd_set_polarity = &rs600_hpd_set_polarity,
+	.ioctl_wait_idle = NULL,
 };
 
 
 /*
  * r520,rv530,rv560,rv570,r580
  */
-void r520_errata(struct radeon_device *rdev);
-void r520_vram_info(struct radeon_device *rdev);
-int r520_mc_init(struct radeon_device *rdev);
-void r520_mc_fini(struct radeon_device *rdev);
-void r520_bandwidth_update(struct radeon_device *rdev);
+int r520_init(struct radeon_device *rdev);
+int r520_resume(struct radeon_device *rdev);
 static struct radeon_asic r520_asic = {
-	.init = &rv515_init,
-	.errata = &r520_errata,
-	.vram_info = &r520_vram_info,
+	.init = &r520_init,
+	.fini = &rv515_fini,
+	.suspend = &rv515_suspend,
+	.resume = &r520_resume,
 	.vga_set_state = &r100_vga_set_state,
 	.gpu_reset = &rv515_gpu_reset,
-	.mc_init = &r520_mc_init,
-	.mc_fini = &r520_mc_fini,
-	.wb_init = &r100_wb_init,
-	.wb_fini = &r100_wb_fini,
-	.gart_init = &rv370_pcie_gart_init,
-	.gart_fini = &rv370_pcie_gart_fini,
-	.gart_enable = &rv370_pcie_gart_enable,
-	.gart_disable = &rv370_pcie_gart_disable,
 	.gart_tlb_flush = &rv370_pcie_gart_tlb_flush,
 	.gart_set_page = &rv370_pcie_gart_set_page,
-	.cp_init = &r100_cp_init,
-	.cp_fini = &r100_cp_fini,
-	.cp_disable = &r100_cp_disable,
 	.cp_commit = &r100_cp_commit,
 	.ring_start = &rv515_ring_start,
 	.ring_test = &r100_ring_test,
 	.ring_ib_execute = &r100_ring_ib_execute,
-	.ib_test = &r100_ib_test,
 	.irq_set = &rs600_irq_set,
 	.irq_process = &rs600_irq_process,
 	.get_vblank_counter = &rs600_get_vblank_counter,
 	.fence_ring_emit = &r300_fence_ring_emit,
 	.cs_parse = &r300_cs_parse,
 	.copy_blit = &r100_copy_blit,
-	.copy_dma = &r300_copy_dma,
+	.copy_dma = &r200_copy_dma,
 	.copy = &r100_copy_blit,
+	.get_engine_clock = &radeon_atom_get_engine_clock,
 	.set_engine_clock = &radeon_atom_set_engine_clock,
+	.get_memory_clock = &radeon_atom_get_memory_clock,
 	.set_memory_clock = &radeon_atom_set_memory_clock,
+	.get_pcie_lanes = &rv370_get_pcie_lanes,
 	.set_pcie_lanes = &rv370_set_pcie_lanes,
 	.set_clock_gating = &radeon_atom_set_clock_gating,
 	.set_surface_reg = r100_set_surface_reg,
 	.clear_surface_reg = r100_clear_surface_reg,
-	.bandwidth_update = &r520_bandwidth_update,
+	.bandwidth_update = &rv515_bandwidth_update,
+	.hpd_init = &rs600_hpd_init,
+	.hpd_fini = &rs600_hpd_fini,
+	.hpd_sense = &rs600_hpd_sense,
+	.hpd_set_polarity = &rs600_hpd_set_polarity,
+	.ioctl_wait_idle = NULL,
 };
 
 /*
@@ -552,51 +593,52 @@ int r600_set_surface_reg(struct radeon_device *rdev, int reg,
 			 uint32_t offset, uint32_t obj_size);
 int r600_clear_surface_reg(struct radeon_device *rdev, int reg);
 void r600_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib);
-int r600_ib_test(struct radeon_device *rdev);
 int r600_ring_test(struct radeon_device *rdev);
 int r600_copy_blit(struct radeon_device *rdev,
 		   uint64_t src_offset, uint64_t dst_offset,
 		   unsigned num_pages, struct radeon_fence *fence);
+void r600_hpd_init(struct radeon_device *rdev);
+void r600_hpd_fini(struct radeon_device *rdev);
+bool r600_hpd_sense(struct radeon_device *rdev, enum radeon_hpd_id hpd);
+void r600_hpd_set_polarity(struct radeon_device *rdev,
+			   enum radeon_hpd_id hpd);
+extern void r600_ioctl_wait_idle(struct radeon_device *rdev, struct radeon_bo *bo);
 
 static struct radeon_asic r600_asic = {
-	.errata = NULL,
 	.init = &r600_init,
 	.fini = &r600_fini,
 	.suspend = &r600_suspend,
 	.resume = &r600_resume,
 	.cp_commit = &r600_cp_commit,
-	.vram_info = NULL,
 	.vga_set_state = &r600_vga_set_state,
 	.gpu_reset = &r600_gpu_reset,
-	.mc_init = NULL,
-	.mc_fini = NULL,
-	.wb_init = &r600_wb_init,
-	.wb_fini = &r600_wb_fini,
-	.gart_enable = NULL,
-	.gart_disable = NULL,
 	.gart_tlb_flush = &r600_pcie_gart_tlb_flush,
 	.gart_set_page = &rs600_gart_set_page,
-	.cp_init = NULL,
-	.cp_fini = NULL,
-	.cp_disable = NULL,
-	.ring_start = NULL,
 	.ring_test = &r600_ring_test,
 	.ring_ib_execute = &r600_ring_ib_execute,
-	.ib_test = &r600_ib_test,
 	.irq_set = &r600_irq_set,
 	.irq_process = &r600_irq_process,
+	.get_vblank_counter = &rs600_get_vblank_counter,
 	.fence_ring_emit = &r600_fence_ring_emit,
 	.cs_parse = &r600_cs_parse,
 	.copy_blit = &r600_copy_blit,
 	.copy_dma = &r600_copy_blit,
 	.copy = &r600_copy_blit,
+	.get_engine_clock = &radeon_atom_get_engine_clock,
 	.set_engine_clock = &radeon_atom_set_engine_clock,
+	.get_memory_clock = &radeon_atom_get_memory_clock,
 	.set_memory_clock = &radeon_atom_set_memory_clock,
+	.get_pcie_lanes = &rv370_get_pcie_lanes,
 	.set_pcie_lanes = NULL,
-	.set_clock_gating = &radeon_atom_set_clock_gating,
+	.set_clock_gating = NULL,
 	.set_surface_reg = r600_set_surface_reg,
 	.clear_surface_reg = r600_clear_surface_reg,
-	.bandwidth_update = &r520_bandwidth_update,
+	.bandwidth_update = &rv515_bandwidth_update,
+	.hpd_init = &r600_hpd_init,
+	.hpd_fini = &r600_hpd_fini,
+	.hpd_sense = &r600_hpd_sense,
+	.hpd_set_polarity = &r600_hpd_set_polarity,
+	.ioctl_wait_idle = r600_ioctl_wait_idle,
 };
 
 /*
@@ -609,44 +651,90 @@ int rv770_resume(struct radeon_device *rdev);
 int rv770_gpu_reset(struct radeon_device *rdev);
 
 static struct radeon_asic rv770_asic = {
-	.errata = NULL,
 	.init = &rv770_init,
 	.fini = &rv770_fini,
 	.suspend = &rv770_suspend,
 	.resume = &rv770_resume,
 	.cp_commit = &r600_cp_commit,
-	.vram_info = NULL,
 	.gpu_reset = &rv770_gpu_reset,
 	.vga_set_state = &r600_vga_set_state,
-	.mc_init = NULL,
-	.mc_fini = NULL,
-	.wb_init = &r600_wb_init,
-	.wb_fini = &r600_wb_fini,
-	.gart_enable = NULL,
-	.gart_disable = NULL,
 	.gart_tlb_flush = &r600_pcie_gart_tlb_flush,
 	.gart_set_page = &rs600_gart_set_page,
-	.cp_init = NULL,
-	.cp_fini = NULL,
-	.cp_disable = NULL,
-	.ring_start = NULL,
 	.ring_test = &r600_ring_test,
 	.ring_ib_execute = &r600_ring_ib_execute,
-	.ib_test = &r600_ib_test,
 	.irq_set = &r600_irq_set,
 	.irq_process = &r600_irq_process,
+	.get_vblank_counter = &rs600_get_vblank_counter,
 	.fence_ring_emit = &r600_fence_ring_emit,
 	.cs_parse = &r600_cs_parse,
 	.copy_blit = &r600_copy_blit,
 	.copy_dma = &r600_copy_blit,
 	.copy = &r600_copy_blit,
+	.get_engine_clock = &radeon_atom_get_engine_clock,
 	.set_engine_clock = &radeon_atom_set_engine_clock,
+	.get_memory_clock = &radeon_atom_get_memory_clock,
 	.set_memory_clock = &radeon_atom_set_memory_clock,
+	.get_pcie_lanes = &rv370_get_pcie_lanes,
 	.set_pcie_lanes = NULL,
 	.set_clock_gating = &radeon_atom_set_clock_gating,
 	.set_surface_reg = r600_set_surface_reg,
 	.clear_surface_reg = r600_clear_surface_reg,
-	.bandwidth_update = &r520_bandwidth_update,
+	.bandwidth_update = &rv515_bandwidth_update,
+	.hpd_init = &r600_hpd_init,
+	.hpd_fini = &r600_hpd_fini,
+	.hpd_sense = &r600_hpd_sense,
+	.hpd_set_polarity = &r600_hpd_set_polarity,
+	.ioctl_wait_idle = r600_ioctl_wait_idle,
+};
+
+/*
+ * evergreen
+ */
+int evergreen_init(struct radeon_device *rdev);
+void evergreen_fini(struct radeon_device *rdev);
+int evergreen_suspend(struct radeon_device *rdev);
+int evergreen_resume(struct radeon_device *rdev);
+int evergreen_gpu_reset(struct radeon_device *rdev);
+void evergreen_bandwidth_update(struct radeon_device *rdev);
+void evergreen_hpd_init(struct radeon_device *rdev);
+void evergreen_hpd_fini(struct radeon_device *rdev);
+bool evergreen_hpd_sense(struct radeon_device *rdev, enum radeon_hpd_id hpd);
+void evergreen_hpd_set_polarity(struct radeon_device *rdev,
+				enum radeon_hpd_id hpd);
+
+static struct radeon_asic evergreen_asic = {
+	.init = &evergreen_init,
+	.fini = &evergreen_fini,
+	.suspend = &evergreen_suspend,
+	.resume = &evergreen_resume,
+	.cp_commit = NULL,
+	.gpu_reset = &evergreen_gpu_reset,
+	.vga_set_state = &r600_vga_set_state,
+	.gart_tlb_flush = &r600_pcie_gart_tlb_flush,
+	.gart_set_page = &rs600_gart_set_page,
+	.ring_test = NULL,
+	.ring_ib_execute = NULL,
+	.irq_set = NULL,
+	.irq_process = NULL,
+	.get_vblank_counter = NULL,
+	.fence_ring_emit = NULL,
+	.cs_parse = NULL,
+	.copy_blit = NULL,
+	.copy_dma = NULL,
+	.copy = NULL,
+	.get_engine_clock = &radeon_atom_get_engine_clock,
+	.set_engine_clock = &radeon_atom_set_engine_clock,
+	.get_memory_clock = &radeon_atom_get_memory_clock,
+	.set_memory_clock = &radeon_atom_set_memory_clock,
+	.set_pcie_lanes = NULL,
+	.set_clock_gating = NULL,
+	.set_surface_reg = r600_set_surface_reg,
+	.clear_surface_reg = r600_clear_surface_reg,
+	.bandwidth_update = &evergreen_bandwidth_update,
+	.hpd_init = &evergreen_hpd_init,
+	.hpd_fini = &evergreen_hpd_fini,
+	.hpd_sense = &evergreen_hpd_sense,
+	.hpd_set_polarity = &evergreen_hpd_set_polarity,
 };
 
 #endif
diff --git a/drivers/gpu/drm/radeon/radeon_atombios.c b/drivers/gpu/drm/radeon/radeon_atombios.c
index 743742128307..93783b15c81d 100644
--- a/drivers/gpu/drm/radeon/radeon_atombios.c
+++ b/drivers/gpu/drm/radeon/radeon_atombios.c
@@ -46,7 +46,9 @@ radeon_add_atom_connector(struct drm_device *dev,
 			  uint32_t supported_device,
 			  int connector_type,
 			  struct radeon_i2c_bus_rec *i2c_bus,
-			  bool linkb, uint32_t igp_lane_info);
+			  bool linkb, uint32_t igp_lane_info,
+			  uint16_t connector_object_id,
+			  struct radeon_hpd *hpd);
 
 /* from radeon_legacy_encoder.c */
 extern void
@@ -59,16 +61,16 @@ union atom_supported_devices {
 	struct _ATOM_SUPPORTED_DEVICES_INFO_2d1 info_2d1;
 };
 
-static inline struct radeon_i2c_bus_rec radeon_lookup_gpio(struct drm_device
-							   *dev, uint8_t id)
+static inline struct radeon_i2c_bus_rec radeon_lookup_i2c_gpio(struct radeon_device *rdev,
+							       uint8_t id)
 {
-	struct radeon_device *rdev = dev->dev_private;
 	struct atom_context *ctx = rdev->mode_info.atom_context;
-	ATOM_GPIO_I2C_ASSIGMENT gpio;
+	ATOM_GPIO_I2C_ASSIGMENT *gpio;
 	struct radeon_i2c_bus_rec i2c;
 	int index = GetIndexIntoMasterTable(DATA, GPIO_I2C_Info);
 	struct _ATOM_GPIO_I2C_INFO *i2c_info;
 	uint16_t data_offset;
+	int i;
 
 	memset(&i2c, 0, sizeof(struct radeon_i2c_bus_rec));
 	i2c.valid = false;
@@ -77,34 +79,129 @@ static inline struct radeon_i2c_bus_rec radeon_lookup_gpio(struct drm_device
 
 	i2c_info = (struct _ATOM_GPIO_I2C_INFO *)(ctx->bios + data_offset);
 
-	gpio = i2c_info->asGPIO_Info[id];
-
-	i2c.mask_clk_reg = le16_to_cpu(gpio.usClkMaskRegisterIndex) * 4;
-	i2c.mask_data_reg = le16_to_cpu(gpio.usDataMaskRegisterIndex) * 4;
-	i2c.put_clk_reg = le16_to_cpu(gpio.usClkEnRegisterIndex) * 4;
-	i2c.put_data_reg = le16_to_cpu(gpio.usDataEnRegisterIndex) * 4;
-	i2c.get_clk_reg = le16_to_cpu(gpio.usClkY_RegisterIndex) * 4;
-	i2c.get_data_reg = le16_to_cpu(gpio.usDataY_RegisterIndex) * 4;
-	i2c.a_clk_reg = le16_to_cpu(gpio.usClkA_RegisterIndex) * 4;
-	i2c.a_data_reg = le16_to_cpu(gpio.usDataA_RegisterIndex) * 4;
-	i2c.mask_clk_mask = (1 << gpio.ucClkMaskShift);
-	i2c.mask_data_mask = (1 << gpio.ucDataMaskShift);
-	i2c.put_clk_mask = (1 << gpio.ucClkEnShift);
-	i2c.put_data_mask = (1 << gpio.ucDataEnShift);
-	i2c.get_clk_mask = (1 << gpio.ucClkY_Shift);
-	i2c.get_data_mask = (1 << gpio.ucDataY_Shift);
-	i2c.a_clk_mask = (1 << gpio.ucClkA_Shift);
-	i2c.a_data_mask = (1 << gpio.ucDataA_Shift);
-	i2c.valid = true;
+
+	for (i = 0; i < ATOM_MAX_SUPPORTED_DEVICE; i++) {
+		gpio = &i2c_info->asGPIO_Info[i];
+
+		if (gpio->sucI2cId.ucAccess == id) {
+			i2c.mask_clk_reg = le16_to_cpu(gpio->usClkMaskRegisterIndex) * 4;
+			i2c.mask_data_reg = le16_to_cpu(gpio->usDataMaskRegisterIndex) * 4;
+			i2c.en_clk_reg = le16_to_cpu(gpio->usClkEnRegisterIndex) * 4;
+			i2c.en_data_reg = le16_to_cpu(gpio->usDataEnRegisterIndex) * 4;
+			i2c.y_clk_reg = le16_to_cpu(gpio->usClkY_RegisterIndex) * 4;
+			i2c.y_data_reg = le16_to_cpu(gpio->usDataY_RegisterIndex) * 4;
+			i2c.a_clk_reg = le16_to_cpu(gpio->usClkA_RegisterIndex) * 4;
+			i2c.a_data_reg = le16_to_cpu(gpio->usDataA_RegisterIndex) * 4;
+			i2c.mask_clk_mask = (1 << gpio->ucClkMaskShift);
+			i2c.mask_data_mask = (1 << gpio->ucDataMaskShift);
+			i2c.en_clk_mask = (1 << gpio->ucClkEnShift);
+			i2c.en_data_mask = (1 << gpio->ucDataEnShift);
+			i2c.y_clk_mask = (1 << gpio->ucClkY_Shift);
+			i2c.y_data_mask = (1 << gpio->ucDataY_Shift);
+			i2c.a_clk_mask = (1 << gpio->ucClkA_Shift);
+			i2c.a_data_mask = (1 << gpio->ucDataA_Shift);
+
+			if (gpio->sucI2cId.sbfAccess.bfHW_Capable)
+				i2c.hw_capable = true;
+			else
+				i2c.hw_capable = false;
+
+			if (gpio->sucI2cId.ucAccess == 0xa0)
+				i2c.mm_i2c = true;
+			else
+				i2c.mm_i2c = false;
+
+			i2c.i2c_id = gpio->sucI2cId.ucAccess;
+
+			i2c.valid = true;
+			break;
+		}
+	}
 
 	return i2c;
 }
 
+static inline struct radeon_gpio_rec radeon_lookup_gpio(struct radeon_device *rdev,
+							u8 id)
+{
+	struct atom_context *ctx = rdev->mode_info.atom_context;
+	struct radeon_gpio_rec gpio;
+	int index = GetIndexIntoMasterTable(DATA, GPIO_Pin_LUT);
+	struct _ATOM_GPIO_PIN_LUT *gpio_info;
+	ATOM_GPIO_PIN_ASSIGNMENT *pin;
+	u16 data_offset, size;
+	int i, num_indices;
+
+	memset(&gpio, 0, sizeof(struct radeon_gpio_rec));
+	gpio.valid = false;
+
+	atom_parse_data_header(ctx, index, &size, NULL, NULL, &data_offset);
+
+	gpio_info = (struct _ATOM_GPIO_PIN_LUT *)(ctx->bios + data_offset);
+
+	num_indices = (size - sizeof(ATOM_COMMON_TABLE_HEADER)) / sizeof(ATOM_GPIO_PIN_ASSIGNMENT);
+
+	for (i = 0; i < num_indices; i++) {
+		pin = &gpio_info->asGPIO_Pin[i];
+		if (id == pin->ucGPIO_ID) {
+			gpio.id = pin->ucGPIO_ID;
+			gpio.reg = pin->usGpioPin_AIndex * 4;
+			gpio.mask = (1 << pin->ucGpioPinBitShift);
+			gpio.valid = true;
+			break;
+		}
+	}
+
+	return gpio;
+}
+
+static struct radeon_hpd radeon_atom_get_hpd_info_from_gpio(struct radeon_device *rdev,
+							    struct radeon_gpio_rec *gpio)
+{
+	struct radeon_hpd hpd;
+	u32 reg;
+
+	if (ASIC_IS_DCE4(rdev))
+		reg = EVERGREEN_DC_GPIO_HPD_A;
+	else
+		reg = AVIVO_DC_GPIO_HPD_A;
+
+	hpd.gpio = *gpio;
+	if (gpio->reg == reg) {
+		switch(gpio->mask) {
+		case (1 << 0):
+			hpd.hpd = RADEON_HPD_1;
+			break;
+		case (1 << 8):
+			hpd.hpd = RADEON_HPD_2;
+			break;
+		case (1 << 16):
+			hpd.hpd = RADEON_HPD_3;
+			break;
+		case (1 << 24):
+			hpd.hpd = RADEON_HPD_4;
+			break;
+		case (1 << 26):
+			hpd.hpd = RADEON_HPD_5;
+			break;
+		case (1 << 28):
+			hpd.hpd = RADEON_HPD_6;
+			break;
+		default:
+			hpd.hpd = RADEON_HPD_NONE;
+			break;
+		}
+	} else
+		hpd.hpd = RADEON_HPD_NONE;
+	return hpd;
+}
+
 static bool radeon_atom_apply_quirks(struct drm_device *dev,
 				     uint32_t supported_device,
 				     int *connector_type,
 				     struct radeon_i2c_bus_rec *i2c_bus,
-				     uint16_t *line_mux)
+				     uint16_t *line_mux,
+				     struct radeon_hpd *hpd)
 {
 
 	/* Asus M2A-VM HDMI board lists the DVI port as HDMI */
@@ -116,6 +213,15 @@ static bool radeon_atom_apply_quirks(struct drm_device *dev,
 			*connector_type = DRM_MODE_CONNECTOR_DVID;
 	}
 
+	/* Asrock RS600 board lists the DVI port as HDMI */
+	if ((dev->pdev->device == 0x7941) &&
+	    (dev->pdev->subsystem_vendor == 0x1849) &&
+	    (dev->pdev->subsystem_device == 0x7941)) {
+		if ((*connector_type == DRM_MODE_CONNECTOR_HDMIA) &&
+		    (supported_device == ATOM_DEVICE_DFP3_SUPPORT))
+			*connector_type = DRM_MODE_CONNECTOR_DVID;
+	}
+
 	/* a-bit f-i90hd - ciaranm on #radeonhd - this board has no DVI */
 	if ((dev->pdev->device == 0x7941) &&
 	    (dev->pdev->subsystem_vendor == 0x147b) &&
@@ -134,6 +240,23 @@ static bool radeon_atom_apply_quirks(struct drm_device *dev,
 		}
 	}
 
+	/* HIS X1300 is DVI+VGA, not DVI+DVI */
+	if ((dev->pdev->device == 0x7146) &&
+	    (dev->pdev->subsystem_vendor == 0x17af) &&
+	    (dev->pdev->subsystem_device == 0x2058)) {
+		if (supported_device == ATOM_DEVICE_DFP1_SUPPORT)
+			return false;
+	}
+
+	/* Gigabyte X1300 is DVI+VGA, not DVI+DVI */
+	if ((dev->pdev->device == 0x7142) &&
+	    (dev->pdev->subsystem_vendor == 0x1458) &&
+	    (dev->pdev->subsystem_device == 0x2134)) {
+		if (supported_device == ATOM_DEVICE_DFP1_SUPPORT)
+			return false;
+	}
+
+
 	/* Funky macbooks */
 	if ((dev->pdev->device == 0x71C5) &&
 	    (dev->pdev->subsystem_vendor == 0x106b) &&
@@ -171,6 +294,24 @@ static bool radeon_atom_apply_quirks(struct drm_device *dev,
 		}
 	}
 
+	/* Acer laptop reports DVI-D as DVI-I */
+	if ((dev->pdev->device == 0x95c4) &&
+	    (dev->pdev->subsystem_vendor == 0x1025) &&
+	    (dev->pdev->subsystem_device == 0x013c)) {
+		if ((*connector_type == DRM_MODE_CONNECTOR_DVII) &&
+		    (supported_device == ATOM_DEVICE_DFP1_SUPPORT))
+			*connector_type = DRM_MODE_CONNECTOR_DVID;
+	}
+
+	/* XFX Pine Group device rv730 reports no VGA DDC lines
+	 * even though they are wired up to record 0x93
+	 */
+	if ((dev->pdev->device == 0x9498) &&
+	    (dev->pdev->subsystem_vendor == 0x1682) &&
+	    (dev->pdev->subsystem_device == 0x2452)) {
+		struct radeon_device *rdev = dev->dev_private;
+		*i2c_bus = radeon_lookup_i2c_gpio(rdev, 0x93);
+	}
 	return true;
 }
 
@@ -193,6 +334,23 @@ const int supported_devices_connector_convert[] = {
 	DRM_MODE_CONNECTOR_DisplayPort
 };
 
+const uint16_t supported_devices_connector_object_id_convert[] = {
+	CONNECTOR_OBJECT_ID_NONE,
+	CONNECTOR_OBJECT_ID_VGA,
+	CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_I, /* not all boards support DL */
+	CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_D, /* not all boards support DL */
+	CONNECTOR_OBJECT_ID_VGA, /* technically DVI-A */
+	CONNECTOR_OBJECT_ID_COMPOSITE,
+	CONNECTOR_OBJECT_ID_SVIDEO,
+	CONNECTOR_OBJECT_ID_LVDS,
+	CONNECTOR_OBJECT_ID_9PIN_DIN,
+	CONNECTOR_OBJECT_ID_9PIN_DIN,
+	CONNECTOR_OBJECT_ID_DISPLAYPORT,
+	CONNECTOR_OBJECT_ID_HDMI_TYPE_A,
+	CONNECTOR_OBJECT_ID_HDMI_TYPE_B,
+	CONNECTOR_OBJECT_ID_SVIDEO
+};
+
 const int object_connector_convert[] = {
 	DRM_MODE_CONNECTOR_Unknown,
 	DRM_MODE_CONNECTOR_DVII,
@@ -213,7 +371,9 @@ const int object_connector_convert[] = {
 	DRM_MODE_CONNECTOR_Unknown,
 	DRM_MODE_CONNECTOR_Unknown,
 	DRM_MODE_CONNECTOR_Unknown,
-	DRM_MODE_CONNECTOR_DisplayPort
+	DRM_MODE_CONNECTOR_DisplayPort,
+	DRM_MODE_CONNECTOR_eDP,
+	DRM_MODE_CONNECTOR_Unknown
 };
 
 bool radeon_get_atom_connector_info_from_object_table(struct drm_device *dev)
@@ -222,16 +382,18 @@ bool radeon_get_atom_connector_info_from_object_table(struct drm_device *dev)
 	struct radeon_mode_info *mode_info = &rdev->mode_info;
 	struct atom_context *ctx = mode_info->atom_context;
 	int index = GetIndexIntoMasterTable(DATA, Object_Header);
-	uint16_t size, data_offset;
-	uint8_t frev, crev, line_mux = 0;
+	u16 size, data_offset;
+	u8 frev, crev;
 	ATOM_CONNECTOR_OBJECT_TABLE *con_obj;
 	ATOM_DISPLAY_OBJECT_PATH_TABLE *path_obj;
 	ATOM_OBJECT_HEADER *obj_header;
 	int i, j, path_size, device_support;
 	int connector_type;
-	uint16_t igp_lane_info, conn_id;
+	u16 igp_lane_info, conn_id, connector_object_id;
 	bool linkb;
 	struct radeon_i2c_bus_rec ddc_bus;
+	struct radeon_gpio_rec gpio;
+	struct radeon_hpd hpd;
 
 	atom_parse_data_header(ctx, index, &size, &frev, &crev, &data_offset);
 
@@ -258,7 +420,6 @@ bool radeon_get_atom_connector_info_from_object_table(struct drm_device *dev)
 		path = (ATOM_DISPLAY_OBJECT_PATH *) addr;
 		path_size += le16_to_cpu(path->usSize);
 		linkb = false;
-
 		if (device_support & le16_to_cpu(path->usDeviceTag)) {
 			uint8_t con_obj_id, con_obj_num, con_obj_type;
 
@@ -272,15 +433,13 @@ bool radeon_get_atom_connector_info_from_object_table(struct drm_device *dev)
 			    (le16_to_cpu(path->usConnObjectId) &
 			     OBJECT_TYPE_MASK) >> OBJECT_TYPE_SHIFT;
 
-			if ((le16_to_cpu(path->usDeviceTag) ==
-			     ATOM_DEVICE_TV1_SUPPORT)
-			    || (le16_to_cpu(path->usDeviceTag) ==
-				ATOM_DEVICE_TV2_SUPPORT)
-			    || (le16_to_cpu(path->usDeviceTag) ==
-				ATOM_DEVICE_CV_SUPPORT))
+			/* TODO CV support */
+			if (le16_to_cpu(path->usDeviceTag) ==
+				ATOM_DEVICE_CV_SUPPORT)
 				continue;
 
-			if ((rdev->family == CHIP_RS780) &&
+			/* IGP chips */
+			if ((rdev->flags & RADEON_IS_IGP) &&
 			    (con_obj_id ==
 			     CONNECTOR_OBJECT_ID_PCIE_CONNECTOR)) {
 				uint16_t igp_offset = 0;
@@ -314,6 +473,7 @@ bool radeon_get_atom_connector_info_from_object_table(struct drm_device *dev)
 						connector_type =
 						    object_connector_convert
 						    [ct];
+						connector_object_id = ct;
 						igp_lane_info =
 						    slot_config & 0xffff;
 					} else
@@ -324,6 +484,7 @@ bool radeon_get_atom_connector_info_from_object_table(struct drm_device *dev)
 				igp_lane_info = 0;
 				connector_type =
 				    object_connector_convert[con_obj_id];
+				connector_object_id = con_obj_id;
 			}
 
 			if (connector_type == DRM_MODE_CONNECTOR_Unknown)
@@ -359,10 +520,9 @@ bool radeon_get_atom_connector_info_from_object_table(struct drm_device *dev)
 				}
 			}
 
-			/* look up gpio for ddc */
+			/* look up gpio for ddc, hpd */
 			if ((le16_to_cpu(path->usDeviceTag) &
-			     (ATOM_DEVICE_TV_SUPPORT | ATOM_DEVICE_CV_SUPPORT))
-			    == 0) {
+			     (ATOM_DEVICE_TV_SUPPORT | ATOM_DEVICE_CV_SUPPORT)) == 0) {
 				for (j = 0; j < con_obj->ucNumberOfObjects; j++) {
 					if (le16_to_cpu(path->usConnObjectId) ==
 					    le16_to_cpu(con_obj->asObjects[j].
@@ -376,21 +536,34 @@ bool radeon_get_atom_connector_info_from_object_table(struct drm_device *dev)
 								 asObjects[j].
 								 usRecordOffset));
 						ATOM_I2C_RECORD *i2c_record;
+						ATOM_HPD_INT_RECORD *hpd_record;
+						ATOM_I2C_ID_CONFIG_ACCESS *i2c_config;
+						hpd.hpd = RADEON_HPD_NONE;
 
 						while (record->ucRecordType > 0
 						       && record->
 						       ucRecordType <=
 						       ATOM_MAX_OBJECT_RECORD_NUMBER) {
-							switch (record->
-								ucRecordType) {
+							switch (record->ucRecordType) {
 							case ATOM_I2C_RECORD_TYPE:
 								i2c_record =
-								    (ATOM_I2C_RECORD
-								     *) record;
-								line_mux =
-								    i2c_record->
-								    sucI2cId.
-								    bfI2C_LineMux;
+								    (ATOM_I2C_RECORD *)
+									record;
+								i2c_config =
+									(ATOM_I2C_ID_CONFIG_ACCESS *)
+									&i2c_record->sucI2cId;
+								ddc_bus = radeon_lookup_i2c_gpio(rdev,
+												 i2c_config->
+												 ucAccess);
+								break;
+							case ATOM_HPD_INT_RECORD_TYPE:
+								hpd_record =
+									(ATOM_HPD_INT_RECORD *)
+									record;
+								gpio = radeon_lookup_gpio(rdev,
+											  hpd_record->ucHPDIntGPIOID);
+								hpd = radeon_atom_get_hpd_info_from_gpio(rdev, &gpio);
+								hpd.plugged_state = hpd_record->ucPlugged_PinState;
 								break;
 							}
 							record =
@@ -403,24 +576,19 @@ bool radeon_get_atom_connector_info_from_object_table(struct drm_device *dev)
 						break;
 					}
 				}
-			} else
-				line_mux = 0;
-
-			if ((le16_to_cpu(path->usDeviceTag) ==
-			     ATOM_DEVICE_TV1_SUPPORT)
-			    || (le16_to_cpu(path->usDeviceTag) ==
-				ATOM_DEVICE_TV2_SUPPORT)
-			    || (le16_to_cpu(path->usDeviceTag) ==
-				ATOM_DEVICE_CV_SUPPORT))
+			} else {
+				hpd.hpd = RADEON_HPD_NONE;
 				ddc_bus.valid = false;
-			else
-				ddc_bus = radeon_lookup_gpio(dev, line_mux);
+			}
+
+			/* needed for aux chan transactions */
+			ddc_bus.hpd_id = hpd.hpd ? (hpd.hpd - 1) : 0;
 
 			conn_id = le16_to_cpu(path->usConnObjectId);
 
 			if (!radeon_atom_apply_quirks
 			    (dev, le16_to_cpu(path->usDeviceTag), &connector_type,
-			     &ddc_bus, &conn_id))
+			     &ddc_bus, &conn_id, &hpd))
 				continue;
 
 			radeon_add_atom_connector(dev,
@@ -428,7 +596,9 @@ bool radeon_get_atom_connector_info_from_object_table(struct drm_device *dev)
 						  le16_to_cpu(path->
 							      usDeviceTag),
 						  connector_type, &ddc_bus,
-						  linkb, igp_lane_info);
+						  linkb, igp_lane_info,
+						  connector_object_id,
+						  &hpd);
 
 		}
 	}
@@ -438,12 +608,52 @@ bool radeon_get_atom_connector_info_from_object_table(struct drm_device *dev)
 	return true;
 }
 
+static uint16_t atombios_get_connector_object_id(struct drm_device *dev,
+						 int connector_type,
+						 uint16_t devices)
+{
+	struct radeon_device *rdev = dev->dev_private;
+
+	if (rdev->flags & RADEON_IS_IGP) {
+		return supported_devices_connector_object_id_convert
+			[connector_type];
+	} else if (((connector_type == DRM_MODE_CONNECTOR_DVII) ||
+		    (connector_type == DRM_MODE_CONNECTOR_DVID)) &&
+		   (devices & ATOM_DEVICE_DFP2_SUPPORT))  {
+		struct radeon_mode_info *mode_info = &rdev->mode_info;
+		struct atom_context *ctx = mode_info->atom_context;
+		int index = GetIndexIntoMasterTable(DATA, XTMDS_Info);
+		uint16_t size, data_offset;
+		uint8_t frev, crev;
+		ATOM_XTMDS_INFO *xtmds;
+
+		atom_parse_data_header(ctx, index, &size, &frev, &crev, &data_offset);
+		xtmds = (ATOM_XTMDS_INFO *)(ctx->bios + data_offset);
+
+		if (xtmds->ucSupportedLink & ATOM_XTMDS_SUPPORTED_DUALLINK) {
+			if (connector_type == DRM_MODE_CONNECTOR_DVII)
+				return CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_I;
+			else
+				return CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_D;
+		} else {
+			if (connector_type == DRM_MODE_CONNECTOR_DVII)
+				return CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_I;
+			else
+				return CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_D;
+		}
+	} else {
+		return supported_devices_connector_object_id_convert
+			[connector_type];
+	}
+}
+
 struct bios_connector {
 	bool valid;
 	uint16_t line_mux;
 	uint16_t devices;
 	int connector_type;
 	struct radeon_i2c_bus_rec ddc_bus;
+	struct radeon_hpd hpd;
 };
 
 bool radeon_get_atom_connector_info_from_supported_devices_table(struct
@@ -459,7 +669,7 @@ bool radeon_get_atom_connector_info_from_supported_devices_table(struct
 	uint16_t device_support;
 	uint8_t dac;
 	union atom_supported_devices *supported_devices;
-	int i, j;
+	int i, j, max_device;
 	struct bios_connector bios_connectors[ATOM_MAX_SUPPORTED_DEVICE];
 
 	atom_parse_data_header(ctx, index, &size, &frev, &crev, &data_offset);
@@ -469,7 +679,12 @@ bool radeon_get_atom_connector_info_from_supported_devices_table(struct
 
 	device_support = le16_to_cpu(supported_devices->info.usDeviceSupport);
 
-	for (i = 0; i < ATOM_MAX_SUPPORTED_DEVICE; i++) {
+	if (frev > 1)
+		max_device = ATOM_MAX_SUPPORTED_DEVICE;
+	else
+		max_device = ATOM_MAX_SUPPORTED_DEVICE_INFO;
+
+	for (i = 0; i < max_device; i++) {
 		ATOM_CONNECTOR_INFO_I2C ci =
 		    supported_devices->info.asConnInfo[i];
 
@@ -495,22 +710,8 @@ bool radeon_get_atom_connector_info_from_supported_devices_table(struct
 
 		dac = ci.sucConnectorInfo.sbfAccess.bfAssociatedDAC;
 
-		if ((rdev->family == CHIP_RS690) ||
-		    (rdev->family == CHIP_RS740)) {
-			if ((i == ATOM_DEVICE_DFP2_INDEX)
-			    && (ci.sucI2cId.sbfAccess.bfI2C_LineMux == 2))
-				bios_connectors[i].line_mux =
-				    ci.sucI2cId.sbfAccess.bfI2C_LineMux + 1;
-			else if ((i == ATOM_DEVICE_DFP3_INDEX)
-				 && (ci.sucI2cId.sbfAccess.bfI2C_LineMux == 1))
-				bios_connectors[i].line_mux =
-				    ci.sucI2cId.sbfAccess.bfI2C_LineMux + 1;
-			else
-				bios_connectors[i].line_mux =
-				    ci.sucI2cId.sbfAccess.bfI2C_LineMux;
-		} else
-			bios_connectors[i].line_mux =
-			    ci.sucI2cId.sbfAccess.bfI2C_LineMux;
+		bios_connectors[i].line_mux =
+			ci.sucI2cId.ucAccess;
 
 		/* give tv unique connector ids */
 		if (i == ATOM_DEVICE_TV1_INDEX) {
@@ -524,8 +725,30 @@ bool radeon_get_atom_connector_info_from_supported_devices_table(struct
 			bios_connectors[i].line_mux = 52;
 		} else
 			bios_connectors[i].ddc_bus =
-			    radeon_lookup_gpio(dev,
-					       bios_connectors[i].line_mux);
+			    radeon_lookup_i2c_gpio(rdev,
+						   bios_connectors[i].line_mux);
+
+		if ((crev > 1) && (frev > 1)) {
+			u8 isb = supported_devices->info_2d1.asIntSrcInfo[i].ucIntSrcBitmap;
+			switch (isb) {
+			case 0x4:
+				bios_connectors[i].hpd.hpd = RADEON_HPD_1;
+				break;
+			case 0xa:
+				bios_connectors[i].hpd.hpd = RADEON_HPD_2;
+				break;
+			default:
+				bios_connectors[i].hpd.hpd = RADEON_HPD_NONE;
+				break;
+			}
+		} else {
+			if (i == ATOM_DEVICE_DFP1_INDEX)
+				bios_connectors[i].hpd.hpd = RADEON_HPD_1;
+			else if (i == ATOM_DEVICE_DFP2_INDEX)
+				bios_connectors[i].hpd.hpd = RADEON_HPD_2;
+			else
+				bios_connectors[i].hpd.hpd = RADEON_HPD_NONE;
+		}
 
 		/* Always set the connector type to VGA for CRT1/CRT2. if they are
 		 * shared with a DVI port, we'll pick up the DVI connector when we
@@ -537,7 +760,8 @@ bool radeon_get_atom_connector_info_from_supported_devices_table(struct
 
 		if (!radeon_atom_apply_quirks
 		    (dev, (1 << i), &bios_connectors[i].connector_type,
-		     &bios_connectors[i].ddc_bus, &bios_connectors[i].line_mux))
+		     &bios_connectors[i].ddc_bus, &bios_connectors[i].line_mux,
+		     &bios_connectors[i].hpd))
 			continue;
 
 		bios_connectors[i].valid = true;
@@ -552,41 +776,42 @@ bool radeon_get_atom_connector_info_from_supported_devices_table(struct
 		else
 			radeon_add_legacy_encoder(dev,
 						  radeon_get_encoder_id(dev,
-									(1 <<
-									 i),
+									(1 << i),
 									dac),
 						  (1 << i));
 	}
 
 	/* combine shared connectors */
-	for (i = 0; i < ATOM_MAX_SUPPORTED_DEVICE; i++) {
+	for (i = 0; i < max_device; i++) {
 		if (bios_connectors[i].valid) {
-			for (j = 0; j < ATOM_MAX_SUPPORTED_DEVICE; j++) {
+			for (j = 0; j < max_device; j++) {
 				if (bios_connectors[j].valid && (i != j)) {
 					if (bios_connectors[i].line_mux ==
 					    bios_connectors[j].line_mux) {
-						if (((bios_connectors[i].
-						      devices &
-						      (ATOM_DEVICE_DFP_SUPPORT))
-						     && (bios_connectors[j].
-							 devices &
-							 (ATOM_DEVICE_CRT_SUPPORT)))
-						    ||
-						    ((bios_connectors[j].
-						      devices &
-						      (ATOM_DEVICE_DFP_SUPPORT))
-						     && (bios_connectors[i].
-							 devices &
-							 (ATOM_DEVICE_CRT_SUPPORT)))) {
-							bios_connectors[i].
-							    devices |=
-							    bios_connectors[j].
-							    devices;
-							bios_connectors[i].
-							    connector_type =
-							    DRM_MODE_CONNECTOR_DVII;
-							bios_connectors[j].
-							    valid = false;
+						/* make sure not to combine LVDS */
+						if (bios_connectors[i].devices & (ATOM_DEVICE_LCD_SUPPORT)) {
+							bios_connectors[i].line_mux = 53;
+							bios_connectors[i].ddc_bus.valid = false;
+							continue;
+						}
+						if (bios_connectors[j].devices & (ATOM_DEVICE_LCD_SUPPORT)) {
+							bios_connectors[j].line_mux = 53;
+							bios_connectors[j].ddc_bus.valid = false;
+							continue;
+						}
+						/* combine analog and digital for DVI-I */
+						if (((bios_connectors[i].devices & (ATOM_DEVICE_DFP_SUPPORT)) &&
+						     (bios_connectors[j].devices & (ATOM_DEVICE_CRT_SUPPORT))) ||
+						    ((bios_connectors[j].devices & (ATOM_DEVICE_DFP_SUPPORT)) &&
+						     (bios_connectors[i].devices & (ATOM_DEVICE_CRT_SUPPORT)))) {
+							bios_connectors[i].devices |=
+								bios_connectors[j].devices;
+							bios_connectors[i].connector_type =
+								DRM_MODE_CONNECTOR_DVII;
+							if (bios_connectors[j].devices & (ATOM_DEVICE_DFP_SUPPORT))
+								bios_connectors[i].hpd =
+									bios_connectors[j].hpd;
+							bios_connectors[j].valid = false;
 						}
 					}
 				}
@@ -595,15 +820,22 @@ bool radeon_get_atom_connector_info_from_supported_devices_table(struct
 	}
 
 	/* add the connectors */
-	for (i = 0; i < ATOM_MAX_SUPPORTED_DEVICE; i++) {
-		if (bios_connectors[i].valid)
+	for (i = 0; i < max_device; i++) {
+		if (bios_connectors[i].valid) {
+			uint16_t connector_object_id =
+				atombios_get_connector_object_id(dev,
+						      bios_connectors[i].connector_type,
+						      bios_connectors[i].devices);
 			radeon_add_atom_connector(dev,
 						  bios_connectors[i].line_mux,
 						  bios_connectors[i].devices,
 						  bios_connectors[i].
 						  connector_type,
 						  &bios_connectors[i].ddc_bus,
-						  false, 0);
+						  false, 0,
+						  connector_object_id,
+						  &bios_connectors[i].hpd);
+		}
 	}
 
 	radeon_link_encoder_connector(dev);
@@ -616,6 +848,7 @@ union firmware_info {
 	ATOM_FIRMWARE_INFO_V1_2 info_12;
 	ATOM_FIRMWARE_INFO_V1_3 info_13;
 	ATOM_FIRMWARE_INFO_V1_4 info_14;
+	ATOM_FIRMWARE_INFO_V2_1 info_21;
 };
 
 bool radeon_atom_get_clock_info(struct drm_device *dev)
@@ -627,6 +860,7 @@ bool radeon_atom_get_clock_info(struct drm_device *dev)
 	uint8_t frev, crev;
 	struct radeon_pll *p1pll = &rdev->clock.p1pll;
 	struct radeon_pll *p2pll = &rdev->clock.p2pll;
+	struct radeon_pll *dcpll = &rdev->clock.dcpll;
 	struct radeon_pll *spll = &rdev->clock.spll;
 	struct radeon_pll *mpll = &rdev->clock.mpll;
 	uint16_t data_offset;
@@ -644,8 +878,12 @@ bool radeon_atom_get_clock_info(struct drm_device *dev)
 		    le16_to_cpu(firmware_info->info.usReferenceClock);
 		p1pll->reference_div = 0;
 
-		p1pll->pll_out_min =
-		    le16_to_cpu(firmware_info->info.usMinPixelClockPLL_Output);
+		if (crev < 2)
+			p1pll->pll_out_min =
+				le16_to_cpu(firmware_info->info.usMinPixelClockPLL_Output);
+		else
+			p1pll->pll_out_min =
+				le32_to_cpu(firmware_info->info_12.ulMinPixelClockPLL_Output);
 		p1pll->pll_out_max =
 		    le32_to_cpu(firmware_info->info.ulMaxPixelClockPLL_Output);
 
@@ -654,6 +892,17 @@ bool radeon_atom_get_clock_info(struct drm_device *dev)
 				p1pll->pll_out_min = 64800;
 			else
 				p1pll->pll_out_min = 20000;
+		} else if (p1pll->pll_out_min > 64800) {
+			/* Limiting the pll output range is a good thing generally as
+			 * it limits the number of possible pll combinations for a given
+			 * frequency presumably to the ones that work best on each card.
+			 * However, certain duallink DVI monitors seem to like
+			 * pll combinations that would be limited by this at least on
+			 * pre-DCE 3.0 r6xx hardware.  This might need to be adjusted per
+			 * family.
+			 */
+			if (!radeon_new_pll)
+				p1pll->pll_out_min = 64800;
 		}
 
 		p1pll->pll_in_min =
@@ -714,8 +963,56 @@ bool radeon_atom_get_clock_info(struct drm_device *dev)
 		rdev->clock.default_mclk =
 		    le32_to_cpu(firmware_info->info.ulDefaultMemoryClock);
 
+		if (ASIC_IS_DCE4(rdev)) {
+			rdev->clock.default_dispclk =
+				le32_to_cpu(firmware_info->info_21.ulDefaultDispEngineClkFreq);
+			if (rdev->clock.default_dispclk == 0)
+				rdev->clock.default_dispclk = 60000; /* 600 Mhz */
+			rdev->clock.dp_extclk =
+				le16_to_cpu(firmware_info->info_21.usUniphyDPModeExtClkFreq);
+		}
+		*dcpll = *p1pll;
+
 		return true;
 	}
+
+	return false;
+}
+
+union igp_info {
+	struct _ATOM_INTEGRATED_SYSTEM_INFO info;
+	struct _ATOM_INTEGRATED_SYSTEM_INFO_V2 info_2;
+};
+
+bool radeon_atombios_sideport_present(struct radeon_device *rdev)
+{
+	struct radeon_mode_info *mode_info = &rdev->mode_info;
+	int index = GetIndexIntoMasterTable(DATA, IntegratedSystemInfo);
+	union igp_info *igp_info;
+	u8 frev, crev;
+	u16 data_offset;
+
+	atom_parse_data_header(mode_info->atom_context, index, NULL, &frev,
+			       &crev, &data_offset);
+
+	igp_info = (union igp_info *)(mode_info->atom_context->bios +
+				      data_offset);
+
+	if (igp_info) {
+		switch (crev) {
+		case 1:
+			if (igp_info->info.ucMemoryType & 0xf0)
+				return true;
+			break;
+		case 2:
+			if (igp_info->info_2.ucMemoryType & 0x0f)
+				return true;
+			break;
+		default:
+			DRM_ERROR("Unsupported IGP table: %d %d\n", frev, crev);
+			break;
+		}
+	}
 	return false;
 }
 
@@ -770,6 +1067,77 @@ bool radeon_atombios_get_tmds_info(struct radeon_encoder *encoder,
 	return false;
 }
 
+static struct radeon_atom_ss *radeon_atombios_get_ss_info(struct
+							  radeon_encoder
+							  *encoder,
+							  int id)
+{
+	struct drm_device *dev = encoder->base.dev;
+	struct radeon_device *rdev = dev->dev_private;
+	struct radeon_mode_info *mode_info = &rdev->mode_info;
+	int index = GetIndexIntoMasterTable(DATA, PPLL_SS_Info);
+	uint16_t data_offset;
+	struct _ATOM_SPREAD_SPECTRUM_INFO *ss_info;
+	uint8_t frev, crev;
+	struct radeon_atom_ss *ss = NULL;
+	int i;
+
+	if (id > ATOM_MAX_SS_ENTRY)
+		return NULL;
+
+	atom_parse_data_header(mode_info->atom_context, index, NULL, &frev,
+			       &crev, &data_offset);
+
+	ss_info =
+	    (struct _ATOM_SPREAD_SPECTRUM_INFO *)(mode_info->atom_context->bios + data_offset);
+
+	if (ss_info) {
+		ss =
+		    kzalloc(sizeof(struct radeon_atom_ss), GFP_KERNEL);
+
+		if (!ss)
+			return NULL;
+
+		for (i = 0; i < ATOM_MAX_SS_ENTRY; i++) {
+			if (ss_info->asSS_Info[i].ucSS_Id == id) {
+				ss->percentage =
+					le16_to_cpu(ss_info->asSS_Info[i].usSpreadSpectrumPercentage);
+				ss->type = ss_info->asSS_Info[i].ucSpreadSpectrumType;
+				ss->step = ss_info->asSS_Info[i].ucSS_Step;
+				ss->delay = ss_info->asSS_Info[i].ucSS_Delay;
+				ss->range = ss_info->asSS_Info[i].ucSS_Range;
+				ss->refdiv = ss_info->asSS_Info[i].ucRecommendedRef_Div;
+				break;
+			}
+		}
+	}
+	return ss;
+}
+
+static void radeon_atom_apply_lvds_quirks(struct drm_device *dev,
+					  struct radeon_encoder_atom_dig *lvds)
+{
+
+	/* Toshiba A300-1BU laptop panel doesn't like new pll divider algo */
+	if ((dev->pdev->device == 0x95c4) &&
+	    (dev->pdev->subsystem_vendor == 0x1179) &&
+	    (dev->pdev->subsystem_device == 0xff50)) {
+		if ((lvds->native_mode.hdisplay == 1280) &&
+		    (lvds->native_mode.vdisplay == 800))
+			lvds->pll_algo = PLL_ALGO_LEGACY;
+	}
+
+	/* Dell Studio 15 laptop panel doesn't like new pll divider algo */
+	if ((dev->pdev->device == 0x95c4) &&
+	    (dev->pdev->subsystem_vendor == 0x1028) &&
+	    (dev->pdev->subsystem_device == 0x029f)) {
+		if ((lvds->native_mode.hdisplay == 1280) &&
+		    (lvds->native_mode.vdisplay == 800))
+			lvds->pll_algo = PLL_ALGO_LEGACY;
+	}
+
+}
+
 union lvds_info {
 	struct _ATOM_LVDS_INFO info;
 	struct _ATOM_LVDS_INFO_V12 info_12;
@@ -783,7 +1151,7 @@ struct radeon_encoder_atom_dig *radeon_atombios_get_lvds_info(struct
 	struct radeon_device *rdev = dev->dev_private;
 	struct radeon_mode_info *mode_info = &rdev->mode_info;
 	int index = GetIndexIntoMasterTable(DATA, LVDS_Info);
-	uint16_t data_offset;
+	uint16_t data_offset, misc;
 	union lvds_info *lvds_info;
 	uint8_t frev, crev;
 	struct radeon_encoder_atom_dig *lvds = NULL;
@@ -801,28 +1169,60 @@ struct radeon_encoder_atom_dig *radeon_atombios_get_lvds_info(struct
 		if (!lvds)
 			return NULL;
 
-		lvds->native_mode.dotclock =
+		lvds->native_mode.clock =
 		    le16_to_cpu(lvds_info->info.sLCDTiming.usPixClk) * 10;
-		lvds->native_mode.panel_xres =
+		lvds->native_mode.hdisplay =
 		    le16_to_cpu(lvds_info->info.sLCDTiming.usHActive);
-		lvds->native_mode.panel_yres =
+		lvds->native_mode.vdisplay =
 		    le16_to_cpu(lvds_info->info.sLCDTiming.usVActive);
-		lvds->native_mode.hblank =
-		    le16_to_cpu(lvds_info->info.sLCDTiming.usHBlanking_Time);
-		lvds->native_mode.hoverplus =
-		    le16_to_cpu(lvds_info->info.sLCDTiming.usHSyncOffset);
-		lvds->native_mode.hsync_width =
-		    le16_to_cpu(lvds_info->info.sLCDTiming.usHSyncWidth);
-		lvds->native_mode.vblank =
-		    le16_to_cpu(lvds_info->info.sLCDTiming.usVBlanking_Time);
-		lvds->native_mode.voverplus =
-		    le16_to_cpu(lvds_info->info.sLCDTiming.usVSyncOffset);
-		lvds->native_mode.vsync_width =
-		    le16_to_cpu(lvds_info->info.sLCDTiming.usVSyncWidth);
+		lvds->native_mode.htotal = lvds->native_mode.hdisplay +
+			le16_to_cpu(lvds_info->info.sLCDTiming.usHBlanking_Time);
+		lvds->native_mode.hsync_start = lvds->native_mode.hdisplay +
+			le16_to_cpu(lvds_info->info.sLCDTiming.usHSyncOffset);
+		lvds->native_mode.hsync_end = lvds->native_mode.hsync_start +
+			le16_to_cpu(lvds_info->info.sLCDTiming.usHSyncWidth);
+		lvds->native_mode.vtotal = lvds->native_mode.vdisplay +
+			le16_to_cpu(lvds_info->info.sLCDTiming.usVBlanking_Time);
+		lvds->native_mode.vsync_start = lvds->native_mode.vdisplay +
+			le16_to_cpu(lvds_info->info.sLCDTiming.usVSyncWidth);
+		lvds->native_mode.vsync_end = lvds->native_mode.vsync_start +
+			le16_to_cpu(lvds_info->info.sLCDTiming.usVSyncWidth);
 		lvds->panel_pwr_delay =
 		    le16_to_cpu(lvds_info->info.usOffDelayInMs);
 		lvds->lvds_misc = lvds_info->info.ucLVDS_Misc;
 
+		misc = le16_to_cpu(lvds_info->info.sLCDTiming.susModeMiscInfo.usAccess);
+		if (misc & ATOM_VSYNC_POLARITY)
+			lvds->native_mode.flags |= DRM_MODE_FLAG_NVSYNC;
+		if (misc & ATOM_HSYNC_POLARITY)
+			lvds->native_mode.flags |= DRM_MODE_FLAG_NHSYNC;
+		if (misc & ATOM_COMPOSITESYNC)
+			lvds->native_mode.flags |= DRM_MODE_FLAG_CSYNC;
+		if (misc & ATOM_INTERLACE)
+			lvds->native_mode.flags |= DRM_MODE_FLAG_INTERLACE;
+		if (misc & ATOM_DOUBLE_CLOCK_MODE)
+			lvds->native_mode.flags |= DRM_MODE_FLAG_DBLSCAN;
+
+		/* set crtc values */
+		drm_mode_set_crtcinfo(&lvds->native_mode, CRTC_INTERLACE_HALVE_V);
+
+		lvds->ss = radeon_atombios_get_ss_info(encoder, lvds_info->info.ucSS_Id);
+
+		if (ASIC_IS_AVIVO(rdev)) {
+			if (radeon_new_pll == 0)
+				lvds->pll_algo = PLL_ALGO_LEGACY;
+			else
+				lvds->pll_algo = PLL_ALGO_NEW;
+		} else {
+			if (radeon_new_pll == 1)
+				lvds->pll_algo = PLL_ALGO_NEW;
+			else
+				lvds->pll_algo = PLL_ALGO_LEGACY;
+		}
+
+		/* LVDS quirks */
+		radeon_atom_apply_lvds_quirks(dev, lvds);
+
 		encoder->native_mode = lvds->native_mode;
 	}
 	return lvds;
@@ -860,8 +1260,7 @@ radeon_atombios_get_primary_dac_info(struct radeon_encoder *encoder)
 }
 
 bool radeon_atom_get_tv_timings(struct radeon_device *rdev, int index,
-				SET_CRTC_TIMING_PARAMETERS_PS_ALLOCATION *crtc_timing,
-				int32_t *pixel_clock)
+				struct drm_display_mode *mode)
 {
 	struct radeon_mode_info *mode_info = &rdev->mode_info;
 	ATOM_ANALOG_TV_INFO *tv_info;
@@ -869,7 +1268,7 @@ bool radeon_atom_get_tv_timings(struct radeon_device *rdev, int index,
 	ATOM_DTD_FORMAT *dtd_timings;
 	int data_index = GetIndexIntoMasterTable(DATA, AnalogTV_Info);
 	u8 frev, crev;
-	uint16_t data_offset;
+	u16 data_offset, misc;
 
 	atom_parse_data_header(mode_info->atom_context, data_index, NULL, &frev, &crev, &data_offset);
 
@@ -879,28 +1278,37 @@ bool radeon_atom_get_tv_timings(struct radeon_device *rdev, int index,
 		if (index > MAX_SUPPORTED_TV_TIMING)
 			return false;
 
-		crtc_timing->usH_Total = le16_to_cpu(tv_info->aModeTimings[index].usCRTC_H_Total);
-		crtc_timing->usH_Disp = le16_to_cpu(tv_info->aModeTimings[index].usCRTC_H_Disp);
-		crtc_timing->usH_SyncStart = le16_to_cpu(tv_info->aModeTimings[index].usCRTC_H_SyncStart);
-		crtc_timing->usH_SyncWidth = le16_to_cpu(tv_info->aModeTimings[index].usCRTC_H_SyncWidth);
-
-		crtc_timing->usV_Total = le16_to_cpu(tv_info->aModeTimings[index].usCRTC_V_Total);
-		crtc_timing->usV_Disp = le16_to_cpu(tv_info->aModeTimings[index].usCRTC_V_Disp);
-		crtc_timing->usV_SyncStart = le16_to_cpu(tv_info->aModeTimings[index].usCRTC_V_SyncStart);
-		crtc_timing->usV_SyncWidth = le16_to_cpu(tv_info->aModeTimings[index].usCRTC_V_SyncWidth);
-
-		crtc_timing->susModeMiscInfo = tv_info->aModeTimings[index].susModeMiscInfo;
-
-		crtc_timing->ucOverscanRight = le16_to_cpu(tv_info->aModeTimings[index].usCRTC_OverscanRight);
-		crtc_timing->ucOverscanLeft = le16_to_cpu(tv_info->aModeTimings[index].usCRTC_OverscanLeft);
-		crtc_timing->ucOverscanBottom = le16_to_cpu(tv_info->aModeTimings[index].usCRTC_OverscanBottom);
-		crtc_timing->ucOverscanTop = le16_to_cpu(tv_info->aModeTimings[index].usCRTC_OverscanTop);
-		*pixel_clock = le16_to_cpu(tv_info->aModeTimings[index].usPixelClock) * 10;
+		mode->crtc_htotal = le16_to_cpu(tv_info->aModeTimings[index].usCRTC_H_Total);
+		mode->crtc_hdisplay = le16_to_cpu(tv_info->aModeTimings[index].usCRTC_H_Disp);
+		mode->crtc_hsync_start = le16_to_cpu(tv_info->aModeTimings[index].usCRTC_H_SyncStart);
+		mode->crtc_hsync_end = le16_to_cpu(tv_info->aModeTimings[index].usCRTC_H_SyncStart) +
+			le16_to_cpu(tv_info->aModeTimings[index].usCRTC_H_SyncWidth);
+
+		mode->crtc_vtotal = le16_to_cpu(tv_info->aModeTimings[index].usCRTC_V_Total);
+		mode->crtc_vdisplay = le16_to_cpu(tv_info->aModeTimings[index].usCRTC_V_Disp);
+		mode->crtc_vsync_start = le16_to_cpu(tv_info->aModeTimings[index].usCRTC_V_SyncStart);
+		mode->crtc_vsync_end = le16_to_cpu(tv_info->aModeTimings[index].usCRTC_V_SyncStart) +
+			le16_to_cpu(tv_info->aModeTimings[index].usCRTC_V_SyncWidth);
+
+		mode->flags = 0;
+		misc = le16_to_cpu(tv_info->aModeTimings[index].susModeMiscInfo.usAccess);
+		if (misc & ATOM_VSYNC_POLARITY)
+			mode->flags |= DRM_MODE_FLAG_NVSYNC;
+		if (misc & ATOM_HSYNC_POLARITY)
+			mode->flags |= DRM_MODE_FLAG_NHSYNC;
+		if (misc & ATOM_COMPOSITESYNC)
+			mode->flags |= DRM_MODE_FLAG_CSYNC;
+		if (misc & ATOM_INTERLACE)
+			mode->flags |= DRM_MODE_FLAG_INTERLACE;
+		if (misc & ATOM_DOUBLE_CLOCK_MODE)
+			mode->flags |= DRM_MODE_FLAG_DBLSCAN;
+
+		mode->clock = le16_to_cpu(tv_info->aModeTimings[index].usPixelClock) * 10;
 
 		if (index == 1) {
 			/* PAL timings appear to have wrong values for totals */
-			crtc_timing->usH_Total -= 1;
-			crtc_timing->usV_Total -= 1;
+			mode->crtc_htotal -= 1;
+			mode->crtc_vtotal -= 1;
 		}
 		break;
 	case 2:
@@ -909,22 +1317,96 @@ bool radeon_atom_get_tv_timings(struct radeon_device *rdev, int index,
 			return false;
 
 		dtd_timings = &tv_info_v1_2->aModeTimings[index];
-		crtc_timing->usH_Total = le16_to_cpu(dtd_timings->usHActive) + le16_to_cpu(dtd_timings->usHBlanking_Time);
-		crtc_timing->usH_Disp = le16_to_cpu(dtd_timings->usHActive);
-		crtc_timing->usH_SyncStart = le16_to_cpu(dtd_timings->usHActive) + le16_to_cpu(dtd_timings->usHSyncOffset);
-		crtc_timing->usH_SyncWidth = le16_to_cpu(dtd_timings->usHSyncWidth);
-		crtc_timing->usV_Total = le16_to_cpu(dtd_timings->usVActive) + le16_to_cpu(dtd_timings->usVBlanking_Time);
-		crtc_timing->usV_Disp = le16_to_cpu(dtd_timings->usVActive);
-		crtc_timing->usV_SyncStart = le16_to_cpu(dtd_timings->usVActive) + le16_to_cpu(dtd_timings->usVSyncOffset);
-		crtc_timing->usV_SyncWidth = le16_to_cpu(dtd_timings->usVSyncWidth);
-
-		crtc_timing->susModeMiscInfo.usAccess = le16_to_cpu(dtd_timings->susModeMiscInfo.usAccess);
-		*pixel_clock = le16_to_cpu(dtd_timings->usPixClk) * 10;
+		mode->crtc_htotal = le16_to_cpu(dtd_timings->usHActive) +
+			le16_to_cpu(dtd_timings->usHBlanking_Time);
+		mode->crtc_hdisplay = le16_to_cpu(dtd_timings->usHActive);
+		mode->crtc_hsync_start = le16_to_cpu(dtd_timings->usHActive) +
+			le16_to_cpu(dtd_timings->usHSyncOffset);
+		mode->crtc_hsync_end = mode->crtc_hsync_start +
+			le16_to_cpu(dtd_timings->usHSyncWidth);
+
+		mode->crtc_vtotal = le16_to_cpu(dtd_timings->usVActive) +
+			le16_to_cpu(dtd_timings->usVBlanking_Time);
+		mode->crtc_vdisplay = le16_to_cpu(dtd_timings->usVActive);
+		mode->crtc_vsync_start = le16_to_cpu(dtd_timings->usVActive) +
+			le16_to_cpu(dtd_timings->usVSyncOffset);
+		mode->crtc_vsync_end = mode->crtc_vsync_start +
+			le16_to_cpu(dtd_timings->usVSyncWidth);
+
+		mode->flags = 0;
+		misc = le16_to_cpu(dtd_timings->susModeMiscInfo.usAccess);
+		if (misc & ATOM_VSYNC_POLARITY)
+			mode->flags |= DRM_MODE_FLAG_NVSYNC;
+		if (misc & ATOM_HSYNC_POLARITY)
+			mode->flags |= DRM_MODE_FLAG_NHSYNC;
+		if (misc & ATOM_COMPOSITESYNC)
+			mode->flags |= DRM_MODE_FLAG_CSYNC;
+		if (misc & ATOM_INTERLACE)
+			mode->flags |= DRM_MODE_FLAG_INTERLACE;
+		if (misc & ATOM_DOUBLE_CLOCK_MODE)
+			mode->flags |= DRM_MODE_FLAG_DBLSCAN;
+
+		mode->clock = le16_to_cpu(dtd_timings->usPixClk) * 10;
 		break;
 	}
 	return true;
 }
 
+enum radeon_tv_std
+radeon_atombios_get_tv_info(struct radeon_device *rdev)
+{
+	struct radeon_mode_info *mode_info = &rdev->mode_info;
+	int index = GetIndexIntoMasterTable(DATA, AnalogTV_Info);
+	uint16_t data_offset;
+	uint8_t frev, crev;
+	struct _ATOM_ANALOG_TV_INFO *tv_info;
+	enum radeon_tv_std tv_std = TV_STD_NTSC;
+
+	atom_parse_data_header(mode_info->atom_context, index, NULL, &frev, &crev, &data_offset);
+
+	tv_info = (struct _ATOM_ANALOG_TV_INFO *)(mode_info->atom_context->bios + data_offset);
+
+	switch (tv_info->ucTV_BootUpDefaultStandard) {
+	case ATOM_TV_NTSC:
+		tv_std = TV_STD_NTSC;
+		DRM_INFO("Default TV standard: NTSC\n");
+		break;
+	case ATOM_TV_NTSCJ:
+		tv_std = TV_STD_NTSC_J;
+		DRM_INFO("Default TV standard: NTSC-J\n");
+		break;
+	case ATOM_TV_PAL:
+		tv_std = TV_STD_PAL;
+		DRM_INFO("Default TV standard: PAL\n");
+		break;
+	case ATOM_TV_PALM:
+		tv_std = TV_STD_PAL_M;
+		DRM_INFO("Default TV standard: PAL-M\n");
+		break;
+	case ATOM_TV_PALN:
+		tv_std = TV_STD_PAL_N;
+		DRM_INFO("Default TV standard: PAL-N\n");
+		break;
+	case ATOM_TV_PALCN:
+		tv_std = TV_STD_PAL_CN;
+		DRM_INFO("Default TV standard: PAL-CN\n");
+		break;
+	case ATOM_TV_PAL60:
+		tv_std = TV_STD_PAL_60;
+		DRM_INFO("Default TV standard: PAL-60\n");
+		break;
+	case ATOM_TV_SECAM:
+		tv_std = TV_STD_SECAM;
+		DRM_INFO("Default TV standard: SECAM\n");
+		break;
+	default:
+		tv_std = TV_STD_NTSC;
+		DRM_INFO("Unknown TV standard; defaulting to NTSC\n");
+		break;
+	}
+	return tv_std;
+}
+
 struct radeon_encoder_tv_dac *
 radeon_atombios_get_tv_dac_info(struct radeon_encoder *encoder)
 {
@@ -960,10 +1442,376 @@ radeon_atombios_get_tv_dac_info(struct radeon_encoder *encoder)
 		dac = dac_info->ucDAC2_NTSC_DAC_Adjustment;
 		tv_dac->ntsc_tvdac_adj = (bg << 16) | (dac << 20);
 
+		tv_dac->tv_std = radeon_atombios_get_tv_info(rdev);
 	}
 	return tv_dac;
 }
 
+union power_info {
+	struct _ATOM_POWERPLAY_INFO info;
+	struct _ATOM_POWERPLAY_INFO_V2 info_2;
+	struct _ATOM_POWERPLAY_INFO_V3 info_3;
+	struct _ATOM_PPLIB_POWERPLAYTABLE info_4;
+};
+
+void radeon_atombios_get_power_modes(struct radeon_device *rdev)
+{
+	struct radeon_mode_info *mode_info = &rdev->mode_info;
+	int index = GetIndexIntoMasterTable(DATA, PowerPlayInfo);
+	u16 data_offset;
+	u8 frev, crev;
+	u32 misc, misc2 = 0, sclk, mclk;
+	union power_info *power_info;
+	struct _ATOM_PPLIB_NONCLOCK_INFO *non_clock_info;
+	struct _ATOM_PPLIB_STATE *power_state;
+	int num_modes = 0, i, j;
+	int state_index = 0, mode_index = 0;
+
+	atom_parse_data_header(mode_info->atom_context, index, NULL, &frev, &crev, &data_offset);
+
+	power_info = (union power_info *)(mode_info->atom_context->bios + data_offset);
+
+	rdev->pm.default_power_state = NULL;
+
+	if (power_info) {
+		if (frev < 4) {
+			num_modes = power_info->info.ucNumOfPowerModeEntries;
+			if (num_modes > ATOM_MAX_NUMBEROF_POWER_BLOCK)
+				num_modes = ATOM_MAX_NUMBEROF_POWER_BLOCK;
+			for (i = 0; i < num_modes; i++) {
+				rdev->pm.power_state[state_index].clock_info[0].voltage.type = VOLTAGE_NONE;
+				switch (frev) {
+				case 1:
+					rdev->pm.power_state[state_index].num_clock_modes = 1;
+					rdev->pm.power_state[state_index].clock_info[0].mclk =
+						le16_to_cpu(power_info->info.asPowerPlayInfo[i].usMemoryClock);
+					rdev->pm.power_state[state_index].clock_info[0].sclk =
+						le16_to_cpu(power_info->info.asPowerPlayInfo[i].usEngineClock);
+					/* skip invalid modes */
+					if ((rdev->pm.power_state[state_index].clock_info[0].mclk == 0) ||
+					    (rdev->pm.power_state[state_index].clock_info[0].sclk == 0))
+						continue;
+					/* skip overclock modes for now */
+					if ((rdev->pm.power_state[state_index].clock_info[0].mclk >
+					     rdev->clock.default_mclk + RADEON_MODE_OVERCLOCK_MARGIN) ||
+					    (rdev->pm.power_state[state_index].clock_info[0].sclk >
+					     rdev->clock.default_sclk + RADEON_MODE_OVERCLOCK_MARGIN))
+						continue;
+					rdev->pm.power_state[state_index].non_clock_info.pcie_lanes =
+						power_info->info.asPowerPlayInfo[i].ucNumPciELanes;
+					misc = le32_to_cpu(power_info->info.asPowerPlayInfo[i].ulMiscInfo);
+					if (misc & ATOM_PM_MISCINFO_VOLTAGE_DROP_SUPPORT) {
+						rdev->pm.power_state[state_index].clock_info[0].voltage.type =
+							VOLTAGE_GPIO;
+						rdev->pm.power_state[state_index].clock_info[0].voltage.gpio =
+							radeon_lookup_gpio(rdev,
+							power_info->info.asPowerPlayInfo[i].ucVoltageDropIndex);
+						if (misc & ATOM_PM_MISCINFO_VOLTAGE_DROP_ACTIVE_HIGH)
+							rdev->pm.power_state[state_index].clock_info[0].voltage.active_high =
+								true;
+						else
+							rdev->pm.power_state[state_index].clock_info[0].voltage.active_high =
+								false;
+					} else if (misc & ATOM_PM_MISCINFO_PROGRAM_VOLTAGE) {
+						rdev->pm.power_state[state_index].clock_info[0].voltage.type =
+							VOLTAGE_VDDC;
+						rdev->pm.power_state[state_index].clock_info[0].voltage.vddc_id =
+							power_info->info.asPowerPlayInfo[i].ucVoltageDropIndex;
+					}
+					/* order matters! */
+					if (misc & ATOM_PM_MISCINFO_POWER_SAVING_MODE)
+						rdev->pm.power_state[state_index].type =
+							POWER_STATE_TYPE_POWERSAVE;
+					if (misc & ATOM_PM_MISCINFO_DEFAULT_DC_STATE_ENTRY_TRUE)
+						rdev->pm.power_state[state_index].type =
+							POWER_STATE_TYPE_BATTERY;
+					if (misc & ATOM_PM_MISCINFO_DEFAULT_LOW_DC_STATE_ENTRY_TRUE)
+						rdev->pm.power_state[state_index].type =
+							POWER_STATE_TYPE_BATTERY;
+					if (misc & ATOM_PM_MISCINFO_LOAD_BALANCE_EN)
+						rdev->pm.power_state[state_index].type =
+							POWER_STATE_TYPE_BALANCED;
+					if (misc & ATOM_PM_MISCINFO_3D_ACCELERATION_EN)
+						rdev->pm.power_state[state_index].type =
+							POWER_STATE_TYPE_PERFORMANCE;
+					if (misc & ATOM_PM_MISCINFO_DRIVER_DEFAULT_MODE) {
+						rdev->pm.power_state[state_index].type =
+							POWER_STATE_TYPE_DEFAULT;
+						rdev->pm.default_power_state = &rdev->pm.power_state[state_index];
+						rdev->pm.power_state[state_index].default_clock_mode =
+							&rdev->pm.power_state[state_index].clock_info[0];
+					}
+					state_index++;
+					break;
+				case 2:
+					rdev->pm.power_state[state_index].num_clock_modes = 1;
+					rdev->pm.power_state[state_index].clock_info[0].mclk =
+						le32_to_cpu(power_info->info_2.asPowerPlayInfo[i].ulMemoryClock);
+					rdev->pm.power_state[state_index].clock_info[0].sclk =
+						le32_to_cpu(power_info->info_2.asPowerPlayInfo[i].ulEngineClock);
+					/* skip invalid modes */
+					if ((rdev->pm.power_state[state_index].clock_info[0].mclk == 0) ||
+					    (rdev->pm.power_state[state_index].clock_info[0].sclk == 0))
+						continue;
+					/* skip overclock modes for now */
+					if ((rdev->pm.power_state[state_index].clock_info[0].mclk >
+					     rdev->clock.default_mclk + RADEON_MODE_OVERCLOCK_MARGIN) ||
+					    (rdev->pm.power_state[state_index].clock_info[0].sclk >
+					     rdev->clock.default_sclk + RADEON_MODE_OVERCLOCK_MARGIN))
+						continue;
+					rdev->pm.power_state[state_index].non_clock_info.pcie_lanes =
+						power_info->info_2.asPowerPlayInfo[i].ucNumPciELanes;
+					misc = le32_to_cpu(power_info->info_2.asPowerPlayInfo[i].ulMiscInfo);
+					misc2 = le32_to_cpu(power_info->info_2.asPowerPlayInfo[i].ulMiscInfo2);
+					if (misc & ATOM_PM_MISCINFO_VOLTAGE_DROP_SUPPORT) {
+						rdev->pm.power_state[state_index].clock_info[0].voltage.type =
+							VOLTAGE_GPIO;
+						rdev->pm.power_state[state_index].clock_info[0].voltage.gpio =
+							radeon_lookup_gpio(rdev,
+							power_info->info_2.asPowerPlayInfo[i].ucVoltageDropIndex);
+						if (misc & ATOM_PM_MISCINFO_VOLTAGE_DROP_ACTIVE_HIGH)
+							rdev->pm.power_state[state_index].clock_info[0].voltage.active_high =
+								true;
+						else
+							rdev->pm.power_state[state_index].clock_info[0].voltage.active_high =
+								false;
+					} else if (misc & ATOM_PM_MISCINFO_PROGRAM_VOLTAGE) {
+						rdev->pm.power_state[state_index].clock_info[0].voltage.type =
+							VOLTAGE_VDDC;
+						rdev->pm.power_state[state_index].clock_info[0].voltage.vddc_id =
+							power_info->info_2.asPowerPlayInfo[i].ucVoltageDropIndex;
+					}
+					/* order matters! */
+					if (misc & ATOM_PM_MISCINFO_POWER_SAVING_MODE)
+						rdev->pm.power_state[state_index].type =
+							POWER_STATE_TYPE_POWERSAVE;
+					if (misc & ATOM_PM_MISCINFO_DEFAULT_DC_STATE_ENTRY_TRUE)
+						rdev->pm.power_state[state_index].type =
+							POWER_STATE_TYPE_BATTERY;
+					if (misc & ATOM_PM_MISCINFO_DEFAULT_LOW_DC_STATE_ENTRY_TRUE)
+						rdev->pm.power_state[state_index].type =
+							POWER_STATE_TYPE_BATTERY;
+					if (misc & ATOM_PM_MISCINFO_LOAD_BALANCE_EN)
+						rdev->pm.power_state[state_index].type =
+							POWER_STATE_TYPE_BALANCED;
+					if (misc & ATOM_PM_MISCINFO_3D_ACCELERATION_EN)
+						rdev->pm.power_state[state_index].type =
+							POWER_STATE_TYPE_PERFORMANCE;
+					if (misc2 & ATOM_PM_MISCINFO2_SYSTEM_AC_LITE_MODE)
+						rdev->pm.power_state[state_index].type =
+							POWER_STATE_TYPE_BALANCED;
+					if (misc & ATOM_PM_MISCINFO_DRIVER_DEFAULT_MODE) {
+						rdev->pm.power_state[state_index].type =
+							POWER_STATE_TYPE_DEFAULT;
+						rdev->pm.default_power_state = &rdev->pm.power_state[state_index];
+						rdev->pm.power_state[state_index].default_clock_mode =
+							&rdev->pm.power_state[state_index].clock_info[0];
+					}
+					state_index++;
+					break;
+				case 3:
+					rdev->pm.power_state[state_index].num_clock_modes = 1;
+					rdev->pm.power_state[state_index].clock_info[0].mclk =
+						le32_to_cpu(power_info->info_3.asPowerPlayInfo[i].ulMemoryClock);
+					rdev->pm.power_state[state_index].clock_info[0].sclk =
+						le32_to_cpu(power_info->info_3.asPowerPlayInfo[i].ulEngineClock);
+					/* skip invalid modes */
+					if ((rdev->pm.power_state[state_index].clock_info[0].mclk == 0) ||
+					    (rdev->pm.power_state[state_index].clock_info[0].sclk == 0))
+						continue;
+					/* skip overclock modes for now */
+					if ((rdev->pm.power_state[state_index].clock_info[0].mclk >
+					     rdev->clock.default_mclk + RADEON_MODE_OVERCLOCK_MARGIN) ||
+					    (rdev->pm.power_state[state_index].clock_info[0].sclk >
+					     rdev->clock.default_sclk + RADEON_MODE_OVERCLOCK_MARGIN))
+						continue;
+					rdev->pm.power_state[state_index].non_clock_info.pcie_lanes =
+						power_info->info_3.asPowerPlayInfo[i].ucNumPciELanes;
+					misc = le32_to_cpu(power_info->info_3.asPowerPlayInfo[i].ulMiscInfo);
+					misc2 = le32_to_cpu(power_info->info_3.asPowerPlayInfo[i].ulMiscInfo2);
+					if (misc & ATOM_PM_MISCINFO_VOLTAGE_DROP_SUPPORT) {
+						rdev->pm.power_state[state_index].clock_info[0].voltage.type =
+							VOLTAGE_GPIO;
+						rdev->pm.power_state[state_index].clock_info[0].voltage.gpio =
+							radeon_lookup_gpio(rdev,
+							power_info->info_3.asPowerPlayInfo[i].ucVoltageDropIndex);
+						if (misc & ATOM_PM_MISCINFO_VOLTAGE_DROP_ACTIVE_HIGH)
+							rdev->pm.power_state[state_index].clock_info[0].voltage.active_high =
+								true;
+						else
+							rdev->pm.power_state[state_index].clock_info[0].voltage.active_high =
+								false;
+					} else if (misc & ATOM_PM_MISCINFO_PROGRAM_VOLTAGE) {
+						rdev->pm.power_state[state_index].clock_info[0].voltage.type =
+							VOLTAGE_VDDC;
+						rdev->pm.power_state[state_index].clock_info[0].voltage.vddc_id =
+							power_info->info_3.asPowerPlayInfo[i].ucVoltageDropIndex;
+						if (misc2 & ATOM_PM_MISCINFO2_VDDCI_DYNAMIC_VOLTAGE_EN) {
+							rdev->pm.power_state[state_index].clock_info[0].voltage.vddci_enabled =
+								true;
+							rdev->pm.power_state[state_index].clock_info[0].voltage.vddci_id =
+							power_info->info_3.asPowerPlayInfo[i].ucVDDCI_VoltageDropIndex;
+						}
+					}
+					/* order matters! */
+					if (misc & ATOM_PM_MISCINFO_POWER_SAVING_MODE)
+						rdev->pm.power_state[state_index].type =
+							POWER_STATE_TYPE_POWERSAVE;
+					if (misc & ATOM_PM_MISCINFO_DEFAULT_DC_STATE_ENTRY_TRUE)
+						rdev->pm.power_state[state_index].type =
+							POWER_STATE_TYPE_BATTERY;
+					if (misc & ATOM_PM_MISCINFO_DEFAULT_LOW_DC_STATE_ENTRY_TRUE)
+						rdev->pm.power_state[state_index].type =
+							POWER_STATE_TYPE_BATTERY;
+					if (misc & ATOM_PM_MISCINFO_LOAD_BALANCE_EN)
+						rdev->pm.power_state[state_index].type =
+							POWER_STATE_TYPE_BALANCED;
+					if (misc & ATOM_PM_MISCINFO_3D_ACCELERATION_EN)
+						rdev->pm.power_state[state_index].type =
+							POWER_STATE_TYPE_PERFORMANCE;
+					if (misc2 & ATOM_PM_MISCINFO2_SYSTEM_AC_LITE_MODE)
+						rdev->pm.power_state[state_index].type =
+							POWER_STATE_TYPE_BALANCED;
+					if (misc & ATOM_PM_MISCINFO_DRIVER_DEFAULT_MODE) {
+						rdev->pm.power_state[state_index].type =
+							POWER_STATE_TYPE_DEFAULT;
+						rdev->pm.default_power_state = &rdev->pm.power_state[state_index];
+						rdev->pm.power_state[state_index].default_clock_mode =
+							&rdev->pm.power_state[state_index].clock_info[0];
+					}
+					state_index++;
+					break;
+				}
+			}
+		} else if (frev == 4) {
+			for (i = 0; i < power_info->info_4.ucNumStates; i++) {
+				mode_index = 0;
+				power_state = (struct _ATOM_PPLIB_STATE *)
+					(mode_info->atom_context->bios +
+					 data_offset +
+					 le16_to_cpu(power_info->info_4.usStateArrayOffset) +
+					 i * power_info->info_4.ucStateEntrySize);
+				non_clock_info = (struct _ATOM_PPLIB_NONCLOCK_INFO *)
+					(mode_info->atom_context->bios +
+					 data_offset +
+					 le16_to_cpu(power_info->info_4.usNonClockInfoArrayOffset) +
+					 (power_state->ucNonClockStateIndex *
+					  power_info->info_4.ucNonClockSize));
+				for (j = 0; j < (power_info->info_4.ucStateEntrySize - 1); j++) {
+					if (rdev->flags & RADEON_IS_IGP) {
+						struct _ATOM_PPLIB_RS780_CLOCK_INFO *clock_info =
+							(struct _ATOM_PPLIB_RS780_CLOCK_INFO *)
+							(mode_info->atom_context->bios +
+							 data_offset +
+							 le16_to_cpu(power_info->info_4.usClockInfoArrayOffset) +
+							 (power_state->ucClockStateIndices[j] *
+							  power_info->info_4.ucClockInfoSize));
+						sclk = le16_to_cpu(clock_info->usLowEngineClockLow);
+						sclk |= clock_info->ucLowEngineClockHigh << 16;
+						rdev->pm.power_state[state_index].clock_info[mode_index].sclk = sclk;
+						/* skip invalid modes */
+						if (rdev->pm.power_state[state_index].clock_info[mode_index].sclk == 0)
+							continue;
+						/* skip overclock modes for now */
+						if (rdev->pm.power_state[state_index].clock_info[mode_index].sclk >
+						    rdev->clock.default_sclk + RADEON_MODE_OVERCLOCK_MARGIN)
+							continue;
+						rdev->pm.power_state[state_index].clock_info[mode_index].voltage.type =
+							VOLTAGE_SW;
+						rdev->pm.power_state[state_index].clock_info[mode_index].voltage.voltage =
+							clock_info->usVDDC;
+						mode_index++;
+					} else {
+						struct _ATOM_PPLIB_R600_CLOCK_INFO *clock_info =
+							(struct _ATOM_PPLIB_R600_CLOCK_INFO *)
+							(mode_info->atom_context->bios +
+							 data_offset +
+							 le16_to_cpu(power_info->info_4.usClockInfoArrayOffset) +
+							 (power_state->ucClockStateIndices[j] *
+							  power_info->info_4.ucClockInfoSize));
+						sclk = le16_to_cpu(clock_info->usEngineClockLow);
+						sclk |= clock_info->ucEngineClockHigh << 16;
+						mclk = le16_to_cpu(clock_info->usMemoryClockLow);
+						mclk |= clock_info->ucMemoryClockHigh << 16;
+						rdev->pm.power_state[state_index].clock_info[mode_index].mclk = mclk;
+						rdev->pm.power_state[state_index].clock_info[mode_index].sclk = sclk;
+						/* skip invalid modes */
+						if ((rdev->pm.power_state[state_index].clock_info[mode_index].mclk == 0) ||
+						    (rdev->pm.power_state[state_index].clock_info[mode_index].sclk == 0))
+							continue;
+						/* skip overclock modes for now */
+						if ((rdev->pm.power_state[state_index].clock_info[mode_index].mclk >
+						     rdev->clock.default_mclk + RADEON_MODE_OVERCLOCK_MARGIN) ||
+						    (rdev->pm.power_state[state_index].clock_info[mode_index].sclk >
+						     rdev->clock.default_sclk + RADEON_MODE_OVERCLOCK_MARGIN))
+							continue;
+						rdev->pm.power_state[state_index].clock_info[mode_index].voltage.type =
+							VOLTAGE_SW;
+						rdev->pm.power_state[state_index].clock_info[mode_index].voltage.voltage =
+							clock_info->usVDDC;
+						mode_index++;
+					}
+				}
+				rdev->pm.power_state[state_index].num_clock_modes = mode_index;
+				if (mode_index) {
+					misc = le32_to_cpu(non_clock_info->ulCapsAndSettings);
+					misc2 = le16_to_cpu(non_clock_info->usClassification);
+					rdev->pm.power_state[state_index].non_clock_info.pcie_lanes =
+						((misc & ATOM_PPLIB_PCIE_LINK_WIDTH_MASK) >>
+						ATOM_PPLIB_PCIE_LINK_WIDTH_SHIFT) + 1;
+					switch (misc2 & ATOM_PPLIB_CLASSIFICATION_UI_MASK) {
+					case ATOM_PPLIB_CLASSIFICATION_UI_BATTERY:
+						rdev->pm.power_state[state_index].type =
+							POWER_STATE_TYPE_BATTERY;
+						break;
+					case ATOM_PPLIB_CLASSIFICATION_UI_BALANCED:
+						rdev->pm.power_state[state_index].type =
+							POWER_STATE_TYPE_BALANCED;
+						break;
+					case ATOM_PPLIB_CLASSIFICATION_UI_PERFORMANCE:
+						rdev->pm.power_state[state_index].type =
+							POWER_STATE_TYPE_PERFORMANCE;
+						break;
+					}
+					if (misc2 & ATOM_PPLIB_CLASSIFICATION_BOOT) {
+						rdev->pm.power_state[state_index].type =
+							POWER_STATE_TYPE_DEFAULT;
+						rdev->pm.default_power_state = &rdev->pm.power_state[state_index];
+						rdev->pm.power_state[state_index].default_clock_mode =
+							&rdev->pm.power_state[state_index].clock_info[mode_index - 1];
+					}
+					state_index++;
+				}
+			}
+		}
+	} else {
+		/* XXX figure out some good default low power mode for cards w/out power tables */
+	}
+
+	if (rdev->pm.default_power_state == NULL) {
+		/* add the default mode */
+		rdev->pm.power_state[state_index].type =
+			POWER_STATE_TYPE_DEFAULT;
+		rdev->pm.power_state[state_index].num_clock_modes = 1;
+		rdev->pm.power_state[state_index].clock_info[0].mclk = rdev->clock.default_mclk;
+		rdev->pm.power_state[state_index].clock_info[0].sclk = rdev->clock.default_sclk;
+		rdev->pm.power_state[state_index].default_clock_mode =
+			&rdev->pm.power_state[state_index].clock_info[0];
+		rdev->pm.power_state[state_index].clock_info[0].voltage.type = VOLTAGE_NONE;
+		if (rdev->asic->get_pcie_lanes)
+			rdev->pm.power_state[state_index].non_clock_info.pcie_lanes = radeon_get_pcie_lanes(rdev);
+		else
+			rdev->pm.power_state[state_index].non_clock_info.pcie_lanes = 16;
+		rdev->pm.default_power_state = &rdev->pm.power_state[state_index];
+		state_index++;
+	}
+	rdev->pm.num_power_states = state_index;
+
+	rdev->pm.current_power_state = rdev->pm.default_power_state;
+	rdev->pm.current_clock_mode =
+		rdev->pm.default_power_state->default_clock_mode;
+}
+
 void radeon_atom_set_clock_gating(struct radeon_device *rdev, int enable)
 {
 	DYNAMIC_CLOCK_GATING_PS_ALLOCATION args;
@@ -974,14 +1822,22 @@ void radeon_atom_set_clock_gating(struct radeon_device *rdev, int enable)
 	atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args);
 }
 
-void radeon_atom_static_pwrmgt_setup(struct radeon_device *rdev, int enable)
+uint32_t radeon_atom_get_engine_clock(struct radeon_device *rdev)
 {
-	ENABLE_ASIC_STATIC_PWR_MGT_PS_ALLOCATION args;
-	int index = GetIndexIntoMasterTable(COMMAND, EnableASIC_StaticPwrMgt);
+	GET_ENGINE_CLOCK_PS_ALLOCATION args;
+	int index = GetIndexIntoMasterTable(COMMAND, GetEngineClock);
 
-	args.ucEnable = enable;
+	atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args);
+	return args.ulReturnEngineClock;
+}
+
+uint32_t radeon_atom_get_memory_clock(struct radeon_device *rdev)
+{
+	GET_MEMORY_CLOCK_PS_ALLOCATION args;
+	int index = GetIndexIntoMasterTable(COMMAND, GetMemoryClock);
 
 	atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args);
+	return args.ulReturnMemoryClock;
 }
 
 void radeon_atom_set_engine_clock(struct radeon_device *rdev,
diff --git a/drivers/gpu/drm/radeon/radeon_atpx_handler.c b/drivers/gpu/drm/radeon/radeon_atpx_handler.c
new file mode 100644
index 000000000000..3f557c4151e0
--- /dev/null
+++ b/drivers/gpu/drm/radeon/radeon_atpx_handler.c
@@ -0,0 +1,257 @@
+/*
+ * Copyright (c) 2010 Red Hat Inc.
+ * Author : Dave Airlie <airlied@redhat.com>
+ *
+ * Licensed under GPLv2
+ *
+ * ATPX support for both Intel/ATI
+ */
+#include <linux/vga_switcheroo.h>
+#include <acpi/acpi.h>
+#include <acpi/acpi_bus.h>
+#include <linux/pci.h>
+
+#define ATPX_VERSION 0
+#define ATPX_GPU_PWR 2
+#define ATPX_MUX_SELECT 3
+
+#define ATPX_INTEGRATED 0
+#define ATPX_DISCRETE 1
+
+#define ATPX_MUX_IGD 0
+#define ATPX_MUX_DISCRETE 1
+
+static struct radeon_atpx_priv {
+	bool atpx_detected;
+	/* handle for device - and atpx */
+	acpi_handle dhandle;
+	acpi_handle atpx_handle;
+	acpi_handle atrm_handle;
+} radeon_atpx_priv;
+
+/* retrieve the ROM in 4k blocks */
+static int radeon_atrm_call(acpi_handle atrm_handle, uint8_t *bios,
+			    int offset, int len)
+{
+	acpi_status status;
+	union acpi_object atrm_arg_elements[2], *obj;
+	struct acpi_object_list atrm_arg;
+	struct acpi_buffer buffer = { ACPI_ALLOCATE_BUFFER, NULL};
+
+	atrm_arg.count = 2;
+	atrm_arg.pointer = &atrm_arg_elements[0];
+
+	atrm_arg_elements[0].type = ACPI_TYPE_INTEGER;
+	atrm_arg_elements[0].integer.value = offset;
+
+	atrm_arg_elements[1].type = ACPI_TYPE_INTEGER;
+	atrm_arg_elements[1].integer.value = len;
+
+	status = acpi_evaluate_object(atrm_handle, NULL, &atrm_arg, &buffer);
+	if (ACPI_FAILURE(status)) {
+		printk("failed to evaluate ATRM got %s\n", acpi_format_exception(status));
+		return -ENODEV;
+	}
+
+	obj = (union acpi_object *)buffer.pointer;
+	memcpy(bios+offset, obj->buffer.pointer, len);
+	kfree(buffer.pointer);
+	return len;
+}
+
+bool radeon_atrm_supported(struct pci_dev *pdev)
+{
+	/* get the discrete ROM only via ATRM */
+	if (!radeon_atpx_priv.atpx_detected)
+		return false;
+
+	if (radeon_atpx_priv.dhandle == DEVICE_ACPI_HANDLE(&pdev->dev))
+		return false;
+	return true;
+}
+
+
+int radeon_atrm_get_bios_chunk(uint8_t *bios, int offset, int len)
+{
+	return radeon_atrm_call(radeon_atpx_priv.atrm_handle, bios, offset, len);
+}
+
+static int radeon_atpx_get_version(acpi_handle handle)
+{
+	acpi_status status;
+	union acpi_object atpx_arg_elements[2], *obj;
+	struct acpi_object_list atpx_arg;
+	struct acpi_buffer buffer = { ACPI_ALLOCATE_BUFFER, NULL };
+
+	atpx_arg.count = 2;
+	atpx_arg.pointer = &atpx_arg_elements[0];
+
+	atpx_arg_elements[0].type = ACPI_TYPE_INTEGER;
+	atpx_arg_elements[0].integer.value = ATPX_VERSION;
+
+	atpx_arg_elements[1].type = ACPI_TYPE_INTEGER;
+	atpx_arg_elements[1].integer.value = ATPX_VERSION;
+
+	status = acpi_evaluate_object(handle, NULL, &atpx_arg, &buffer);
+	if (ACPI_FAILURE(status)) {
+		printk("%s: failed to call ATPX: %s\n", __func__, acpi_format_exception(status));
+		return -ENOSYS;
+	}
+	obj = (union acpi_object *)buffer.pointer;
+	if (obj && (obj->type == ACPI_TYPE_BUFFER))
+		printk(KERN_INFO "radeon atpx: version is %d\n", *((u8 *)(obj->buffer.pointer) + 2));
+	kfree(buffer.pointer);
+	return 0;
+}
+
+static int radeon_atpx_execute(acpi_handle handle, int cmd_id, u16 value)
+{
+	acpi_status status;
+	union acpi_object atpx_arg_elements[2];
+	struct acpi_object_list atpx_arg;
+	struct acpi_buffer buffer = { ACPI_ALLOCATE_BUFFER, NULL };
+	uint8_t buf[4] = {0};
+
+	if (!handle)
+		return -EINVAL;
+
+	atpx_arg.count = 2;
+	atpx_arg.pointer = &atpx_arg_elements[0];
+
+	atpx_arg_elements[0].type = ACPI_TYPE_INTEGER;
+	atpx_arg_elements[0].integer.value = cmd_id;
+
+	buf[2] = value & 0xff;
+	buf[3] = (value >> 8) & 0xff;
+
+	atpx_arg_elements[1].type = ACPI_TYPE_BUFFER;
+	atpx_arg_elements[1].buffer.length = 4;
+	atpx_arg_elements[1].buffer.pointer = buf;
+
+	status = acpi_evaluate_object(handle, NULL, &atpx_arg, &buffer);
+	if (ACPI_FAILURE(status)) {
+		printk("%s: failed to call ATPX: %s\n", __func__, acpi_format_exception(status));
+		return -ENOSYS;
+	}
+	kfree(buffer.pointer);
+
+	return 0;
+}
+
+static int radeon_atpx_set_discrete_state(acpi_handle handle, int state)
+{
+	return radeon_atpx_execute(handle, ATPX_GPU_PWR, state);
+}
+
+static int radeon_atpx_switch_mux(acpi_handle handle, int mux_id)
+{
+	return radeon_atpx_execute(handle, ATPX_MUX_SELECT, mux_id);
+}
+
+
+static int radeon_atpx_switchto(enum vga_switcheroo_client_id id)
+{
+	if (id == VGA_SWITCHEROO_IGD)
+		radeon_atpx_switch_mux(radeon_atpx_priv.atpx_handle, 0);
+	else
+		radeon_atpx_switch_mux(radeon_atpx_priv.atpx_handle, 1);
+	return 0;
+}
+
+static int radeon_atpx_power_state(enum vga_switcheroo_client_id id,
+				   enum vga_switcheroo_state state)
+{
+	/* on w500 ACPI can't change intel gpu state */
+	if (id == VGA_SWITCHEROO_IGD)
+		return 0;
+
+	radeon_atpx_set_discrete_state(radeon_atpx_priv.atpx_handle, state);
+	return 0;
+}
+
+static bool radeon_atpx_pci_probe_handle(struct pci_dev *pdev)
+{
+	acpi_handle dhandle, atpx_handle, atrm_handle;
+	acpi_status status;
+
+	dhandle = DEVICE_ACPI_HANDLE(&pdev->dev);
+	if (!dhandle)
+		return false;
+
+	status = acpi_get_handle(dhandle, "ATPX", &atpx_handle);
+	if (ACPI_FAILURE(status))
+		return false;
+
+	status = acpi_get_handle(dhandle, "ATRM", &atrm_handle);
+	if (ACPI_FAILURE(status))
+		return false;
+
+	radeon_atpx_priv.dhandle = dhandle;
+	radeon_atpx_priv.atpx_handle = atpx_handle;
+	radeon_atpx_priv.atrm_handle = atrm_handle;
+	return true;
+}
+
+static int radeon_atpx_init(void)
+{
+	/* set up the ATPX handle */
+
+	radeon_atpx_get_version(radeon_atpx_priv.atpx_handle);
+	return 0;
+}
+
+static int radeon_atpx_get_client_id(struct pci_dev *pdev)
+{
+	if (radeon_atpx_priv.dhandle == DEVICE_ACPI_HANDLE(&pdev->dev))
+		return VGA_SWITCHEROO_IGD;
+	else
+		return VGA_SWITCHEROO_DIS;
+}
+
+static struct vga_switcheroo_handler radeon_atpx_handler = {
+	.switchto = radeon_atpx_switchto,
+	.power_state = radeon_atpx_power_state,
+	.init = radeon_atpx_init,
+	.get_client_id = radeon_atpx_get_client_id,
+};
+
+static bool radeon_atpx_detect(void)
+{
+	char acpi_method_name[255] = { 0 };
+	struct acpi_buffer buffer = {sizeof(acpi_method_name), acpi_method_name};
+	struct pci_dev *pdev = NULL;
+	bool has_atpx = false;
+	int vga_count = 0;
+
+	while ((pdev = pci_get_class(PCI_CLASS_DISPLAY_VGA << 8, pdev)) != NULL) {
+		vga_count++;
+
+		has_atpx |= (radeon_atpx_pci_probe_handle(pdev) == true);
+	}
+
+	if (has_atpx && vga_count == 2) {
+		acpi_get_name(radeon_atpx_priv.atpx_handle, ACPI_FULL_PATHNAME, &buffer);
+		printk(KERN_INFO "VGA switcheroo: detected switching method %s handle\n",
+		       acpi_method_name);
+		radeon_atpx_priv.atpx_detected = true;
+		return true;
+	}
+	return false;
+}
+
+void radeon_register_atpx_handler(void)
+{
+	bool r;
+
+	/* detect if we have any ATPX + 2 VGA in the system */
+	r = radeon_atpx_detect();
+	if (!r)
+		return;
+
+	vga_switcheroo_register_handler(&radeon_atpx_handler);
+}
+
+void radeon_unregister_atpx_handler(void)
+{
+	vga_switcheroo_unregister_handler();
+}
diff --git a/drivers/gpu/drm/radeon/radeon_benchmark.c b/drivers/gpu/drm/radeon/radeon_benchmark.c
index 2e938f7496fb..7932dc4d6b90 100644
--- a/drivers/gpu/drm/radeon/radeon_benchmark.c
+++ b/drivers/gpu/drm/radeon/radeon_benchmark.c
@@ -29,8 +29,8 @@
 void radeon_benchmark_move(struct radeon_device *rdev, unsigned bsize,
 			   unsigned sdomain, unsigned ddomain)
 {
-	struct radeon_object *dobj = NULL;
-	struct radeon_object *sobj = NULL;
+	struct radeon_bo *dobj = NULL;
+	struct radeon_bo *sobj = NULL;
 	struct radeon_fence *fence = NULL;
 	uint64_t saddr, daddr;
 	unsigned long start_jiffies;
@@ -41,54 +41,73 @@ void radeon_benchmark_move(struct radeon_device *rdev, unsigned bsize,
 
 	size = bsize;
 	n = 1024;
-	r = radeon_object_create(rdev, NULL, size, true, sdomain, false, &sobj);
+	r = radeon_bo_create(rdev, NULL, size, true, sdomain, &sobj);
 	if (r) {
 		goto out_cleanup;
 	}
-	r = radeon_object_pin(sobj, sdomain, &saddr);
+	r = radeon_bo_reserve(sobj, false);
+	if (unlikely(r != 0))
+		goto out_cleanup;
+	r = radeon_bo_pin(sobj, sdomain, &saddr);
+	radeon_bo_unreserve(sobj);
 	if (r) {
 		goto out_cleanup;
 	}
-	r = radeon_object_create(rdev, NULL, size, true, ddomain, false, &dobj);
+	r = radeon_bo_create(rdev, NULL, size, true, ddomain, &dobj);
 	if (r) {
 		goto out_cleanup;
 	}
-	r = radeon_object_pin(dobj, ddomain, &daddr);
+	r = radeon_bo_reserve(dobj, false);
+	if (unlikely(r != 0))
+		goto out_cleanup;
+	r = radeon_bo_pin(dobj, ddomain, &daddr);
+	radeon_bo_unreserve(dobj);
 	if (r) {
 		goto out_cleanup;
 	}
-	start_jiffies = jiffies;
-	for (i = 0; i < n; i++) {
-		r = radeon_fence_create(rdev, &fence);
-		if (r) {
-			goto out_cleanup;
+
+	/* r100 doesn't have dma engine so skip the test */
+	if (rdev->asic->copy_dma) {
+
+		start_jiffies = jiffies;
+		for (i = 0; i < n; i++) {
+			r = radeon_fence_create(rdev, &fence);
+			if (r) {
+				goto out_cleanup;
+			}
+
+			r = radeon_copy_dma(rdev, saddr, daddr,
+					size / RADEON_GPU_PAGE_SIZE, fence);
+
+			if (r) {
+				goto out_cleanup;
+			}
+			r = radeon_fence_wait(fence, false);
+			if (r) {
+				goto out_cleanup;
+			}
+			radeon_fence_unref(&fence);
 		}
-		r = radeon_copy_dma(rdev, saddr, daddr, size / 4096, fence);
-		if (r) {
-			goto out_cleanup;
+		end_jiffies = jiffies;
+		time = end_jiffies - start_jiffies;
+		time = jiffies_to_msecs(time);
+		if (time > 0) {
+			i = ((n * size) >> 10) / time;
+			printk(KERN_INFO "radeon: dma %u bo moves of %ukb from"
+					" %d to %d in %lums (%ukb/ms %ukb/s %uM/s)\n",
+					n, size >> 10,
+					sdomain, ddomain, time,
+					i, i * 1000, (i * 1000) / 1024);
 		}
-		r = radeon_fence_wait(fence, false);
-		if (r) {
-			goto out_cleanup;
-		}
-		radeon_fence_unref(&fence);
-	}
-	end_jiffies = jiffies;
-	time = end_jiffies - start_jiffies;
-	time = jiffies_to_msecs(time);
-	if (time > 0) {
-		i = ((n * size) >> 10) / time;
-		printk(KERN_INFO "radeon: dma %u bo moves of %ukb from %d to %d"
-		       " in %lums (%ukb/ms %ukb/s %uM/s)\n", n, size >> 10,
-		       sdomain, ddomain, time, i, i * 1000, (i * 1000) / 1024);
 	}
+
 	start_jiffies = jiffies;
 	for (i = 0; i < n; i++) {
 		r = radeon_fence_create(rdev, &fence);
 		if (r) {
 			goto out_cleanup;
 		}
-		r = radeon_copy_blit(rdev, saddr, daddr, size / 4096, fence);
+		r = radeon_copy_blit(rdev, saddr, daddr, size / RADEON_GPU_PAGE_SIZE, fence);
 		if (r) {
 			goto out_cleanup;
 		}
@@ -109,12 +128,20 @@ void radeon_benchmark_move(struct radeon_device *rdev, unsigned bsize,
 	}
 out_cleanup:
 	if (sobj) {
-		radeon_object_unpin(sobj);
-		radeon_object_unref(&sobj);
+		r = radeon_bo_reserve(sobj, false);
+		if (likely(r == 0)) {
+			radeon_bo_unpin(sobj);
+			radeon_bo_unreserve(sobj);
+		}
+		radeon_bo_unref(&sobj);
 	}
 	if (dobj) {
-		radeon_object_unpin(dobj);
-		radeon_object_unref(&dobj);
+		r = radeon_bo_reserve(dobj, false);
+		if (likely(r == 0)) {
+			radeon_bo_unpin(dobj);
+			radeon_bo_unreserve(dobj);
+		}
+		radeon_bo_unref(&dobj);
 	}
 	if (fence) {
 		radeon_fence_unref(&fence);
diff --git a/drivers/gpu/drm/radeon/radeon_bios.c b/drivers/gpu/drm/radeon/radeon_bios.c
index 96e37a6e7ce4..557240460526 100644
--- a/drivers/gpu/drm/radeon/radeon_bios.c
+++ b/drivers/gpu/drm/radeon/radeon_bios.c
@@ -30,15 +30,51 @@
 #include "radeon.h"
 #include "atom.h"
 
+#include <linux/vga_switcheroo.h>
 /*
  * BIOS.
  */
+
+/* If you boot an IGP board with a discrete card as the primary,
+ * the IGP rom is not accessible via the rom bar as the IGP rom is
+ * part of the system bios.  On boot, the system bios puts a
+ * copy of the igp rom at the start of vram if a discrete card is
+ * present.
+ */
+static bool igp_read_bios_from_vram(struct radeon_device *rdev)
+{
+	uint8_t __iomem *bios;
+	resource_size_t vram_base;
+	resource_size_t size = 256 * 1024; /* ??? */
+
+	rdev->bios = NULL;
+	vram_base = drm_get_resource_start(rdev->ddev, 0);
+	bios = ioremap(vram_base, size);
+	if (!bios) {
+		return false;
+	}
+
+	if (size == 0 || bios[0] != 0x55 || bios[1] != 0xaa) {
+		iounmap(bios);
+		return false;
+	}
+	rdev->bios = kmalloc(size, GFP_KERNEL);
+	if (rdev->bios == NULL) {
+		iounmap(bios);
+		return false;
+	}
+	memcpy_fromio(rdev->bios, bios, size);
+	iounmap(bios);
+	return true;
+}
+
 static bool radeon_read_bios(struct radeon_device *rdev)
 {
 	uint8_t __iomem *bios;
 	size_t size;
 
 	rdev->bios = NULL;
+	/* XXX: some cards may return 0 for rom size? ddx has a workaround */
 	bios = pci_map_rom(rdev->pdev, &size);
 	if (!bios) {
 		return false;
@@ -58,6 +94,38 @@ static bool radeon_read_bios(struct radeon_device *rdev)
 	return true;
 }
 
+/* ATRM is used to get the BIOS on the discrete cards in
+ * dual-gpu systems.
+ */
+static bool radeon_atrm_get_bios(struct radeon_device *rdev)
+{
+	int ret;
+	int size = 64 * 1024;
+	int i;
+
+	if (!radeon_atrm_supported(rdev->pdev))
+		return false;
+
+	rdev->bios = kmalloc(size, GFP_KERNEL);
+	if (!rdev->bios) {
+		DRM_ERROR("Unable to allocate bios\n");
+		return false;
+	}
+
+	for (i = 0; i < size / ATRM_BIOS_PAGE; i++) {
+		ret = radeon_atrm_get_bios_chunk(rdev->bios,
+						 (i * ATRM_BIOS_PAGE),
+						 ATRM_BIOS_PAGE);
+		if (ret <= 0)
+			break;
+	}
+
+	if (i == 0 || rdev->bios[0] != 0x55 || rdev->bios[1] != 0xaa) {
+		kfree(rdev->bios);
+		return false;
+	}
+	return true;
+}
 static bool r700_read_disabled_bios(struct radeon_device *rdev)
 {
 	uint32_t viph_control;
@@ -341,7 +409,9 @@ static bool legacy_read_disabled_bios(struct radeon_device *rdev)
 
 static bool radeon_read_disabled_bios(struct radeon_device *rdev)
 {
-	if (rdev->family >= CHIP_RV770)
+	if (rdev->flags & RADEON_IS_IGP)
+		return igp_read_bios_from_vram(rdev);
+	else if (rdev->family >= CHIP_RV770)
 		return r700_read_disabled_bios(rdev);
 	else if (rdev->family >= CHIP_R600)
 		return r600_read_disabled_bios(rdev);
@@ -351,12 +421,17 @@ static bool radeon_read_disabled_bios(struct radeon_device *rdev)
 		return legacy_read_disabled_bios(rdev);
 }
 
+
 bool radeon_get_bios(struct radeon_device *rdev)
 {
 	bool r;
 	uint16_t tmp;
 
-	r = radeon_read_bios(rdev);
+	r = radeon_atrm_get_bios(rdev);
+	if (r == false)
+		r = igp_read_bios_from_vram(rdev);
+	if (r == false)
+		r = radeon_read_bios(rdev);
 	if (r == false) {
 		r = radeon_read_disabled_bios(rdev);
 	}
@@ -366,6 +441,13 @@ bool radeon_get_bios(struct radeon_device *rdev)
 		return false;
 	}
 	if (rdev->bios[0] != 0x55 || rdev->bios[1] != 0xaa) {
+		printk("BIOS signature incorrect %x %x\n", rdev->bios[0], rdev->bios[1]);
+		goto free_bios;
+	}
+
+	tmp = RBIOS16(0x18);
+	if (RBIOS8(tmp + 0x14) != 0x0) {
+		DRM_INFO("Not an x86 BIOS ROM, not using.\n");
 		goto free_bios;
 	}
 
diff --git a/drivers/gpu/drm/radeon/radeon_clocks.c b/drivers/gpu/drm/radeon/radeon_clocks.c
index 152eef13197a..f64936cc4dd9 100644
--- a/drivers/gpu/drm/radeon/radeon_clocks.c
+++ b/drivers/gpu/drm/radeon/radeon_clocks.c
@@ -32,7 +32,7 @@
 #include "atom.h"
 
 /* 10 khz */
-static uint32_t radeon_legacy_get_engine_clock(struct radeon_device *rdev)
+uint32_t radeon_legacy_get_engine_clock(struct radeon_device *rdev)
 {
 	struct radeon_pll *spll = &rdev->clock.spll;
 	uint32_t fb_div, ref_div, post_div, sclk;
@@ -44,6 +44,10 @@ static uint32_t radeon_legacy_get_engine_clock(struct radeon_device *rdev)
 
 	ref_div =
 	    RREG32_PLL(RADEON_M_SPLL_REF_FB_DIV) & RADEON_M_SPLL_REF_DIV_MASK;
+
+	if (ref_div == 0)
+		return 0;
+
 	sclk = fb_div / ref_div;
 
 	post_div = RREG32_PLL(RADEON_SCLK_CNTL) & RADEON_SCLK_SRC_SEL_MASK;
@@ -52,13 +56,13 @@ static uint32_t radeon_legacy_get_engine_clock(struct radeon_device *rdev)
 	else if (post_div == 3)
 		sclk >>= 2;
 	else if (post_div == 4)
-		sclk >>= 4;
+		sclk >>= 3;
 
 	return sclk;
 }
 
 /* 10 khz */
-static uint32_t radeon_legacy_get_memory_clock(struct radeon_device *rdev)
+uint32_t radeon_legacy_get_memory_clock(struct radeon_device *rdev)
 {
 	struct radeon_pll *mpll = &rdev->clock.mpll;
 	uint32_t fb_div, ref_div, post_div, mclk;
@@ -70,6 +74,10 @@ static uint32_t radeon_legacy_get_memory_clock(struct radeon_device *rdev)
 
 	ref_div =
 	    RREG32_PLL(RADEON_M_SPLL_REF_FB_DIV) & RADEON_M_SPLL_REF_DIV_MASK;
+
+	if (ref_div == 0)
+		return 0;
+
 	mclk = fb_div / ref_div;
 
 	post_div = RREG32_PLL(RADEON_MCLK_CNTL) & 0x7;
@@ -78,7 +86,7 @@ static uint32_t radeon_legacy_get_memory_clock(struct radeon_device *rdev)
 	else if (post_div == 3)
 		mclk >>= 2;
 	else if (post_div == 4)
-		mclk >>= 4;
+		mclk >>= 3;
 
 	return mclk;
 }
@@ -88,6 +96,7 @@ void radeon_get_clock_info(struct drm_device *dev)
 	struct radeon_device *rdev = dev->dev_private;
 	struct radeon_pll *p1pll = &rdev->clock.p1pll;
 	struct radeon_pll *p2pll = &rdev->clock.p2pll;
+	struct radeon_pll *dcpll = &rdev->clock.dcpll;
 	struct radeon_pll *spll = &rdev->clock.spll;
 	struct radeon_pll *mpll = &rdev->clock.mpll;
 	int ret;
@@ -98,8 +107,19 @@ void radeon_get_clock_info(struct drm_device *dev)
 		ret = radeon_combios_get_clock_info(dev);
 
 	if (ret) {
-		if (p1pll->reference_div < 2)
-			p1pll->reference_div = 12;
+		if (p1pll->reference_div < 2) {
+			if (!ASIC_IS_AVIVO(rdev)) {
+				u32 tmp = RREG32_PLL(RADEON_PPLL_REF_DIV);
+				if (ASIC_IS_R300(rdev))
+					p1pll->reference_div =
+						(tmp & R300_PPLL_REF_DIV_ACC_MASK) >> R300_PPLL_REF_DIV_ACC_SHIFT;
+				else
+					p1pll->reference_div = tmp & RADEON_PPLL_REF_DIV_MASK;
+				if (p1pll->reference_div < 2)
+					p1pll->reference_div = 12;
+			} else
+				p1pll->reference_div = 12;
+		}
 		if (p2pll->reference_div < 2)
 			p2pll->reference_div = 12;
 		if (rdev->family < CHIP_RS600) {
@@ -185,6 +205,17 @@ void radeon_get_clock_info(struct drm_device *dev)
 		p2pll->max_frac_feedback_div = 0;
 	}
 
+	/* dcpll is DCE4 only */
+	dcpll->min_post_div = 2;
+	dcpll->max_post_div = 0x7f;
+	dcpll->min_frac_feedback_div = 0;
+	dcpll->max_frac_feedback_div = 9;
+	dcpll->min_ref_div = 2;
+	dcpll->max_ref_div = 0x3ff;
+	dcpll->min_feedback_div = 4;
+	dcpll->max_feedback_div = 0xfff;
+	dcpll->best_vco = 0;
+
 	p1pll->min_ref_div = 2;
 	p1pll->max_ref_div = 0x3ff;
 	p1pll->min_feedback_div = 4;
@@ -411,7 +442,7 @@ void radeon_legacy_set_clock_gating(struct radeon_device *rdev, int enable)
 					R300_PIXCLK_TRANS_ALWAYS_ONb |
 					R300_PIXCLK_TVO_ALWAYS_ONb |
 					R300_P2G2CLK_ALWAYS_ONb |
-					R300_P2G2CLK_ALWAYS_ONb);
+					R300_P2G2CLK_DAC_ALWAYS_ONb);
 				WREG32_PLL(RADEON_PIXCLKS_CNTL, tmp);
 			} else if (rdev->family >= CHIP_RV350) {
 				tmp = RREG32_PLL(R300_SCLK_CNTL2);
@@ -464,7 +495,7 @@ void radeon_legacy_set_clock_gating(struct radeon_device *rdev, int enable)
 					R300_PIXCLK_TRANS_ALWAYS_ONb |
 					R300_PIXCLK_TVO_ALWAYS_ONb |
 					R300_P2G2CLK_ALWAYS_ONb |
-					R300_P2G2CLK_ALWAYS_ONb);
+					R300_P2G2CLK_DAC_ALWAYS_ONb);
 				WREG32_PLL(RADEON_PIXCLKS_CNTL, tmp);
 
 				tmp = RREG32_PLL(RADEON_MCLK_MISC);
@@ -654,7 +685,7 @@ void radeon_legacy_set_clock_gating(struct radeon_device *rdev, int enable)
 				 R300_PIXCLK_TRANS_ALWAYS_ONb |
 				 R300_PIXCLK_TVO_ALWAYS_ONb |
 				 R300_P2G2CLK_ALWAYS_ONb |
-				 R300_P2G2CLK_ALWAYS_ONb |
+				 R300_P2G2CLK_DAC_ALWAYS_ONb |
 				 R300_DISP_DAC_PIXCLK_DAC2_BLANK_OFF);
 			WREG32_PLL(RADEON_PIXCLKS_CNTL, tmp);
 		} else if (rdev->family >= CHIP_RV350) {
@@ -705,7 +736,7 @@ void radeon_legacy_set_clock_gating(struct radeon_device *rdev, int enable)
 				 R300_PIXCLK_TRANS_ALWAYS_ONb |
 				 R300_PIXCLK_TVO_ALWAYS_ONb |
 				 R300_P2G2CLK_ALWAYS_ONb |
-				 R300_P2G2CLK_ALWAYS_ONb |
+				 R300_P2G2CLK_DAC_ALWAYS_ONb |
 				 R300_DISP_DAC_PIXCLK_DAC2_BLANK_OFF);
 			WREG32_PLL(RADEON_PIXCLKS_CNTL, tmp);
 		} else {
@@ -827,8 +858,10 @@ int radeon_static_clocks_init(struct drm_device *dev)
 	/* XXX make sure engine is idle */
 
 	if (radeon_dynclks != -1) {
-		if (radeon_dynclks)
-			radeon_set_clock_gating(rdev, 1);
+		if (radeon_dynclks) {
+			if (rdev->asic->set_clock_gating)
+				radeon_set_clock_gating(rdev, 1);
+		}
 	}
 	radeon_apply_clock_quirks(rdev);
 	return 0;
diff --git a/drivers/gpu/drm/radeon/radeon_combios.c b/drivers/gpu/drm/radeon/radeon_combios.c
index 748265a105b3..e9ea38ece375 100644
--- a/drivers/gpu/drm/radeon/radeon_combios.c
+++ b/drivers/gpu/drm/radeon/radeon_combios.c
@@ -49,7 +49,9 @@ radeon_add_legacy_connector(struct drm_device *dev,
 			    uint32_t connector_id,
 			    uint32_t supported_device,
 			    int connector_type,
-			    struct radeon_i2c_bus_rec *i2c_bus);
+			    struct radeon_i2c_bus_rec *i2c_bus,
+			    uint16_t connector_object_id,
+			    struct radeon_hpd *hpd);
 
 /* from radeon_legacy_encoder.c */
 extern void
@@ -148,6 +150,9 @@ static uint16_t combios_get_table_offset(struct drm_device *dev,
 	int rev;
 	uint16_t offset = 0, check_offset;
 
+	if (!rdev->bios)
+		return 0;
+
 	switch (table) {
 		/* absolute offset tables */
 	case COMBIOS_ASIC_INIT_1_TABLE:
@@ -441,38 +446,164 @@ static uint16_t combios_get_table_offset(struct drm_device *dev,
 
 }
 
-struct radeon_i2c_bus_rec combios_setup_i2c_bus(int ddc_line)
+bool radeon_combios_check_hardcoded_edid(struct radeon_device *rdev)
+{
+	int edid_info;
+	struct edid *edid;
+	edid_info = combios_get_table_offset(rdev->ddev, COMBIOS_HARDCODED_EDID_TABLE);
+	if (!edid_info)
+		return false;
+
+	edid = kmalloc(EDID_LENGTH * (DRM_MAX_EDID_EXT_NUM + 1),
+		       GFP_KERNEL);
+	if (edid == NULL)
+		return false;
+
+	memcpy((unsigned char *)edid,
+	       (unsigned char *)(rdev->bios + edid_info), EDID_LENGTH);
+
+	if (!drm_edid_is_valid(edid)) {
+		kfree(edid);
+		return false;
+	}
+
+	rdev->mode_info.bios_hardcoded_edid = edid;
+	return true;
+}
+
+struct edid *
+radeon_combios_get_hardcoded_edid(struct radeon_device *rdev)
+{
+	if (rdev->mode_info.bios_hardcoded_edid)
+		return rdev->mode_info.bios_hardcoded_edid;
+	return NULL;
+}
+
+static struct radeon_i2c_bus_rec combios_setup_i2c_bus(struct radeon_device *rdev,
+						       int ddc_line)
 {
 	struct radeon_i2c_bus_rec i2c;
 
-	i2c.mask_clk_mask = RADEON_GPIO_EN_1;
-	i2c.mask_data_mask = RADEON_GPIO_EN_0;
-	i2c.a_clk_mask = RADEON_GPIO_A_1;
-	i2c.a_data_mask = RADEON_GPIO_A_0;
-	i2c.put_clk_mask = RADEON_GPIO_EN_1;
-	i2c.put_data_mask = RADEON_GPIO_EN_0;
-	i2c.get_clk_mask = RADEON_GPIO_Y_1;
-	i2c.get_data_mask = RADEON_GPIO_Y_0;
-	if ((ddc_line == RADEON_LCD_GPIO_MASK) ||
-	    (ddc_line == RADEON_MDGPIO_EN_REG)) {
-		i2c.mask_clk_reg = ddc_line;
-		i2c.mask_data_reg = ddc_line;
-		i2c.a_clk_reg = ddc_line;
-		i2c.a_data_reg = ddc_line;
-		i2c.put_clk_reg = ddc_line;
-		i2c.put_data_reg = ddc_line;
-		i2c.get_clk_reg = ddc_line + 4;
-		i2c.get_data_reg = ddc_line + 4;
+	if (ddc_line == RADEON_GPIOPAD_MASK) {
+		i2c.mask_clk_reg = RADEON_GPIOPAD_MASK;
+		i2c.mask_data_reg = RADEON_GPIOPAD_MASK;
+		i2c.a_clk_reg = RADEON_GPIOPAD_A;
+		i2c.a_data_reg = RADEON_GPIOPAD_A;
+		i2c.en_clk_reg = RADEON_GPIOPAD_EN;
+		i2c.en_data_reg = RADEON_GPIOPAD_EN;
+		i2c.y_clk_reg = RADEON_GPIOPAD_Y;
+		i2c.y_data_reg = RADEON_GPIOPAD_Y;
+	} else if (ddc_line == RADEON_MDGPIO_MASK) {
+		i2c.mask_clk_reg = RADEON_MDGPIO_MASK;
+		i2c.mask_data_reg = RADEON_MDGPIO_MASK;
+		i2c.a_clk_reg = RADEON_MDGPIO_A;
+		i2c.a_data_reg = RADEON_MDGPIO_A;
+		i2c.en_clk_reg = RADEON_MDGPIO_EN;
+		i2c.en_data_reg = RADEON_MDGPIO_EN;
+		i2c.y_clk_reg = RADEON_MDGPIO_Y;
+		i2c.y_data_reg = RADEON_MDGPIO_Y;
 	} else {
+		i2c.mask_clk_mask = RADEON_GPIO_EN_1;
+		i2c.mask_data_mask = RADEON_GPIO_EN_0;
+		i2c.a_clk_mask = RADEON_GPIO_A_1;
+		i2c.a_data_mask = RADEON_GPIO_A_0;
+		i2c.en_clk_mask = RADEON_GPIO_EN_1;
+		i2c.en_data_mask = RADEON_GPIO_EN_0;
+		i2c.y_clk_mask = RADEON_GPIO_Y_1;
+		i2c.y_data_mask = RADEON_GPIO_Y_0;
+
 		i2c.mask_clk_reg = ddc_line;
 		i2c.mask_data_reg = ddc_line;
 		i2c.a_clk_reg = ddc_line;
 		i2c.a_data_reg = ddc_line;
-		i2c.put_clk_reg = ddc_line;
-		i2c.put_data_reg = ddc_line;
-		i2c.get_clk_reg = ddc_line;
-		i2c.get_data_reg = ddc_line;
+		i2c.en_clk_reg = ddc_line;
+		i2c.en_data_reg = ddc_line;
+		i2c.y_clk_reg = ddc_line;
+		i2c.y_data_reg = ddc_line;
+	}
+
+	switch (rdev->family) {
+	case CHIP_R100:
+	case CHIP_RV100:
+	case CHIP_RS100:
+	case CHIP_RV200:
+	case CHIP_RS200:
+	case CHIP_RS300:
+		switch (ddc_line) {
+		case RADEON_GPIO_DVI_DDC:
+			/* in theory this should be hw capable,
+			 * but it doesn't seem to work
+			 */
+			i2c.hw_capable = false;
+			break;
+		default:
+			i2c.hw_capable = false;
+			break;
+		}
+		break;
+	case CHIP_R200:
+		switch (ddc_line) {
+		case RADEON_GPIO_DVI_DDC:
+		case RADEON_GPIO_MONID:
+			i2c.hw_capable = true;
+			break;
+		default:
+			i2c.hw_capable = false;
+			break;
+		}
+		break;
+	case CHIP_RV250:
+	case CHIP_RV280:
+		switch (ddc_line) {
+		case RADEON_GPIO_VGA_DDC:
+		case RADEON_GPIO_DVI_DDC:
+		case RADEON_GPIO_CRT2_DDC:
+			i2c.hw_capable = true;
+			break;
+		default:
+			i2c.hw_capable = false;
+			break;
+		}
+		break;
+	case CHIP_R300:
+	case CHIP_R350:
+		switch (ddc_line) {
+		case RADEON_GPIO_VGA_DDC:
+		case RADEON_GPIO_DVI_DDC:
+			i2c.hw_capable = true;
+			break;
+		default:
+			i2c.hw_capable = false;
+			break;
+		}
+		break;
+	case CHIP_RV350:
+	case CHIP_RV380:
+	case CHIP_RS400:
+	case CHIP_RS480:
+		switch (ddc_line) {
+		case RADEON_GPIO_VGA_DDC:
+		case RADEON_GPIO_DVI_DDC:
+			i2c.hw_capable = true;
+			break;
+		case RADEON_GPIO_MONID:
+			/* hw i2c on RADEON_GPIO_MONID doesn't seem to work
+			 * reliably on some pre-r4xx hardware; not sure why.
+			 */
+			i2c.hw_capable = false;
+			break;
+		default:
+			i2c.hw_capable = false;
+			break;
+		}
+		break;
+	default:
+		i2c.hw_capable = false;
+		break;
 	}
+	i2c.mm_i2c = false;
+	i2c.i2c_id = 0;
+	i2c.hpd_id = 0;
 
 	if (ddc_line)
 		i2c.valid = true;
@@ -493,9 +624,6 @@ bool radeon_combios_get_clock_info(struct drm_device *dev)
 	int8_t rev;
 	uint16_t sclk, mclk;
 
-	if (rdev->bios == NULL)
-		return NULL;
-
 	pll_info = combios_get_table_offset(dev, COMBIOS_PLL_INFO_TABLE);
 	if (pll_info) {
 		rev = RBIOS8(pll_info);
@@ -561,6 +689,48 @@ bool radeon_combios_get_clock_info(struct drm_device *dev)
 	return false;
 }
 
+bool radeon_combios_sideport_present(struct radeon_device *rdev)
+{
+	struct drm_device *dev = rdev->ddev;
+	u16 igp_info;
+
+	igp_info = combios_get_table_offset(dev, COMBIOS_INTEGRATED_SYSTEM_INFO_TABLE);
+
+	if (igp_info) {
+		if (RBIOS16(igp_info + 0x4))
+			return true;
+	}
+	return false;
+}
+
+static const uint32_t default_primarydac_adj[CHIP_LAST] = {
+	0x00000808,		/* r100  */
+	0x00000808,		/* rv100 */
+	0x00000808,		/* rs100 */
+	0x00000808,		/* rv200 */
+	0x00000808,		/* rs200 */
+	0x00000808,		/* r200  */
+	0x00000808,		/* rv250 */
+	0x00000000,		/* rs300 */
+	0x00000808,		/* rv280 */
+	0x00000808,		/* r300  */
+	0x00000808,		/* r350  */
+	0x00000808,		/* rv350 */
+	0x00000808,		/* rv380 */
+	0x00000808,		/* r420  */
+	0x00000808,		/* r423  */
+	0x00000808,		/* rv410 */
+	0x00000000,		/* rs400 */
+	0x00000000,		/* rs480 */
+};
+
+static void radeon_legacy_get_primary_dac_info_from_table(struct radeon_device *rdev,
+							  struct radeon_encoder_primary_dac *p_dac)
+{
+	p_dac->ps2_pdac_adj = default_primarydac_adj[rdev->family];
+	return;
+}
+
 struct radeon_encoder_primary_dac *radeon_combios_get_primary_dac_info(struct
 								       radeon_encoder
 								       *encoder)
@@ -570,20 +740,17 @@ struct radeon_encoder_primary_dac *radeon_combios_get_primary_dac_info(struct
 	uint16_t dac_info;
 	uint8_t rev, bg, dac;
 	struct radeon_encoder_primary_dac *p_dac = NULL;
+	int found = 0;
 
-	if (rdev->bios == NULL)
+	p_dac = kzalloc(sizeof(struct radeon_encoder_primary_dac),
+			GFP_KERNEL);
+
+	if (!p_dac)
 		return NULL;
 
 	/* check CRT table */
 	dac_info = combios_get_table_offset(dev, COMBIOS_CRT_INFO_TABLE);
 	if (dac_info) {
-		p_dac =
-		    kzalloc(sizeof(struct radeon_encoder_primary_dac),
-			    GFP_KERNEL);
-
-		if (!p_dac)
-			return NULL;
-
 		rev = RBIOS8(dac_info) & 0x3;
 		if (rev < 2) {
 			bg = RBIOS8(dac_info + 0x2) & 0xf;
@@ -594,17 +761,19 @@ struct radeon_encoder_primary_dac *radeon_combios_get_primary_dac_info(struct
 			dac = RBIOS8(dac_info + 0x3) & 0xf;
 			p_dac->ps2_pdac_adj = (bg << 8) | (dac);
 		}
-
+		found = 1;
 	}
 
+	if (!found) /* fallback to defaults */
+		radeon_legacy_get_primary_dac_info_from_table(rdev, p_dac);
+
 	return p_dac;
 }
 
-static enum radeon_tv_std
-radeon_combios_get_tv_info(struct radeon_encoder *encoder)
+enum radeon_tv_std
+radeon_combios_get_tv_info(struct radeon_device *rdev)
 {
-	struct drm_device *dev = encoder->base.dev;
-	struct radeon_device *rdev = dev->dev_private;
+	struct drm_device *dev = rdev->ddev;
 	uint16_t tv_info;
 	enum radeon_tv_std tv_std = TV_STD_NTSC;
 
@@ -711,9 +880,6 @@ struct radeon_encoder_tv_dac *radeon_combios_get_tv_dac_info(struct
 	if (!tv_dac)
 		return NULL;
 
-	if (rdev->bios == NULL)
-		goto out;
-
 	/* first check TV table */
 	dac_info = combios_get_table_offset(dev, COMBIOS_TV_INFO_TABLE);
 	if (dac_info) {
@@ -745,7 +911,7 @@ struct radeon_encoder_tv_dac *radeon_combios_get_tv_dac_info(struct
 			tv_dac->ntsc_tvdac_adj = (bg << 16) | (dac << 20);
 			found = 1;
 		}
-		tv_dac->tv_std = radeon_combios_get_tv_info(encoder);
+		tv_dac->tv_std = radeon_combios_get_tv_info(rdev);
 	}
 	if (!found) {
 		/* then check CRT table */
@@ -775,7 +941,6 @@ struct radeon_encoder_tv_dac *radeon_combios_get_tv_dac_info(struct
 		}
 	}
 
-out:
 	if (!found) /* fallback to defaults */
 		radeon_legacy_get_tv_dac_info_from_table(rdev, tv_dac);
 
@@ -808,25 +973,25 @@ static struct radeon_encoder_lvds *radeon_legacy_get_lvds_info_from_regs(struct
 	lvds->panel_blon_delay = (lvds_ss_gen_cntl >> RADEON_LVDS_PWRSEQ_DELAY2_SHIFT) & 0xf;
 
 	if (fp_vert_stretch & RADEON_VERT_STRETCH_ENABLE)
-		lvds->native_mode.panel_yres =
+		lvds->native_mode.vdisplay =
 		    ((fp_vert_stretch & RADEON_VERT_PANEL_SIZE) >>
 		     RADEON_VERT_PANEL_SHIFT) + 1;
 	else
-		lvds->native_mode.panel_yres =
+		lvds->native_mode.vdisplay =
 		    (RREG32(RADEON_CRTC_V_TOTAL_DISP) >> 16) + 1;
 
 	if (fp_horz_stretch & RADEON_HORZ_STRETCH_ENABLE)
-		lvds->native_mode.panel_xres =
+		lvds->native_mode.hdisplay =
 		    (((fp_horz_stretch & RADEON_HORZ_PANEL_SIZE) >>
 		      RADEON_HORZ_PANEL_SHIFT) + 1) * 8;
 	else
-		lvds->native_mode.panel_xres =
+		lvds->native_mode.hdisplay =
 		    ((RREG32(RADEON_CRTC_H_TOTAL_DISP) >> 16) + 1) * 8;
 
-	if ((lvds->native_mode.panel_xres < 640) ||
-	    (lvds->native_mode.panel_yres < 480)) {
-		lvds->native_mode.panel_xres = 640;
-		lvds->native_mode.panel_yres = 480;
+	if ((lvds->native_mode.hdisplay < 640) ||
+	    (lvds->native_mode.vdisplay < 480)) {
+		lvds->native_mode.hdisplay = 640;
+		lvds->native_mode.vdisplay = 480;
 	}
 
 	ppll_div_sel = RREG8(RADEON_CLOCK_CNTL_INDEX + 1) & 0x3;
@@ -846,8 +1011,8 @@ static struct radeon_encoder_lvds *radeon_legacy_get_lvds_info_from_regs(struct
 	lvds->panel_vcc_delay = 200;
 
 	DRM_INFO("Panel info derived from registers\n");
-	DRM_INFO("Panel Size %dx%d\n", lvds->native_mode.panel_xres,
-		 lvds->native_mode.panel_yres);
+	DRM_INFO("Panel Size %dx%d\n", lvds->native_mode.hdisplay,
+		 lvds->native_mode.vdisplay);
 
 	return lvds;
 }
@@ -863,11 +1028,6 @@ struct radeon_encoder_lvds *radeon_combios_get_lvds_info(struct radeon_encoder
 	int tmp, i;
 	struct radeon_encoder_lvds *lvds = NULL;
 
-	if (rdev->bios == NULL) {
-		lvds = radeon_legacy_get_lvds_info_from_regs(rdev);
-		goto out;
-	}
-
 	lcd_info = combios_get_table_offset(dev, COMBIOS_LCD_INFO_TABLE);
 
 	if (lcd_info) {
@@ -882,15 +1042,14 @@ struct radeon_encoder_lvds *radeon_combios_get_lvds_info(struct radeon_encoder
 
 		DRM_INFO("Panel ID String: %s\n", stmp);
 
-		lvds->native_mode.panel_xres = RBIOS16(lcd_info + 0x19);
-		lvds->native_mode.panel_yres = RBIOS16(lcd_info + 0x1b);
+		lvds->native_mode.hdisplay = RBIOS16(lcd_info + 0x19);
+		lvds->native_mode.vdisplay = RBIOS16(lcd_info + 0x1b);
 
-		DRM_INFO("Panel Size %dx%d\n", lvds->native_mode.panel_xres,
-			 lvds->native_mode.panel_yres);
+		DRM_INFO("Panel Size %dx%d\n", lvds->native_mode.hdisplay,
+			 lvds->native_mode.vdisplay);
 
 		lvds->panel_vcc_delay = RBIOS16(lcd_info + 0x2c);
-		if (lvds->panel_vcc_delay > 2000 || lvds->panel_vcc_delay < 0)
-			lvds->panel_vcc_delay = 2000;
+		lvds->panel_vcc_delay = min_t(u16, lvds->panel_vcc_delay, 2000);
 
 		lvds->panel_pwr_delay = RBIOS8(lcd_info + 0x24);
 		lvds->panel_digon_delay = RBIOS16(lcd_info + 0x38) & 0xf;
@@ -944,34 +1103,32 @@ struct radeon_encoder_lvds *radeon_combios_get_lvds_info(struct radeon_encoder
 			if (tmp == 0)
 				break;
 
-			if ((RBIOS16(tmp) == lvds->native_mode.panel_xres) &&
+			if ((RBIOS16(tmp) == lvds->native_mode.hdisplay) &&
 			    (RBIOS16(tmp + 2) ==
-			     lvds->native_mode.panel_yres)) {
-				lvds->native_mode.hblank =
-				    (RBIOS16(tmp + 17) - RBIOS16(tmp + 19)) * 8;
-				lvds->native_mode.hoverplus =
-				    (RBIOS16(tmp + 21) - RBIOS16(tmp + 19) -
-				     1) * 8;
-				lvds->native_mode.hsync_width =
-				    RBIOS8(tmp + 23) * 8;
-
-				lvds->native_mode.vblank = (RBIOS16(tmp + 24) -
-							    RBIOS16(tmp + 26));
-				lvds->native_mode.voverplus =
-				    ((RBIOS16(tmp + 28) & 0x7ff) -
-				     RBIOS16(tmp + 26));
-				lvds->native_mode.vsync_width =
-				    ((RBIOS16(tmp + 28) & 0xf800) >> 11);
-				lvds->native_mode.dotclock =
-				    RBIOS16(tmp + 9) * 10;
+			     lvds->native_mode.vdisplay)) {
+				lvds->native_mode.htotal = RBIOS16(tmp + 17) * 8;
+				lvds->native_mode.hsync_start = RBIOS16(tmp + 21) * 8;
+				lvds->native_mode.hsync_end = (RBIOS8(tmp + 23) +
+							       RBIOS16(tmp + 21)) * 8;
+
+				lvds->native_mode.vtotal = RBIOS16(tmp + 24);
+				lvds->native_mode.vsync_start = RBIOS16(tmp + 28) & 0x7ff;
+				lvds->native_mode.vsync_end =
+					((RBIOS16(tmp + 28) & 0xf800) >> 11) +
+					(RBIOS16(tmp + 28) & 0x7ff);
+
+				lvds->native_mode.clock = RBIOS16(tmp + 9) * 10;
 				lvds->native_mode.flags = 0;
+				/* set crtc values */
+				drm_mode_set_crtcinfo(&lvds->native_mode, CRTC_INTERLACE_HALVE_V);
+
 			}
 		}
 	} else {
 		DRM_INFO("No panel info found in BIOS\n");
 		lvds = radeon_legacy_get_lvds_info_from_regs(rdev);
 	}
-out:
+
 	if (lvds)
 		encoder->native_mode = lvds->native_mode;
 	return lvds;
@@ -994,8 +1151,8 @@ static const struct radeon_tmds_pll default_tmds_pll[CHIP_LAST][4] = {
 	{{0xffffffff, 0xb01cb}, {0, 0}, {0, 0}, {0, 0}},	/* CHIP_R420  */
 	{{0xffffffff, 0xb01cb}, {0, 0}, {0, 0}, {0, 0}},	/* CHIP_R423  */
 	{{0xffffffff, 0xb01cb}, {0, 0}, {0, 0}, {0, 0}},	/* CHIP_RV410 */
-	{{15000, 0xb0155}, {0xffffffff, 0xb01cb}, {0, 0}, {0, 0}},	/* CHIP_RS400 */
-	{{15000, 0xb0155}, {0xffffffff, 0xb01cb}, {0, 0}, {0, 0}},	/* CHIP_RS480 */
+	{ {0, 0}, {0, 0}, {0, 0}, {0, 0} },	/* CHIP_RS400 */
+	{ {0, 0}, {0, 0}, {0, 0}, {0, 0} },	/* CHIP_RS480 */
 };
 
 bool radeon_legacy_get_tmds_info_from_table(struct radeon_encoder *encoder,
@@ -1023,13 +1180,9 @@ bool radeon_legacy_get_tmds_info_from_combios(struct radeon_encoder *encoder,
 	int i, n;
 	uint8_t ver;
 
-	if (rdev->bios == NULL)
-		return false;
-
 	tmds_info = combios_get_table_offset(dev, COMBIOS_DFP_INFO_TABLE);
 
 	if (tmds_info) {
-
 		ver = RBIOS8(tmds_info);
 		DRM_INFO("DFP table revision: %d\n", ver);
 		if (ver == 3) {
@@ -1064,96 +1217,184 @@ bool radeon_legacy_get_tmds_info_from_combios(struct radeon_encoder *encoder,
 					  tmds->tmds_pll[i].value);
 			}
 		}
-	} else
+	} else {
 		DRM_INFO("No TMDS info found in BIOS\n");
+		return false;
+	}
 	return true;
 }
 
-struct radeon_encoder_int_tmds *radeon_combios_get_tmds_info(struct radeon_encoder *encoder)
+bool radeon_legacy_get_ext_tmds_info_from_table(struct radeon_encoder *encoder,
+						struct radeon_encoder_ext_tmds *tmds)
 {
-	struct radeon_encoder_int_tmds *tmds = NULL;
-	bool ret;
-
-	tmds = kzalloc(sizeof(struct radeon_encoder_int_tmds), GFP_KERNEL);
+	struct drm_device *dev = encoder->base.dev;
+	struct radeon_device *rdev = dev->dev_private;
+	struct radeon_i2c_bus_rec i2c_bus;
 
-	if (!tmds)
-		return NULL;
+	/* default for macs */
+	i2c_bus = combios_setup_i2c_bus(rdev, RADEON_GPIO_MONID);
+	tmds->i2c_bus = radeon_i2c_create(dev, &i2c_bus, "DVO");
 
-	ret = radeon_legacy_get_tmds_info_from_combios(encoder, tmds);
-	if (ret == false)
-		radeon_legacy_get_tmds_info_from_table(encoder, tmds);
+	/* XXX some macs have duallink chips */
+	switch (rdev->mode_info.connector_table) {
+	case CT_POWERBOOK_EXTERNAL:
+	case CT_MINI_EXTERNAL:
+	default:
+		tmds->dvo_chip = DVO_SIL164;
+		tmds->slave_addr = 0x70 >> 1; /* 7 bit addressing */
+		break;
+	}
 
-	return tmds;
+	return true;
 }
 
-void radeon_combios_get_ext_tmds_info(struct radeon_encoder *encoder)
+bool radeon_legacy_get_ext_tmds_info_from_combios(struct radeon_encoder *encoder,
+						  struct radeon_encoder_ext_tmds *tmds)
 {
 	struct drm_device *dev = encoder->base.dev;
 	struct radeon_device *rdev = dev->dev_private;
-	uint16_t ext_tmds_info;
-	uint8_t ver;
+	uint16_t offset;
+	uint8_t ver, id, blocks, clk, data;
+	int i;
+	enum radeon_combios_ddc gpio;
+	struct radeon_i2c_bus_rec i2c_bus;
 
-	if (rdev->bios == NULL)
-		return;
+	tmds->i2c_bus = NULL;
+	if (rdev->flags & RADEON_IS_IGP) {
+		offset = combios_get_table_offset(dev, COMBIOS_I2C_INFO_TABLE);
+		if (offset) {
+			ver = RBIOS8(offset);
+			DRM_INFO("GPIO Table revision: %d\n", ver);
+			blocks = RBIOS8(offset + 2);
+			for (i = 0; i < blocks; i++) {
+				id = RBIOS8(offset + 3 + (i * 5) + 0);
+				if (id == 136) {
+					clk = RBIOS8(offset + 3 + (i * 5) + 3);
+					data = RBIOS8(offset + 3 + (i * 5) + 4);
+					i2c_bus.valid = true;
+					i2c_bus.mask_clk_mask = (1 << clk);
+					i2c_bus.mask_data_mask = (1 << data);
+					i2c_bus.a_clk_mask = (1 << clk);
+					i2c_bus.a_data_mask = (1 << data);
+					i2c_bus.en_clk_mask = (1 << clk);
+					i2c_bus.en_data_mask = (1 << data);
+					i2c_bus.y_clk_mask = (1 << clk);
+					i2c_bus.y_data_mask = (1 << data);
+					i2c_bus.mask_clk_reg = RADEON_GPIOPAD_MASK;
+					i2c_bus.mask_data_reg = RADEON_GPIOPAD_MASK;
+					i2c_bus.a_clk_reg = RADEON_GPIOPAD_A;
+					i2c_bus.a_data_reg = RADEON_GPIOPAD_A;
+					i2c_bus.en_clk_reg = RADEON_GPIOPAD_EN;
+					i2c_bus.en_data_reg = RADEON_GPIOPAD_EN;
+					i2c_bus.y_clk_reg = RADEON_GPIOPAD_Y;
+					i2c_bus.y_data_reg = RADEON_GPIOPAD_Y;
+					tmds->i2c_bus = radeon_i2c_create(dev, &i2c_bus, "DVO");
+					tmds->dvo_chip = DVO_SIL164;
+					tmds->slave_addr = 0x70 >> 1; /* 7 bit addressing */
+					break;
+				}
+			}
+		}
+	} else {
+		offset = combios_get_table_offset(dev, COMBIOS_EXT_TMDS_INFO_TABLE);
+		if (offset) {
+			ver = RBIOS8(offset);
+			DRM_INFO("External TMDS Table revision: %d\n", ver);
+			tmds->slave_addr = RBIOS8(offset + 4 + 2);
+			tmds->slave_addr >>= 1; /* 7 bit addressing */
+			gpio = RBIOS8(offset + 4 + 3);
+			switch (gpio) {
+			case DDC_MONID:
+				i2c_bus = combios_setup_i2c_bus(rdev, RADEON_GPIO_MONID);
+				tmds->i2c_bus = radeon_i2c_create(dev, &i2c_bus, "DVO");
+				break;
+			case DDC_DVI:
+				i2c_bus = combios_setup_i2c_bus(rdev, RADEON_GPIO_DVI_DDC);
+				tmds->i2c_bus = radeon_i2c_create(dev, &i2c_bus, "DVO");
+				break;
+			case DDC_VGA:
+				i2c_bus = combios_setup_i2c_bus(rdev, RADEON_GPIO_VGA_DDC);
+				tmds->i2c_bus = radeon_i2c_create(dev, &i2c_bus, "DVO");
+				break;
+			case DDC_CRT2:
+				/* R3xx+ chips don't have GPIO_CRT2_DDC gpio pad */
+				if (rdev->family >= CHIP_R300)
+					i2c_bus = combios_setup_i2c_bus(rdev, RADEON_GPIO_MONID);
+				else
+					i2c_bus = combios_setup_i2c_bus(rdev, RADEON_GPIO_CRT2_DDC);
+				tmds->i2c_bus = radeon_i2c_create(dev, &i2c_bus, "DVO");
+				break;
+			case DDC_LCD: /* MM i2c */
+				i2c_bus.valid = true;
+				i2c_bus.hw_capable = true;
+				i2c_bus.mm_i2c = true;
+				tmds->i2c_bus = radeon_i2c_create(dev, &i2c_bus, "DVO");
+				break;
+			default:
+				DRM_ERROR("Unsupported gpio %d\n", gpio);
+				break;
+			}
+		}
+	}
 
-	ext_tmds_info =
-	    combios_get_table_offset(dev, COMBIOS_EXT_TMDS_INFO_TABLE);
-	if (ext_tmds_info) {
-		ver = RBIOS8(ext_tmds_info);
-		DRM_INFO("External TMDS Table revision: %d\n", ver);
-		// TODO
+	if (!tmds->i2c_bus) {
+		DRM_INFO("No valid Ext TMDS info found in BIOS\n");
+		return false;
 	}
+
+	return true;
 }
 
 bool radeon_get_legacy_connector_info_from_table(struct drm_device *dev)
 {
 	struct radeon_device *rdev = dev->dev_private;
 	struct radeon_i2c_bus_rec ddc_i2c;
+	struct radeon_hpd hpd;
 
 	rdev->mode_info.connector_table = radeon_connector_table;
 	if (rdev->mode_info.connector_table == CT_NONE) {
 #ifdef CONFIG_PPC_PMAC
-		if (machine_is_compatible("PowerBook3,3")) {
+		if (of_machine_is_compatible("PowerBook3,3")) {
 			/* powerbook with VGA */
 			rdev->mode_info.connector_table = CT_POWERBOOK_VGA;
-		} else if (machine_is_compatible("PowerBook3,4") ||
-			   machine_is_compatible("PowerBook3,5")) {
+		} else if (of_machine_is_compatible("PowerBook3,4") ||
+			   of_machine_is_compatible("PowerBook3,5")) {
 			/* powerbook with internal tmds */
 			rdev->mode_info.connector_table = CT_POWERBOOK_INTERNAL;
-		} else if (machine_is_compatible("PowerBook5,1") ||
-			   machine_is_compatible("PowerBook5,2") ||
-			   machine_is_compatible("PowerBook5,3") ||
-			   machine_is_compatible("PowerBook5,4") ||
-			   machine_is_compatible("PowerBook5,5")) {
+		} else if (of_machine_is_compatible("PowerBook5,1") ||
+			   of_machine_is_compatible("PowerBook5,2") ||
+			   of_machine_is_compatible("PowerBook5,3") ||
+			   of_machine_is_compatible("PowerBook5,4") ||
+			   of_machine_is_compatible("PowerBook5,5")) {
 			/* powerbook with external single link tmds (sil164) */
 			rdev->mode_info.connector_table = CT_POWERBOOK_EXTERNAL;
-		} else if (machine_is_compatible("PowerBook5,6")) {
+		} else if (of_machine_is_compatible("PowerBook5,6")) {
 			/* powerbook with external dual or single link tmds */
 			rdev->mode_info.connector_table = CT_POWERBOOK_EXTERNAL;
-		} else if (machine_is_compatible("PowerBook5,7") ||
-			   machine_is_compatible("PowerBook5,8") ||
-			   machine_is_compatible("PowerBook5,9")) {
+		} else if (of_machine_is_compatible("PowerBook5,7") ||
+			   of_machine_is_compatible("PowerBook5,8") ||
+			   of_machine_is_compatible("PowerBook5,9")) {
 			/* PowerBook6,2 ? */
 			/* powerbook with external dual link tmds (sil1178?) */
 			rdev->mode_info.connector_table = CT_POWERBOOK_EXTERNAL;
-		} else if (machine_is_compatible("PowerBook4,1") ||
-			   machine_is_compatible("PowerBook4,2") ||
-			   machine_is_compatible("PowerBook4,3") ||
-			   machine_is_compatible("PowerBook6,3") ||
-			   machine_is_compatible("PowerBook6,5") ||
-			   machine_is_compatible("PowerBook6,7")) {
+		} else if (of_machine_is_compatible("PowerBook4,1") ||
+			   of_machine_is_compatible("PowerBook4,2") ||
+			   of_machine_is_compatible("PowerBook4,3") ||
+			   of_machine_is_compatible("PowerBook6,3") ||
+			   of_machine_is_compatible("PowerBook6,5") ||
+			   of_machine_is_compatible("PowerBook6,7")) {
 			/* ibook */
 			rdev->mode_info.connector_table = CT_IBOOK;
-		} else if (machine_is_compatible("PowerMac4,4")) {
+		} else if (of_machine_is_compatible("PowerMac4,4")) {
 			/* emac */
 			rdev->mode_info.connector_table = CT_EMAC;
-		} else if (machine_is_compatible("PowerMac10,1")) {
+		} else if (of_machine_is_compatible("PowerMac10,1")) {
 			/* mini with internal tmds */
 			rdev->mode_info.connector_table = CT_MINI_INTERNAL;
-		} else if (machine_is_compatible("PowerMac10,2")) {
+		} else if (of_machine_is_compatible("PowerMac10,2")) {
 			/* mini with external tmds */
 			rdev->mode_info.connector_table = CT_MINI_EXTERNAL;
-		} else if (machine_is_compatible("PowerMac12,1")) {
+		} else if (of_machine_is_compatible("PowerMac12,1")) {
 			/* PowerMac8,1 ? */
 			/* imac g5 isight */
 			rdev->mode_info.connector_table = CT_IMAC_G5_ISIGHT;
@@ -1169,7 +1410,8 @@ bool radeon_get_legacy_connector_info_from_table(struct drm_device *dev)
 		/* these are the most common settings */
 		if (rdev->flags & RADEON_SINGLE_CRTC) {
 			/* VGA - primary dac */
-			ddc_i2c = combios_setup_i2c_bus(RADEON_GPIO_VGA_DDC);
+			ddc_i2c = combios_setup_i2c_bus(rdev, RADEON_GPIO_VGA_DDC);
+			hpd.hpd = RADEON_HPD_NONE;
 			radeon_add_legacy_encoder(dev,
 						  radeon_get_encoder_id(dev,
 									ATOM_DEVICE_CRT1_SUPPORT,
@@ -1178,10 +1420,13 @@ bool radeon_get_legacy_connector_info_from_table(struct drm_device *dev)
 			radeon_add_legacy_connector(dev, 0,
 						    ATOM_DEVICE_CRT1_SUPPORT,
 						    DRM_MODE_CONNECTOR_VGA,
-						    &ddc_i2c);
+						    &ddc_i2c,
+						    CONNECTOR_OBJECT_ID_VGA,
+						    &hpd);
 		} else if (rdev->flags & RADEON_IS_MOBILITY) {
 			/* LVDS */
-			ddc_i2c = combios_setup_i2c_bus(RADEON_LCD_GPIO_MASK);
+			ddc_i2c = combios_setup_i2c_bus(rdev, 0);
+			hpd.hpd = RADEON_HPD_NONE;
 			radeon_add_legacy_encoder(dev,
 						  radeon_get_encoder_id(dev,
 									ATOM_DEVICE_LCD1_SUPPORT,
@@ -1190,10 +1435,13 @@ bool radeon_get_legacy_connector_info_from_table(struct drm_device *dev)
 			radeon_add_legacy_connector(dev, 0,
 						    ATOM_DEVICE_LCD1_SUPPORT,
 						    DRM_MODE_CONNECTOR_LVDS,
-						    &ddc_i2c);
+						    &ddc_i2c,
+						    CONNECTOR_OBJECT_ID_LVDS,
+						    &hpd);
 
 			/* VGA - primary dac */
-			ddc_i2c = combios_setup_i2c_bus(RADEON_GPIO_VGA_DDC);
+			ddc_i2c = combios_setup_i2c_bus(rdev, RADEON_GPIO_VGA_DDC);
+			hpd.hpd = RADEON_HPD_NONE;
 			radeon_add_legacy_encoder(dev,
 						  radeon_get_encoder_id(dev,
 									ATOM_DEVICE_CRT1_SUPPORT,
@@ -1202,10 +1450,13 @@ bool radeon_get_legacy_connector_info_from_table(struct drm_device *dev)
 			radeon_add_legacy_connector(dev, 1,
 						    ATOM_DEVICE_CRT1_SUPPORT,
 						    DRM_MODE_CONNECTOR_VGA,
-						    &ddc_i2c);
+						    &ddc_i2c,
+						    CONNECTOR_OBJECT_ID_VGA,
+						    &hpd);
 		} else {
 			/* DVI-I - tv dac, int tmds */
-			ddc_i2c = combios_setup_i2c_bus(RADEON_GPIO_DVI_DDC);
+			ddc_i2c = combios_setup_i2c_bus(rdev, RADEON_GPIO_DVI_DDC);
+			hpd.hpd = RADEON_HPD_1;
 			radeon_add_legacy_encoder(dev,
 						  radeon_get_encoder_id(dev,
 									ATOM_DEVICE_DFP1_SUPPORT,
@@ -1220,10 +1471,13 @@ bool radeon_get_legacy_connector_info_from_table(struct drm_device *dev)
 						    ATOM_DEVICE_DFP1_SUPPORT |
 						    ATOM_DEVICE_CRT2_SUPPORT,
 						    DRM_MODE_CONNECTOR_DVII,
-						    &ddc_i2c);
+						    &ddc_i2c,
+						    CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_I,
+						    &hpd);
 
 			/* VGA - primary dac */
-			ddc_i2c = combios_setup_i2c_bus(RADEON_GPIO_VGA_DDC);
+			ddc_i2c = combios_setup_i2c_bus(rdev, RADEON_GPIO_VGA_DDC);
+			hpd.hpd = RADEON_HPD_NONE;
 			radeon_add_legacy_encoder(dev,
 						  radeon_get_encoder_id(dev,
 									ATOM_DEVICE_CRT1_SUPPORT,
@@ -1232,11 +1486,15 @@ bool radeon_get_legacy_connector_info_from_table(struct drm_device *dev)
 			radeon_add_legacy_connector(dev, 1,
 						    ATOM_DEVICE_CRT1_SUPPORT,
 						    DRM_MODE_CONNECTOR_VGA,
-						    &ddc_i2c);
+						    &ddc_i2c,
+						    CONNECTOR_OBJECT_ID_VGA,
+						    &hpd);
 		}
 
 		if (rdev->family != CHIP_R100 && rdev->family != CHIP_R200) {
 			/* TV - tv dac */
+			ddc_i2c.valid = false;
+			hpd.hpd = RADEON_HPD_NONE;
 			radeon_add_legacy_encoder(dev,
 						  radeon_get_encoder_id(dev,
 									ATOM_DEVICE_TV1_SUPPORT,
@@ -1245,31 +1503,41 @@ bool radeon_get_legacy_connector_info_from_table(struct drm_device *dev)
 			radeon_add_legacy_connector(dev, 2,
 						    ATOM_DEVICE_TV1_SUPPORT,
 						    DRM_MODE_CONNECTOR_SVIDEO,
-						    &ddc_i2c);
+						    &ddc_i2c,
+						    CONNECTOR_OBJECT_ID_SVIDEO,
+						    &hpd);
 		}
 		break;
 	case CT_IBOOK:
 		DRM_INFO("Connector Table: %d (ibook)\n",
 			 rdev->mode_info.connector_table);
 		/* LVDS */
-		ddc_i2c = combios_setup_i2c_bus(RADEON_GPIO_DVI_DDC);
+		ddc_i2c = combios_setup_i2c_bus(rdev, RADEON_GPIO_DVI_DDC);
+		hpd.hpd = RADEON_HPD_NONE;
 		radeon_add_legacy_encoder(dev,
 					  radeon_get_encoder_id(dev,
 								ATOM_DEVICE_LCD1_SUPPORT,
 								0),
 					  ATOM_DEVICE_LCD1_SUPPORT);
 		radeon_add_legacy_connector(dev, 0, ATOM_DEVICE_LCD1_SUPPORT,
-					    DRM_MODE_CONNECTOR_LVDS, &ddc_i2c);
+					    DRM_MODE_CONNECTOR_LVDS, &ddc_i2c,
+					    CONNECTOR_OBJECT_ID_LVDS,
+					    &hpd);
 		/* VGA - TV DAC */
-		ddc_i2c = combios_setup_i2c_bus(RADEON_GPIO_VGA_DDC);
+		ddc_i2c = combios_setup_i2c_bus(rdev, RADEON_GPIO_VGA_DDC);
+		hpd.hpd = RADEON_HPD_NONE;
 		radeon_add_legacy_encoder(dev,
 					  radeon_get_encoder_id(dev,
 								ATOM_DEVICE_CRT2_SUPPORT,
 								2),
 					  ATOM_DEVICE_CRT2_SUPPORT);
 		radeon_add_legacy_connector(dev, 1, ATOM_DEVICE_CRT2_SUPPORT,
-					    DRM_MODE_CONNECTOR_VGA, &ddc_i2c);
+					    DRM_MODE_CONNECTOR_VGA, &ddc_i2c,
+					    CONNECTOR_OBJECT_ID_VGA,
+					    &hpd);
 		/* TV - TV DAC */
+		ddc_i2c.valid = false;
+		hpd.hpd = RADEON_HPD_NONE;
 		radeon_add_legacy_encoder(dev,
 					  radeon_get_encoder_id(dev,
 								ATOM_DEVICE_TV1_SUPPORT,
@@ -1277,22 +1545,28 @@ bool radeon_get_legacy_connector_info_from_table(struct drm_device *dev)
 					  ATOM_DEVICE_TV1_SUPPORT);
 		radeon_add_legacy_connector(dev, 2, ATOM_DEVICE_TV1_SUPPORT,
 					    DRM_MODE_CONNECTOR_SVIDEO,
-					    &ddc_i2c);
+					    &ddc_i2c,
+					    CONNECTOR_OBJECT_ID_SVIDEO,
+					    &hpd);
 		break;
 	case CT_POWERBOOK_EXTERNAL:
 		DRM_INFO("Connector Table: %d (powerbook external tmds)\n",
 			 rdev->mode_info.connector_table);
 		/* LVDS */
-		ddc_i2c = combios_setup_i2c_bus(RADEON_GPIO_DVI_DDC);
+		ddc_i2c = combios_setup_i2c_bus(rdev, RADEON_GPIO_DVI_DDC);
+		hpd.hpd = RADEON_HPD_NONE;
 		radeon_add_legacy_encoder(dev,
 					  radeon_get_encoder_id(dev,
 								ATOM_DEVICE_LCD1_SUPPORT,
 								0),
 					  ATOM_DEVICE_LCD1_SUPPORT);
 		radeon_add_legacy_connector(dev, 0, ATOM_DEVICE_LCD1_SUPPORT,
-					    DRM_MODE_CONNECTOR_LVDS, &ddc_i2c);
+					    DRM_MODE_CONNECTOR_LVDS, &ddc_i2c,
+					    CONNECTOR_OBJECT_ID_LVDS,
+					    &hpd);
 		/* DVI-I - primary dac, ext tmds */
-		ddc_i2c = combios_setup_i2c_bus(RADEON_GPIO_VGA_DDC);
+		ddc_i2c = combios_setup_i2c_bus(rdev, RADEON_GPIO_VGA_DDC);
+		hpd.hpd = RADEON_HPD_2; /* ??? */
 		radeon_add_legacy_encoder(dev,
 					  radeon_get_encoder_id(dev,
 								ATOM_DEVICE_DFP2_SUPPORT,
@@ -1303,11 +1577,16 @@ bool radeon_get_legacy_connector_info_from_table(struct drm_device *dev)
 								ATOM_DEVICE_CRT1_SUPPORT,
 								1),
 					  ATOM_DEVICE_CRT1_SUPPORT);
+		/* XXX some are SL */
 		radeon_add_legacy_connector(dev, 1,
 					    ATOM_DEVICE_DFP2_SUPPORT |
 					    ATOM_DEVICE_CRT1_SUPPORT,
-					    DRM_MODE_CONNECTOR_DVII, &ddc_i2c);
+					    DRM_MODE_CONNECTOR_DVII, &ddc_i2c,
+					    CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_I,
+					    &hpd);
 		/* TV - TV DAC */
+		ddc_i2c.valid = false;
+		hpd.hpd = RADEON_HPD_NONE;
 		radeon_add_legacy_encoder(dev,
 					  radeon_get_encoder_id(dev,
 								ATOM_DEVICE_TV1_SUPPORT,
@@ -1315,22 +1594,28 @@ bool radeon_get_legacy_connector_info_from_table(struct drm_device *dev)
 					  ATOM_DEVICE_TV1_SUPPORT);
 		radeon_add_legacy_connector(dev, 2, ATOM_DEVICE_TV1_SUPPORT,
 					    DRM_MODE_CONNECTOR_SVIDEO,
-					    &ddc_i2c);
+					    &ddc_i2c,
+					    CONNECTOR_OBJECT_ID_SVIDEO,
+					    &hpd);
 		break;
 	case CT_POWERBOOK_INTERNAL:
 		DRM_INFO("Connector Table: %d (powerbook internal tmds)\n",
 			 rdev->mode_info.connector_table);
 		/* LVDS */
-		ddc_i2c = combios_setup_i2c_bus(RADEON_GPIO_DVI_DDC);
+		ddc_i2c = combios_setup_i2c_bus(rdev, RADEON_GPIO_DVI_DDC);
+		hpd.hpd = RADEON_HPD_NONE;
 		radeon_add_legacy_encoder(dev,
 					  radeon_get_encoder_id(dev,
 								ATOM_DEVICE_LCD1_SUPPORT,
 								0),
 					  ATOM_DEVICE_LCD1_SUPPORT);
 		radeon_add_legacy_connector(dev, 0, ATOM_DEVICE_LCD1_SUPPORT,
-					    DRM_MODE_CONNECTOR_LVDS, &ddc_i2c);
+					    DRM_MODE_CONNECTOR_LVDS, &ddc_i2c,
+					    CONNECTOR_OBJECT_ID_LVDS,
+					    &hpd);
 		/* DVI-I - primary dac, int tmds */
-		ddc_i2c = combios_setup_i2c_bus(RADEON_GPIO_VGA_DDC);
+		ddc_i2c = combios_setup_i2c_bus(rdev, RADEON_GPIO_VGA_DDC);
+		hpd.hpd = RADEON_HPD_1; /* ??? */
 		radeon_add_legacy_encoder(dev,
 					  radeon_get_encoder_id(dev,
 								ATOM_DEVICE_DFP1_SUPPORT,
@@ -1344,8 +1629,12 @@ bool radeon_get_legacy_connector_info_from_table(struct drm_device *dev)
 		radeon_add_legacy_connector(dev, 1,
 					    ATOM_DEVICE_DFP1_SUPPORT |
 					    ATOM_DEVICE_CRT1_SUPPORT,
-					    DRM_MODE_CONNECTOR_DVII, &ddc_i2c);
+					    DRM_MODE_CONNECTOR_DVII, &ddc_i2c,
+					    CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_I,
+					    &hpd);
 		/* TV - TV DAC */
+		ddc_i2c.valid = false;
+		hpd.hpd = RADEON_HPD_NONE;
 		radeon_add_legacy_encoder(dev,
 					  radeon_get_encoder_id(dev,
 								ATOM_DEVICE_TV1_SUPPORT,
@@ -1353,30 +1642,40 @@ bool radeon_get_legacy_connector_info_from_table(struct drm_device *dev)
 					  ATOM_DEVICE_TV1_SUPPORT);
 		radeon_add_legacy_connector(dev, 2, ATOM_DEVICE_TV1_SUPPORT,
 					    DRM_MODE_CONNECTOR_SVIDEO,
-					    &ddc_i2c);
+					    &ddc_i2c,
+					    CONNECTOR_OBJECT_ID_SVIDEO,
+					    &hpd);
 		break;
 	case CT_POWERBOOK_VGA:
 		DRM_INFO("Connector Table: %d (powerbook vga)\n",
 			 rdev->mode_info.connector_table);
 		/* LVDS */
-		ddc_i2c = combios_setup_i2c_bus(RADEON_GPIO_DVI_DDC);
+		ddc_i2c = combios_setup_i2c_bus(rdev, RADEON_GPIO_DVI_DDC);
+		hpd.hpd = RADEON_HPD_NONE;
 		radeon_add_legacy_encoder(dev,
 					  radeon_get_encoder_id(dev,
 								ATOM_DEVICE_LCD1_SUPPORT,
 								0),
 					  ATOM_DEVICE_LCD1_SUPPORT);
 		radeon_add_legacy_connector(dev, 0, ATOM_DEVICE_LCD1_SUPPORT,
-					    DRM_MODE_CONNECTOR_LVDS, &ddc_i2c);
+					    DRM_MODE_CONNECTOR_LVDS, &ddc_i2c,
+					    CONNECTOR_OBJECT_ID_LVDS,
+					    &hpd);
 		/* VGA - primary dac */
-		ddc_i2c = combios_setup_i2c_bus(RADEON_GPIO_VGA_DDC);
+		ddc_i2c = combios_setup_i2c_bus(rdev, RADEON_GPIO_VGA_DDC);
+		hpd.hpd = RADEON_HPD_NONE;
 		radeon_add_legacy_encoder(dev,
 					  radeon_get_encoder_id(dev,
 								ATOM_DEVICE_CRT1_SUPPORT,
 								1),
 					  ATOM_DEVICE_CRT1_SUPPORT);
 		radeon_add_legacy_connector(dev, 1, ATOM_DEVICE_CRT1_SUPPORT,
-					    DRM_MODE_CONNECTOR_VGA, &ddc_i2c);
+					    DRM_MODE_CONNECTOR_VGA, &ddc_i2c,
+					    CONNECTOR_OBJECT_ID_VGA,
+					    &hpd);
 		/* TV - TV DAC */
+		ddc_i2c.valid = false;
+		hpd.hpd = RADEON_HPD_NONE;
 		radeon_add_legacy_encoder(dev,
 					  radeon_get_encoder_id(dev,
 								ATOM_DEVICE_TV1_SUPPORT,
@@ -1384,13 +1683,16 @@ bool radeon_get_legacy_connector_info_from_table(struct drm_device *dev)
 					  ATOM_DEVICE_TV1_SUPPORT);
 		radeon_add_legacy_connector(dev, 2, ATOM_DEVICE_TV1_SUPPORT,
 					    DRM_MODE_CONNECTOR_SVIDEO,
-					    &ddc_i2c);
+					    &ddc_i2c,
+					    CONNECTOR_OBJECT_ID_SVIDEO,
+					    &hpd);
 		break;
 	case CT_MINI_EXTERNAL:
 		DRM_INFO("Connector Table: %d (mini external tmds)\n",
 			 rdev->mode_info.connector_table);
 		/* DVI-I - tv dac, ext tmds */
-		ddc_i2c = combios_setup_i2c_bus(RADEON_GPIO_CRT2_DDC);
+		ddc_i2c = combios_setup_i2c_bus(rdev, RADEON_GPIO_CRT2_DDC);
+		hpd.hpd = RADEON_HPD_2; /* ??? */
 		radeon_add_legacy_encoder(dev,
 					  radeon_get_encoder_id(dev,
 								ATOM_DEVICE_DFP2_SUPPORT,
@@ -1401,11 +1703,16 @@ bool radeon_get_legacy_connector_info_from_table(struct drm_device *dev)
 								ATOM_DEVICE_CRT2_SUPPORT,
 								2),
 					  ATOM_DEVICE_CRT2_SUPPORT);
+		/* XXX are any DL? */
 		radeon_add_legacy_connector(dev, 0,
 					    ATOM_DEVICE_DFP2_SUPPORT |
 					    ATOM_DEVICE_CRT2_SUPPORT,
-					    DRM_MODE_CONNECTOR_DVII, &ddc_i2c);
+					    DRM_MODE_CONNECTOR_DVII, &ddc_i2c,
+					    CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_I,
+					    &hpd);
 		/* TV - TV DAC */
+		ddc_i2c.valid = false;
+		hpd.hpd = RADEON_HPD_NONE;
 		radeon_add_legacy_encoder(dev,
 					  radeon_get_encoder_id(dev,
 								ATOM_DEVICE_TV1_SUPPORT,
@@ -1413,13 +1720,16 @@ bool radeon_get_legacy_connector_info_from_table(struct drm_device *dev)
 					  ATOM_DEVICE_TV1_SUPPORT);
 		radeon_add_legacy_connector(dev, 1, ATOM_DEVICE_TV1_SUPPORT,
 					    DRM_MODE_CONNECTOR_SVIDEO,
-					    &ddc_i2c);
+					    &ddc_i2c,
+					    CONNECTOR_OBJECT_ID_SVIDEO,
+					    &hpd);
 		break;
 	case CT_MINI_INTERNAL:
 		DRM_INFO("Connector Table: %d (mini internal tmds)\n",
 			 rdev->mode_info.connector_table);
 		/* DVI-I - tv dac, int tmds */
-		ddc_i2c = combios_setup_i2c_bus(RADEON_GPIO_CRT2_DDC);
+		ddc_i2c = combios_setup_i2c_bus(rdev, RADEON_GPIO_CRT2_DDC);
+		hpd.hpd = RADEON_HPD_1; /* ??? */
 		radeon_add_legacy_encoder(dev,
 					  radeon_get_encoder_id(dev,
 								ATOM_DEVICE_DFP1_SUPPORT,
@@ -1433,8 +1743,12 @@ bool radeon_get_legacy_connector_info_from_table(struct drm_device *dev)
 		radeon_add_legacy_connector(dev, 0,
 					    ATOM_DEVICE_DFP1_SUPPORT |
 					    ATOM_DEVICE_CRT2_SUPPORT,
-					    DRM_MODE_CONNECTOR_DVII, &ddc_i2c);
+					    DRM_MODE_CONNECTOR_DVII, &ddc_i2c,
+					    CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_I,
+					    &hpd);
 		/* TV - TV DAC */
+		ddc_i2c.valid = false;
+		hpd.hpd = RADEON_HPD_NONE;
 		radeon_add_legacy_encoder(dev,
 					  radeon_get_encoder_id(dev,
 								ATOM_DEVICE_TV1_SUPPORT,
@@ -1442,30 +1756,40 @@ bool radeon_get_legacy_connector_info_from_table(struct drm_device *dev)
 					  ATOM_DEVICE_TV1_SUPPORT);
 		radeon_add_legacy_connector(dev, 1, ATOM_DEVICE_TV1_SUPPORT,
 					    DRM_MODE_CONNECTOR_SVIDEO,
-					    &ddc_i2c);
+					    &ddc_i2c,
+					    CONNECTOR_OBJECT_ID_SVIDEO,
+					    &hpd);
 		break;
 	case CT_IMAC_G5_ISIGHT:
 		DRM_INFO("Connector Table: %d (imac g5 isight)\n",
 			 rdev->mode_info.connector_table);
 		/* DVI-D - int tmds */
-		ddc_i2c = combios_setup_i2c_bus(RADEON_GPIO_MONID);
+		ddc_i2c = combios_setup_i2c_bus(rdev, RADEON_GPIO_MONID);
+		hpd.hpd = RADEON_HPD_1; /* ??? */
 		radeon_add_legacy_encoder(dev,
 					  radeon_get_encoder_id(dev,
 								ATOM_DEVICE_DFP1_SUPPORT,
 								0),
 					  ATOM_DEVICE_DFP1_SUPPORT);
 		radeon_add_legacy_connector(dev, 0, ATOM_DEVICE_DFP1_SUPPORT,
-					    DRM_MODE_CONNECTOR_DVID, &ddc_i2c);
+					    DRM_MODE_CONNECTOR_DVID, &ddc_i2c,
+					    CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_D,
+					    &hpd);
 		/* VGA - tv dac */
-		ddc_i2c = combios_setup_i2c_bus(RADEON_GPIO_DVI_DDC);
+		ddc_i2c = combios_setup_i2c_bus(rdev, RADEON_GPIO_DVI_DDC);
+		hpd.hpd = RADEON_HPD_NONE;
 		radeon_add_legacy_encoder(dev,
 					  radeon_get_encoder_id(dev,
 								ATOM_DEVICE_CRT2_SUPPORT,
 								2),
 					  ATOM_DEVICE_CRT2_SUPPORT);
 		radeon_add_legacy_connector(dev, 1, ATOM_DEVICE_CRT2_SUPPORT,
-					    DRM_MODE_CONNECTOR_VGA, &ddc_i2c);
+					    DRM_MODE_CONNECTOR_VGA, &ddc_i2c,
+					    CONNECTOR_OBJECT_ID_VGA,
+					    &hpd);
 		/* TV - TV DAC */
+		ddc_i2c.valid = false;
+		hpd.hpd = RADEON_HPD_NONE;
 		radeon_add_legacy_encoder(dev,
 					  radeon_get_encoder_id(dev,
 								ATOM_DEVICE_TV1_SUPPORT,
@@ -1473,30 +1797,40 @@ bool radeon_get_legacy_connector_info_from_table(struct drm_device *dev)
 					  ATOM_DEVICE_TV1_SUPPORT);
 		radeon_add_legacy_connector(dev, 2, ATOM_DEVICE_TV1_SUPPORT,
 					    DRM_MODE_CONNECTOR_SVIDEO,
-					    &ddc_i2c);
+					    &ddc_i2c,
+					    CONNECTOR_OBJECT_ID_SVIDEO,
+					    &hpd);
 		break;
 	case CT_EMAC:
 		DRM_INFO("Connector Table: %d (emac)\n",
 			 rdev->mode_info.connector_table);
 		/* VGA - primary dac */
-		ddc_i2c = combios_setup_i2c_bus(RADEON_GPIO_VGA_DDC);
+		ddc_i2c = combios_setup_i2c_bus(rdev, RADEON_GPIO_VGA_DDC);
+		hpd.hpd = RADEON_HPD_NONE;
 		radeon_add_legacy_encoder(dev,
 					  radeon_get_encoder_id(dev,
 								ATOM_DEVICE_CRT1_SUPPORT,
 								1),
 					  ATOM_DEVICE_CRT1_SUPPORT);
 		radeon_add_legacy_connector(dev, 0, ATOM_DEVICE_CRT1_SUPPORT,
-					    DRM_MODE_CONNECTOR_VGA, &ddc_i2c);
+					    DRM_MODE_CONNECTOR_VGA, &ddc_i2c,
+					    CONNECTOR_OBJECT_ID_VGA,
+					    &hpd);
 		/* VGA - tv dac */
-		ddc_i2c = combios_setup_i2c_bus(RADEON_GPIO_CRT2_DDC);
+		ddc_i2c = combios_setup_i2c_bus(rdev, RADEON_GPIO_CRT2_DDC);
+		hpd.hpd = RADEON_HPD_NONE;
 		radeon_add_legacy_encoder(dev,
 					  radeon_get_encoder_id(dev,
 								ATOM_DEVICE_CRT2_SUPPORT,
 								2),
 					  ATOM_DEVICE_CRT2_SUPPORT);
 		radeon_add_legacy_connector(dev, 1, ATOM_DEVICE_CRT2_SUPPORT,
-					    DRM_MODE_CONNECTOR_VGA, &ddc_i2c);
+					    DRM_MODE_CONNECTOR_VGA, &ddc_i2c,
+					    CONNECTOR_OBJECT_ID_VGA,
+					    &hpd);
 		/* TV - TV DAC */
+		ddc_i2c.valid = false;
+		hpd.hpd = RADEON_HPD_NONE;
 		radeon_add_legacy_encoder(dev,
 					  radeon_get_encoder_id(dev,
 								ATOM_DEVICE_TV1_SUPPORT,
@@ -1504,7 +1838,9 @@ bool radeon_get_legacy_connector_info_from_table(struct drm_device *dev)
 					  ATOM_DEVICE_TV1_SUPPORT);
 		radeon_add_legacy_connector(dev, 2, ATOM_DEVICE_TV1_SUPPORT,
 					    DRM_MODE_CONNECTOR_SVIDEO,
-					    &ddc_i2c);
+					    &ddc_i2c,
+					    CONNECTOR_OBJECT_ID_SVIDEO,
+					    &hpd);
 		break;
 	default:
 		DRM_INFO("Connector table: %d (invalid)\n",
@@ -1521,7 +1857,8 @@ static bool radeon_apply_legacy_quirks(struct drm_device *dev,
 				       int bios_index,
 				       enum radeon_combios_connector
 				       *legacy_connector,
-				       struct radeon_i2c_bus_rec *ddc_i2c)
+				       struct radeon_i2c_bus_rec *ddc_i2c,
+				       struct radeon_hpd *hpd)
 {
 	struct radeon_device *rdev = dev->dev_private;
 
@@ -1529,29 +1866,26 @@ static bool radeon_apply_legacy_quirks(struct drm_device *dev,
 	if ((rdev->family == CHIP_RS400 ||
 	     rdev->family == CHIP_RS480) &&
 	    ddc_i2c->mask_clk_reg == RADEON_GPIO_CRT2_DDC)
-		*ddc_i2c = combios_setup_i2c_bus(RADEON_GPIO_MONID);
+		*ddc_i2c = combios_setup_i2c_bus(rdev, RADEON_GPIO_MONID);
 	else if ((rdev->family == CHIP_RS400 ||
 		  rdev->family == CHIP_RS480) &&
 		 ddc_i2c->mask_clk_reg == RADEON_GPIO_MONID) {
-		ddc_i2c->valid = true;
+		*ddc_i2c = combios_setup_i2c_bus(rdev, RADEON_GPIOPAD_MASK);
 		ddc_i2c->mask_clk_mask = (0x20 << 8);
 		ddc_i2c->mask_data_mask = 0x80;
 		ddc_i2c->a_clk_mask = (0x20 << 8);
 		ddc_i2c->a_data_mask = 0x80;
-		ddc_i2c->put_clk_mask = (0x20 << 8);
-		ddc_i2c->put_data_mask = 0x80;
-		ddc_i2c->get_clk_mask = (0x20 << 8);
-		ddc_i2c->get_data_mask = 0x80;
-		ddc_i2c->mask_clk_reg = RADEON_GPIOPAD_MASK;
-		ddc_i2c->mask_data_reg = RADEON_GPIOPAD_MASK;
-		ddc_i2c->a_clk_reg = RADEON_GPIOPAD_A;
-		ddc_i2c->a_data_reg = RADEON_GPIOPAD_A;
-		ddc_i2c->put_clk_reg = RADEON_GPIOPAD_EN;
-		ddc_i2c->put_data_reg = RADEON_GPIOPAD_EN;
-		ddc_i2c->get_clk_reg = RADEON_LCD_GPIO_Y_REG;
-		ddc_i2c->get_data_reg = RADEON_LCD_GPIO_Y_REG;
+		ddc_i2c->en_clk_mask = (0x20 << 8);
+		ddc_i2c->en_data_mask = 0x80;
+		ddc_i2c->y_clk_mask = (0x20 << 8);
+		ddc_i2c->y_data_mask = 0x80;
 	}
 
+	/* R3xx+ chips don't have GPIO_CRT2_DDC gpio pad */
+	if ((rdev->family >= CHIP_R300) &&
+	    ddc_i2c->mask_clk_reg == RADEON_GPIO_CRT2_DDC)
+		*ddc_i2c = combios_setup_i2c_bus(rdev, RADEON_GPIO_DVI_DDC);
+
 	/* Certain IBM chipset RN50s have a BIOS reporting two VGAs,
 	   one with VGA DDC and one with CRT2 DDC. - kill the CRT2 DDC one */
 	if (dev->pdev->device == 0x515e &&
@@ -1581,18 +1915,74 @@ static bool radeon_apply_legacy_quirks(struct drm_device *dev,
 	return true;
 }
 
+static bool radeon_apply_legacy_tv_quirks(struct drm_device *dev)
+{
+	/* Acer 5102 has non-existent TV port */
+	if (dev->pdev->device == 0x5975 &&
+	    dev->pdev->subsystem_vendor == 0x1025 &&
+	    dev->pdev->subsystem_device == 0x009f)
+		return false;
+
+	/* HP dc5750 has non-existent TV port */
+	if (dev->pdev->device == 0x5974 &&
+	    dev->pdev->subsystem_vendor == 0x103c &&
+	    dev->pdev->subsystem_device == 0x280a)
+		return false;
+
+	/* MSI S270 has non-existent TV port */
+	if (dev->pdev->device == 0x5955 &&
+	    dev->pdev->subsystem_vendor == 0x1462 &&
+	    dev->pdev->subsystem_device == 0x0131)
+		return false;
+
+	return true;
+}
+
+static uint16_t combios_check_dl_dvi(struct drm_device *dev, int is_dvi_d)
+{
+	struct radeon_device *rdev = dev->dev_private;
+	uint32_t ext_tmds_info;
+
+	if (rdev->flags & RADEON_IS_IGP) {
+		if (is_dvi_d)
+			return CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_D;
+		else
+			return CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_I;
+	}
+	ext_tmds_info = combios_get_table_offset(dev, COMBIOS_EXT_TMDS_INFO_TABLE);
+	if (ext_tmds_info) {
+		uint8_t rev = RBIOS8(ext_tmds_info);
+		uint8_t flags = RBIOS8(ext_tmds_info + 4 + 5);
+		if (rev >= 3) {
+			if (is_dvi_d)
+				return CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_D;
+			else
+				return CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_I;
+		} else {
+			if (flags & 1) {
+				if (is_dvi_d)
+					return CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_D;
+				else
+					return CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_I;
+			}
+		}
+	}
+	if (is_dvi_d)
+		return CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_D;
+	else
+		return CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_I;
+}
+
 bool radeon_get_legacy_connector_info_from_bios(struct drm_device *dev)
 {
 	struct radeon_device *rdev = dev->dev_private;
 	uint32_t conn_info, entry, devices;
-	uint16_t tmp;
+	uint16_t tmp, connector_object_id;
 	enum radeon_combios_ddc ddc_type;
 	enum radeon_combios_connector connector;
 	int i = 0;
 	struct radeon_i2c_bus_rec ddc_i2c;
-
-	if (rdev->bios == NULL)
-		return false;
+	struct radeon_hpd hpd;
 
 	conn_info = combios_get_table_offset(dev, COMBIOS_CONNECTOR_INFO_TABLE);
 	if (conn_info) {
@@ -1610,26 +2000,41 @@ bool radeon_get_legacy_connector_info_from_bios(struct drm_device *dev)
 			switch (ddc_type) {
 			case DDC_MONID:
 				ddc_i2c =
-				    combios_setup_i2c_bus(RADEON_GPIO_MONID);
+					combios_setup_i2c_bus(rdev, RADEON_GPIO_MONID);
 				break;
 			case DDC_DVI:
 				ddc_i2c =
-				    combios_setup_i2c_bus(RADEON_GPIO_DVI_DDC);
+					combios_setup_i2c_bus(rdev, RADEON_GPIO_DVI_DDC);
 				break;
 			case DDC_VGA:
 				ddc_i2c =
-				    combios_setup_i2c_bus(RADEON_GPIO_VGA_DDC);
+					combios_setup_i2c_bus(rdev, RADEON_GPIO_VGA_DDC);
 				break;
 			case DDC_CRT2:
 				ddc_i2c =
-				    combios_setup_i2c_bus(RADEON_GPIO_CRT2_DDC);
+					combios_setup_i2c_bus(rdev, RADEON_GPIO_CRT2_DDC);
+				break;
+			default:
+				break;
+			}
+
+			switch (connector) {
+			case CONNECTOR_PROPRIETARY_LEGACY:
+			case CONNECTOR_DVI_I_LEGACY:
+			case CONNECTOR_DVI_D_LEGACY:
+				if ((tmp >> 4) & 0x1)
+					hpd.hpd = RADEON_HPD_2;
+				else
+					hpd.hpd = RADEON_HPD_1;
 				break;
 			default:
+				hpd.hpd = RADEON_HPD_NONE;
 				break;
 			}
 
-			radeon_apply_legacy_quirks(dev, i, &connector,
-						   &ddc_i2c);
+			if (!radeon_apply_legacy_quirks(dev, i, &connector,
+							&ddc_i2c, &hpd))
+				continue;
 
 			switch (connector) {
 			case CONNECTOR_PROPRIETARY_LEGACY:
@@ -1644,7 +2049,9 @@ bool radeon_get_legacy_connector_info_from_bios(struct drm_device *dev)
 				radeon_add_legacy_connector(dev, i, devices,
 							    legacy_connector_convert
 							    [connector],
-							    &ddc_i2c);
+							    &ddc_i2c,
+							    CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_D,
+							    &hpd);
 				break;
 			case CONNECTOR_CRT_LEGACY:
 				if (tmp & 0x1) {
@@ -1669,7 +2076,9 @@ bool radeon_get_legacy_connector_info_from_bios(struct drm_device *dev)
 							    devices,
 							    legacy_connector_convert
 							    [connector],
-							    &ddc_i2c);
+							    &ddc_i2c,
+							    CONNECTOR_OBJECT_ID_VGA,
+							    &hpd);
 				break;
 			case CONNECTOR_DVI_I_LEGACY:
 				devices = 0;
@@ -1698,6 +2107,7 @@ bool radeon_get_legacy_connector_info_from_bios(struct drm_device *dev)
 								   ATOM_DEVICE_DFP2_SUPPORT,
 								   0),
 								  ATOM_DEVICE_DFP2_SUPPORT);
+					connector_object_id = combios_check_dl_dvi(dev, 0);
 				} else {
 					devices |= ATOM_DEVICE_DFP1_SUPPORT;
 					radeon_add_legacy_encoder(dev,
@@ -1706,19 +2116,25 @@ bool radeon_get_legacy_connector_info_from_bios(struct drm_device *dev)
 								   ATOM_DEVICE_DFP1_SUPPORT,
 								   0),
 								  ATOM_DEVICE_DFP1_SUPPORT);
+					connector_object_id = CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_I;
 				}
 				radeon_add_legacy_connector(dev,
 							    i,
 							    devices,
 							    legacy_connector_convert
 							    [connector],
-							    &ddc_i2c);
+							    &ddc_i2c,
+							    connector_object_id,
+							    &hpd);
 				break;
 			case CONNECTOR_DVI_D_LEGACY:
-				if ((tmp >> 4) & 0x1)
+				if ((tmp >> 4) & 0x1) {
 					devices = ATOM_DEVICE_DFP2_SUPPORT;
-				else
+					connector_object_id = combios_check_dl_dvi(dev, 1);
+				} else {
 					devices = ATOM_DEVICE_DFP1_SUPPORT;
+					connector_object_id = CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_I;
+				}
 				radeon_add_legacy_encoder(dev,
 							  radeon_get_encoder_id
 							  (dev, devices, 0),
@@ -1726,7 +2142,9 @@ bool radeon_get_legacy_connector_info_from_bios(struct drm_device *dev)
 				radeon_add_legacy_connector(dev, i, devices,
 							    legacy_connector_convert
 							    [connector],
-							    &ddc_i2c);
+							    &ddc_i2c,
+							    connector_object_id,
+							    &hpd);
 				break;
 			case CONNECTOR_CTV_LEGACY:
 			case CONNECTOR_STV_LEGACY:
@@ -1740,7 +2158,9 @@ bool radeon_get_legacy_connector_info_from_bios(struct drm_device *dev)
 							    ATOM_DEVICE_TV1_SUPPORT,
 							    legacy_connector_convert
 							    [connector],
-							    &ddc_i2c);
+							    &ddc_i2c,
+							    CONNECTOR_OBJECT_ID_SVIDEO,
+							    &hpd);
 				break;
 			default:
 				DRM_ERROR("Unknown connector type: %d\n",
@@ -1766,16 +2186,39 @@ bool radeon_get_legacy_connector_info_from_bios(struct drm_device *dev)
 									0),
 						  ATOM_DEVICE_DFP1_SUPPORT);
 
-			ddc_i2c = combios_setup_i2c_bus(RADEON_GPIO_DVI_DDC);
+			ddc_i2c = combios_setup_i2c_bus(rdev, RADEON_GPIO_DVI_DDC);
+			hpd.hpd = RADEON_HPD_NONE;
 			radeon_add_legacy_connector(dev,
 						    0,
 						    ATOM_DEVICE_CRT1_SUPPORT |
 						    ATOM_DEVICE_DFP1_SUPPORT,
 						    DRM_MODE_CONNECTOR_DVII,
-						    &ddc_i2c);
+						    &ddc_i2c,
+						    CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_I,
+						    &hpd);
 		} else {
-			DRM_DEBUG("No connector info found\n");
-			return false;
+			uint16_t crt_info =
+				combios_get_table_offset(dev, COMBIOS_CRT_INFO_TABLE);
+			DRM_DEBUG("Found CRT table, assuming VGA connector\n");
+			if (crt_info) {
+				radeon_add_legacy_encoder(dev,
+							  radeon_get_encoder_id(dev,
+										ATOM_DEVICE_CRT1_SUPPORT,
+										1),
+							  ATOM_DEVICE_CRT1_SUPPORT);
+				ddc_i2c = combios_setup_i2c_bus(rdev, RADEON_GPIO_VGA_DDC);
+				hpd.hpd = RADEON_HPD_NONE;
+				radeon_add_legacy_connector(dev,
+							    0,
+							    ATOM_DEVICE_CRT1_SUPPORT,
+							    DRM_MODE_CONNECTOR_VGA,
+							    &ddc_i2c,
+							    CONNECTOR_OBJECT_ID_VGA,
+							    &hpd);
+			} else {
+				DRM_DEBUG("No connector info found\n");
+				return false;
+			}
 		}
 	}
 
@@ -1799,27 +2242,27 @@ bool radeon_get_legacy_connector_info_from_bios(struct drm_device *dev)
 				case DDC_MONID:
 					ddc_i2c =
 					    combios_setup_i2c_bus
-					    (RADEON_GPIO_MONID);
+						(rdev, RADEON_GPIO_MONID);
 					break;
 				case DDC_DVI:
 					ddc_i2c =
 					    combios_setup_i2c_bus
-					    (RADEON_GPIO_DVI_DDC);
+						(rdev, RADEON_GPIO_DVI_DDC);
 					break;
 				case DDC_VGA:
 					ddc_i2c =
 					    combios_setup_i2c_bus
-					    (RADEON_GPIO_VGA_DDC);
+						(rdev, RADEON_GPIO_VGA_DDC);
 					break;
 				case DDC_CRT2:
 					ddc_i2c =
 					    combios_setup_i2c_bus
-					    (RADEON_GPIO_CRT2_DDC);
+						(rdev, RADEON_GPIO_CRT2_DDC);
 					break;
 				case DDC_LCD:
 					ddc_i2c =
 					    combios_setup_i2c_bus
-					    (RADEON_LCD_GPIO_MASK);
+						(rdev, RADEON_GPIOPAD_MASK);
 					ddc_i2c.mask_clk_mask =
 					    RBIOS32(lcd_ddc_info + 3);
 					ddc_i2c.mask_data_mask =
@@ -1828,19 +2271,19 @@ bool radeon_get_legacy_connector_info_from_bios(struct drm_device *dev)
 					    RBIOS32(lcd_ddc_info + 3);
 					ddc_i2c.a_data_mask =
 					    RBIOS32(lcd_ddc_info + 7);
-					ddc_i2c.put_clk_mask =
+					ddc_i2c.en_clk_mask =
 					    RBIOS32(lcd_ddc_info + 3);
-					ddc_i2c.put_data_mask =
+					ddc_i2c.en_data_mask =
 					    RBIOS32(lcd_ddc_info + 7);
-					ddc_i2c.get_clk_mask =
+					ddc_i2c.y_clk_mask =
 					    RBIOS32(lcd_ddc_info + 3);
-					ddc_i2c.get_data_mask =
+					ddc_i2c.y_data_mask =
 					    RBIOS32(lcd_ddc_info + 7);
 					break;
 				case DDC_GPIO:
 					ddc_i2c =
 					    combios_setup_i2c_bus
-					    (RADEON_MDGPIO_EN_REG);
+						(rdev, RADEON_MDGPIO_MASK);
 					ddc_i2c.mask_clk_mask =
 					    RBIOS32(lcd_ddc_info + 3);
 					ddc_i2c.mask_data_mask =
@@ -1849,13 +2292,13 @@ bool radeon_get_legacy_connector_info_from_bios(struct drm_device *dev)
 					    RBIOS32(lcd_ddc_info + 3);
 					ddc_i2c.a_data_mask =
 					    RBIOS32(lcd_ddc_info + 7);
-					ddc_i2c.put_clk_mask =
+					ddc_i2c.en_clk_mask =
 					    RBIOS32(lcd_ddc_info + 3);
-					ddc_i2c.put_data_mask =
+					ddc_i2c.en_data_mask =
 					    RBIOS32(lcd_ddc_info + 7);
-					ddc_i2c.get_clk_mask =
+					ddc_i2c.y_clk_mask =
 					    RBIOS32(lcd_ddc_info + 3);
-					ddc_i2c.get_data_mask =
+					ddc_i2c.y_data_mask =
 					    RBIOS32(lcd_ddc_info + 7);
 					break;
 				default:
@@ -1866,11 +2309,14 @@ bool radeon_get_legacy_connector_info_from_bios(struct drm_device *dev)
 			} else
 				ddc_i2c.valid = false;
 
+			hpd.hpd = RADEON_HPD_NONE;
 			radeon_add_legacy_connector(dev,
 						    5,
 						    ATOM_DEVICE_LCD1_SUPPORT,
 						    DRM_MODE_CONNECTOR_LVDS,
-						    &ddc_i2c);
+						    &ddc_i2c,
+						    CONNECTOR_OBJECT_ID_LVDS,
+						    &hpd);
 		}
 	}
 
@@ -1880,16 +2326,21 @@ bool radeon_get_legacy_connector_info_from_bios(struct drm_device *dev)
 		    combios_get_table_offset(dev, COMBIOS_TV_INFO_TABLE);
 		if (tv_info) {
 			if (RBIOS8(tv_info + 6) == 'T') {
-				radeon_add_legacy_encoder(dev,
-							  radeon_get_encoder_id
-							  (dev,
-							   ATOM_DEVICE_TV1_SUPPORT,
-							   2),
-							  ATOM_DEVICE_TV1_SUPPORT);
-				radeon_add_legacy_connector(dev, 6,
-							    ATOM_DEVICE_TV1_SUPPORT,
-							    DRM_MODE_CONNECTOR_SVIDEO,
-							    &ddc_i2c);
+				if (radeon_apply_legacy_tv_quirks(dev)) {
+					hpd.hpd = RADEON_HPD_NONE;
+					radeon_add_legacy_encoder(dev,
+								  radeon_get_encoder_id
+								  (dev,
+								   ATOM_DEVICE_TV1_SUPPORT,
+								   2),
+								  ATOM_DEVICE_TV1_SUPPORT);
+					radeon_add_legacy_connector(dev, 6,
+								    ATOM_DEVICE_TV1_SUPPORT,
+								    DRM_MODE_CONNECTOR_SVIDEO,
+								    &ddc_i2c,
+								    CONNECTOR_OBJECT_ID_SVIDEO,
+								    &hpd);
+				}
 			}
 		}
 	}
@@ -1899,6 +2350,293 @@ bool radeon_get_legacy_connector_info_from_bios(struct drm_device *dev)
 	return true;
 }
 
+void radeon_combios_get_power_modes(struct radeon_device *rdev)
+{
+	struct drm_device *dev = rdev->ddev;
+	u16 offset, misc, misc2 = 0;
+	u8 rev, blocks, tmp;
+	int state_index = 0;
+
+	rdev->pm.default_power_state = NULL;
+
+	if (rdev->flags & RADEON_IS_MOBILITY) {
+		offset = combios_get_table_offset(dev, COMBIOS_POWERPLAY_INFO_TABLE);
+		if (offset) {
+			rev = RBIOS8(offset);
+			blocks = RBIOS8(offset + 0x2);
+			/* power mode 0 tends to be the only valid one */
+			rdev->pm.power_state[state_index].num_clock_modes = 1;
+			rdev->pm.power_state[state_index].clock_info[0].mclk = RBIOS32(offset + 0x5 + 0x2);
+			rdev->pm.power_state[state_index].clock_info[0].sclk = RBIOS32(offset + 0x5 + 0x6);
+			if ((rdev->pm.power_state[state_index].clock_info[0].mclk == 0) ||
+			    (rdev->pm.power_state[state_index].clock_info[0].sclk == 0))
+				goto default_mode;
+			/* skip overclock modes for now */
+			if ((rdev->pm.power_state[state_index].clock_info[0].mclk >
+			     rdev->clock.default_mclk + RADEON_MODE_OVERCLOCK_MARGIN) ||
+			    (rdev->pm.power_state[state_index].clock_info[0].sclk >
+			     rdev->clock.default_sclk + RADEON_MODE_OVERCLOCK_MARGIN))
+				goto default_mode;
+			rdev->pm.power_state[state_index].type =
+				POWER_STATE_TYPE_BATTERY;
+			misc = RBIOS16(offset + 0x5 + 0x0);
+			if (rev > 4)
+				misc2 = RBIOS16(offset + 0x5 + 0xe);
+			if (misc & 0x4) {
+				rdev->pm.power_state[state_index].clock_info[0].voltage.type = VOLTAGE_GPIO;
+				if (misc & 0x8)
+					rdev->pm.power_state[state_index].clock_info[0].voltage.active_high =
+						true;
+				else
+					rdev->pm.power_state[state_index].clock_info[0].voltage.active_high =
+						false;
+				rdev->pm.power_state[state_index].clock_info[0].voltage.gpio.valid = true;
+				if (rev < 6) {
+					rdev->pm.power_state[state_index].clock_info[0].voltage.gpio.reg =
+						RBIOS16(offset + 0x5 + 0xb) * 4;
+					tmp = RBIOS8(offset + 0x5 + 0xd);
+					rdev->pm.power_state[state_index].clock_info[0].voltage.gpio.mask = (1 << tmp);
+				} else {
+					u8 entries = RBIOS8(offset + 0x5 + 0xb);
+					u16 voltage_table_offset = RBIOS16(offset + 0x5 + 0xc);
+					if (entries && voltage_table_offset) {
+						rdev->pm.power_state[state_index].clock_info[0].voltage.gpio.reg =
+							RBIOS16(voltage_table_offset) * 4;
+						tmp = RBIOS8(voltage_table_offset + 0x2);
+						rdev->pm.power_state[state_index].clock_info[0].voltage.gpio.mask = (1 << tmp);
+					} else
+						rdev->pm.power_state[state_index].clock_info[0].voltage.gpio.valid = false;
+				}
+				switch ((misc2 & 0x700) >> 8) {
+				case 0:
+				default:
+					rdev->pm.power_state[state_index].clock_info[0].voltage.delay = 0;
+					break;
+				case 1:
+					rdev->pm.power_state[state_index].clock_info[0].voltage.delay = 33;
+					break;
+				case 2:
+					rdev->pm.power_state[state_index].clock_info[0].voltage.delay = 66;
+					break;
+				case 3:
+					rdev->pm.power_state[state_index].clock_info[0].voltage.delay = 99;
+					break;
+				case 4:
+					rdev->pm.power_state[state_index].clock_info[0].voltage.delay = 132;
+					break;
+				}
+			} else
+				rdev->pm.power_state[state_index].clock_info[0].voltage.type = VOLTAGE_NONE;
+			if (rev > 6)
+				rdev->pm.power_state[state_index].non_clock_info.pcie_lanes =
+					RBIOS8(offset + 0x5 + 0x10);
+			state_index++;
+		} else {
+			/* XXX figure out some good default low power mode for mobility cards w/out power tables */
+		}
+	} else {
+		/* XXX figure out some good default low power mode for desktop cards */
+	}
+
+default_mode:
+	/* add the default mode */
+	rdev->pm.power_state[state_index].type =
+		POWER_STATE_TYPE_DEFAULT;
+	rdev->pm.power_state[state_index].num_clock_modes = 1;
+	rdev->pm.power_state[state_index].clock_info[0].mclk = rdev->clock.default_mclk;
+	rdev->pm.power_state[state_index].clock_info[0].sclk = rdev->clock.default_sclk;
+	rdev->pm.power_state[state_index].default_clock_mode = &rdev->pm.power_state[state_index].clock_info[0];
+	rdev->pm.power_state[state_index].clock_info[0].voltage.type = VOLTAGE_NONE;
+	if (rdev->asic->get_pcie_lanes)
+		rdev->pm.power_state[state_index].non_clock_info.pcie_lanes = radeon_get_pcie_lanes(rdev);
+	else
+		rdev->pm.power_state[state_index].non_clock_info.pcie_lanes = 16;
+	rdev->pm.default_power_state = &rdev->pm.power_state[state_index];
+	rdev->pm.num_power_states = state_index + 1;
+
+	rdev->pm.current_power_state = rdev->pm.default_power_state;
+	rdev->pm.current_clock_mode =
+		rdev->pm.default_power_state->default_clock_mode;
+}
+
+void radeon_external_tmds_setup(struct drm_encoder *encoder)
+{
+	struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
+	struct radeon_encoder_ext_tmds *tmds = radeon_encoder->enc_priv;
+
+	if (!tmds)
+		return;
+
+	switch (tmds->dvo_chip) {
+	case DVO_SIL164:
+		/* sil 164 */
+		radeon_i2c_put_byte(tmds->i2c_bus,
+				    tmds->slave_addr,
+				    0x08, 0x30);
+		radeon_i2c_put_byte(tmds->i2c_bus,
+				       tmds->slave_addr,
+				       0x09, 0x00);
+		radeon_i2c_put_byte(tmds->i2c_bus,
+				    tmds->slave_addr,
+				    0x0a, 0x90);
+		radeon_i2c_put_byte(tmds->i2c_bus,
+				    tmds->slave_addr,
+				    0x0c, 0x89);
+		radeon_i2c_put_byte(tmds->i2c_bus,
+				       tmds->slave_addr,
+				       0x08, 0x3b);
+		break;
+	case DVO_SIL1178:
+		/* sil 1178 - untested */
+		/*
+		 * 0x0f, 0x44
+		 * 0x0f, 0x4c
+		 * 0x0e, 0x01
+		 * 0x0a, 0x80
+		 * 0x09, 0x30
+		 * 0x0c, 0xc9
+		 * 0x0d, 0x70
+		 * 0x08, 0x32
+		 * 0x08, 0x33
+		 */
+		break;
+	default:
+		break;
+	}
+
+}
+
+bool radeon_combios_external_tmds_setup(struct drm_encoder *encoder)
+{
+	struct drm_device *dev = encoder->dev;
+	struct radeon_device *rdev = dev->dev_private;
+	struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
+	uint16_t offset;
+	uint8_t blocks, slave_addr, rev;
+	uint32_t index, id;
+	uint32_t reg, val, and_mask, or_mask;
+	struct radeon_encoder_ext_tmds *tmds = radeon_encoder->enc_priv;
+
+	if (!tmds)
+		return false;
+
+	if (rdev->flags & RADEON_IS_IGP) {
+		offset = combios_get_table_offset(dev, COMBIOS_TMDS_POWER_ON_TABLE);
+		rev = RBIOS8(offset);
+		if (offset) {
+			rev = RBIOS8(offset);
+			if (rev > 1) {
+				blocks = RBIOS8(offset + 3);
+				index = offset + 4;
+				while (blocks > 0) {
+					id = RBIOS16(index);
+					index += 2;
+					switch (id >> 13) {
+					case 0:
+						reg = (id & 0x1fff) * 4;
+						val = RBIOS32(index);
+						index += 4;
+						WREG32(reg, val);
+						break;
+					case 2:
+						reg = (id & 0x1fff) * 4;
+						and_mask = RBIOS32(index);
+						index += 4;
+						or_mask = RBIOS32(index);
+						index += 4;
+						val = RREG32(reg);
+						val = (val & and_mask) | or_mask;
+						WREG32(reg, val);
+						break;
+					case 3:
+						val = RBIOS16(index);
+						index += 2;
+						udelay(val);
+						break;
+					case 4:
+						val = RBIOS16(index);
+						index += 2;
+						udelay(val * 1000);
+						break;
+					case 6:
+						slave_addr = id & 0xff;
+						slave_addr >>= 1; /* 7 bit addressing */
+						index++;
+						reg = RBIOS8(index);
+						index++;
+						val = RBIOS8(index);
+						index++;
+						radeon_i2c_put_byte(tmds->i2c_bus,
+								    slave_addr,
+								    reg, val);
+						break;
+					default:
+						DRM_ERROR("Unknown id %d\n", id >> 13);
+						break;
+					}
+					blocks--;
+				}
+				return true;
+			}
+		}
+	} else {
+		offset = combios_get_table_offset(dev, COMBIOS_EXT_TMDS_INFO_TABLE);
+		if (offset) {
+			index = offset + 10;
+			id = RBIOS16(index);
+			while (id != 0xffff) {
+				index += 2;
+				switch (id >> 13) {
+				case 0:
+					reg = (id & 0x1fff) * 4;
+					val = RBIOS32(index);
+					WREG32(reg, val);
+					break;
+				case 2:
+					reg = (id & 0x1fff) * 4;
+					and_mask = RBIOS32(index);
+					index += 4;
+					or_mask = RBIOS32(index);
+					index += 4;
+					val = RREG32(reg);
+					val = (val & and_mask) | or_mask;
+					WREG32(reg, val);
+					break;
+				case 4:
+					val = RBIOS16(index);
+					index += 2;
+					udelay(val);
+					break;
+				case 5:
+					reg = id & 0x1fff;
+					and_mask = RBIOS32(index);
+					index += 4;
+					or_mask = RBIOS32(index);
+					index += 4;
+					val = RREG32_PLL(reg);
+					val = (val & and_mask) | or_mask;
+					WREG32_PLL(reg, val);
+					break;
+				case 6:
+					reg = id & 0x1fff;
+					val = RBIOS8(index);
+					index += 1;
+					radeon_i2c_put_byte(tmds->i2c_bus,
+							    tmds->slave_addr,
+							    reg, val);
+					break;
+				default:
+					DRM_ERROR("Unknown id %d\n", id >> 13);
+					break;
+				}
+				id = RBIOS16(index);
+			}
+			return true;
+		}
+	}
+	return false;
+}
+
 static void combios_parse_mmio_table(struct drm_device *dev, uint16_t offset)
 {
 	struct radeon_device *rdev = dev->dev_private;
diff --git a/drivers/gpu/drm/radeon/radeon_connectors.c b/drivers/gpu/drm/radeon/radeon_connectors.c
index af1d551f1a8f..ee0083f982d8 100644
--- a/drivers/gpu/drm/radeon/radeon_connectors.c
+++ b/drivers/gpu/drm/radeon/radeon_connectors.c
@@ -26,6 +26,7 @@
 #include "drmP.h"
 #include "drm_edid.h"
 #include "drm_crtc_helper.h"
+#include "drm_fb_helper.h"
 #include "radeon_drm.h"
 #include "radeon.h"
 #include "atom.h"
@@ -39,6 +40,28 @@ radeon_atombios_connected_scratch_regs(struct drm_connector *connector,
 				       struct drm_encoder *encoder,
 				       bool connected);
 
+void radeon_connector_hotplug(struct drm_connector *connector)
+{
+	struct drm_device *dev = connector->dev;
+	struct radeon_device *rdev = dev->dev_private;
+	struct radeon_connector *radeon_connector = to_radeon_connector(connector);
+
+	if (radeon_connector->hpd.hpd != RADEON_HPD_NONE)
+		radeon_hpd_set_polarity(rdev, radeon_connector->hpd.hpd);
+
+	if ((connector->connector_type == DRM_MODE_CONNECTOR_DisplayPort) ||
+	    (connector->connector_type == DRM_MODE_CONNECTOR_eDP)) {
+		if ((radeon_dp_getsinktype(radeon_connector) == CONNECTOR_OBJECT_ID_DISPLAYPORT) ||
+		    (radeon_dp_getsinktype(radeon_connector) == CONNECTOR_OBJECT_ID_eDP)) {
+			if (radeon_dp_needs_link_train(radeon_connector)) {
+				if (connector->encoder)
+					dp_link_train(connector->encoder, connector);
+			}
+		}
+	}
+
+}
+
 static void radeon_property_change_mode(struct drm_encoder *encoder)
 {
 	struct drm_crtc *crtc = encoder->crtc;
@@ -177,29 +200,28 @@ static struct drm_display_mode *radeon_fp_native_mode(struct drm_encoder *encode
 	struct drm_device *dev = encoder->dev;
 	struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
 	struct drm_display_mode *mode = NULL;
-	struct radeon_native_mode *native_mode = &radeon_encoder->native_mode;
-
-	if (native_mode->panel_xres != 0 &&
-	    native_mode->panel_yres != 0 &&
-	    native_mode->dotclock != 0) {
-		mode = drm_mode_create(dev);
-
-		mode->hdisplay = native_mode->panel_xres;
-		mode->vdisplay = native_mode->panel_yres;
-
-		mode->htotal = mode->hdisplay + native_mode->hblank;
-		mode->hsync_start = mode->hdisplay + native_mode->hoverplus;
-		mode->hsync_end = mode->hsync_start + native_mode->hsync_width;
-		mode->vtotal = mode->vdisplay + native_mode->vblank;
-		mode->vsync_start = mode->vdisplay + native_mode->voverplus;
-		mode->vsync_end = mode->vsync_start + native_mode->vsync_width;
-		mode->clock = native_mode->dotclock;
-		mode->flags = 0;
+	struct drm_display_mode *native_mode = &radeon_encoder->native_mode;
 
+	if (native_mode->hdisplay != 0 &&
+	    native_mode->vdisplay != 0 &&
+	    native_mode->clock != 0) {
+		mode = drm_mode_duplicate(dev, native_mode);
 		mode->type = DRM_MODE_TYPE_PREFERRED | DRM_MODE_TYPE_DRIVER;
 		drm_mode_set_name(mode);
 
 		DRM_DEBUG("Adding native panel mode %s\n", mode->name);
+	} else if (native_mode->hdisplay != 0 &&
+		   native_mode->vdisplay != 0) {
+		/* mac laptops without an edid */
+		/* Note that this is not necessarily the exact panel mode,
+		 * but an approximation based on the cvt formula.  For these
+		 * systems we should ideally read the mode info out of the
+		 * registers or add a mode table, but this works and is much
+		 * simpler.
+		 */
+		mode = drm_cvt_mode(dev, native_mode->hdisplay, native_mode->vdisplay, 60, true, false, false);
+		mode->type = DRM_MODE_TYPE_PREFERRED | DRM_MODE_TYPE_DRIVER;
+		DRM_DEBUG("Adding cvt approximation of native panel mode %s\n", mode->name);
 	}
 	return mode;
 }
@@ -209,7 +231,7 @@ static void radeon_add_common_modes(struct drm_encoder *encoder, struct drm_conn
 	struct drm_device *dev = encoder->dev;
 	struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
 	struct drm_display_mode *mode = NULL;
-	struct radeon_native_mode *native_mode = &radeon_encoder->native_mode;
+	struct drm_display_mode *native_mode = &radeon_encoder->native_mode;
 	int i;
 	struct mode_size {
 		int w;
@@ -235,17 +257,22 @@ static void radeon_add_common_modes(struct drm_encoder *encoder, struct drm_conn
 	};
 
 	for (i = 0; i < 17; i++) {
+		if (radeon_encoder->devices & (ATOM_DEVICE_TV_SUPPORT)) {
+			if (common_modes[i].w > 1024 ||
+			    common_modes[i].h > 768)
+				continue;
+		}
 		if (radeon_encoder->devices & (ATOM_DEVICE_LCD_SUPPORT)) {
-			if (common_modes[i].w > native_mode->panel_xres ||
-			    common_modes[i].h > native_mode->panel_yres ||
-			    (common_modes[i].w == native_mode->panel_xres &&
-			     common_modes[i].h == native_mode->panel_yres))
+			if (common_modes[i].w > native_mode->hdisplay ||
+			    common_modes[i].h > native_mode->vdisplay ||
+			    (common_modes[i].w == native_mode->hdisplay &&
+			     common_modes[i].h == native_mode->vdisplay))
 				continue;
 		}
 		if (common_modes[i].w < 320 || common_modes[i].h < 200)
 			continue;
 
-		mode = drm_cvt_mode(dev, common_modes[i].w, common_modes[i].h, 60, false, false);
+		mode = drm_cvt_mode(dev, common_modes[i].w, common_modes[i].h, 60, false, false, false);
 		drm_mode_probed_add(connector, mode);
 	}
 }
@@ -343,28 +370,23 @@ static void radeon_fixup_lvds_native_mode(struct drm_encoder *encoder,
 					  struct drm_connector *connector)
 {
 	struct radeon_encoder *radeon_encoder =	to_radeon_encoder(encoder);
-	struct radeon_native_mode *native_mode = &radeon_encoder->native_mode;
+	struct drm_display_mode *native_mode = &radeon_encoder->native_mode;
 
 	/* Try to get native mode details from EDID if necessary */
-	if (!native_mode->dotclock) {
+	if (!native_mode->clock) {
 		struct drm_display_mode *t, *mode;
 
 		list_for_each_entry_safe(mode, t, &connector->probed_modes, head) {
-			if (mode->hdisplay == native_mode->panel_xres &&
-			    mode->vdisplay == native_mode->panel_yres) {
-				native_mode->hblank = mode->htotal - mode->hdisplay;
-				native_mode->hoverplus = mode->hsync_start - mode->hdisplay;
-				native_mode->hsync_width = mode->hsync_end - mode->hsync_start;
-				native_mode->vblank = mode->vtotal - mode->vdisplay;
-				native_mode->voverplus = mode->vsync_start - mode->vdisplay;
-				native_mode->vsync_width = mode->vsync_end - mode->vsync_start;
-				native_mode->dotclock = mode->clock;
+			if (mode->hdisplay == native_mode->hdisplay &&
+			    mode->vdisplay == native_mode->vdisplay) {
+				*native_mode = *mode;
+				drm_mode_set_crtcinfo(native_mode, CRTC_INTERLACE_HALVE_V);
 				DRM_INFO("Determined LVDS native mode details from EDID\n");
 				break;
 			}
 		}
 	}
-	if (!native_mode->dotclock) {
+	if (!native_mode->clock) {
 		DRM_INFO("No LVDS native mode details, disabling RMX\n");
 		radeon_encoder->rmx_type = RMX_OFF;
 	}
@@ -409,13 +431,62 @@ static int radeon_lvds_get_modes(struct drm_connector *connector)
 static int radeon_lvds_mode_valid(struct drm_connector *connector,
 				  struct drm_display_mode *mode)
 {
+	struct drm_encoder *encoder = radeon_best_single_encoder(connector);
+
+	if ((mode->hdisplay < 320) || (mode->vdisplay < 240))
+		return MODE_PANEL;
+
+	if (encoder) {
+		struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
+		struct drm_display_mode *native_mode = &radeon_encoder->native_mode;
+
+		/* AVIVO hardware supports downscaling modes larger than the panel
+		 * to the panel size, but I'm not sure this is desirable.
+		 */
+		if ((mode->hdisplay > native_mode->hdisplay) ||
+		    (mode->vdisplay > native_mode->vdisplay))
+			return MODE_PANEL;
+
+		/* if scaling is disabled, block non-native modes */
+		if (radeon_encoder->rmx_type == RMX_OFF) {
+			if ((mode->hdisplay != native_mode->hdisplay) ||
+			    (mode->vdisplay != native_mode->vdisplay))
+				return MODE_PANEL;
+		}
+	}
+
 	return MODE_OK;
 }
 
 static enum drm_connector_status radeon_lvds_detect(struct drm_connector *connector)
 {
-	enum drm_connector_status ret = connector_status_connected;
+	struct radeon_connector *radeon_connector = to_radeon_connector(connector);
+	struct drm_encoder *encoder = radeon_best_single_encoder(connector);
+	enum drm_connector_status ret = connector_status_disconnected;
+
+	if (encoder) {
+		struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
+		struct drm_display_mode *native_mode = &radeon_encoder->native_mode;
+
+		/* check if panel is valid */
+		if (native_mode->hdisplay >= 320 && native_mode->vdisplay >= 240)
+			ret = connector_status_connected;
+
+	}
+
+	/* check for edid as well */
+	if (radeon_connector->edid)
+		ret = connector_status_connected;
+	else {
+		if (radeon_connector->ddc_bus) {
+			radeon_connector->edid = drm_get_edid(&radeon_connector->base,
+							      &radeon_connector->ddc_bus->adapter);
+			if (radeon_connector->edid)
+				ret = connector_status_connected;
+		}
+	}
 	/* check acpi lid status ??? */
+
 	radeon_connector_update_scratch_regs(connector, ret);
 	return ret;
 }
@@ -426,6 +497,8 @@ static void radeon_connector_destroy(struct drm_connector *connector)
 
 	if (radeon_connector->ddc_bus)
 		radeon_i2c_destroy(radeon_connector->ddc_bus);
+	if (radeon_connector->edid)
+		kfree(radeon_connector->edid);
 	kfree(radeon_connector->con_priv);
 	drm_sysfs_connector_remove(connector);
 	drm_connector_cleanup(connector);
@@ -495,6 +568,8 @@ static int radeon_vga_get_modes(struct drm_connector *connector)
 static int radeon_vga_mode_valid(struct drm_connector *connector,
 				  struct drm_display_mode *mode)
 {
+	/* XXX check mode bandwidth */
+	/* XXX verify against max DAC output frequency */
 	return MODE_OK;
 }
 
@@ -503,20 +578,41 @@ static enum drm_connector_status radeon_vga_detect(struct drm_connector *connect
 	struct radeon_connector *radeon_connector = to_radeon_connector(connector);
 	struct drm_encoder *encoder;
 	struct drm_encoder_helper_funcs *encoder_funcs;
-	bool dret;
+	bool dret = false;
 	enum drm_connector_status ret = connector_status_disconnected;
 
 	encoder = radeon_best_single_encoder(connector);
 	if (!encoder)
 		ret = connector_status_disconnected;
 
-	radeon_i2c_do_lock(radeon_connector, 1);
-	dret = radeon_ddc_probe(radeon_connector);
-	radeon_i2c_do_lock(radeon_connector, 0);
-	if (dret)
-		ret = connector_status_connected;
-	else {
-		if (radeon_connector->dac_load_detect) {
+	if (radeon_connector->ddc_bus)
+		dret = radeon_ddc_probe(radeon_connector);
+	if (dret) {
+		if (radeon_connector->edid) {
+			kfree(radeon_connector->edid);
+			radeon_connector->edid = NULL;
+		}
+		radeon_connector->edid = drm_get_edid(&radeon_connector->base, &radeon_connector->ddc_bus->adapter);
+
+		if (!radeon_connector->edid) {
+			DRM_ERROR("%s: probed a monitor but no|invalid EDID\n",
+					drm_get_connector_name(connector));
+			ret = connector_status_connected;
+		} else {
+			radeon_connector->use_digital = !!(radeon_connector->edid->input & DRM_EDID_INPUT_DIGITAL);
+
+			/* some oems have boards with separate digital and analog connectors
+			 * with a shared ddc line (often vga + hdmi)
+			 */
+			if (radeon_connector->use_digital && radeon_connector->shared_ddc) {
+				kfree(radeon_connector->edid);
+				radeon_connector->edid = NULL;
+				ret = connector_status_disconnected;
+			} else
+				ret = connector_status_connected;
+		}
+	} else {
+		if (radeon_connector->dac_load_detect && encoder) {
 			encoder_funcs = encoder->helper_private;
 			ret = encoder_funcs->detect(encoder, connector);
 		}
@@ -559,7 +655,7 @@ static int radeon_tv_get_modes(struct drm_connector *connector)
 		radeon_add_common_modes(encoder, connector);
 	else {
 		/* only 800x600 is supported right now on pre-avivo chips */
-		tv_mode = drm_cvt_mode(dev, 800, 600, 60, false, false);
+		tv_mode = drm_cvt_mode(dev, 800, 600, 60, false, false, false);
 		tv_mode->type = DRM_MODE_TYPE_DRIVER | DRM_MODE_TYPE_PREFERRED;
 		drm_mode_probed_add(connector, tv_mode);
 	}
@@ -569,6 +665,8 @@ static int radeon_tv_get_modes(struct drm_connector *connector)
 static int radeon_tv_mode_valid(struct drm_connector *connector,
 				struct drm_display_mode *mode)
 {
+	if ((mode->hdisplay > 1024) || (mode->vdisplay > 768))
+		return MODE_CLOCK_RANGE;
 	return MODE_OK;
 }
 
@@ -637,26 +735,65 @@ static enum drm_connector_status radeon_dvi_detect(struct drm_connector *connect
 	struct drm_mode_object *obj;
 	int i;
 	enum drm_connector_status ret = connector_status_disconnected;
-	bool dret;
+	bool dret = false;
 
-	radeon_i2c_do_lock(radeon_connector, 1);
-	dret = radeon_ddc_probe(radeon_connector);
-	radeon_i2c_do_lock(radeon_connector, 0);
+	if (radeon_connector->ddc_bus)
+		dret = radeon_ddc_probe(radeon_connector);
 	if (dret) {
-		radeon_i2c_do_lock(radeon_connector, 1);
+		if (radeon_connector->edid) {
+			kfree(radeon_connector->edid);
+			radeon_connector->edid = NULL;
+		}
 		radeon_connector->edid = drm_get_edid(&radeon_connector->base, &radeon_connector->ddc_bus->adapter);
-		radeon_i2c_do_lock(radeon_connector, 0);
 
 		if (!radeon_connector->edid) {
-			DRM_ERROR("DDC responded but not EDID found for %s\n",
-				  drm_get_connector_name(connector));
+			DRM_ERROR("%s: probed a monitor but no|invalid EDID\n",
+					drm_get_connector_name(connector));
 		} else {
 			radeon_connector->use_digital = !!(radeon_connector->edid->input & DRM_EDID_INPUT_DIGITAL);
 
-			/* if this isn't a digital monitor
-			   then we need to make sure we don't have any
-			   TV conflicts */
-			ret = connector_status_connected;
+			/* some oems have boards with separate digital and analog connectors
+			 * with a shared ddc line (often vga + hdmi)
+			 */
+			if ((!radeon_connector->use_digital) && radeon_connector->shared_ddc) {
+				kfree(radeon_connector->edid);
+				radeon_connector->edid = NULL;
+				ret = connector_status_disconnected;
+			} else
+				ret = connector_status_connected;
+
+			/* multiple connectors on the same encoder with the same ddc line
+			 * This tends to be HDMI and DVI on the same encoder with the
+			 * same ddc line.  If the edid says HDMI, consider the HDMI port
+			 * connected and the DVI port disconnected.  If the edid doesn't
+			 * say HDMI, vice versa.
+			 */
+			if (radeon_connector->shared_ddc && (ret == connector_status_connected)) {
+				struct drm_device *dev = connector->dev;
+				struct drm_connector *list_connector;
+				struct radeon_connector *list_radeon_connector;
+				list_for_each_entry(list_connector, &dev->mode_config.connector_list, head) {
+					if (connector == list_connector)
+						continue;
+					list_radeon_connector = to_radeon_connector(list_connector);
+					if (radeon_connector->devices == list_radeon_connector->devices) {
+						if (drm_detect_hdmi_monitor(radeon_connector->edid)) {
+							if (connector->connector_type == DRM_MODE_CONNECTOR_DVID) {
+								kfree(radeon_connector->edid);
+								radeon_connector->edid = NULL;
+								ret = connector_status_disconnected;
+							}
+						} else {
+							if ((connector->connector_type == DRM_MODE_CONNECTOR_HDMIA) ||
+							    (connector->connector_type == DRM_MODE_CONNECTOR_HDMIB)) {
+								kfree(radeon_connector->edid);
+								radeon_connector->edid = NULL;
+								ret = connector_status_disconnected;
+							}
+						}
+					}
+				}
+			}
 		}
 	}
 
@@ -743,9 +880,44 @@ struct drm_encoder *radeon_dvi_encoder(struct drm_connector *connector)
 	return NULL;
 }
 
+static void radeon_dvi_force(struct drm_connector *connector)
+{
+	struct radeon_connector *radeon_connector = to_radeon_connector(connector);
+	if (connector->force == DRM_FORCE_ON)
+		radeon_connector->use_digital = false;
+	if (connector->force == DRM_FORCE_ON_DIGITAL)
+		radeon_connector->use_digital = true;
+}
+
+static int radeon_dvi_mode_valid(struct drm_connector *connector,
+				  struct drm_display_mode *mode)
+{
+	struct drm_device *dev = connector->dev;
+	struct radeon_device *rdev = dev->dev_private;
+	struct radeon_connector *radeon_connector = to_radeon_connector(connector);
+
+	/* XXX check mode bandwidth */
+
+	/* clocks over 135 MHz have heat issues with DVI on RV100 */
+	if (radeon_connector->use_digital &&
+	    (rdev->family == CHIP_RV100) &&
+	    (mode->clock > 135000))
+		return MODE_CLOCK_HIGH;
+
+	if (radeon_connector->use_digital && (mode->clock > 165000)) {
+		if ((radeon_connector->connector_object_id == CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_I) ||
+		    (radeon_connector->connector_object_id == CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_D) ||
+		    (radeon_connector->connector_object_id == CONNECTOR_OBJECT_ID_HDMI_TYPE_B))
+			return MODE_OK;
+		else
+			return MODE_CLOCK_HIGH;
+	}
+	return MODE_OK;
+}
+
 struct drm_connector_helper_funcs radeon_dvi_connector_helper_funcs = {
 	.get_modes = radeon_dvi_get_modes,
-	.mode_valid = radeon_vga_mode_valid,
+	.mode_valid = radeon_dvi_mode_valid,
 	.best_encoder = radeon_dvi_encoder,
 };
 
@@ -755,6 +927,92 @@ struct drm_connector_funcs radeon_dvi_connector_funcs = {
 	.fill_modes = drm_helper_probe_single_connector_modes,
 	.set_property = radeon_connector_set_property,
 	.destroy = radeon_connector_destroy,
+	.force = radeon_dvi_force,
+};
+
+static void radeon_dp_connector_destroy(struct drm_connector *connector)
+{
+	struct radeon_connector *radeon_connector = to_radeon_connector(connector);
+	struct radeon_connector_atom_dig *radeon_dig_connector = radeon_connector->con_priv;
+
+	if (radeon_connector->ddc_bus)
+		radeon_i2c_destroy(radeon_connector->ddc_bus);
+	if (radeon_connector->edid)
+		kfree(radeon_connector->edid);
+	if (radeon_dig_connector->dp_i2c_bus)
+		radeon_i2c_destroy_dp(radeon_dig_connector->dp_i2c_bus);
+	kfree(radeon_connector->con_priv);
+	drm_sysfs_connector_remove(connector);
+	drm_connector_cleanup(connector);
+	kfree(connector);
+}
+
+static int radeon_dp_get_modes(struct drm_connector *connector)
+{
+	struct radeon_connector *radeon_connector = to_radeon_connector(connector);
+	int ret;
+
+	ret = radeon_ddc_get_modes(radeon_connector);
+	return ret;
+}
+
+static enum drm_connector_status radeon_dp_detect(struct drm_connector *connector)
+{
+	struct radeon_connector *radeon_connector = to_radeon_connector(connector);
+	enum drm_connector_status ret = connector_status_disconnected;
+	struct radeon_connector_atom_dig *radeon_dig_connector = radeon_connector->con_priv;
+	u8 sink_type;
+
+	if (radeon_connector->edid) {
+		kfree(radeon_connector->edid);
+		radeon_connector->edid = NULL;
+	}
+
+	sink_type = radeon_dp_getsinktype(radeon_connector);
+	if ((sink_type == CONNECTOR_OBJECT_ID_DISPLAYPORT) ||
+	    (sink_type == CONNECTOR_OBJECT_ID_eDP)) {
+		if (radeon_dp_getdpcd(radeon_connector)) {
+			radeon_dig_connector->dp_sink_type = sink_type;
+			ret = connector_status_connected;
+		}
+	} else {
+		if (radeon_ddc_probe(radeon_connector)) {
+			radeon_dig_connector->dp_sink_type = sink_type;
+			ret = connector_status_connected;
+		}
+	}
+
+	return ret;
+}
+
+static int radeon_dp_mode_valid(struct drm_connector *connector,
+				  struct drm_display_mode *mode)
+{
+	struct radeon_connector *radeon_connector = to_radeon_connector(connector);
+	struct radeon_connector_atom_dig *radeon_dig_connector = radeon_connector->con_priv;
+
+	/* XXX check mode bandwidth */
+
+	if ((radeon_dig_connector->dp_sink_type == CONNECTOR_OBJECT_ID_DISPLAYPORT) ||
+	    (radeon_dig_connector->dp_sink_type == CONNECTOR_OBJECT_ID_eDP))
+		return radeon_dp_mode_valid_helper(radeon_connector, mode);
+	else
+		return MODE_OK;
+}
+
+struct drm_connector_helper_funcs radeon_dp_connector_helper_funcs = {
+	.get_modes = radeon_dp_get_modes,
+	.mode_valid = radeon_dp_mode_valid,
+	.best_encoder = radeon_dvi_encoder,
+};
+
+struct drm_connector_funcs radeon_dp_connector_funcs = {
+	.dpms = drm_helper_connector_dpms,
+	.detect = radeon_dp_detect,
+	.fill_modes = drm_helper_probe_single_connector_modes,
+	.set_property = radeon_connector_set_property,
+	.destroy = radeon_dp_connector_destroy,
+	.force = radeon_dvi_force,
 };
 
 void
@@ -764,13 +1022,17 @@ radeon_add_atom_connector(struct drm_device *dev,
 			  int connector_type,
 			  struct radeon_i2c_bus_rec *i2c_bus,
 			  bool linkb,
-			  uint32_t igp_lane_info)
+			  uint32_t igp_lane_info,
+			  uint16_t connector_object_id,
+			  struct radeon_hpd *hpd)
 {
 	struct radeon_device *rdev = dev->dev_private;
 	struct drm_connector *connector;
 	struct radeon_connector *radeon_connector;
 	struct radeon_connector_atom_dig *radeon_dig_connector;
 	uint32_t subpixel_order = SubPixelNone;
+	bool shared_ddc = false;
+	int ret;
 
 	/* fixme - tv/cv/din */
 	if (connector_type == DRM_MODE_CONNECTOR_Unknown)
@@ -783,6 +1045,12 @@ radeon_add_atom_connector(struct drm_device *dev,
 			radeon_connector->devices |= supported_device;
 			return;
 		}
+		if (radeon_connector->ddc_bus && i2c_bus->valid) {
+			if (radeon_connector->ddc_bus->rec.i2c_id == i2c_bus->i2c_id) {
+				radeon_connector->shared_ddc = true;
+				shared_ddc = true;
+			}
+		}
 	}
 
 	radeon_connector = kzalloc(sizeof(struct radeon_connector), GFP_KERNEL);
@@ -793,27 +1061,36 @@ radeon_add_atom_connector(struct drm_device *dev,
 
 	radeon_connector->connector_id = connector_id;
 	radeon_connector->devices = supported_device;
+	radeon_connector->shared_ddc = shared_ddc;
+	radeon_connector->connector_object_id = connector_object_id;
+	radeon_connector->hpd = *hpd;
 	switch (connector_type) {
 	case DRM_MODE_CONNECTOR_VGA:
 		drm_connector_init(dev, &radeon_connector->base, &radeon_vga_connector_funcs, connector_type);
-		drm_connector_helper_add(&radeon_connector->base, &radeon_vga_connector_helper_funcs);
+		ret = drm_connector_helper_add(&radeon_connector->base, &radeon_vga_connector_helper_funcs);
+		if (ret)
+			goto failed;
 		if (i2c_bus->valid) {
 			radeon_connector->ddc_bus = radeon_i2c_create(dev, i2c_bus, "VGA");
 			if (!radeon_connector->ddc_bus)
 				goto failed;
 		}
+		radeon_connector->dac_load_detect = true;
 		drm_connector_attach_property(&radeon_connector->base,
 					      rdev->mode_info.load_detect_property,
 					      1);
 		break;
 	case DRM_MODE_CONNECTOR_DVIA:
 		drm_connector_init(dev, &radeon_connector->base, &radeon_vga_connector_funcs, connector_type);
-		drm_connector_helper_add(&radeon_connector->base, &radeon_vga_connector_helper_funcs);
+		ret = drm_connector_helper_add(&radeon_connector->base, &radeon_vga_connector_helper_funcs);
+		if (ret)
+			goto failed;
 		if (i2c_bus->valid) {
 			radeon_connector->ddc_bus = radeon_i2c_create(dev, i2c_bus, "DVI");
 			if (!radeon_connector->ddc_bus)
 				goto failed;
 		}
+		radeon_connector->dac_load_detect = true;
 		drm_connector_attach_property(&radeon_connector->base,
 					      rdev->mode_info.load_detect_property,
 					      1);
@@ -827,7 +1104,9 @@ radeon_add_atom_connector(struct drm_device *dev,
 		radeon_dig_connector->igp_lane_info = igp_lane_info;
 		radeon_connector->con_priv = radeon_dig_connector;
 		drm_connector_init(dev, &radeon_connector->base, &radeon_dvi_connector_funcs, connector_type);
-		drm_connector_helper_add(&radeon_connector->base, &radeon_dvi_connector_helper_funcs);
+		ret = drm_connector_helper_add(&radeon_connector->base, &radeon_dvi_connector_helper_funcs);
+		if (ret)
+			goto failed;
 		if (i2c_bus->valid) {
 			radeon_connector->ddc_bus = radeon_i2c_create(dev, i2c_bus, "DVI");
 			if (!radeon_connector->ddc_bus)
@@ -837,9 +1116,12 @@ radeon_add_atom_connector(struct drm_device *dev,
 		drm_connector_attach_property(&radeon_connector->base,
 					      rdev->mode_info.coherent_mode_property,
 					      1);
-		drm_connector_attach_property(&radeon_connector->base,
-					      rdev->mode_info.load_detect_property,
-					      1);
+		if (connector_type == DRM_MODE_CONNECTOR_DVII) {
+			radeon_connector->dac_load_detect = true;
+			drm_connector_attach_property(&radeon_connector->base,
+						      rdev->mode_info.load_detect_property,
+						      1);
+		}
 		break;
 	case DRM_MODE_CONNECTOR_HDMIA:
 	case DRM_MODE_CONNECTOR_HDMIB:
@@ -850,7 +1132,9 @@ radeon_add_atom_connector(struct drm_device *dev,
 		radeon_dig_connector->igp_lane_info = igp_lane_info;
 		radeon_connector->con_priv = radeon_dig_connector;
 		drm_connector_init(dev, &radeon_connector->base, &radeon_dvi_connector_funcs, connector_type);
-		drm_connector_helper_add(&radeon_connector->base, &radeon_dvi_connector_helper_funcs);
+		ret = drm_connector_helper_add(&radeon_connector->base, &radeon_dvi_connector_helper_funcs);
+		if (ret)
+			goto failed;
 		if (i2c_bus->valid) {
 			radeon_connector->ddc_bus = radeon_i2c_create(dev, i2c_bus, "HDMI");
 			if (!radeon_connector->ddc_bus)
@@ -862,31 +1146,53 @@ radeon_add_atom_connector(struct drm_device *dev,
 		subpixel_order = SubPixelHorizontalRGB;
 		break;
 	case DRM_MODE_CONNECTOR_DisplayPort:
+	case DRM_MODE_CONNECTOR_eDP:
 		radeon_dig_connector = kzalloc(sizeof(struct radeon_connector_atom_dig), GFP_KERNEL);
 		if (!radeon_dig_connector)
 			goto failed;
 		radeon_dig_connector->linkb = linkb;
 		radeon_dig_connector->igp_lane_info = igp_lane_info;
 		radeon_connector->con_priv = radeon_dig_connector;
-		drm_connector_init(dev, &radeon_connector->base, &radeon_dvi_connector_funcs, connector_type);
-		drm_connector_helper_add(&radeon_connector->base, &radeon_dvi_connector_helper_funcs);
+		drm_connector_init(dev, &radeon_connector->base, &radeon_dp_connector_funcs, connector_type);
+		ret = drm_connector_helper_add(&radeon_connector->base, &radeon_dp_connector_helper_funcs);
+		if (ret)
+			goto failed;
 		if (i2c_bus->valid) {
-			radeon_connector->ddc_bus = radeon_i2c_create(dev, i2c_bus, "DP");
+			/* add DP i2c bus */
+			if (connector_type == DRM_MODE_CONNECTOR_eDP)
+				radeon_dig_connector->dp_i2c_bus = radeon_i2c_create_dp(dev, i2c_bus, "eDP-auxch");
+			else
+				radeon_dig_connector->dp_i2c_bus = radeon_i2c_create_dp(dev, i2c_bus, "DP-auxch");
+			if (!radeon_dig_connector->dp_i2c_bus)
+				goto failed;
+			if (connector_type == DRM_MODE_CONNECTOR_eDP)
+				radeon_connector->ddc_bus = radeon_i2c_create(dev, i2c_bus, "eDP");
+			else
+				radeon_connector->ddc_bus = radeon_i2c_create(dev, i2c_bus, "DP");
 			if (!radeon_connector->ddc_bus)
 				goto failed;
 		}
 		subpixel_order = SubPixelHorizontalRGB;
+		drm_connector_attach_property(&radeon_connector->base,
+					      rdev->mode_info.coherent_mode_property,
+					      1);
 		break;
 	case DRM_MODE_CONNECTOR_SVIDEO:
 	case DRM_MODE_CONNECTOR_Composite:
 	case DRM_MODE_CONNECTOR_9PinDIN:
 		if (radeon_tv == 1) {
 			drm_connector_init(dev, &radeon_connector->base, &radeon_tv_connector_funcs, connector_type);
-			drm_connector_helper_add(&radeon_connector->base, &radeon_tv_connector_helper_funcs);
+			ret = drm_connector_helper_add(&radeon_connector->base, &radeon_tv_connector_helper_funcs);
+			if (ret)
+				goto failed;
+			radeon_connector->dac_load_detect = true;
+			drm_connector_attach_property(&radeon_connector->base,
+						      rdev->mode_info.load_detect_property,
+						      1);
+			drm_connector_attach_property(&radeon_connector->base,
+						      rdev->mode_info.tv_std_property,
+						      radeon_atombios_get_tv_info(rdev));
 		}
-		drm_connector_attach_property(&radeon_connector->base,
-					      rdev->mode_info.load_detect_property,
-					      1);
 		break;
 	case DRM_MODE_CONNECTOR_LVDS:
 		radeon_dig_connector = kzalloc(sizeof(struct radeon_connector_atom_dig), GFP_KERNEL);
@@ -896,13 +1202,14 @@ radeon_add_atom_connector(struct drm_device *dev,
 		radeon_dig_connector->igp_lane_info = igp_lane_info;
 		radeon_connector->con_priv = radeon_dig_connector;
 		drm_connector_init(dev, &radeon_connector->base, &radeon_lvds_connector_funcs, connector_type);
-		drm_connector_helper_add(&radeon_connector->base, &radeon_lvds_connector_helper_funcs);
+		ret = drm_connector_helper_add(&radeon_connector->base, &radeon_lvds_connector_helper_funcs);
+		if (ret)
+			goto failed;
 		if (i2c_bus->valid) {
 			radeon_connector->ddc_bus = radeon_i2c_create(dev, i2c_bus, "LVDS");
 			if (!radeon_connector->ddc_bus)
 				goto failed;
 		}
-		drm_mode_create_scaling_mode_property(dev);
 		drm_connector_attach_property(&radeon_connector->base,
 					      dev->mode_config.scaling_mode_property,
 					      DRM_MODE_SCALE_FULLSCREEN);
@@ -926,12 +1233,15 @@ radeon_add_legacy_connector(struct drm_device *dev,
 			    uint32_t connector_id,
 			    uint32_t supported_device,
 			    int connector_type,
-			    struct radeon_i2c_bus_rec *i2c_bus)
+			    struct radeon_i2c_bus_rec *i2c_bus,
+			    uint16_t connector_object_id,
+			    struct radeon_hpd *hpd)
 {
 	struct radeon_device *rdev = dev->dev_private;
 	struct drm_connector *connector;
 	struct radeon_connector *radeon_connector;
 	uint32_t subpixel_order = SubPixelNone;
+	int ret;
 
 	/* fixme - tv/cv/din */
 	if (connector_type == DRM_MODE_CONNECTOR_Unknown)
@@ -954,27 +1264,35 @@ radeon_add_legacy_connector(struct drm_device *dev,
 
 	radeon_connector->connector_id = connector_id;
 	radeon_connector->devices = supported_device;
+	radeon_connector->connector_object_id = connector_object_id;
+	radeon_connector->hpd = *hpd;
 	switch (connector_type) {
 	case DRM_MODE_CONNECTOR_VGA:
 		drm_connector_init(dev, &radeon_connector->base, &radeon_vga_connector_funcs, connector_type);
-		drm_connector_helper_add(&radeon_connector->base, &radeon_vga_connector_helper_funcs);
+		ret = drm_connector_helper_add(&radeon_connector->base, &radeon_vga_connector_helper_funcs);
+		if (ret)
+			goto failed;
 		if (i2c_bus->valid) {
 			radeon_connector->ddc_bus = radeon_i2c_create(dev, i2c_bus, "VGA");
 			if (!radeon_connector->ddc_bus)
 				goto failed;
 		}
+		radeon_connector->dac_load_detect = true;
 		drm_connector_attach_property(&radeon_connector->base,
 					      rdev->mode_info.load_detect_property,
 					      1);
 		break;
 	case DRM_MODE_CONNECTOR_DVIA:
 		drm_connector_init(dev, &radeon_connector->base, &radeon_vga_connector_funcs, connector_type);
-		drm_connector_helper_add(&radeon_connector->base, &radeon_vga_connector_helper_funcs);
+		ret = drm_connector_helper_add(&radeon_connector->base, &radeon_vga_connector_helper_funcs);
+		if (ret)
+			goto failed;
 		if (i2c_bus->valid) {
 			radeon_connector->ddc_bus = radeon_i2c_create(dev, i2c_bus, "DVI");
 			if (!radeon_connector->ddc_bus)
 				goto failed;
 		}
+		radeon_connector->dac_load_detect = true;
 		drm_connector_attach_property(&radeon_connector->base,
 					      rdev->mode_info.load_detect_property,
 					      1);
@@ -982,11 +1300,14 @@ radeon_add_legacy_connector(struct drm_device *dev,
 	case DRM_MODE_CONNECTOR_DVII:
 	case DRM_MODE_CONNECTOR_DVID:
 		drm_connector_init(dev, &radeon_connector->base, &radeon_dvi_connector_funcs, connector_type);
-		drm_connector_helper_add(&radeon_connector->base, &radeon_dvi_connector_helper_funcs);
+		ret = drm_connector_helper_add(&radeon_connector->base, &radeon_dvi_connector_helper_funcs);
+		if (ret)
+			goto failed;
 		if (i2c_bus->valid) {
 			radeon_connector->ddc_bus = radeon_i2c_create(dev, i2c_bus, "DVI");
 			if (!radeon_connector->ddc_bus)
 				goto failed;
+			radeon_connector->dac_load_detect = true;
 			drm_connector_attach_property(&radeon_connector->base,
 						      rdev->mode_info.load_detect_property,
 						      1);
@@ -998,15 +1319,30 @@ radeon_add_legacy_connector(struct drm_device *dev,
 	case DRM_MODE_CONNECTOR_9PinDIN:
 		if (radeon_tv == 1) {
 			drm_connector_init(dev, &radeon_connector->base, &radeon_tv_connector_funcs, connector_type);
-			drm_connector_helper_add(&radeon_connector->base, &radeon_tv_connector_helper_funcs);
+			ret = drm_connector_helper_add(&radeon_connector->base, &radeon_tv_connector_helper_funcs);
+			if (ret)
+				goto failed;
+			radeon_connector->dac_load_detect = true;
+			/* RS400,RC410,RS480 chipset seems to report a lot
+			 * of false positive on load detect, we haven't yet
+			 * found a way to make load detect reliable on those
+			 * chipset, thus just disable it for TV.
+			 */
+			if (rdev->family == CHIP_RS400 || rdev->family == CHIP_RS480)
+				radeon_connector->dac_load_detect = false;
 			drm_connector_attach_property(&radeon_connector->base,
 						      rdev->mode_info.load_detect_property,
-						      1);
+						      radeon_connector->dac_load_detect);
+			drm_connector_attach_property(&radeon_connector->base,
+						      rdev->mode_info.tv_std_property,
+						      radeon_combios_get_tv_info(rdev));
 		}
 		break;
 	case DRM_MODE_CONNECTOR_LVDS:
 		drm_connector_init(dev, &radeon_connector->base, &radeon_lvds_connector_funcs, connector_type);
-		drm_connector_helper_add(&radeon_connector->base, &radeon_lvds_connector_helper_funcs);
+		ret = drm_connector_helper_add(&radeon_connector->base, &radeon_lvds_connector_helper_funcs);
+		if (ret)
+			goto failed;
 		if (i2c_bus->valid) {
 			radeon_connector->ddc_bus = radeon_i2c_create(dev, i2c_bus, "LVDS");
 			if (!radeon_connector->ddc_bus)
diff --git a/drivers/gpu/drm/radeon/radeon_cp.c b/drivers/gpu/drm/radeon/radeon_cp.c
index 4f7afc79dd82..dc6eba6b96dd 100644
--- a/drivers/gpu/drm/radeon/radeon_cp.c
+++ b/drivers/gpu/drm/radeon/radeon_cp.c
@@ -1644,6 +1644,7 @@ static int radeon_do_resume_cp(struct drm_device *dev, struct drm_file *file_pri
 	radeon_cp_load_microcode(dev_priv);
 	radeon_cp_init_ring_buffer(dev, dev_priv, file_priv);
 
+	dev_priv->have_z_offset = 0;
 	radeon_do_engine_reset(dev);
 	radeon_irq_set_state(dev, RADEON_SW_INT_ENABLE, 1);
 
@@ -1941,8 +1942,8 @@ struct drm_buf *radeon_freelist_get(struct drm_device * dev)
 	for (t = 0; t < dev_priv->usec_timeout; t++) {
 		u32 done_age = GET_SCRATCH(dev_priv, 1);
 		DRM_DEBUG("done_age = %d\n", done_age);
-		for (i = start; i < dma->buf_count; i++) {
-			buf = dma->buflist[i];
+		for (i = 0; i < dma->buf_count; i++) {
+			buf = dma->buflist[start];
 			buf_priv = buf->dev_private;
 			if (buf->file_priv == NULL || (buf->pending &&
 						       buf_priv->age <=
@@ -1951,7 +1952,8 @@ struct drm_buf *radeon_freelist_get(struct drm_device * dev)
 				buf->pending = 0;
 				return buf;
 			}
-			start = 0;
+			if (++start >= dma->buf_count)
+				start = 0;
 		}
 
 		if (t) {
@@ -1960,47 +1962,9 @@ struct drm_buf *radeon_freelist_get(struct drm_device * dev)
 		}
 	}
 
-	DRM_DEBUG("returning NULL!\n");
 	return NULL;
 }
 
-#if 0
-struct drm_buf *radeon_freelist_get(struct drm_device * dev)
-{
-	struct drm_device_dma *dma = dev->dma;
-	drm_radeon_private_t *dev_priv = dev->dev_private;
-	drm_radeon_buf_priv_t *buf_priv;
-	struct drm_buf *buf;
-	int i, t;
-	int start;
-	u32 done_age;
-
-	done_age = radeon_read_ring_rptr(dev_priv, RADEON_SCRATCHOFF(1));
-	if (++dev_priv->last_buf >= dma->buf_count)
-		dev_priv->last_buf = 0;
-
-	start = dev_priv->last_buf;
-	dev_priv->stats.freelist_loops++;
-
-	for (t = 0; t < 2; t++) {
-		for (i = start; i < dma->buf_count; i++) {
-			buf = dma->buflist[i];
-			buf_priv = buf->dev_private;
-			if (buf->file_priv == 0 || (buf->pending &&
-						    buf_priv->age <=
-						    done_age)) {
-				dev_priv->stats.requested_bufs++;
-				buf->pending = 0;
-				return buf;
-			}
-		}
-		start = 0;
-	}
-
-	return NULL;
-}
-#endif
-
 void radeon_freelist_reset(struct drm_device * dev)
 {
 	struct drm_device_dma *dma = dev->dma;
@@ -2182,6 +2146,7 @@ int radeon_master_create(struct drm_device *dev, struct drm_master *master)
 			 &master_priv->sarea);
 	if (ret) {
 		DRM_ERROR("SAREA setup failed\n");
+		kfree(master_priv);
 		return ret;
 	}
 	master_priv->sarea_priv = master_priv->sarea->handle + sizeof(struct drm_sarea);
diff --git a/drivers/gpu/drm/radeon/radeon_cs.c b/drivers/gpu/drm/radeon/radeon_cs.c
index 12f5990c2d2a..70ba02ed7723 100644
--- a/drivers/gpu/drm/radeon/radeon_cs.c
+++ b/drivers/gpu/drm/radeon/radeon_cs.c
@@ -76,17 +76,17 @@ int radeon_cs_parser_relocs(struct radeon_cs_parser *p)
 			}
 			p->relocs_ptr[i] = &p->relocs[i];
 			p->relocs[i].robj = p->relocs[i].gobj->driver_private;
-			p->relocs[i].lobj.robj = p->relocs[i].robj;
+			p->relocs[i].lobj.bo = p->relocs[i].robj;
 			p->relocs[i].lobj.rdomain = r->read_domains;
 			p->relocs[i].lobj.wdomain = r->write_domain;
 			p->relocs[i].handle = r->handle;
 			p->relocs[i].flags = r->flags;
 			INIT_LIST_HEAD(&p->relocs[i].lobj.list);
-			radeon_object_list_add_object(&p->relocs[i].lobj,
-						      &p->validated);
+			radeon_bo_list_add_object(&p->relocs[i].lobj,
+						&p->validated);
 		}
 	}
-	return radeon_object_list_validate(&p->validated, p->ib->fence);
+	return radeon_bo_list_validate(&p->validated);
 }
 
 int radeon_cs_parser_init(struct radeon_cs_parser *p, void *data)
@@ -142,15 +142,31 @@ int radeon_cs_parser_init(struct radeon_cs_parser *p, void *data)
 		}
 
 		p->chunks[i].length_dw = user_chunk.length_dw;
-		cdata = (uint32_t *)(unsigned long)user_chunk.chunk_data;
+		p->chunks[i].user_ptr = (void __user *)(unsigned long)user_chunk.chunk_data;
 
-		size = p->chunks[i].length_dw * sizeof(uint32_t);
-		p->chunks[i].kdata = kmalloc(size, GFP_KERNEL);
-		if (p->chunks[i].kdata == NULL) {
-			return -ENOMEM;
-		}
-		if (DRM_COPY_FROM_USER(p->chunks[i].kdata, cdata, size)) {
-			return -EFAULT;
+		cdata = (uint32_t *)(unsigned long)user_chunk.chunk_data;
+		if (p->chunks[i].chunk_id != RADEON_CHUNK_ID_IB) {
+			size = p->chunks[i].length_dw * sizeof(uint32_t);
+			p->chunks[i].kdata = kmalloc(size, GFP_KERNEL);
+			if (p->chunks[i].kdata == NULL) {
+				return -ENOMEM;
+			}
+			if (DRM_COPY_FROM_USER(p->chunks[i].kdata,
+					       p->chunks[i].user_ptr, size)) {
+				return -EFAULT;
+			}
+		} else {
+			p->chunks[i].kpage[0] = kmalloc(PAGE_SIZE, GFP_KERNEL);
+			p->chunks[i].kpage[1] = kmalloc(PAGE_SIZE, GFP_KERNEL);
+			if (p->chunks[i].kpage[0] == NULL || p->chunks[i].kpage[1] == NULL) {
+				kfree(p->chunks[i].kpage[0]);
+				kfree(p->chunks[i].kpage[1]);
+				return -ENOMEM;
+			}
+			p->chunks[i].kpage_idx[0] = -1;
+			p->chunks[i].kpage_idx[1] = -1;
+			p->chunks[i].last_copied_page = -1;
+			p->chunks[i].last_page_index = ((p->chunks[i].length_dw * 4) - 1) / PAGE_SIZE;
 		}
 	}
 	if (p->chunks[p->chunk_ib_idx].length_dw > (16 * 1024)) {
@@ -173,23 +189,21 @@ static void radeon_cs_parser_fini(struct radeon_cs_parser *parser, int error)
 {
 	unsigned i;
 
-	if (error) {
-		radeon_object_list_unvalidate(&parser->validated);
-	} else {
-		radeon_object_list_clean(&parser->validated);
+	if (!error && parser->ib) {
+		radeon_bo_list_fence(&parser->validated, parser->ib->fence);
 	}
+	radeon_bo_list_unreserve(&parser->validated);
 	for (i = 0; i < parser->nrelocs; i++) {
-		if (parser->relocs[i].gobj) {
-			mutex_lock(&parser->rdev->ddev->struct_mutex);
-			drm_gem_object_unreference(parser->relocs[i].gobj);
-			mutex_unlock(&parser->rdev->ddev->struct_mutex);
-		}
+		if (parser->relocs[i].gobj)
+			drm_gem_object_unreference_unlocked(parser->relocs[i].gobj);
 	}
 	kfree(parser->track);
 	kfree(parser->relocs);
 	kfree(parser->relocs_ptr);
 	for (i = 0; i < parser->nchunks; i++) {
 		kfree(parser->chunks[i].kdata);
+		kfree(parser->chunks[i].kpage[0]);
+		kfree(parser->chunks[i].kpage[1]);
 	}
 	kfree(parser->chunks);
 	kfree(parser->chunks_array);
@@ -212,6 +226,7 @@ int radeon_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 	memset(&parser, 0, sizeof(struct radeon_cs_parser));
 	parser.filp = filp;
 	parser.rdev = rdev;
+	parser.dev = rdev->dev;
 	r = radeon_cs_parser_init(&parser, data);
 	if (r) {
 		DRM_ERROR("Failed to initialize parser !\n");
@@ -238,8 +253,14 @@ int radeon_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 	 * uncached). */
 	ib_chunk = &parser.chunks[parser.chunk_ib_idx];
 	parser.ib->length_dw = ib_chunk->length_dw;
-	memcpy((void *)parser.ib->ptr, ib_chunk->kdata, ib_chunk->length_dw*4);
 	r = radeon_cs_parse(&parser);
+	if (r || parser.parser_error) {
+		DRM_ERROR("Invalid command stream !\n");
+		radeon_cs_parser_fini(&parser, r);
+		mutex_unlock(&rdev->cs_mutex);
+		return r;
+	}
+	r = radeon_cs_finish_pages(&parser);
 	if (r) {
 		DRM_ERROR("Invalid command stream !\n");
 		radeon_cs_parser_fini(&parser, r);
@@ -254,3 +275,64 @@ int radeon_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 	mutex_unlock(&rdev->cs_mutex);
 	return r;
 }
+
+int radeon_cs_finish_pages(struct radeon_cs_parser *p)
+{
+	struct radeon_cs_chunk *ibc = &p->chunks[p->chunk_ib_idx];
+	int i;
+	int size = PAGE_SIZE;
+
+	for (i = ibc->last_copied_page + 1; i <= ibc->last_page_index; i++) {
+		if (i == ibc->last_page_index) {
+			size = (ibc->length_dw * 4) % PAGE_SIZE;
+			if (size == 0)
+				size = PAGE_SIZE;
+		}
+		
+		if (DRM_COPY_FROM_USER(p->ib->ptr + (i * (PAGE_SIZE/4)),
+				       ibc->user_ptr + (i * PAGE_SIZE),
+				       size))
+			return -EFAULT;
+	}
+	return 0;
+}
+
+int radeon_cs_update_pages(struct radeon_cs_parser *p, int pg_idx)
+{
+	int new_page;
+	struct radeon_cs_chunk *ibc = &p->chunks[p->chunk_ib_idx];
+	int i;
+	int size = PAGE_SIZE;
+
+	for (i = ibc->last_copied_page + 1; i < pg_idx; i++) {
+		if (DRM_COPY_FROM_USER(p->ib->ptr + (i * (PAGE_SIZE/4)),
+				       ibc->user_ptr + (i * PAGE_SIZE),
+				       PAGE_SIZE)) {
+			p->parser_error = -EFAULT;
+			return 0;
+		}
+	}
+
+	new_page = ibc->kpage_idx[0] < ibc->kpage_idx[1] ? 0 : 1;
+
+	if (pg_idx == ibc->last_page_index) {
+		size = (ibc->length_dw * 4) % PAGE_SIZE;
+			if (size == 0)
+				size = PAGE_SIZE;
+	}
+
+	if (DRM_COPY_FROM_USER(ibc->kpage[new_page],
+			       ibc->user_ptr + (pg_idx * PAGE_SIZE),
+			       size)) {
+		p->parser_error = -EFAULT;
+		return 0;
+	}
+
+	/* copy to IB here */
+	memcpy((void *)(p->ib->ptr+(pg_idx*(PAGE_SIZE/4))), ibc->kpage[new_page], size);
+
+	ibc->last_copied_page = pg_idx;
+	ibc->kpage_idx[new_page] = pg_idx;
+
+	return new_page;
+}
diff --git a/drivers/gpu/drm/radeon/radeon_cursor.c b/drivers/gpu/drm/radeon/radeon_cursor.c
index b13c79e38bc0..b7023fff89eb 100644
--- a/drivers/gpu/drm/radeon/radeon_cursor.c
+++ b/drivers/gpu/drm/radeon/radeon_cursor.c
@@ -36,7 +36,14 @@ static void radeon_lock_cursor(struct drm_crtc *crtc, bool lock)
 	struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc);
 	uint32_t cur_lock;
 
-	if (ASIC_IS_AVIVO(rdev)) {
+	if (ASIC_IS_DCE4(rdev)) {
+		cur_lock = RREG32(EVERGREEN_CUR_UPDATE + radeon_crtc->crtc_offset);
+		if (lock)
+			cur_lock |= EVERGREEN_CURSOR_UPDATE_LOCK;
+		else
+			cur_lock &= ~EVERGREEN_CURSOR_UPDATE_LOCK;
+		WREG32(EVERGREEN_CUR_UPDATE + radeon_crtc->crtc_offset, cur_lock);
+	} else if (ASIC_IS_AVIVO(rdev)) {
 		cur_lock = RREG32(AVIVO_D1CUR_UPDATE + radeon_crtc->crtc_offset);
 		if (lock)
 			cur_lock |= AVIVO_D1CURSOR_UPDATE_LOCK;
@@ -58,7 +65,10 @@ static void radeon_hide_cursor(struct drm_crtc *crtc)
 	struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc);
 	struct radeon_device *rdev = crtc->dev->dev_private;
 
-	if (ASIC_IS_AVIVO(rdev)) {
+	if (ASIC_IS_DCE4(rdev)) {
+		WREG32(RADEON_MM_INDEX, EVERGREEN_CUR_CONTROL + radeon_crtc->crtc_offset);
+		WREG32(RADEON_MM_DATA, EVERGREEN_CURSOR_MODE(EVERGREEN_CURSOR_24_8_PRE_MULT));
+	} else if (ASIC_IS_AVIVO(rdev)) {
 		WREG32(RADEON_MM_INDEX, AVIVO_D1CUR_CONTROL + radeon_crtc->crtc_offset);
 		WREG32(RADEON_MM_DATA, (AVIVO_D1CURSOR_MODE_24BPP << AVIVO_D1CURSOR_MODE_SHIFT));
 	} else {
@@ -81,10 +91,14 @@ static void radeon_show_cursor(struct drm_crtc *crtc)
 	struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc);
 	struct radeon_device *rdev = crtc->dev->dev_private;
 
-	if (ASIC_IS_AVIVO(rdev)) {
+	if (ASIC_IS_DCE4(rdev)) {
+		WREG32(RADEON_MM_INDEX, EVERGREEN_CUR_CONTROL + radeon_crtc->crtc_offset);
+		WREG32(RADEON_MM_DATA, EVERGREEN_CURSOR_EN |
+		       EVERGREEN_CURSOR_MODE(EVERGREEN_CURSOR_24_8_PRE_MULT));
+	} else if (ASIC_IS_AVIVO(rdev)) {
 		WREG32(RADEON_MM_INDEX, AVIVO_D1CUR_CONTROL + radeon_crtc->crtc_offset);
 		WREG32(RADEON_MM_DATA, AVIVO_D1CURSOR_EN |
-			     (AVIVO_D1CURSOR_MODE_24BPP << AVIVO_D1CURSOR_MODE_SHIFT));
+		       (AVIVO_D1CURSOR_MODE_24BPP << AVIVO_D1CURSOR_MODE_SHIFT));
 	} else {
 		switch (radeon_crtc->crtc_id) {
 		case 0:
@@ -109,9 +123,18 @@ static void radeon_set_cursor(struct drm_crtc *crtc, struct drm_gem_object *obj,
 	struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc);
 	struct radeon_device *rdev = crtc->dev->dev_private;
 
-	if (ASIC_IS_AVIVO(rdev))
+	if (ASIC_IS_DCE4(rdev)) {
+		WREG32(EVERGREEN_CUR_SURFACE_ADDRESS_HIGH + radeon_crtc->crtc_offset, 0);
+		WREG32(EVERGREEN_CUR_SURFACE_ADDRESS + radeon_crtc->crtc_offset, gpu_addr);
+	} else if (ASIC_IS_AVIVO(rdev)) {
+		if (rdev->family >= CHIP_RV770) {
+			if (radeon_crtc->crtc_id)
+				WREG32(R700_D2CUR_SURFACE_ADDRESS_HIGH, 0);
+			else
+				WREG32(R700_D1CUR_SURFACE_ADDRESS_HIGH, 0);
+		}
 		WREG32(AVIVO_D1CUR_SURFACE_ADDRESS + radeon_crtc->crtc_offset, gpu_addr);
-	else {
+	} else {
 		radeon_crtc->legacy_cursor_offset = gpu_addr - radeon_crtc->legacy_display_base_addr;
 		/* offset is from DISP(2)_BASE_ADDRESS */
 		WREG32(RADEON_CUR_OFFSET + radeon_crtc->crtc_offset, radeon_crtc->legacy_cursor_offset);
@@ -163,17 +186,13 @@ int radeon_crtc_cursor_set(struct drm_crtc *crtc,
 unpin:
 	if (radeon_crtc->cursor_bo) {
 		radeon_gem_object_unpin(radeon_crtc->cursor_bo);
-		mutex_lock(&crtc->dev->struct_mutex);
-		drm_gem_object_unreference(radeon_crtc->cursor_bo);
-		mutex_unlock(&crtc->dev->struct_mutex);
+		drm_gem_object_unreference_unlocked(radeon_crtc->cursor_bo);
 	}
 
 	radeon_crtc->cursor_bo = obj;
 	return 0;
 fail:
-	mutex_lock(&crtc->dev->struct_mutex);
-	drm_gem_object_unreference(obj);
-	mutex_unlock(&crtc->dev->struct_mutex);
+	drm_gem_object_unreference_unlocked(obj);
 
 	return 0;
 }
@@ -195,7 +214,20 @@ int radeon_crtc_cursor_move(struct drm_crtc *crtc,
 		yorigin = CURSOR_HEIGHT - 1;
 
 	radeon_lock_cursor(crtc, true);
-	if (ASIC_IS_AVIVO(rdev)) {
+	if (ASIC_IS_DCE4(rdev)) {
+		/* cursors are offset into the total surface */
+		x += crtc->x;
+		y += crtc->y;
+		DRM_DEBUG("x %d y %d c->x %d c->y %d\n", x, y, crtc->x, crtc->y);
+
+		/* XXX: check if evergreen has the same issues as avivo chips */
+		WREG32(EVERGREEN_CUR_POSITION + radeon_crtc->crtc_offset,
+		       ((xorigin ? 0 : x) << 16) |
+		       (yorigin ? 0 : y));
+		WREG32(EVERGREEN_CUR_HOT_SPOT + radeon_crtc->crtc_offset, (xorigin << 16) | yorigin);
+		WREG32(EVERGREEN_CUR_SIZE + radeon_crtc->crtc_offset,
+		       ((radeon_crtc->cursor_width - 1) << 16) | (radeon_crtc->cursor_height - 1));
+	} else if (ASIC_IS_AVIVO(rdev)) {
 		int w = radeon_crtc->cursor_width;
 		int i = 0;
 		struct drm_crtc *crtc_p;
diff --git a/drivers/gpu/drm/radeon/radeon_device.c b/drivers/gpu/drm/radeon/radeon_device.c
index daf5db780956..e28e4ed5f720 100644
--- a/drivers/gpu/drm/radeon/radeon_device.c
+++ b/drivers/gpu/drm/radeon/radeon_device.c
@@ -30,6 +30,7 @@
 #include <drm/drm_crtc_helper.h>
 #include <drm/radeon_drm.h>
 #include <linux/vgaarb.h>
+#include <linux/vga_switcheroo.h>
 #include "radeon_reg.h"
 #include "radeon.h"
 #include "radeon_asic.h"
@@ -44,10 +45,11 @@ void radeon_surface_init(struct radeon_device *rdev)
 	if (rdev->family < CHIP_R600) {
 		int i;
 
-		for (i = 0; i < 8; i++) {
-			WREG32(RADEON_SURFACE0_INFO +
-			       i * (RADEON_SURFACE1_INFO - RADEON_SURFACE0_INFO),
-			       0);
+		for (i = 0; i < RADEON_GEM_MAX_SURFACES; i++) {
+			if (rdev->surface_regs[i].bo)
+				radeon_bo_get_surface_reg(rdev->surface_regs[i].bo);
+			else
+				radeon_clear_surface_reg(rdev, i);
 		}
 		/* enable surfaces */
 		WREG32(RADEON_SURFACE_CNTL, 0);
@@ -99,80 +101,103 @@ void radeon_scratch_free(struct radeon_device *rdev, uint32_t reg)
 	}
 }
 
-/*
- * MC common functions
+/**
+ * radeon_vram_location - try to find VRAM location
+ * @rdev: radeon device structure holding all necessary informations
+ * @mc: memory controller structure holding memory informations
+ * @base: base address at which to put VRAM
+ *
+ * Function will place try to place VRAM at base address provided
+ * as parameter (which is so far either PCI aperture address or
+ * for IGP TOM base address).
+ *
+ * If there is not enough space to fit the unvisible VRAM in the 32bits
+ * address space then we limit the VRAM size to the aperture.
+ *
+ * If we are using AGP and if the AGP aperture doesn't allow us to have
+ * room for all the VRAM than we restrict the VRAM to the PCI aperture
+ * size and print a warning.
+ *
+ * This function will never fails, worst case are limiting VRAM.
+ *
+ * Note: GTT start, end, size should be initialized before calling this
+ * function on AGP platform.
+ *
+ * Note: We don't explictly enforce VRAM start to be aligned on VRAM size,
+ * this shouldn't be a problem as we are using the PCI aperture as a reference.
+ * Otherwise this would be needed for rv280, all r3xx, and all r4xx, but
+ * not IGP.
+ *
+ * Note: we use mc_vram_size as on some board we need to program the mc to
+ * cover the whole aperture even if VRAM size is inferior to aperture size
+ * Novell bug 204882 + along with lots of ubuntu ones
+ *
+ * Note: when limiting vram it's safe to overwritte real_vram_size because
+ * we are not in case where real_vram_size is inferior to mc_vram_size (ie
+ * note afected by bogus hw of Novell bug 204882 + along with lots of ubuntu
+ * ones)
+ *
+ * Note: IGP TOM addr should be the same as the aperture addr, we don't
+ * explicitly check for that thought.
+ *
+ * FIXME: when reducing VRAM size align new size on power of 2.
  */
-int radeon_mc_setup(struct radeon_device *rdev)
+void radeon_vram_location(struct radeon_device *rdev, struct radeon_mc *mc, u64 base)
+{
+	mc->vram_start = base;
+	if (mc->mc_vram_size > (0xFFFFFFFF - base + 1)) {
+		dev_warn(rdev->dev, "limiting VRAM to PCI aperture size\n");
+		mc->real_vram_size = mc->aper_size;
+		mc->mc_vram_size = mc->aper_size;
+	}
+	mc->vram_end = mc->vram_start + mc->mc_vram_size - 1;
+	if (rdev->flags & RADEON_IS_AGP && mc->vram_end > mc->gtt_start && mc->vram_end <= mc->gtt_end) {
+		dev_warn(rdev->dev, "limiting VRAM to PCI aperture size\n");
+		mc->real_vram_size = mc->aper_size;
+		mc->mc_vram_size = mc->aper_size;
+	}
+	mc->vram_end = mc->vram_start + mc->mc_vram_size - 1;
+	dev_info(rdev->dev, "VRAM: %lluM 0x%08llX - 0x%08llX (%lluM used)\n",
+			mc->mc_vram_size >> 20, mc->vram_start,
+			mc->vram_end, mc->real_vram_size >> 20);
+}
+
+/**
+ * radeon_gtt_location - try to find GTT location
+ * @rdev: radeon device structure holding all necessary informations
+ * @mc: memory controller structure holding memory informations
+ *
+ * Function will place try to place GTT before or after VRAM.
+ *
+ * If GTT size is bigger than space left then we ajust GTT size.
+ * Thus function will never fails.
+ *
+ * FIXME: when reducing GTT size align new size on power of 2.
+ */
+void radeon_gtt_location(struct radeon_device *rdev, struct radeon_mc *mc)
 {
-	uint32_t tmp;
+	u64 size_af, size_bf;
 
-	/* Some chips have an "issue" with the memory controller, the
-	 * location must be aligned to the size. We just align it down,
-	 * too bad if we walk over the top of system memory, we don't
-	 * use DMA without a remapped anyway.
-	 * Affected chips are rv280, all r3xx, and all r4xx, but not IGP
-	 */
-	/* FGLRX seems to setup like this, VRAM a 0, then GART.
-	 */
-	/*
-	 * Note: from R6xx the address space is 40bits but here we only
-	 * use 32bits (still have to see a card which would exhaust 4G
-	 * address space).
-	 */
-	if (rdev->mc.vram_location != 0xFFFFFFFFUL) {
-		/* vram location was already setup try to put gtt after
-		 * if it fits */
-		tmp = rdev->mc.vram_location + rdev->mc.mc_vram_size;
-		tmp = (tmp + rdev->mc.gtt_size - 1) & ~(rdev->mc.gtt_size - 1);
-		if ((0xFFFFFFFFUL - tmp) >= rdev->mc.gtt_size) {
-			rdev->mc.gtt_location = tmp;
-		} else {
-			if (rdev->mc.gtt_size >= rdev->mc.vram_location) {
-				printk(KERN_ERR "[drm] GTT too big to fit "
-				       "before or after vram location.\n");
-				return -EINVAL;
-			}
-			rdev->mc.gtt_location = 0;
-		}
-	} else if (rdev->mc.gtt_location != 0xFFFFFFFFUL) {
-		/* gtt location was already setup try to put vram before
-		 * if it fits */
-		if (rdev->mc.mc_vram_size < rdev->mc.gtt_location) {
-			rdev->mc.vram_location = 0;
-		} else {
-			tmp = rdev->mc.gtt_location + rdev->mc.gtt_size;
-			tmp += (rdev->mc.mc_vram_size - 1);
-			tmp &= ~(rdev->mc.mc_vram_size - 1);
-			if ((0xFFFFFFFFUL - tmp) >= rdev->mc.mc_vram_size) {
-				rdev->mc.vram_location = tmp;
-			} else {
-				printk(KERN_ERR "[drm] vram too big to fit "
-				       "before or after GTT location.\n");
-				return -EINVAL;
-			}
+	size_af = 0xFFFFFFFF - mc->vram_end;
+	size_bf = mc->vram_start;
+	if (size_bf > size_af) {
+		if (mc->gtt_size > size_bf) {
+			dev_warn(rdev->dev, "limiting GTT\n");
+			mc->gtt_size = size_bf;
 		}
+		mc->gtt_start = mc->vram_start - mc->gtt_size;
 	} else {
-		rdev->mc.vram_location = 0;
-		tmp = rdev->mc.mc_vram_size;
-		tmp = (tmp + rdev->mc.gtt_size - 1) & ~(rdev->mc.gtt_size - 1);
-		rdev->mc.gtt_location = tmp;
-	}
-	rdev->mc.vram_start = rdev->mc.vram_location;
-	rdev->mc.vram_end = rdev->mc.vram_location + rdev->mc.mc_vram_size - 1;
-	rdev->mc.gtt_start = rdev->mc.gtt_location;
-	rdev->mc.gtt_end = rdev->mc.gtt_location + rdev->mc.gtt_size - 1;
-	DRM_INFO("radeon: VRAM %uM\n", (unsigned)(rdev->mc.mc_vram_size >> 20));
-	DRM_INFO("radeon: VRAM from 0x%08X to 0x%08X\n",
-		 (unsigned)rdev->mc.vram_location,
-		 (unsigned)(rdev->mc.vram_location + rdev->mc.mc_vram_size - 1));
-	DRM_INFO("radeon: GTT %uM\n", (unsigned)(rdev->mc.gtt_size >> 20));
-	DRM_INFO("radeon: GTT from 0x%08X to 0x%08X\n",
-		 (unsigned)rdev->mc.gtt_location,
-		 (unsigned)(rdev->mc.gtt_location + rdev->mc.gtt_size - 1));
-	return 0;
+		if (mc->gtt_size > size_af) {
+			dev_warn(rdev->dev, "limiting GTT\n");
+			mc->gtt_size = size_af;
+		}
+		mc->gtt_start = mc->vram_end + 1;
+	}
+	mc->gtt_end = mc->gtt_start + mc->gtt_size - 1;
+	dev_info(rdev->dev, "GTT: %lluM 0x%08llX - 0x%08llX\n",
+			mc->gtt_size >> 20, mc->gtt_start, mc->gtt_end);
 }
 
-
 /*
  * GPU helpers function.
  */
@@ -181,7 +206,16 @@ bool radeon_card_posted(struct radeon_device *rdev)
 	uint32_t reg;
 
 	/* first check CRTCs */
-	if (ASIC_IS_AVIVO(rdev)) {
+	if (ASIC_IS_DCE4(rdev)) {
+		reg = RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC0_REGISTER_OFFSET) |
+			RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC1_REGISTER_OFFSET) |
+			RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC2_REGISTER_OFFSET) |
+			RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC3_REGISTER_OFFSET) |
+			RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC4_REGISTER_OFFSET) |
+			RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC5_REGISTER_OFFSET);
+		if (reg & EVERGREEN_CRTC_MASTER_EN)
+			return true;
+	} else if (ASIC_IS_AVIVO(rdev)) {
 		reg = RREG32(AVIVO_D1CRTC_CONTROL) |
 		      RREG32(AVIVO_D2CRTC_CONTROL);
 		if (reg & AVIVO_CRTC_EN) {
@@ -208,8 +242,28 @@ bool radeon_card_posted(struct radeon_device *rdev)
 
 }
 
+bool radeon_boot_test_post_card(struct radeon_device *rdev)
+{
+	if (radeon_card_posted(rdev))
+		return true;
+
+	if (rdev->bios) {
+		DRM_INFO("GPU not posted. posting now...\n");
+		if (rdev->is_atom_bios)
+			atom_asic_init(rdev->mode_info.atom_context);
+		else
+			radeon_combios_asic_init(rdev->ddev);
+		return true;
+	} else {
+		dev_err(rdev->dev, "Card not posted and no BIOS - ignoring\n");
+		return false;
+	}
+}
+
 int radeon_dummy_page_init(struct radeon_device *rdev)
 {
+	if (rdev->dummy_page.page)
+		return 0;
 	rdev->dummy_page.page = alloc_page(GFP_DMA32 | GFP_KERNEL | __GFP_ZERO);
 	if (rdev->dummy_page.page == NULL)
 		return -ENOMEM;
@@ -291,7 +345,7 @@ void radeon_register_accessor_init(struct radeon_device *rdev)
 		rdev->mc_rreg = &rs600_mc_rreg;
 		rdev->mc_wreg = &rs600_mc_wreg;
 	}
-	if (rdev->family >= CHIP_R600) {
+	if ((rdev->family >= CHIP_R600) && (rdev->family <= CHIP_RV740)) {
 		rdev->pciep_rreg = &r600_pciep_rreg;
 		rdev->pciep_wreg = &r600_pciep_wreg;
 	}
@@ -310,25 +364,22 @@ int radeon_asic_init(struct radeon_device *rdev)
 	case CHIP_RS100:
 	case CHIP_RV200:
 	case CHIP_RS200:
+		rdev->asic = &r100_asic;
+		break;
 	case CHIP_R200:
 	case CHIP_RV250:
 	case CHIP_RS300:
 	case CHIP_RV280:
-		rdev->asic = &r100_asic;
+		rdev->asic = &r200_asic;
 		break;
 	case CHIP_R300:
 	case CHIP_R350:
 	case CHIP_RV350:
 	case CHIP_RV380:
-		rdev->asic = &r300_asic;
-		if (rdev->flags & RADEON_IS_PCIE) {
-			rdev->asic->gart_init = &rv370_pcie_gart_init;
-			rdev->asic->gart_fini = &rv370_pcie_gart_fini;
-			rdev->asic->gart_enable = &rv370_pcie_gart_enable;
-			rdev->asic->gart_disable = &rv370_pcie_gart_disable;
-			rdev->asic->gart_tlb_flush = &rv370_pcie_gart_tlb_flush;
-			rdev->asic->gart_set_page = &rv370_pcie_gart_set_page;
-		}
+		if (rdev->flags & RADEON_IS_PCIE)
+			rdev->asic = &r300_asic_pcie;
+		else
+			rdev->asic = &r300_asic;
 		break;
 	case CHIP_R420:
 	case CHIP_R423:
@@ -372,10 +423,23 @@ int radeon_asic_init(struct radeon_device *rdev)
 	case CHIP_RV740:
 		rdev->asic = &rv770_asic;
 		break;
+	case CHIP_CEDAR:
+	case CHIP_REDWOOD:
+	case CHIP_JUNIPER:
+	case CHIP_CYPRESS:
+	case CHIP_HEMLOCK:
+		rdev->asic = &evergreen_asic;
+		break;
 	default:
 		/* FIXME: not supported yet */
 		return -EINVAL;
 	}
+
+	if (rdev->flags & RADEON_IS_IGP) {
+		rdev->asic->get_memory_clock = NULL;
+		rdev->asic->set_memory_clock = NULL;
+	}
+
 	return 0;
 }
 
@@ -448,27 +512,37 @@ static uint32_t cail_reg_read(struct card_info *info, uint32_t reg)
 	return r;
 }
 
-static struct card_info atom_card_info = {
-	.dev = NULL,
-	.reg_read = cail_reg_read,
-	.reg_write = cail_reg_write,
-	.mc_read = cail_mc_read,
-	.mc_write = cail_mc_write,
-	.pll_read = cail_pll_read,
-	.pll_write = cail_pll_write,
-};
-
 int radeon_atombios_init(struct radeon_device *rdev)
 {
-	atom_card_info.dev = rdev->ddev;
-	rdev->mode_info.atom_context = atom_parse(&atom_card_info, rdev->bios);
+	struct card_info *atom_card_info =
+	    kzalloc(sizeof(struct card_info), GFP_KERNEL);
+
+	if (!atom_card_info)
+		return -ENOMEM;
+
+	rdev->mode_info.atom_card_info = atom_card_info;
+	atom_card_info->dev = rdev->ddev;
+	atom_card_info->reg_read = cail_reg_read;
+	atom_card_info->reg_write = cail_reg_write;
+	atom_card_info->mc_read = cail_mc_read;
+	atom_card_info->mc_write = cail_mc_write;
+	atom_card_info->pll_read = cail_pll_read;
+	atom_card_info->pll_write = cail_pll_write;
+
+	rdev->mode_info.atom_context = atom_parse(atom_card_info, rdev->bios);
+	mutex_init(&rdev->mode_info.atom_context->mutex);
 	radeon_atom_initialize_bios_scratch_regs(rdev->ddev);
+	atom_allocate_fb_scratch(rdev->mode_info.atom_context);
 	return 0;
 }
 
 void radeon_atombios_fini(struct radeon_device *rdev)
 {
-	kfree(rdev->mode_info.atom_context);
+	if (rdev->mode_info.atom_context) {
+		kfree(rdev->mode_info.atom_context->scratch);
+		kfree(rdev->mode_info.atom_context);
+	}
+	kfree(rdev->mode_info.atom_card_info);
 }
 
 int radeon_combios_init(struct radeon_device *rdev)
@@ -485,7 +559,6 @@ void radeon_combios_fini(struct radeon_device *rdev)
 static unsigned int radeon_vga_set_decode(void *cookie, bool state)
 {
 	struct radeon_device *rdev = cookie;
-
 	radeon_vga_set_state(rdev, state);
 	if (state)
 		return VGA_RSRC_LEGACY_IO | VGA_RSRC_LEGACY_MEM |
@@ -493,9 +566,126 @@ static unsigned int radeon_vga_set_decode(void *cookie, bool state)
 	else
 		return VGA_RSRC_NORMAL_IO | VGA_RSRC_NORMAL_MEM;
 }
-/*
- * Radeon device.
- */
+
+void radeon_agp_disable(struct radeon_device *rdev)
+{
+	rdev->flags &= ~RADEON_IS_AGP;
+	if (rdev->family >= CHIP_R600) {
+		DRM_INFO("Forcing AGP to PCIE mode\n");
+		rdev->flags |= RADEON_IS_PCIE;
+	} else if (rdev->family >= CHIP_RV515 ||
+			rdev->family == CHIP_RV380 ||
+			rdev->family == CHIP_RV410 ||
+			rdev->family == CHIP_R423) {
+		DRM_INFO("Forcing AGP to PCIE mode\n");
+		rdev->flags |= RADEON_IS_PCIE;
+		rdev->asic->gart_tlb_flush = &rv370_pcie_gart_tlb_flush;
+		rdev->asic->gart_set_page = &rv370_pcie_gart_set_page;
+	} else {
+		DRM_INFO("Forcing AGP to PCI mode\n");
+		rdev->flags |= RADEON_IS_PCI;
+		rdev->asic->gart_tlb_flush = &r100_pci_gart_tlb_flush;
+		rdev->asic->gart_set_page = &r100_pci_gart_set_page;
+	}
+	rdev->mc.gtt_size = radeon_gart_size * 1024 * 1024;
+}
+
+void radeon_check_arguments(struct radeon_device *rdev)
+{
+	/* vramlimit must be a power of two */
+	switch (radeon_vram_limit) {
+	case 0:
+	case 4:
+	case 8:
+	case 16:
+	case 32:
+	case 64:
+	case 128:
+	case 256:
+	case 512:
+	case 1024:
+	case 2048:
+	case 4096:
+		break;
+	default:
+		dev_warn(rdev->dev, "vram limit (%d) must be a power of 2\n",
+				radeon_vram_limit);
+		radeon_vram_limit = 0;
+		break;
+	}
+	radeon_vram_limit = radeon_vram_limit << 20;
+	/* gtt size must be power of two and greater or equal to 32M */
+	switch (radeon_gart_size) {
+	case 4:
+	case 8:
+	case 16:
+		dev_warn(rdev->dev, "gart size (%d) too small forcing to 512M\n",
+				radeon_gart_size);
+		radeon_gart_size = 512;
+		break;
+	case 32:
+	case 64:
+	case 128:
+	case 256:
+	case 512:
+	case 1024:
+	case 2048:
+	case 4096:
+		break;
+	default:
+		dev_warn(rdev->dev, "gart size (%d) must be a power of 2\n",
+				radeon_gart_size);
+		radeon_gart_size = 512;
+		break;
+	}
+	rdev->mc.gtt_size = radeon_gart_size * 1024 * 1024;
+	/* AGP mode can only be -1, 1, 2, 4, 8 */
+	switch (radeon_agpmode) {
+	case -1:
+	case 0:
+	case 1:
+	case 2:
+	case 4:
+	case 8:
+		break;
+	default:
+		dev_warn(rdev->dev, "invalid AGP mode %d (valid mode: "
+				"-1, 0, 1, 2, 4, 8)\n", radeon_agpmode);
+		radeon_agpmode = 0;
+		break;
+	}
+}
+
+static void radeon_switcheroo_set_state(struct pci_dev *pdev, enum vga_switcheroo_state state)
+{
+	struct drm_device *dev = pci_get_drvdata(pdev);
+	struct radeon_device *rdev = dev->dev_private;
+	pm_message_t pmm = { .event = PM_EVENT_SUSPEND };
+	if (state == VGA_SWITCHEROO_ON) {
+		printk(KERN_INFO "radeon: switched on\n");
+		/* don't suspend or resume card normally */
+		rdev->powered_down = false;
+		radeon_resume_kms(dev);
+	} else {
+		printk(KERN_INFO "radeon: switched off\n");
+		radeon_suspend_kms(dev, pmm);
+		/* don't suspend or resume card normally */
+		rdev->powered_down = true;
+	}
+}
+
+static bool radeon_switcheroo_can_switch(struct pci_dev *pdev)
+{
+	struct drm_device *dev = pci_get_drvdata(pdev);
+	bool can_switch;
+
+	spin_lock(&dev->count_lock);
+	can_switch = (dev->open_count == 0);
+	spin_unlock(&dev->count_lock);
+	return can_switch;
+}
+
+
 int radeon_device_init(struct radeon_device *rdev,
 		       struct drm_device *ddev,
 		       struct pci_dev *pdev,
@@ -521,39 +711,28 @@ int radeon_device_init(struct radeon_device *rdev,
 	mutex_init(&rdev->cs_mutex);
 	mutex_init(&rdev->ib_pool.mutex);
 	mutex_init(&rdev->cp.mutex);
+	mutex_init(&rdev->dc_hw_i2c_mutex);
+	if (rdev->family >= CHIP_R600)
+		spin_lock_init(&rdev->ih.lock);
+	mutex_init(&rdev->gem.mutex);
+	mutex_init(&rdev->pm.mutex);
 	rwlock_init(&rdev->fence_drv.lock);
 	INIT_LIST_HEAD(&rdev->gem.objects);
+	init_waitqueue_head(&rdev->irq.vblank_queue);
+
+	/* setup workqueue */
+	rdev->wq = create_workqueue("radeon");
+	if (rdev->wq == NULL)
+		return -ENOMEM;
 
 	/* Set asic functions */
 	r = radeon_asic_init(rdev);
-	if (r) {
+	if (r)
 		return r;
-	}
+	radeon_check_arguments(rdev);
 
-	if (radeon_agpmode == -1) {
-		rdev->flags &= ~RADEON_IS_AGP;
-		if (rdev->family >= CHIP_RV515 ||
-		    rdev->family == CHIP_RV380 ||
-		    rdev->family == CHIP_RV410 ||
-		    rdev->family == CHIP_R423) {
-			DRM_INFO("Forcing AGP to PCIE mode\n");
-			rdev->flags |= RADEON_IS_PCIE;
-			rdev->asic->gart_init = &rv370_pcie_gart_init;
-			rdev->asic->gart_fini = &rv370_pcie_gart_fini;
-			rdev->asic->gart_enable = &rv370_pcie_gart_enable;
-			rdev->asic->gart_disable = &rv370_pcie_gart_disable;
-			rdev->asic->gart_tlb_flush = &rv370_pcie_gart_tlb_flush;
-			rdev->asic->gart_set_page = &rv370_pcie_gart_set_page;
-		} else {
-			DRM_INFO("Forcing AGP to PCI mode\n");
-			rdev->flags |= RADEON_IS_PCI;
-			rdev->asic->gart_init = &r100_pci_gart_init;
-			rdev->asic->gart_fini = &r100_pci_gart_fini;
-			rdev->asic->gart_enable = &r100_pci_gart_enable;
-			rdev->asic->gart_disable = &r100_pci_gart_disable;
-			rdev->asic->gart_tlb_flush = &r100_pci_gart_tlb_flush;
-			rdev->asic->gart_set_page = &r100_pci_gart_set_page;
-		}
+	if (rdev->flags & RADEON_IS_AGP && radeon_agpmode == -1) {
+		radeon_agp_disable(rdev);
 	}
 
 	/* set DMA mask + need_dma32 flags.
@@ -585,111 +764,29 @@ int radeon_device_init(struct radeon_device *rdev,
 	DRM_INFO("register mmio base: 0x%08X\n", (uint32_t)rdev->rmmio_base);
 	DRM_INFO("register mmio size: %u\n", (unsigned)rdev->rmmio_size);
 
-	rdev->new_init_path = false;
-	r = radeon_init(rdev);
-	if (r) {
-		return r;
-	}
-
 	/* if we have > 1 VGA cards, then disable the radeon VGA resources */
-	r = vga_client_register(rdev->pdev, rdev, NULL, radeon_vga_set_decode);
-	if (r) {
-		return -EINVAL;
-	}
-
-	if (!rdev->new_init_path) {
-		/* Setup errata flags */
-		radeon_errata(rdev);
-		/* Initialize scratch registers */
-		radeon_scratch_init(rdev);
-		/* Initialize surface registers */
-		radeon_surface_init(rdev);
+	/* this will fail for cards that aren't VGA class devices, just
+	 * ignore it */
+	vga_client_register(rdev->pdev, rdev, NULL, radeon_vga_set_decode);
+	vga_switcheroo_register_client(rdev->pdev,
+				       radeon_switcheroo_set_state,
+				       radeon_switcheroo_can_switch);
 
-		/* BIOS*/
-		if (!radeon_get_bios(rdev)) {
-			if (ASIC_IS_AVIVO(rdev))
-				return -EINVAL;
-		}
-		if (rdev->is_atom_bios) {
-			r = radeon_atombios_init(rdev);
-			if (r) {
-				return r;
-			}
-		} else {
-			r = radeon_combios_init(rdev);
-			if (r) {
-				return r;
-			}
-		}
-		/* Reset gpu before posting otherwise ATOM will enter infinite loop */
-		if (radeon_gpu_reset(rdev)) {
-			/* FIXME: what do we want to do here ? */
-		}
-		/* check if cards are posted or not */
-		if (!radeon_card_posted(rdev) && rdev->bios) {
-			DRM_INFO("GPU not posted. posting now...\n");
-			if (rdev->is_atom_bios) {
-				atom_asic_init(rdev->mode_info.atom_context);
-			} else {
-				radeon_combios_asic_init(rdev->ddev);
-			}
-		}
-		/* Get clock & vram information */
-		radeon_get_clock_info(rdev->ddev);
-		radeon_vram_info(rdev);
-		/* Initialize clocks */
-		r = radeon_clocks_init(rdev);
-		if (r) {
-			return r;
-		}
+	r = radeon_init(rdev);
+	if (r)
+		return r;
 
-		/* Initialize memory controller (also test AGP) */
-		r = radeon_mc_init(rdev);
-		if (r) {
-			return r;
-		}
-		/* Fence driver */
-		r = radeon_fence_driver_init(rdev);
-		if (r) {
-			return r;
-		}
-		r = radeon_irq_kms_init(rdev);
-		if (r) {
-			return r;
-		}
-		/* Memory manager */
-		r = radeon_object_init(rdev);
-		if (r) {
-			return r;
-		}
-		r = radeon_gpu_gart_init(rdev);
+	if (rdev->flags & RADEON_IS_AGP && !rdev->accel_working) {
+		/* Acceleration not working on AGP card try again
+		 * with fallback to PCI or PCIE GART
+		 */
+		radeon_gpu_reset(rdev);
+		radeon_fini(rdev);
+		radeon_agp_disable(rdev);
+		r = radeon_init(rdev);
 		if (r)
 			return r;
-		/* Initialize GART (initialize after TTM so we can allocate
-		 * memory through TTM but finalize after TTM) */
-		r = radeon_gart_enable(rdev);
-		if (r)
-			return 0;
-			r = radeon_gem_init(rdev);
-		if (r)
-			return 0;
-
-		/* 1M ring buffer */
-		r = radeon_cp_init(rdev, 1024 * 1024);
-		if (r)
-			return 0;
-		r = radeon_wb_init(rdev);
-		if (r)
-			DRM_ERROR("radeon: failled initializing WB (%d).\n", r);
-		r = radeon_ib_pool_init(rdev);
-		if (r)
-			return 0;
-		r = radeon_ib_test(rdev);
-		if (r)
-			return 0;
-		rdev->accel_working = true;
 	}
-	DRM_INFO("radeon: kernel modesetting successfully initialized.\n");
 	if (radeon_testing) {
 		radeon_test_moves(rdev);
 	}
@@ -703,32 +800,10 @@ void radeon_device_fini(struct radeon_device *rdev)
 {
 	DRM_INFO("radeon: finishing device.\n");
 	rdev->shutdown = true;
-	/* Order matter so becarefull if you rearrange anythings */
-	if (!rdev->new_init_path) {
-		radeon_ib_pool_fini(rdev);
-		radeon_cp_fini(rdev);
-		radeon_wb_fini(rdev);
-		radeon_gpu_gart_fini(rdev);
-		radeon_gem_fini(rdev);
-		radeon_mc_fini(rdev);
-#if __OS_HAS_AGP
-		radeon_agp_fini(rdev);
-#endif
-		radeon_irq_kms_fini(rdev);
-		vga_client_register(rdev->pdev, NULL, NULL, NULL);
-		radeon_fence_driver_fini(rdev);
-		radeon_clocks_fini(rdev);
-		radeon_object_fini(rdev);
-		if (rdev->is_atom_bios) {
-			radeon_atombios_fini(rdev);
-		} else {
-			radeon_combios_fini(rdev);
-		}
-		kfree(rdev->bios);
-		rdev->bios = NULL;
-	} else {
-		radeon_fini(rdev);
-	}
+	radeon_fini(rdev);
+	destroy_workqueue(rdev->wq);
+	vga_switcheroo_unregister_client(rdev->pdev);
+	vga_client_register(rdev->pdev, NULL, NULL, NULL);
 	iounmap(rdev->rmmio);
 	rdev->rmmio = NULL;
 }
@@ -739,45 +814,48 @@ void radeon_device_fini(struct radeon_device *rdev)
  */
 int radeon_suspend_kms(struct drm_device *dev, pm_message_t state)
 {
-	struct radeon_device *rdev = dev->dev_private;
+	struct radeon_device *rdev;
 	struct drm_crtc *crtc;
+	int r;
 
-	if (dev == NULL || rdev == NULL) {
+	if (dev == NULL || dev->dev_private == NULL) {
 		return -ENODEV;
 	}
 	if (state.event == PM_EVENT_PRETHAW) {
 		return 0;
 	}
+	rdev = dev->dev_private;
+
+	if (rdev->powered_down)
+		return 0;
 	/* unpin the front buffers */
 	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
 		struct radeon_framebuffer *rfb = to_radeon_framebuffer(crtc->fb);
-		struct radeon_object *robj;
+		struct radeon_bo *robj;
 
 		if (rfb == NULL || rfb->obj == NULL) {
 			continue;
 		}
 		robj = rfb->obj->driver_private;
-		if (robj != rdev->fbdev_robj) {
-			radeon_object_unpin(robj);
+		if (robj != rdev->fbdev_rbo) {
+			r = radeon_bo_reserve(robj, false);
+			if (unlikely(r == 0)) {
+				radeon_bo_unpin(robj);
+				radeon_bo_unreserve(robj);
+			}
 		}
 	}
 	/* evict vram memory */
-	radeon_object_evict_vram(rdev);
+	radeon_bo_evict_vram(rdev);
 	/* wait for gpu to finish processing current batch */
 	radeon_fence_wait_last(rdev);
 
 	radeon_save_bios_scratch_regs(rdev);
 
-	if (!rdev->new_init_path) {
-		radeon_cp_disable(rdev);
-		radeon_gart_disable(rdev);
-		rdev->irq.sw_int = false;
-		radeon_irq_set(rdev);
-	} else {
-		radeon_suspend(rdev);
-	}
+	radeon_suspend(rdev);
+	radeon_hpd_fini(rdev);
 	/* evict remaining vram memory */
-	radeon_object_evict_vram(rdev);
+	radeon_bo_evict_vram(rdev);
 
 	pci_save_state(dev->pdev);
 	if (state.event == PM_EVENT_SUSPEND) {
@@ -794,7 +872,9 @@ int radeon_suspend_kms(struct drm_device *dev, pm_message_t state)
 int radeon_resume_kms(struct drm_device *dev)
 {
 	struct radeon_device *rdev = dev->dev_private;
-	int r;
+
+	if (rdev->powered_down)
+		return 0;
 
 	acquire_console_sem();
 	pci_set_power_state(dev->pdev, PCI_D0);
@@ -804,47 +884,15 @@ int radeon_resume_kms(struct drm_device *dev)
 		return -1;
 	}
 	pci_set_master(dev->pdev);
-	/* Reset gpu before posting otherwise ATOM will enter infinite loop */
-	if (!rdev->new_init_path) {
-		if (radeon_gpu_reset(rdev)) {
-			/* FIXME: what do we want to do here ? */
-		}
-		/* post card */
-		if (rdev->is_atom_bios) {
-			atom_asic_init(rdev->mode_info.atom_context);
-		} else {
-			radeon_combios_asic_init(rdev->ddev);
-		}
-		/* Initialize clocks */
-		r = radeon_clocks_init(rdev);
-		if (r) {
-			release_console_sem();
-			return r;
-		}
-		/* Enable IRQ */
-		rdev->irq.sw_int = true;
-		radeon_irq_set(rdev);
-		/* Initialize GPU Memory Controller */
-		r = radeon_mc_init(rdev);
-		if (r) {
-			goto out;
-		}
-		r = radeon_gart_enable(rdev);
-		if (r) {
-			goto out;
-		}
-		r = radeon_cp_init(rdev, rdev->cp.ring_size);
-		if (r) {
-			goto out;
-		}
-	} else {
-		radeon_resume(rdev);
-	}
-out:
+	/* resume AGP if in use */
+	radeon_agp_resume(rdev);
+	radeon_resume(rdev);
 	radeon_restore_bios_scratch_regs(rdev);
 	fb_set_suspend(rdev->fbdev_info, 0);
 	release_console_sem();
 
+	/* reset hpd state */
+	radeon_hpd_init(rdev);
 	/* blat the mode back in */
 	drm_helper_resume_force_mode(dev);
 	return 0;
diff --git a/drivers/gpu/drm/radeon/radeon_display.c b/drivers/gpu/drm/radeon/radeon_display.c
index 5d8141b13765..ba8d806dcf39 100644
--- a/drivers/gpu/drm/radeon/radeon_display.c
+++ b/drivers/gpu/drm/radeon/radeon_display.c
@@ -68,6 +68,36 @@ static void avivo_crtc_load_lut(struct drm_crtc *crtc)
 	WREG32(AVIVO_D1GRPH_LUT_SEL + radeon_crtc->crtc_offset, radeon_crtc->crtc_id);
 }
 
+static void evergreen_crtc_load_lut(struct drm_crtc *crtc)
+{
+	struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc);
+	struct drm_device *dev = crtc->dev;
+	struct radeon_device *rdev = dev->dev_private;
+	int i;
+
+	DRM_DEBUG("%d\n", radeon_crtc->crtc_id);
+	WREG32(EVERGREEN_DC_LUT_CONTROL + radeon_crtc->crtc_offset, 0);
+
+	WREG32(EVERGREEN_DC_LUT_BLACK_OFFSET_BLUE + radeon_crtc->crtc_offset, 0);
+	WREG32(EVERGREEN_DC_LUT_BLACK_OFFSET_GREEN + radeon_crtc->crtc_offset, 0);
+	WREG32(EVERGREEN_DC_LUT_BLACK_OFFSET_RED + radeon_crtc->crtc_offset, 0);
+
+	WREG32(EVERGREEN_DC_LUT_WHITE_OFFSET_BLUE + radeon_crtc->crtc_offset, 0xffff);
+	WREG32(EVERGREEN_DC_LUT_WHITE_OFFSET_GREEN + radeon_crtc->crtc_offset, 0xffff);
+	WREG32(EVERGREEN_DC_LUT_WHITE_OFFSET_RED + radeon_crtc->crtc_offset, 0xffff);
+
+	WREG32(EVERGREEN_DC_LUT_RW_MODE, radeon_crtc->crtc_id);
+	WREG32(EVERGREEN_DC_LUT_WRITE_EN_MASK, 0x00000007);
+
+	WREG32(EVERGREEN_DC_LUT_RW_INDEX, 0);
+	for (i = 0; i < 256; i++) {
+		WREG32(EVERGREEN_DC_LUT_30_COLOR,
+		       (radeon_crtc->lut_r[i] << 20) |
+		       (radeon_crtc->lut_g[i] << 10) |
+		       (radeon_crtc->lut_b[i] << 0));
+	}
+}
+
 static void legacy_crtc_load_lut(struct drm_crtc *crtc)
 {
 	struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc);
@@ -100,57 +130,52 @@ void radeon_crtc_load_lut(struct drm_crtc *crtc)
 	if (!crtc->enabled)
 		return;
 
-	if (ASIC_IS_AVIVO(rdev))
+	if (ASIC_IS_DCE4(rdev))
+		evergreen_crtc_load_lut(crtc);
+	else if (ASIC_IS_AVIVO(rdev))
 		avivo_crtc_load_lut(crtc);
 	else
 		legacy_crtc_load_lut(crtc);
 }
 
-/** Sets the color ramps on behalf of RandR */
+/** Sets the color ramps on behalf of fbcon */
 void radeon_crtc_fb_gamma_set(struct drm_crtc *crtc, u16 red, u16 green,
 			      u16 blue, int regno)
 {
 	struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc);
 
-	if (regno == 0)
-		DRM_DEBUG("gamma set %d\n", radeon_crtc->crtc_id);
 	radeon_crtc->lut_r[regno] = red >> 6;
 	radeon_crtc->lut_g[regno] = green >> 6;
 	radeon_crtc->lut_b[regno] = blue >> 6;
 }
 
+/** Gets the color ramps on behalf of fbcon */
+void radeon_crtc_fb_gamma_get(struct drm_crtc *crtc, u16 *red, u16 *green,
+			      u16 *blue, int regno)
+{
+	struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc);
+
+	*red = radeon_crtc->lut_r[regno] << 6;
+	*green = radeon_crtc->lut_g[regno] << 6;
+	*blue = radeon_crtc->lut_b[regno] << 6;
+}
+
 static void radeon_crtc_gamma_set(struct drm_crtc *crtc, u16 *red, u16 *green,
 				  u16 *blue, uint32_t size)
 {
 	struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc);
-	int i, j;
+	int i;
 
 	if (size != 256) {
 		return;
 	}
-	if (crtc->fb == NULL) {
-		return;
-	}
 
-	if (crtc->fb->depth == 16) {
-		for (i = 0; i < 64; i++) {
-			if (i <= 31) {
-				for (j = 0; j < 8; j++) {
-					radeon_crtc->lut_r[i * 8 + j] = red[i] >> 6;
-					radeon_crtc->lut_b[i * 8 + j] = blue[i] >> 6;
-				}
-			}
-			for (j = 0; j < 4; j++)
-				radeon_crtc->lut_g[i * 4 + j] = green[i] >> 6;
-		}
-	} else {
-		for (i = 0; i < 256; i++) {
-			radeon_crtc->lut_r[i] = red[i] >> 6;
-			radeon_crtc->lut_g[i] = green[i] >> 6;
-			radeon_crtc->lut_b[i] = blue[i] >> 6;
-		}
+	/* userspace palettes are always correct as is */
+	for (i = 0; i < 256; i++) {
+		radeon_crtc->lut_r[i] = red[i] >> 6;
+		radeon_crtc->lut_g[i] = green[i] >> 6;
+		radeon_crtc->lut_b[i] = blue[i] >> 6;
 	}
-
 	radeon_crtc_load_lut(crtc);
 }
 
@@ -241,7 +266,7 @@ static const char *encoder_names[34] = {
 	"INTERNAL_UNIPHY2",
 };
 
-static const char *connector_names[13] = {
+static const char *connector_names[15] = {
 	"Unknown",
 	"VGA",
 	"DVI-I",
@@ -255,6 +280,18 @@ static const char *connector_names[13] = {
 	"DisplayPort",
 	"HDMI-A",
 	"HDMI-B",
+	"TV",
+	"eDP",
+};
+
+static const char *hpd_names[7] = {
+	"NONE",
+	"HPD1",
+	"HPD2",
+	"HPD3",
+	"HPD4",
+	"HPD5",
+	"HPD6",
 };
 
 static void radeon_print_display_setup(struct drm_device *dev)
@@ -271,16 +308,27 @@ static void radeon_print_display_setup(struct drm_device *dev)
 		radeon_connector = to_radeon_connector(connector);
 		DRM_INFO("Connector %d:\n", i);
 		DRM_INFO("  %s\n", connector_names[connector->connector_type]);
-		if (radeon_connector->ddc_bus)
+		if (radeon_connector->hpd.hpd != RADEON_HPD_NONE)
+			DRM_INFO("  %s\n", hpd_names[radeon_connector->hpd.hpd]);
+		if (radeon_connector->ddc_bus) {
 			DRM_INFO("  DDC: 0x%x 0x%x 0x%x 0x%x 0x%x 0x%x 0x%x 0x%x\n",
 				 radeon_connector->ddc_bus->rec.mask_clk_reg,
 				 radeon_connector->ddc_bus->rec.mask_data_reg,
 				 radeon_connector->ddc_bus->rec.a_clk_reg,
 				 radeon_connector->ddc_bus->rec.a_data_reg,
-				 radeon_connector->ddc_bus->rec.put_clk_reg,
-				 radeon_connector->ddc_bus->rec.put_data_reg,
-				 radeon_connector->ddc_bus->rec.get_clk_reg,
-				 radeon_connector->ddc_bus->rec.get_data_reg);
+				 radeon_connector->ddc_bus->rec.en_clk_reg,
+				 radeon_connector->ddc_bus->rec.en_data_reg,
+				 radeon_connector->ddc_bus->rec.y_clk_reg,
+				 radeon_connector->ddc_bus->rec.y_data_reg);
+		} else {
+			if (connector->connector_type == DRM_MODE_CONNECTOR_VGA ||
+			    connector->connector_type == DRM_MODE_CONNECTOR_DVII ||
+			    connector->connector_type == DRM_MODE_CONNECTOR_DVID ||
+			    connector->connector_type == DRM_MODE_CONNECTOR_DVIA ||
+			    connector->connector_type == DRM_MODE_CONNECTOR_HDMIA ||
+			    connector->connector_type == DRM_MODE_CONNECTOR_HDMIB)
+				DRM_INFO("  DDC: no ddc bus - possible BIOS bug - please report to xorg-driver-ati@lists.x.org\n");
+		}
 		DRM_INFO("  Encoders:\n");
 		list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) {
 			radeon_encoder = to_radeon_encoder(encoder);
@@ -324,13 +372,17 @@ static bool radeon_setup_enc_conn(struct drm_device *dev)
 				ret = radeon_get_atom_connector_info_from_object_table(dev);
 			else
 				ret = radeon_get_atom_connector_info_from_supported_devices_table(dev);
-		} else
+		} else {
 			ret = radeon_get_legacy_connector_info_from_bios(dev);
+			if (ret == false)
+				ret = radeon_get_legacy_connector_info_from_table(dev);
+		}
 	} else {
 		if (!ASIC_IS_AVIVO(rdev))
 			ret = radeon_get_legacy_connector_info_from_table(dev);
 	}
 	if (ret) {
+		radeon_setup_encoder_clones(dev);
 		radeon_print_display_setup(dev);
 		list_for_each_entry(drm_connector, &dev->mode_config.connector_list, head)
 			radeon_ddc_dump(drm_connector);
@@ -341,27 +393,28 @@ static bool radeon_setup_enc_conn(struct drm_device *dev)
 
 int radeon_ddc_get_modes(struct radeon_connector *radeon_connector)
 {
-	struct edid *edid;
+	struct drm_device *dev = radeon_connector->base.dev;
+	struct radeon_device *rdev = dev->dev_private;
 	int ret = 0;
 
+	if ((radeon_connector->base.connector_type == DRM_MODE_CONNECTOR_DisplayPort) ||
+	    (radeon_connector->base.connector_type == DRM_MODE_CONNECTOR_eDP)) {
+		struct radeon_connector_atom_dig *dig = radeon_connector->con_priv;
+		if ((dig->dp_sink_type == CONNECTOR_OBJECT_ID_DISPLAYPORT ||
+		     dig->dp_sink_type == CONNECTOR_OBJECT_ID_eDP) && dig->dp_i2c_bus)
+			radeon_connector->edid = drm_get_edid(&radeon_connector->base, &dig->dp_i2c_bus->adapter);
+	}
 	if (!radeon_connector->ddc_bus)
 		return -1;
 	if (!radeon_connector->edid) {
-		radeon_i2c_do_lock(radeon_connector, 1);
-		edid = drm_get_edid(&radeon_connector->base, &radeon_connector->ddc_bus->adapter);
-		radeon_i2c_do_lock(radeon_connector, 0);
-	} else
-		edid = radeon_connector->edid;
-
-	if (edid) {
-		/* update digital bits here */
-		if (edid->input & DRM_EDID_INPUT_DIGITAL)
-			radeon_connector->use_digital = 1;
-		else
-			radeon_connector->use_digital = 0;
-		drm_mode_connector_update_edid_property(&radeon_connector->base, edid);
-		ret = drm_add_edid_modes(&radeon_connector->base, edid);
-		kfree(edid);
+		radeon_connector->edid = drm_get_edid(&radeon_connector->base, &radeon_connector->ddc_bus->adapter);
+	}
+	/* some servers provide a hardcoded edid in rom for KVMs */
+	if (!radeon_connector->edid)
+		radeon_connector->edid = radeon_combios_get_hardcoded_edid(rdev);
+	if (radeon_connector->edid) {
+		drm_mode_connector_update_edid_property(&radeon_connector->base, radeon_connector->edid);
+		ret = drm_add_edid_modes(&radeon_connector->base, radeon_connector->edid);
 		return ret;
 	}
 	drm_mode_connector_update_edid_property(&radeon_connector->base, NULL);
@@ -376,9 +429,7 @@ static int radeon_ddc_dump(struct drm_connector *connector)
 
 	if (!radeon_connector->ddc_bus)
 		return -1;
-	radeon_i2c_do_lock(radeon_connector, 1);
 	edid = drm_get_edid(connector, &radeon_connector->ddc_bus->adapter);
-	radeon_i2c_do_lock(radeon_connector, 0);
 	if (edid) {
 		kfree(edid);
 	}
@@ -395,17 +446,18 @@ static inline uint32_t radeon_div(uint64_t n, uint32_t d)
 	return n;
 }
 
-void radeon_compute_pll(struct radeon_pll *pll,
-			uint64_t freq,
-			uint32_t *dot_clock_p,
-			uint32_t *fb_div_p,
-			uint32_t *frac_fb_div_p,
-			uint32_t *ref_div_p,
-			uint32_t *post_div_p,
-			int flags)
+static void radeon_compute_pll_legacy(struct radeon_pll *pll,
+				      uint64_t freq,
+				      uint32_t *dot_clock_p,
+				      uint32_t *fb_div_p,
+				      uint32_t *frac_fb_div_p,
+				      uint32_t *ref_div_p,
+				      uint32_t *post_div_p)
 {
 	uint32_t min_ref_div = pll->min_ref_div;
 	uint32_t max_ref_div = pll->max_ref_div;
+	uint32_t min_post_div = pll->min_post_div;
+	uint32_t max_post_div = pll->max_post_div;
 	uint32_t min_fractional_feed_div = 0;
 	uint32_t max_fractional_feed_div = 0;
 	uint32_t best_vco = pll->best_vco;
@@ -421,7 +473,7 @@ void radeon_compute_pll(struct radeon_pll *pll,
 	DRM_DEBUG("PLL freq %llu %u %u\n", freq, pll->min_ref_div, pll->max_ref_div);
 	freq = freq * 1000;
 
-	if (flags & RADEON_PLL_USE_REF_DIV)
+	if (pll->flags & RADEON_PLL_USE_REF_DIV)
 		min_ref_div = max_ref_div = pll->reference_div;
 	else {
 		while (min_ref_div < max_ref_div-1) {
@@ -436,19 +488,22 @@ void radeon_compute_pll(struct radeon_pll *pll,
 		}
 	}
 
-	if (flags & RADEON_PLL_USE_FRAC_FB_DIV) {
+	if (pll->flags & RADEON_PLL_USE_POST_DIV)
+		min_post_div = max_post_div = pll->post_div;
+
+	if (pll->flags & RADEON_PLL_USE_FRAC_FB_DIV) {
 		min_fractional_feed_div = pll->min_frac_feedback_div;
 		max_fractional_feed_div = pll->max_frac_feedback_div;
 	}
 
-	for (post_div = pll->min_post_div; post_div <= pll->max_post_div; ++post_div) {
+	for (post_div = min_post_div; post_div <= max_post_div; ++post_div) {
 		uint32_t ref_div;
 
-		if ((flags & RADEON_PLL_NO_ODD_POST_DIV) && (post_div & 1))
+		if ((pll->flags & RADEON_PLL_NO_ODD_POST_DIV) && (post_div & 1))
 			continue;
 
 		/* legacy radeons only have a few post_divs */
-		if (flags & RADEON_PLL_LEGACY) {
+		if (pll->flags & RADEON_PLL_LEGACY) {
 			if ((post_div == 5) ||
 			    (post_div == 7) ||
 			    (post_div == 9) ||
@@ -495,7 +550,7 @@ void radeon_compute_pll(struct radeon_pll *pll,
 					tmp += (uint64_t)pll->reference_freq * 1000 * frac_feedback_div;
 					current_freq = radeon_div(tmp, ref_div * post_div);
 
-					if (flags & RADEON_PLL_PREFER_CLOSEST_LOWER) {
+					if (pll->flags & RADEON_PLL_PREFER_CLOSEST_LOWER) {
 						error = freq - current_freq;
 						error = error < 0 ? 0xffffffff : error;
 					} else
@@ -522,12 +577,12 @@ void radeon_compute_pll(struct radeon_pll *pll,
 							best_freq = current_freq;
 							best_error = error;
 							best_vco_diff = vco_diff;
-						} else if (((flags & RADEON_PLL_PREFER_LOW_REF_DIV) && (ref_div < best_ref_div)) ||
-							   ((flags & RADEON_PLL_PREFER_HIGH_REF_DIV) && (ref_div > best_ref_div)) ||
-							   ((flags & RADEON_PLL_PREFER_LOW_FB_DIV) && (feedback_div < best_feedback_div)) ||
-							   ((flags & RADEON_PLL_PREFER_HIGH_FB_DIV) && (feedback_div > best_feedback_div)) ||
-							   ((flags & RADEON_PLL_PREFER_LOW_POST_DIV) && (post_div < best_post_div)) ||
-							   ((flags & RADEON_PLL_PREFER_HIGH_POST_DIV) && (post_div > best_post_div))) {
+						} else if (((pll->flags & RADEON_PLL_PREFER_LOW_REF_DIV) && (ref_div < best_ref_div)) ||
+							   ((pll->flags & RADEON_PLL_PREFER_HIGH_REF_DIV) && (ref_div > best_ref_div)) ||
+							   ((pll->flags & RADEON_PLL_PREFER_LOW_FB_DIV) && (feedback_div < best_feedback_div)) ||
+							   ((pll->flags & RADEON_PLL_PREFER_HIGH_FB_DIV) && (feedback_div > best_feedback_div)) ||
+							   ((pll->flags & RADEON_PLL_PREFER_LOW_POST_DIV) && (post_div < best_post_div)) ||
+							   ((pll->flags & RADEON_PLL_PREFER_HIGH_POST_DIV) && (post_div > best_post_div))) {
 							best_post_div = post_div;
 							best_ref_div = ref_div;
 							best_feedback_div = feedback_div;
@@ -557,6 +612,196 @@ void radeon_compute_pll(struct radeon_pll *pll,
 	*post_div_p = best_post_div;
 }
 
+static bool
+calc_fb_div(struct radeon_pll *pll,
+	    uint32_t freq,
+            uint32_t post_div,
+            uint32_t ref_div,
+            uint32_t *fb_div,
+            uint32_t *fb_div_frac)
+{
+	fixed20_12 feedback_divider, a, b;
+	u32 vco_freq;
+
+	vco_freq = freq * post_div;
+	/* feedback_divider = vco_freq * ref_div / pll->reference_freq; */
+	a.full = rfixed_const(pll->reference_freq);
+	feedback_divider.full = rfixed_const(vco_freq);
+	feedback_divider.full = rfixed_div(feedback_divider, a);
+	a.full = rfixed_const(ref_div);
+	feedback_divider.full = rfixed_mul(feedback_divider, a);
+
+	if (pll->flags & RADEON_PLL_USE_FRAC_FB_DIV) {
+		/* feedback_divider = floor((feedback_divider * 10.0) + 0.5) * 0.1; */
+		a.full = rfixed_const(10);
+		feedback_divider.full = rfixed_mul(feedback_divider, a);
+		feedback_divider.full += rfixed_const_half(0);
+		feedback_divider.full = rfixed_floor(feedback_divider);
+		feedback_divider.full = rfixed_div(feedback_divider, a);
+
+		/* *fb_div = floor(feedback_divider); */
+		a.full = rfixed_floor(feedback_divider);
+		*fb_div = rfixed_trunc(a);
+		/* *fb_div_frac = fmod(feedback_divider, 1.0) * 10.0; */
+		a.full = rfixed_const(10);
+		b.full = rfixed_mul(feedback_divider, a);
+
+		feedback_divider.full = rfixed_floor(feedback_divider);
+		feedback_divider.full = rfixed_mul(feedback_divider, a);
+		feedback_divider.full = b.full - feedback_divider.full;
+		*fb_div_frac = rfixed_trunc(feedback_divider);
+	} else {
+		/* *fb_div = floor(feedback_divider + 0.5); */
+		feedback_divider.full += rfixed_const_half(0);
+		feedback_divider.full = rfixed_floor(feedback_divider);
+
+		*fb_div = rfixed_trunc(feedback_divider);
+		*fb_div_frac = 0;
+	}
+
+	if (((*fb_div) < pll->min_feedback_div) || ((*fb_div) > pll->max_feedback_div))
+		return false;
+	else
+		return true;
+}
+
+static bool
+calc_fb_ref_div(struct radeon_pll *pll,
+		uint32_t freq,
+		uint32_t post_div,
+		uint32_t *fb_div,
+                uint32_t *fb_div_frac,
+                uint32_t *ref_div)
+{
+	fixed20_12 ffreq, max_error, error, pll_out, a;
+	u32 vco;
+
+	ffreq.full = rfixed_const(freq);
+	/* max_error = ffreq * 0.0025; */
+	a.full = rfixed_const(400);
+	max_error.full = rfixed_div(ffreq, a);
+
+	for ((*ref_div) = pll->min_ref_div; (*ref_div) < pll->max_ref_div; ++(*ref_div)) {
+		if (calc_fb_div(pll, freq, post_div, (*ref_div), fb_div, fb_div_frac)) {
+			vco = pll->reference_freq * (((*fb_div) * 10) + (*fb_div_frac));
+			vco = vco / ((*ref_div) * 10);
+
+			if ((vco < pll->pll_out_min) || (vco > pll->pll_out_max))
+				continue;
+
+			/* pll_out = vco / post_div; */
+			a.full = rfixed_const(post_div);
+			pll_out.full = rfixed_const(vco);
+			pll_out.full = rfixed_div(pll_out, a);
+
+			if (pll_out.full >= ffreq.full) {
+				error.full = pll_out.full - ffreq.full;
+				if (error.full <= max_error.full)
+					return true;
+			}
+		}
+	}
+	return false;
+}
+
+static void radeon_compute_pll_new(struct radeon_pll *pll,
+				   uint64_t freq,
+				   uint32_t *dot_clock_p,
+				   uint32_t *fb_div_p,
+				   uint32_t *frac_fb_div_p,
+				   uint32_t *ref_div_p,
+				   uint32_t *post_div_p)
+{
+	u32 fb_div = 0, fb_div_frac = 0, post_div = 0, ref_div = 0;
+	u32 best_freq = 0, vco_frequency;
+
+	/* freq = freq / 10; */
+	do_div(freq, 10);
+
+	if (pll->flags & RADEON_PLL_USE_POST_DIV) {
+		post_div = pll->post_div;
+		if ((post_div < pll->min_post_div) || (post_div > pll->max_post_div))
+			goto done;
+
+		vco_frequency = freq * post_div;
+		if ((vco_frequency < pll->pll_out_min) || (vco_frequency > pll->pll_out_max))
+			goto done;
+
+		if (pll->flags & RADEON_PLL_USE_REF_DIV) {
+			ref_div = pll->reference_div;
+			if ((ref_div < pll->min_ref_div) || (ref_div > pll->max_ref_div))
+				goto done;
+			if (!calc_fb_div(pll, freq, post_div, ref_div, &fb_div, &fb_div_frac))
+				goto done;
+		}
+	} else {
+		for (post_div = pll->max_post_div; post_div >= pll->min_post_div; --post_div) {
+			if (pll->flags & RADEON_PLL_LEGACY) {
+				if ((post_div == 5) ||
+				    (post_div == 7) ||
+				    (post_div == 9) ||
+				    (post_div == 10) ||
+				    (post_div == 11))
+					continue;
+			}
+
+			if ((pll->flags & RADEON_PLL_NO_ODD_POST_DIV) && (post_div & 1))
+				continue;
+
+			vco_frequency = freq * post_div;
+			if ((vco_frequency < pll->pll_out_min) || (vco_frequency > pll->pll_out_max))
+				continue;
+			if (pll->flags & RADEON_PLL_USE_REF_DIV) {
+				ref_div = pll->reference_div;
+				if ((ref_div < pll->min_ref_div) || (ref_div > pll->max_ref_div))
+					goto done;
+				if (calc_fb_div(pll, freq, post_div, ref_div, &fb_div, &fb_div_frac))
+					break;
+			} else {
+				if (calc_fb_ref_div(pll, freq, post_div, &fb_div, &fb_div_frac, &ref_div))
+					break;
+			}
+		}
+	}
+
+	best_freq = pll->reference_freq * 10 * fb_div;
+	best_freq += pll->reference_freq * fb_div_frac;
+	best_freq = best_freq / (ref_div * post_div);
+
+done:
+	if (best_freq == 0)
+		DRM_ERROR("Couldn't find valid PLL dividers\n");
+
+	*dot_clock_p = best_freq / 10;
+	*fb_div_p = fb_div;
+	*frac_fb_div_p = fb_div_frac;
+	*ref_div_p = ref_div;
+	*post_div_p = post_div;
+
+	DRM_DEBUG("%u %d.%d, %d, %d\n", *dot_clock_p, *fb_div_p, *frac_fb_div_p, *ref_div_p, *post_div_p);
+}
+
+void radeon_compute_pll(struct radeon_pll *pll,
+			uint64_t freq,
+			uint32_t *dot_clock_p,
+			uint32_t *fb_div_p,
+			uint32_t *frac_fb_div_p,
+			uint32_t *ref_div_p,
+			uint32_t *post_div_p)
+{
+	switch (pll->algo) {
+	case PLL_ALGO_NEW:
+		radeon_compute_pll_new(pll, freq, dot_clock_p, fb_div_p,
+				       frac_fb_div_p, ref_div_p, post_div_p);
+		break;
+	case PLL_ALGO_LEGACY:
+	default:
+		radeon_compute_pll_legacy(pll, freq, dot_clock_p, fb_div_p,
+					  frac_fb_div_p, ref_div_p, post_div_p);
+		break;
+	}
+}
+
 static void radeon_user_framebuffer_destroy(struct drm_framebuffer *fb)
 {
 	struct radeon_framebuffer *radeon_fb = to_radeon_framebuffer(fb);
@@ -565,12 +810,8 @@ static void radeon_user_framebuffer_destroy(struct drm_framebuffer *fb)
 	if (fb->fbdev)
 		radeonfb_remove(dev, fb);
 
-	if (radeon_fb->obj) {
-		radeon_gem_object_unpin(radeon_fb->obj);
-		mutex_lock(&dev->struct_mutex);
-		drm_gem_object_unreference(radeon_fb->obj);
-		mutex_unlock(&dev->struct_mutex);
-	}
+	if (radeon_fb->obj)
+		drm_gem_object_unreference_unlocked(radeon_fb->obj);
 	drm_framebuffer_cleanup(fb);
 	kfree(radeon_fb);
 }
@@ -614,7 +855,11 @@ radeon_user_framebuffer_create(struct drm_device *dev,
 	struct drm_gem_object *obj;
 
 	obj = drm_gem_object_lookup(dev, file_priv, mode_cmd->handle);
-
+	if (obj ==  NULL) {
+		dev_err(&dev->pdev->dev, "No GEM object associated to handle 0x%08X, "
+			"can't create framebuffer\n", mode_cmd->handle);
+		return NULL;
+	}
 	return radeon_framebuffer_create(dev, mode_cmd, obj);
 }
 
@@ -644,7 +889,7 @@ static struct drm_prop_enum_list radeon_tv_std_enum_list[] =
 	{ TV_STD_SECAM, "secam" },
 };
 
-int radeon_modeset_create_props(struct radeon_device *rdev)
+static int radeon_modeset_create_props(struct radeon_device *rdev)
 {
 	int i, sz;
 
@@ -657,7 +902,7 @@ int radeon_modeset_create_props(struct radeon_device *rdev)
 			return -ENOMEM;
 
 		rdev->mode_info.coherent_mode_property->values[0] = 0;
-		rdev->mode_info.coherent_mode_property->values[0] = 1;
+		rdev->mode_info.coherent_mode_property->values[1] = 1;
 	}
 
 	if (!ASIC_IS_AVIVO(rdev)) {
@@ -681,7 +926,7 @@ int radeon_modeset_create_props(struct radeon_device *rdev)
 	if (!rdev->mode_info.load_detect_property)
 		return -ENOMEM;
 	rdev->mode_info.load_detect_property->values[0] = 0;
-	rdev->mode_info.load_detect_property->values[0] = 1;
+	rdev->mode_info.load_detect_property->values[1] = 1;
 
 	drm_mode_create_scaling_mode_property(rdev->ddev);
 
@@ -702,7 +947,7 @@ int radeon_modeset_create_props(struct radeon_device *rdev)
 
 int radeon_modeset_init(struct radeon_device *rdev)
 {
-	int num_crtc = 2, i;
+	int i;
 	int ret;
 
 	drm_mode_config_init(rdev->ddev);
@@ -724,8 +969,24 @@ int radeon_modeset_init(struct radeon_device *rdev)
 	if (ret) {
 		return ret;
 	}
-	/* allocate crtcs - TODO single crtc */
-	for (i = 0; i < num_crtc; i++) {
+
+	/* check combios for a valid hardcoded EDID - Sun servers */
+	if (!rdev->is_atom_bios) {
+		/* check for hardcoded EDID in BIOS */
+		radeon_combios_check_hardcoded_edid(rdev);
+	}
+
+	if (rdev->flags & RADEON_SINGLE_CRTC)
+		rdev->num_crtc = 1;
+	else {
+		if (ASIC_IS_DCE4(rdev))
+			rdev->num_crtc = 6;
+		else
+			rdev->num_crtc = 2;
+	}
+
+	/* allocate crtcs */
+	for (i = 0; i < rdev->num_crtc; i++) {
 		radeon_crtc_init(rdev->ddev, i);
 	}
 
@@ -734,13 +995,18 @@ int radeon_modeset_init(struct radeon_device *rdev)
 	if (!ret) {
 		return ret;
 	}
+	/* initialize hpd */
+	radeon_hpd_init(rdev);
 	drm_helper_initial_config(rdev->ddev);
 	return 0;
 }
 
 void radeon_modeset_fini(struct radeon_device *rdev)
 {
+	kfree(rdev->mode_info.bios_hardcoded_edid);
+
 	if (rdev->mode_info.mode_config_initialized) {
+		radeon_hpd_fini(rdev);
 		drm_mode_config_cleanup(rdev->ddev);
 		rdev->mode_info.mode_config_initialized = false;
 	}
@@ -761,10 +1027,18 @@ bool radeon_crtc_scaling_mode_fixup(struct drm_crtc *crtc,
 		if (encoder->crtc != crtc)
 			continue;
 		if (first) {
-			radeon_crtc->rmx_type = radeon_encoder->rmx_type;
+			/* set scaling */
+			if (radeon_encoder->rmx_type == RMX_OFF)
+				radeon_crtc->rmx_type = RMX_OFF;
+			else if (mode->hdisplay < radeon_encoder->native_mode.hdisplay ||
+				 mode->vdisplay < radeon_encoder->native_mode.vdisplay)
+				radeon_crtc->rmx_type = radeon_encoder->rmx_type;
+			else
+				radeon_crtc->rmx_type = RMX_OFF;
+			/* copy native mode */
 			memcpy(&radeon_crtc->native_mode,
-				&radeon_encoder->native_mode,
-				sizeof(struct radeon_native_mode));
+			       &radeon_encoder->native_mode,
+				sizeof(struct drm_display_mode));
 			first = false;
 		} else {
 			if (radeon_crtc->rmx_type != radeon_encoder->rmx_type) {
@@ -782,10 +1056,10 @@ bool radeon_crtc_scaling_mode_fixup(struct drm_crtc *crtc,
 	if (radeon_crtc->rmx_type != RMX_OFF) {
 		fixed20_12 a, b;
 		a.full = rfixed_const(crtc->mode.vdisplay);
-		b.full = rfixed_const(radeon_crtc->native_mode.panel_xres);
+		b.full = rfixed_const(radeon_crtc->native_mode.hdisplay);
 		radeon_crtc->vsc.full = rfixed_div(a, b);
 		a.full = rfixed_const(crtc->mode.hdisplay);
-		b.full = rfixed_const(radeon_crtc->native_mode.panel_yres);
+		b.full = rfixed_const(radeon_crtc->native_mode.vdisplay);
 		radeon_crtc->hsc.full = rfixed_div(a, b);
 	} else {
 		radeon_crtc->vsc.full = rfixed_const(1);
diff --git a/drivers/gpu/drm/radeon/radeon_drv.c b/drivers/gpu/drm/radeon/radeon_drv.c
index 50fce498910c..6eec0ece6a6c 100644
--- a/drivers/gpu/drm/radeon/radeon_drv.c
+++ b/drivers/gpu/drm/radeon/radeon_drv.c
@@ -40,9 +40,11 @@
 
 /*
  * KMS wrapper.
+ * - 2.0.0 - initial interface
+ * - 2.1.0 - add square tiling interface
  */
 #define KMS_DRIVER_MAJOR	2
-#define KMS_DRIVER_MINOR	0
+#define KMS_DRIVER_MINOR	1
 #define KMS_DRIVER_PATCHLEVEL	0
 int radeon_driver_load_kms(struct drm_device *dev, unsigned long flags);
 int radeon_driver_unload_kms(struct drm_device *dev);
@@ -62,9 +64,6 @@ void radeon_driver_irq_preinstall_kms(struct drm_device *dev);
 int radeon_driver_irq_postinstall_kms(struct drm_device *dev);
 void radeon_driver_irq_uninstall_kms(struct drm_device *dev);
 irqreturn_t radeon_driver_irq_handler_kms(DRM_IRQ_ARGS);
-int radeon_master_create_kms(struct drm_device *dev, struct drm_master *master);
-void radeon_master_destroy_kms(struct drm_device *dev,
-			       struct drm_master *master);
 int radeon_dma_ioctl_kms(struct drm_device *dev, void *data,
 			 struct drm_file *file_priv);
 int radeon_gem_object_init(struct drm_gem_object *obj);
@@ -89,6 +88,9 @@ int radeon_benchmarking = 0;
 int radeon_testing = 0;
 int radeon_connector_table = 0;
 int radeon_tv = 1;
+int radeon_new_pll = -1;
+int radeon_dynpm = -1;
+int radeon_audio = 1;
 
 MODULE_PARM_DESC(no_wb, "Disable AGP writeback for scratch registers");
 module_param_named(no_wb, radeon_no_wb, int, 0444);
@@ -123,6 +125,15 @@ module_param_named(connector_table, radeon_connector_table, int, 0444);
 MODULE_PARM_DESC(tv, "TV enable (0 = disable)");
 module_param_named(tv, radeon_tv, int, 0444);
 
+MODULE_PARM_DESC(new_pll, "Select new PLL code");
+module_param_named(new_pll, radeon_new_pll, int, 0444);
+
+MODULE_PARM_DESC(dynpm, "Disable/Enable dynamic power management (1 = enable)");
+module_param_named(dynpm, radeon_dynpm, int, 0444);
+
+MODULE_PARM_DESC(audio, "Audio enable (0 = disable)");
+module_param_named(audio, radeon_audio, int, 0444);
+
 static int radeon_suspend(struct drm_device *dev, pm_message_t state)
 {
 	drm_radeon_private_t *dev_priv = dev->dev_private;
@@ -191,7 +202,7 @@ static struct drm_driver driver_old = {
 		 .owner = THIS_MODULE,
 		 .open = drm_open,
 		 .release = drm_release,
-		 .ioctl = drm_ioctl,
+		 .unlocked_ioctl = drm_ioctl,
 		 .mmap = drm_mmap,
 		 .poll = drm_poll,
 		 .fasync = drm_fasync,
@@ -260,8 +271,6 @@ static struct drm_driver kms_driver = {
 	.get_vblank_counter = radeon_get_vblank_counter_kms,
 	.enable_vblank = radeon_enable_vblank_kms,
 	.disable_vblank = radeon_disable_vblank_kms,
-	.master_create = radeon_master_create_kms,
-	.master_destroy = radeon_master_destroy_kms,
 #if defined(CONFIG_DEBUG_FS)
 	.debugfs_init = radeon_debugfs_init,
 	.debugfs_cleanup = radeon_debugfs_cleanup,
@@ -281,7 +290,7 @@ static struct drm_driver kms_driver = {
 		 .owner = THIS_MODULE,
 		 .open = drm_open,
 		 .release = drm_release,
-		 .ioctl = drm_ioctl,
+		 .unlocked_ioctl = drm_ioctl,
 		 .mmap = radeon_mmap,
 		 .poll = drm_poll,
 		 .fasync = drm_fasync,
@@ -336,6 +345,7 @@ static int __init radeon_init(void)
 		driver = &kms_driver;
 		driver->driver_features |= DRIVER_MODESET;
 		driver->num_ioctls = radeon_max_kms_ioctl;
+		radeon_register_atpx_handler();
 	}
 	/* if the vga console setting is enabled still
 	 * let modprobe override it */
@@ -345,6 +355,7 @@ static int __init radeon_init(void)
 static void __exit radeon_exit(void)
 {
 	drm_exit(driver);
+	radeon_unregister_atpx_handler();
 }
 
 module_init(radeon_init);
diff --git a/drivers/gpu/drm/radeon/radeon_drv.h b/drivers/gpu/drm/radeon/radeon_drv.h
index 350962e0f346..ec55f2b23c22 100644
--- a/drivers/gpu/drm/radeon/radeon_drv.h
+++ b/drivers/gpu/drm/radeon/radeon_drv.h
@@ -106,9 +106,10 @@
  * 1.29- R500 3D cmd buffer support
  * 1.30- Add support for occlusion queries
  * 1.31- Add support for num Z pipes from GET_PARAM
+ * 1.32- fixes for rv740 setup
  */
 #define DRIVER_MAJOR		1
-#define DRIVER_MINOR		31
+#define DRIVER_MINOR		32
 #define DRIVER_PATCHLEVEL	0
 
 enum radeon_cp_microcode_version {
@@ -267,6 +268,8 @@ typedef struct drm_radeon_private {
 
 	u32 scratch_ages[5];
 
+	int have_z_offset;
+
 	/* starting from here on, data is preserved accross an open */
 	uint32_t flags;		/* see radeon_chip_flags */
 	resource_size_t fb_aper_offset;
@@ -294,6 +297,9 @@ typedef struct drm_radeon_private {
 	int r700_sc_prim_fifo_size;
 	int r700_sc_hiz_tile_fifo_size;
 	int r700_sc_earlyz_tile_fifo_fize;
+	int r600_group_size;
+	int r600_npipes;
+	int r600_nbanks;
 
 	struct mutex cs_mutex;
 	u32 cs_id_scnt;
@@ -309,9 +315,11 @@ typedef struct drm_radeon_buf_priv {
 	u32 age;
 } drm_radeon_buf_priv_t;
 
+struct drm_buffer;
+
 typedef struct drm_radeon_kcmd_buffer {
 	int bufsz;
-	char *buf;
+	struct drm_buffer *buffer;
 	int nbox;
 	struct drm_clip_rect __user *boxes;
 } drm_radeon_kcmd_buffer_t;
@@ -454,6 +462,15 @@ extern void r600_blit_swap(struct drm_device *dev,
 			   int sx, int sy, int dx, int dy,
 			   int w, int h, int src_pitch, int dst_pitch, int cpp);
 
+/* atpx handler */
+#if defined(CONFIG_VGA_SWITCHEROO)
+void radeon_register_atpx_handler(void);
+void radeon_unregister_atpx_handler(void);
+#else
+static inline void radeon_register_atpx_handler(void) {}
+static inline void radeon_unregister_atpx_handler(void) {}
+#endif
+
 /* Flags for stats.boxes
  */
 #define RADEON_BOX_DMA_IDLE      0x1
@@ -1104,7 +1121,6 @@ extern u32 radeon_get_scratch(drm_radeon_private_t *dev_priv, int index);
 #       define R600_IT_WAIT_REG_MEM             0x00003C00
 #       define R600_IT_MEM_WRITE                0x00003D00
 #       define R600_IT_INDIRECT_BUFFER          0x00003200
-#       define R600_IT_CP_INTERRUPT             0x00004000
 #       define R600_IT_SURFACE_SYNC             0x00004300
 #              define R600_CB0_DEST_BASE_ENA    (1 << 6)
 #              define R600_TC_ACTION_ENA        (1 << 23)
@@ -2122,4 +2138,32 @@ extern void radeon_commit_ring(drm_radeon_private_t *dev_priv);
 	write &= mask;						\
 } while (0)
 
+/**
+ * Copy given number of dwords from drm buffer to the ring buffer.
+ */
+#define OUT_RING_DRM_BUFFER(buf, sz) do {				\
+	int _size = (sz) * 4;						\
+	struct drm_buffer *_buf = (buf);				\
+	int _part_size;							\
+	while (_size > 0) {						\
+		_part_size = _size;					\
+									\
+		if (write + _part_size/4 > mask)			\
+			_part_size = ((mask + 1) - write)*4;		\
+									\
+		if (drm_buffer_index(_buf) + _part_size > PAGE_SIZE)	\
+			_part_size = PAGE_SIZE - drm_buffer_index(_buf);\
+									\
+									\
+									\
+		memcpy(ring + write, &_buf->data[drm_buffer_page(_buf)]	\
+			[drm_buffer_index(_buf)], _part_size);		\
+									\
+		_size -= _part_size;					\
+		write = (write + _part_size/4) & mask;			\
+		drm_buffer_advance(_buf, _part_size);			\
+	}								\
+} while (0)
+
+
 #endif				/* __RADEON_DRV_H__ */
diff --git a/drivers/gpu/drm/radeon/radeon_encoders.c b/drivers/gpu/drm/radeon/radeon_encoders.c
index 621646752cd2..bc926ea0a530 100644
--- a/drivers/gpu/drm/radeon/radeon_encoders.c
+++ b/drivers/gpu/drm/radeon/radeon_encoders.c
@@ -31,6 +31,55 @@
 
 extern int atom_debug;
 
+/* evil but including atombios.h is much worse */
+bool radeon_atom_get_tv_timings(struct radeon_device *rdev, int index,
+				struct drm_display_mode *mode);
+
+static uint32_t radeon_encoder_clones(struct drm_encoder *encoder)
+{
+	struct drm_device *dev = encoder->dev;
+	struct radeon_device *rdev = dev->dev_private;
+	struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
+	struct drm_encoder *clone_encoder;
+	uint32_t index_mask = 0;
+	int count;
+
+	/* DIG routing gets problematic */
+	if (rdev->family >= CHIP_R600)
+		return index_mask;
+	/* LVDS/TV are too wacky */
+	if (radeon_encoder->devices & ATOM_DEVICE_LCD_SUPPORT)
+		return index_mask;
+	/* DVO requires 2x ppll clocks depending on tmds chip */
+	if (radeon_encoder->devices & ATOM_DEVICE_DFP2_SUPPORT)
+		return index_mask;
+
+	count = -1;
+	list_for_each_entry(clone_encoder, &dev->mode_config.encoder_list, head) {
+		struct radeon_encoder *radeon_clone = to_radeon_encoder(clone_encoder);
+		count++;
+
+		if (clone_encoder == encoder)
+			continue;
+		if (radeon_clone->devices & (ATOM_DEVICE_LCD_SUPPORT))
+			continue;
+		if (radeon_clone->devices & ATOM_DEVICE_DFP2_SUPPORT)
+			continue;
+		else
+			index_mask |= (1 << count);
+	}
+	return index_mask;
+}
+
+void radeon_setup_encoder_clones(struct drm_device *dev)
+{
+	struct drm_encoder *encoder;
+
+	list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) {
+		encoder->possible_clones = radeon_encoder_clones(encoder);
+	}
+}
+
 uint32_t
 radeon_get_encoder_id(struct drm_device *dev, uint32_t supported_device, uint8_t dac)
 {
@@ -107,6 +156,26 @@ radeon_get_encoder_id(struct drm_device *dev, uint32_t supported_device, uint8_t
 	return ret;
 }
 
+static inline bool radeon_encoder_is_digital(struct drm_encoder *encoder)
+{
+	struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
+	switch (radeon_encoder->encoder_id) {
+	case ENCODER_OBJECT_ID_INTERNAL_LVDS:
+	case ENCODER_OBJECT_ID_INTERNAL_TMDS1:
+	case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_TMDS1:
+	case ENCODER_OBJECT_ID_INTERNAL_LVTM1:
+	case ENCODER_OBJECT_ID_INTERNAL_DVO1:
+	case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DVO1:
+	case ENCODER_OBJECT_ID_INTERNAL_DDI:
+	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY:
+	case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_LVTMA:
+	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1:
+	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2:
+		return true;
+	default:
+		return false;
+	}
+}
 void
 radeon_link_encoder_connector(struct drm_device *dev)
 {
@@ -153,83 +222,91 @@ radeon_get_connector_for_encoder(struct drm_encoder *encoder)
 
 	list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
 		radeon_connector = to_radeon_connector(connector);
-		if (radeon_encoder->devices & radeon_connector->devices)
+		if (radeon_encoder->active_device & radeon_connector->devices)
 			return connector;
 	}
 	return NULL;
 }
 
-/* used for both atom and legacy */
-void radeon_rmx_mode_fixup(struct drm_encoder *encoder,
-			   struct drm_display_mode *mode,
-			   struct drm_display_mode *adjusted_mode)
+static struct radeon_connector_atom_dig *
+radeon_get_atom_connector_priv_from_encoder(struct drm_encoder *encoder)
 {
-	struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
 	struct drm_device *dev = encoder->dev;
 	struct radeon_device *rdev = dev->dev_private;
-	struct radeon_native_mode *native_mode = &radeon_encoder->native_mode;
-
-	if (mode->hdisplay < native_mode->panel_xres ||
-	    mode->vdisplay < native_mode->panel_yres) {
-		if (ASIC_IS_AVIVO(rdev)) {
-			adjusted_mode->hdisplay = native_mode->panel_xres;
-			adjusted_mode->vdisplay = native_mode->panel_yres;
-			adjusted_mode->htotal = native_mode->panel_xres + native_mode->hblank;
-			adjusted_mode->hsync_start = native_mode->panel_xres + native_mode->hoverplus;
-			adjusted_mode->hsync_end = adjusted_mode->hsync_start + native_mode->hsync_width;
-			adjusted_mode->vtotal = native_mode->panel_yres + native_mode->vblank;
-			adjusted_mode->vsync_start = native_mode->panel_yres + native_mode->voverplus;
-			adjusted_mode->vsync_end = adjusted_mode->vsync_start + native_mode->vsync_width;
-			/* update crtc values */
-			drm_mode_set_crtcinfo(adjusted_mode, CRTC_INTERLACE_HALVE_V);
-			/* adjust crtc values */
-			adjusted_mode->crtc_hdisplay = native_mode->panel_xres;
-			adjusted_mode->crtc_vdisplay = native_mode->panel_yres;
-			adjusted_mode->crtc_htotal = adjusted_mode->crtc_hdisplay + native_mode->hblank;
-			adjusted_mode->crtc_hsync_start = adjusted_mode->crtc_hdisplay + native_mode->hoverplus;
-			adjusted_mode->crtc_hsync_end = adjusted_mode->crtc_hsync_start + native_mode->hsync_width;
-			adjusted_mode->crtc_vtotal = adjusted_mode->crtc_vdisplay + native_mode->vblank;
-			adjusted_mode->crtc_vsync_start = adjusted_mode->crtc_vdisplay + native_mode->voverplus;
-			adjusted_mode->crtc_vsync_end = adjusted_mode->crtc_vsync_start + native_mode->vsync_width;
-		} else {
-			adjusted_mode->htotal = native_mode->panel_xres + native_mode->hblank;
-			adjusted_mode->hsync_start = native_mode->panel_xres + native_mode->hoverplus;
-			adjusted_mode->hsync_end = adjusted_mode->hsync_start + native_mode->hsync_width;
-			adjusted_mode->vtotal = native_mode->panel_yres + native_mode->vblank;
-			adjusted_mode->vsync_start = native_mode->panel_yres + native_mode->voverplus;
-			adjusted_mode->vsync_end = adjusted_mode->vsync_start + native_mode->vsync_width;
-			/* update crtc values */
-			drm_mode_set_crtcinfo(adjusted_mode, CRTC_INTERLACE_HALVE_V);
-			/* adjust crtc values */
-			adjusted_mode->crtc_htotal = adjusted_mode->crtc_hdisplay + native_mode->hblank;
-			adjusted_mode->crtc_hsync_start = adjusted_mode->crtc_hdisplay + native_mode->hoverplus;
-			adjusted_mode->crtc_hsync_end = adjusted_mode->crtc_hsync_start + native_mode->hsync_width;
-			adjusted_mode->crtc_vtotal = adjusted_mode->crtc_vdisplay + native_mode->vblank;
-			adjusted_mode->crtc_vsync_start = adjusted_mode->crtc_vdisplay + native_mode->voverplus;
-			adjusted_mode->crtc_vsync_end = adjusted_mode->crtc_vsync_start + native_mode->vsync_width;
-		}
-		adjusted_mode->flags = native_mode->flags;
-		adjusted_mode->clock = native_mode->dotclock;
-	}
-}
+	struct drm_connector *connector;
+	struct radeon_connector *radeon_connector;
+	struct radeon_connector_atom_dig *dig_connector;
 
+	if (!rdev->is_atom_bios)
+		return NULL;
+
+	connector = radeon_get_connector_for_encoder(encoder);
+	if (!connector)
+		return NULL;
+
+	radeon_connector = to_radeon_connector(connector);
+
+	if (!radeon_connector->con_priv)
+		return NULL;
+
+	dig_connector = radeon_connector->con_priv;
+
+	return dig_connector;
+}
 
 static bool radeon_atom_mode_fixup(struct drm_encoder *encoder,
 				   struct drm_display_mode *mode,
 				   struct drm_display_mode *adjusted_mode)
 {
 	struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
+	struct drm_device *dev = encoder->dev;
+	struct radeon_device *rdev = dev->dev_private;
 
-	drm_mode_set_crtcinfo(adjusted_mode, 0);
+	/* adjust pm to upcoming mode change */
+	radeon_pm_compute_clocks(rdev);
 
-	if (radeon_encoder->rmx_type != RMX_OFF)
-		radeon_rmx_mode_fixup(encoder, mode, adjusted_mode);
+	/* set the active encoder to connector routing */
+	radeon_encoder_set_active_device(encoder);
+	drm_mode_set_crtcinfo(adjusted_mode, 0);
 
 	/* hw bug */
 	if ((mode->flags & DRM_MODE_FLAG_INTERLACE)
 	    && (mode->crtc_vsync_start < (mode->crtc_vdisplay + 2)))
 		adjusted_mode->crtc_vsync_start = adjusted_mode->crtc_vdisplay + 2;
 
+	/* get the native mode for LVDS */
+	if (radeon_encoder->active_device & (ATOM_DEVICE_LCD_SUPPORT)) {
+		struct drm_display_mode *native_mode = &radeon_encoder->native_mode;
+		int mode_id = adjusted_mode->base.id;
+		*adjusted_mode = *native_mode;
+		if (!ASIC_IS_AVIVO(rdev)) {
+			adjusted_mode->hdisplay = mode->hdisplay;
+			adjusted_mode->vdisplay = mode->vdisplay;
+			adjusted_mode->crtc_hdisplay = mode->hdisplay;
+			adjusted_mode->crtc_vdisplay = mode->vdisplay;
+		}
+		adjusted_mode->base.id = mode_id;
+	}
+
+	/* get the native mode for TV */
+	if (radeon_encoder->active_device & (ATOM_DEVICE_TV_SUPPORT)) {
+		struct radeon_encoder_atom_dac *tv_dac = radeon_encoder->enc_priv;
+		if (tv_dac) {
+			if (tv_dac->tv_std == TV_STD_NTSC ||
+			    tv_dac->tv_std == TV_STD_NTSC_J ||
+			    tv_dac->tv_std == TV_STD_PAL_M)
+				radeon_atom_get_tv_timings(rdev, 0, adjusted_mode);
+			else
+				radeon_atom_get_tv_timings(rdev, 1, adjusted_mode);
+		}
+	}
+
+	if (ASIC_IS_DCE3(rdev) &&
+	    (radeon_encoder->active_device & (ATOM_DEVICE_DFP_SUPPORT))) {
+		struct drm_connector *connector = radeon_get_connector_for_encoder(encoder);
+		radeon_dp_set_link_config(connector, mode);
+	}
+
 	return true;
 }
 
@@ -404,35 +481,25 @@ union lvds_encoder_control {
 	LVDS_ENCODER_CONTROL_PS_ALLOCATION_V2 v2;
 };
 
-static void
+void
 atombios_digital_setup(struct drm_encoder *encoder, int action)
 {
 	struct drm_device *dev = encoder->dev;
 	struct radeon_device *rdev = dev->dev_private;
 	struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
+	struct radeon_encoder_atom_dig *dig = radeon_encoder->enc_priv;
+	struct radeon_connector_atom_dig *dig_connector =
+		radeon_get_atom_connector_priv_from_encoder(encoder);
 	union lvds_encoder_control args;
 	int index = 0;
+	int hdmi_detected = 0;
 	uint8_t frev, crev;
-	struct radeon_encoder_atom_dig *dig;
-	struct drm_connector *connector;
-	struct radeon_connector *radeon_connector;
-	struct radeon_connector_atom_dig *dig_connector;
 
-	connector = radeon_get_connector_for_encoder(encoder);
-	if (!connector)
+	if (!dig || !dig_connector)
 		return;
 
-	radeon_connector = to_radeon_connector(connector);
-
-	if (!radeon_encoder->enc_priv)
-		return;
-
-	dig = radeon_encoder->enc_priv;
-
-	if (!radeon_connector->con_priv)
-		return;
-
-	dig_connector = radeon_connector->con_priv;
+	if (atombios_get_encoder_mode(encoder) == ATOM_ENCODER_MODE_HDMI)
+		hdmi_detected = 1;
 
 	memset(&args, 0, sizeof(args));
 
@@ -461,13 +528,13 @@ atombios_digital_setup(struct drm_encoder *encoder, int action)
 		case 1:
 			args.v1.ucMisc = 0;
 			args.v1.ucAction = action;
-			if (drm_detect_hdmi_monitor((struct edid *)connector->edid_blob_ptr))
+			if (hdmi_detected)
 				args.v1.ucMisc |= PANEL_ENCODER_MISC_HDMI_TYPE;
 			args.v1.usPixelClock = cpu_to_le16(radeon_encoder->pixel_clock / 10);
 			if (radeon_encoder->devices & (ATOM_DEVICE_LCD_SUPPORT)) {
-				if (dig->lvds_misc & (1 << 0))
+				if (dig->lvds_misc & ATOM_PANEL_MISC_DUAL)
 					args.v1.ucMisc |= PANEL_ENCODER_MISC_DUAL;
-				if (dig->lvds_misc & (1 << 1))
+				if (dig->lvds_misc & ATOM_PANEL_MISC_888RGB)
 					args.v1.ucMisc |= (1 << 1);
 			} else {
 				if (dig_connector->linkb)
@@ -486,7 +553,7 @@ atombios_digital_setup(struct drm_encoder *encoder, int action)
 				if (dig->coherent_mode)
 					args.v2.ucMisc |= PANEL_ENCODER_MISC_COHERENT;
 			}
-			if (drm_detect_hdmi_monitor((struct edid *)connector->edid_blob_ptr))
+			if (hdmi_detected)
 				args.v2.ucMisc |= PANEL_ENCODER_MISC_HDMI_TYPE;
 			args.v2.usPixelClock = cpu_to_le16(radeon_encoder->pixel_clock / 10);
 			args.v2.ucTruncate = 0;
@@ -494,18 +561,18 @@ atombios_digital_setup(struct drm_encoder *encoder, int action)
 			args.v2.ucTemporal = 0;
 			args.v2.ucFRC = 0;
 			if (radeon_encoder->devices & (ATOM_DEVICE_LCD_SUPPORT)) {
-				if (dig->lvds_misc & (1 << 0))
+				if (dig->lvds_misc & ATOM_PANEL_MISC_DUAL)
 					args.v2.ucMisc |= PANEL_ENCODER_MISC_DUAL;
-				if (dig->lvds_misc & (1 << 5)) {
+				if (dig->lvds_misc & ATOM_PANEL_MISC_SPATIAL) {
 					args.v2.ucSpatial = PANEL_ENCODER_SPATIAL_DITHER_EN;
-					if (dig->lvds_misc & (1 << 1))
+					if (dig->lvds_misc & ATOM_PANEL_MISC_888RGB)
 						args.v2.ucSpatial |= PANEL_ENCODER_SPATIAL_DITHER_DEPTH;
 				}
-				if (dig->lvds_misc & (1 << 6)) {
+				if (dig->lvds_misc & ATOM_PANEL_MISC_TEMPORAL) {
 					args.v2.ucTemporal = PANEL_ENCODER_TEMPORAL_DITHER_EN;
-					if (dig->lvds_misc & (1 << 1))
+					if (dig->lvds_misc & ATOM_PANEL_MISC_888RGB)
 						args.v2.ucTemporal |= PANEL_ENCODER_TEMPORAL_DITHER_DEPTH;
-					if (((dig->lvds_misc >> 2) & 0x3) == 2)
+					if (((dig->lvds_misc >> ATOM_PANEL_MISC_GREY_LEVEL_SHIFT) & 0x3) == 2)
 						args.v2.ucTemporal |= PANEL_ENCODER_TEMPORAL_LEVEL_4;
 				}
 			} else {
@@ -526,7 +593,7 @@ atombios_digital_setup(struct drm_encoder *encoder, int action)
 	}
 
 	atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args);
-
+	r600_hdmi_enable(encoder, hdmi_detected);
 }
 
 int
@@ -534,6 +601,7 @@ atombios_get_encoder_mode(struct drm_encoder *encoder)
 {
 	struct drm_connector *connector;
 	struct radeon_connector *radeon_connector;
+	struct radeon_connector_atom_dig *dig_connector;
 
 	connector = radeon_get_connector_for_encoder(encoder);
 	if (!connector)
@@ -544,7 +612,7 @@ atombios_get_encoder_mode(struct drm_encoder *encoder)
 	switch (connector->connector_type) {
 	case DRM_MODE_CONNECTOR_DVII:
 	case DRM_MODE_CONNECTOR_HDMIB: /* HDMI-B is basically DL-DVI; analog works fine */
-		if (drm_detect_hdmi_monitor((struct edid *)connector->edid_blob_ptr))
+		if (drm_detect_hdmi_monitor(radeon_connector->edid))
 			return ATOM_ENCODER_MODE_HDMI;
 		else if (radeon_connector->use_digital)
 			return ATOM_ENCODER_MODE_DVI;
@@ -554,7 +622,7 @@ atombios_get_encoder_mode(struct drm_encoder *encoder)
 	case DRM_MODE_CONNECTOR_DVID:
 	case DRM_MODE_CONNECTOR_HDMIA:
 	default:
-		if (drm_detect_hdmi_monitor((struct edid *)connector->edid_blob_ptr))
+		if (drm_detect_hdmi_monitor(radeon_connector->edid))
 			return ATOM_ENCODER_MODE_HDMI;
 		else
 			return ATOM_ENCODER_MODE_DVI;
@@ -563,21 +631,23 @@ atombios_get_encoder_mode(struct drm_encoder *encoder)
 		return ATOM_ENCODER_MODE_LVDS;
 		break;
 	case DRM_MODE_CONNECTOR_DisplayPort:
-		/*if (radeon_output->MonType == MT_DP)
-		  return ATOM_ENCODER_MODE_DP;
-		  else*/
-		if (drm_detect_hdmi_monitor((struct edid *)connector->edid_blob_ptr))
+	case DRM_MODE_CONNECTOR_eDP:
+		dig_connector = radeon_connector->con_priv;
+		if ((dig_connector->dp_sink_type == CONNECTOR_OBJECT_ID_DISPLAYPORT) ||
+		    (dig_connector->dp_sink_type == CONNECTOR_OBJECT_ID_eDP))
+			return ATOM_ENCODER_MODE_DP;
+		else if (drm_detect_hdmi_monitor(radeon_connector->edid))
 			return ATOM_ENCODER_MODE_HDMI;
 		else
 			return ATOM_ENCODER_MODE_DVI;
 		break;
-	case CONNECTOR_DVI_A:
-	case CONNECTOR_VGA:
+	case DRM_MODE_CONNECTOR_DVIA:
+	case DRM_MODE_CONNECTOR_VGA:
 		return ATOM_ENCODER_MODE_CRT;
 		break;
-	case CONNECTOR_STV:
-	case CONNECTOR_CTV:
-	case CONNECTOR_DIN:
+	case DRM_MODE_CONNECTOR_Composite:
+	case DRM_MODE_CONNECTOR_SVIDEO:
+	case DRM_MODE_CONNECTOR_9PinDIN:
 		/* fix me */
 		return ATOM_ENCODER_MODE_TV;
 		/*return ATOM_ENCODER_MODE_CV;*/
@@ -585,98 +655,114 @@ atombios_get_encoder_mode(struct drm_encoder *encoder)
 	}
 }
 
-static void
+/*
+ * DIG Encoder/Transmitter Setup
+ *
+ * DCE 3.0/3.1
+ * - 2 DIG transmitter blocks. UNIPHY (links A and B) and LVTMA.
+ * Supports up to 3 digital outputs
+ * - 2 DIG encoder blocks.
+ * DIG1 can drive UNIPHY link A or link B
+ * DIG2 can drive UNIPHY link B or LVTMA
+ *
+ * DCE 3.2
+ * - 3 DIG transmitter blocks. UNIPHY0/1/2 (links A and B).
+ * Supports up to 5 digital outputs
+ * - 2 DIG encoder blocks.
+ * DIG1/2 can drive UNIPHY0/1/2 link A or link B
+ *
+ * DCE 4.0
+ * - 3 DIG transmitter blocks UNPHY0/1/2 (links A and B).
+ * Supports up to 6 digital outputs
+ * - 6 DIG encoder blocks.
+ * - DIG to PHY mapping is hardcoded
+ * DIG1 drives UNIPHY0 link A, A+B
+ * DIG2 drives UNIPHY0 link B
+ * DIG3 drives UNIPHY1 link A, A+B
+ * DIG4 drives UNIPHY1 link B
+ * DIG5 drives UNIPHY2 link A, A+B
+ * DIG6 drives UNIPHY2 link B
+ *
+ * Routing
+ * crtc -> dig encoder -> UNIPHY/LVTMA (1 or 2 links)
+ * Examples:
+ * crtc0 -> dig2 -> LVTMA   links A+B -> TMDS/HDMI
+ * crtc1 -> dig1 -> UNIPHY0 link  B   -> DP
+ * crtc0 -> dig1 -> UNIPHY2 link  A   -> LVDS
+ * crtc1 -> dig2 -> UNIPHY1 link  B+A -> TMDS/HDMI
+ */
+
+union dig_encoder_control {
+	DIG_ENCODER_CONTROL_PS_ALLOCATION v1;
+	DIG_ENCODER_CONTROL_PARAMETERS_V2 v2;
+	DIG_ENCODER_CONTROL_PARAMETERS_V3 v3;
+};
+
+void
 atombios_dig_encoder_setup(struct drm_encoder *encoder, int action)
 {
 	struct drm_device *dev = encoder->dev;
 	struct radeon_device *rdev = dev->dev_private;
 	struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
-	DIG_ENCODER_CONTROL_PS_ALLOCATION args;
+	struct radeon_encoder_atom_dig *dig = radeon_encoder->enc_priv;
+	struct radeon_connector_atom_dig *dig_connector =
+		radeon_get_atom_connector_priv_from_encoder(encoder);
+	union dig_encoder_control args;
 	int index = 0, num = 0;
 	uint8_t frev, crev;
-	struct radeon_encoder_atom_dig *dig;
-	struct drm_connector *connector;
-	struct radeon_connector *radeon_connector;
-	struct radeon_connector_atom_dig *dig_connector;
-
-	connector = radeon_get_connector_for_encoder(encoder);
-	if (!connector)
-		return;
-
-	radeon_connector = to_radeon_connector(connector);
-
-	if (!radeon_connector->con_priv)
-		return;
 
-	dig_connector = radeon_connector->con_priv;
-
-	if (!radeon_encoder->enc_priv)
+	if (!dig || !dig_connector)
 		return;
 
-	dig = radeon_encoder->enc_priv;
-
 	memset(&args, 0, sizeof(args));
 
-	if (ASIC_IS_DCE32(rdev)) {
-		if (dig->dig_block)
+	if (ASIC_IS_DCE4(rdev))
+		index = GetIndexIntoMasterTable(COMMAND, DIGxEncoderControl);
+	else {
+		if (dig->dig_encoder)
 			index = GetIndexIntoMasterTable(COMMAND, DIG2EncoderControl);
 		else
 			index = GetIndexIntoMasterTable(COMMAND, DIG1EncoderControl);
-		num = dig->dig_block + 1;
-	} else {
-		switch (radeon_encoder->encoder_id) {
-		case ENCODER_OBJECT_ID_INTERNAL_UNIPHY:
-			index = GetIndexIntoMasterTable(COMMAND, DIG1EncoderControl);
-			num = 1;
-			break;
-		case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_LVTMA:
-			index = GetIndexIntoMasterTable(COMMAND, DIG2EncoderControl);
-			num = 2;
-			break;
-		}
 	}
+	num = dig->dig_encoder + 1;
 
 	atom_parse_cmd_header(rdev->mode_info.atom_context, index, &frev, &crev);
 
-	args.ucAction = action;
-	args.usPixelClock = cpu_to_le16(radeon_encoder->pixel_clock / 10);
+	args.v1.ucAction = action;
+	args.v1.usPixelClock = cpu_to_le16(radeon_encoder->pixel_clock / 10);
+	args.v1.ucEncoderMode = atombios_get_encoder_mode(encoder);
+
+	if (args.v1.ucEncoderMode == ATOM_ENCODER_MODE_DP) {
+		if (dig_connector->dp_clock == 270000)
+			args.v1.ucConfig |= ATOM_ENCODER_CONFIG_DPLINKRATE_2_70GHZ;
+		args.v1.ucLaneNum = dig_connector->dp_lane_count;
+	} else if (radeon_encoder->pixel_clock > 165000)
+		args.v1.ucLaneNum = 8;
+	else
+		args.v1.ucLaneNum = 4;
 
-	if (ASIC_IS_DCE32(rdev)) {
+	if (ASIC_IS_DCE4(rdev)) {
+		args.v3.acConfig.ucDigSel = dig->dig_encoder;
+		args.v3.ucBitPerColor = PANEL_8BIT_PER_COLOR;
+	} else {
 		switch (radeon_encoder->encoder_id) {
 		case ENCODER_OBJECT_ID_INTERNAL_UNIPHY:
-			args.ucConfig = ATOM_ENCODER_CONFIG_V2_TRANSMITTER1;
+			args.v1.ucConfig = ATOM_ENCODER_CONFIG_V2_TRANSMITTER1;
 			break;
 		case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1:
-			args.ucConfig = ATOM_ENCODER_CONFIG_V2_TRANSMITTER2;
+		case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_LVTMA:
+			args.v1.ucConfig = ATOM_ENCODER_CONFIG_V2_TRANSMITTER2;
 			break;
 		case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2:
-			args.ucConfig = ATOM_ENCODER_CONFIG_V2_TRANSMITTER3;
+			args.v1.ucConfig = ATOM_ENCODER_CONFIG_V2_TRANSMITTER3;
 			break;
 		}
-	} else {
-		switch (radeon_encoder->encoder_id) {
-		case ENCODER_OBJECT_ID_INTERNAL_UNIPHY:
-			args.ucConfig = ATOM_ENCODER_CONFIG_TRANSMITTER1;
-			break;
-		case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_LVTMA:
-			args.ucConfig = ATOM_ENCODER_CONFIG_TRANSMITTER2;
-			break;
-		}
-	}
-
-	if (radeon_encoder->pixel_clock > 165000) {
-		args.ucConfig |= ATOM_ENCODER_CONFIG_LINKA_B;
-		args.ucLaneNum = 8;
-	} else {
 		if (dig_connector->linkb)
-			args.ucConfig |= ATOM_ENCODER_CONFIG_LINKB;
+			args.v1.ucConfig |= ATOM_ENCODER_CONFIG_LINKB;
 		else
-			args.ucConfig |= ATOM_ENCODER_CONFIG_LINKA;
-		args.ucLaneNum = 4;
+			args.v1.ucConfig |= ATOM_ENCODER_CONFIG_LINKA;
 	}
 
-	args.ucEncoderMode = atombios_get_encoder_mode(encoder);
-
 	atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args);
 
 }
@@ -684,41 +770,38 @@ atombios_dig_encoder_setup(struct drm_encoder *encoder, int action)
 union dig_transmitter_control {
 	DIG_TRANSMITTER_CONTROL_PS_ALLOCATION v1;
 	DIG_TRANSMITTER_CONTROL_PARAMETERS_V2 v2;
+	DIG_TRANSMITTER_CONTROL_PARAMETERS_V3 v3;
 };
 
-static void
-atombios_dig_transmitter_setup(struct drm_encoder *encoder, int action)
+void
+atombios_dig_transmitter_setup(struct drm_encoder *encoder, int action, uint8_t lane_num, uint8_t lane_set)
 {
 	struct drm_device *dev = encoder->dev;
 	struct radeon_device *rdev = dev->dev_private;
 	struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
+	struct radeon_encoder_atom_dig *dig = radeon_encoder->enc_priv;
+	struct radeon_connector_atom_dig *dig_connector =
+		radeon_get_atom_connector_priv_from_encoder(encoder);
+	struct drm_connector *connector;
+	struct radeon_connector *radeon_connector;
 	union dig_transmitter_control args;
 	int index = 0, num = 0;
 	uint8_t frev, crev;
-	struct radeon_encoder_atom_dig *dig;
-	struct drm_connector *connector;
-	struct radeon_connector *radeon_connector;
-	struct radeon_connector_atom_dig *dig_connector;
+	bool is_dp = false;
+	int pll_id = 0;
 
-	connector = radeon_get_connector_for_encoder(encoder);
-	if (!connector)
+	if (!dig || !dig_connector)
 		return;
 
+	connector = radeon_get_connector_for_encoder(encoder);
 	radeon_connector = to_radeon_connector(connector);
 
-	if (!radeon_encoder->enc_priv)
-		return;
-
-	dig = radeon_encoder->enc_priv;
-
-	if (!radeon_connector->con_priv)
-		return;
-
-	dig_connector = radeon_connector->con_priv;
+	if (atombios_get_encoder_mode(encoder) == ATOM_ENCODER_MODE_DP)
+		is_dp = true;
 
 	memset(&args, 0, sizeof(args));
 
-	if (ASIC_IS_DCE32(rdev))
+	if (ASIC_IS_DCE32(rdev) || ASIC_IS_DCE4(rdev))
 		index = GetIndexIntoMasterTable(COMMAND, UNIPHYTransmitterControl);
 	else {
 		switch (radeon_encoder->encoder_id) {
@@ -734,16 +817,72 @@ atombios_dig_transmitter_setup(struct drm_encoder *encoder, int action)
 	atom_parse_cmd_header(rdev->mode_info.atom_context, index, &frev, &crev);
 
 	args.v1.ucAction = action;
+	if (action == ATOM_TRANSMITTER_ACTION_INIT) {
+		args.v1.usInitInfo = radeon_connector->connector_object_id;
+	} else if (action == ATOM_TRANSMITTER_ACTION_SETUP_VSEMPH) {
+		args.v1.asMode.ucLaneSel = lane_num;
+		args.v1.asMode.ucLaneSet = lane_set;
+	} else {
+		if (is_dp)
+			args.v1.usPixelClock =
+				cpu_to_le16(dig_connector->dp_clock / 10);
+		else if (radeon_encoder->pixel_clock > 165000)
+			args.v1.usPixelClock = cpu_to_le16((radeon_encoder->pixel_clock / 2) / 10);
+		else
+			args.v1.usPixelClock = cpu_to_le16(radeon_encoder->pixel_clock / 10);
+	}
+	if (ASIC_IS_DCE4(rdev)) {
+		if (is_dp)
+			args.v3.ucLaneNum = dig_connector->dp_lane_count;
+		else if (radeon_encoder->pixel_clock > 165000)
+			args.v3.ucLaneNum = 8;
+		else
+			args.v3.ucLaneNum = 4;
 
-	if (ASIC_IS_DCE32(rdev)) {
-		if (radeon_encoder->pixel_clock > 165000) {
-			args.v2.usPixelClock = cpu_to_le16((radeon_encoder->pixel_clock * 10 * 2) / 100);
-			args.v2.acConfig.fDualLinkConnector = 1;
-		} else {
-			args.v2.usPixelClock = cpu_to_le16((radeon_encoder->pixel_clock * 10 * 4) / 100);
+		if (dig_connector->linkb) {
+			args.v3.acConfig.ucLinkSel = 1;
+			args.v3.acConfig.ucEncoderSel = 1;
 		}
-		if (dig->dig_block)
+
+		/* Select the PLL for the PHY
+		 * DP PHY should be clocked from external src if there is
+		 * one.
+		 */
+		if (encoder->crtc) {
+			struct radeon_crtc *radeon_crtc = to_radeon_crtc(encoder->crtc);
+			pll_id = radeon_crtc->pll_id;
+		}
+		if (is_dp && rdev->clock.dp_extclk)
+			args.v3.acConfig.ucRefClkSource = 2; /* external src */
+		else
+			args.v3.acConfig.ucRefClkSource = pll_id;
+
+		switch (radeon_encoder->encoder_id) {
+		case ENCODER_OBJECT_ID_INTERNAL_UNIPHY:
+			args.v3.acConfig.ucTransmitterSel = 0;
+			num = 0;
+			break;
+		case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1:
+			args.v3.acConfig.ucTransmitterSel = 1;
+			num = 1;
+			break;
+		case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2:
+			args.v3.acConfig.ucTransmitterSel = 2;
+			num = 2;
+			break;
+		}
+
+		if (is_dp)
+			args.v3.acConfig.fCoherentMode = 1; /* DP requires coherent */
+		else if (radeon_encoder->devices & (ATOM_DEVICE_DFP_SUPPORT)) {
+			if (dig->coherent_mode)
+				args.v3.acConfig.fCoherentMode = 1;
+		}
+	} else if (ASIC_IS_DCE32(rdev)) {
+		if (dig->dig_encoder == 1)
 			args.v2.acConfig.ucEncoderSel = 1;
+		if (dig_connector->linkb)
+			args.v2.acConfig.ucLinkSel = 1;
 
 		switch (radeon_encoder->encoder_id) {
 		case ENCODER_OBJECT_ID_INTERNAL_UNIPHY:
@@ -760,27 +899,29 @@ atombios_dig_transmitter_setup(struct drm_encoder *encoder, int action)
 			break;
 		}
 
-		if (radeon_encoder->devices & (ATOM_DEVICE_DFP_SUPPORT)) {
+		if (is_dp)
+			args.v2.acConfig.fCoherentMode = 1;
+		else if (radeon_encoder->devices & (ATOM_DEVICE_DFP_SUPPORT)) {
 			if (dig->coherent_mode)
 				args.v2.acConfig.fCoherentMode = 1;
 		}
 	} else {
 		args.v1.ucConfig = ATOM_TRANSMITTER_CONFIG_CLKSRC_PPLL;
-		args.v1.usPixelClock = cpu_to_le16((radeon_encoder->pixel_clock) / 10);
+
+		if (dig->dig_encoder)
+			args.v1.ucConfig |= ATOM_TRANSMITTER_CONFIG_DIG2_ENCODER;
+		else
+			args.v1.ucConfig |= ATOM_TRANSMITTER_CONFIG_DIG1_ENCODER;
 
 		switch (radeon_encoder->encoder_id) {
 		case ENCODER_OBJECT_ID_INTERNAL_UNIPHY:
-			args.v1.ucConfig |= ATOM_TRANSMITTER_CONFIG_DIG1_ENCODER;
 			if (rdev->flags & RADEON_IS_IGP) {
 				if (radeon_encoder->pixel_clock > 165000) {
-					args.v1.ucConfig |= (ATOM_TRANSMITTER_CONFIG_8LANE_LINK |
-							     ATOM_TRANSMITTER_CONFIG_LINKA_B);
 					if (dig_connector->igp_lane_info & 0x3)
 						args.v1.ucConfig |= ATOM_TRANSMITTER_CONFIG_LANE_0_7;
 					else if (dig_connector->igp_lane_info & 0xc)
 						args.v1.ucConfig |= ATOM_TRANSMITTER_CONFIG_LANE_8_15;
 				} else {
-					args.v1.ucConfig |= ATOM_TRANSMITTER_CONFIG_LINKA;
 					if (dig_connector->igp_lane_info & 0x1)
 						args.v1.ucConfig |= ATOM_TRANSMITTER_CONFIG_LANE_0_3;
 					else if (dig_connector->igp_lane_info & 0x2)
@@ -790,42 +931,27 @@ atombios_dig_transmitter_setup(struct drm_encoder *encoder, int action)
 					else if (dig_connector->igp_lane_info & 0x8)
 						args.v1.ucConfig |= ATOM_TRANSMITTER_CONFIG_LANE_12_15;
 				}
-			} else {
-				if (radeon_encoder->pixel_clock > 165000)
-					args.v1.ucConfig |= (ATOM_TRANSMITTER_CONFIG_8LANE_LINK |
-							     ATOM_TRANSMITTER_CONFIG_LINKA_B |
-							     ATOM_TRANSMITTER_CONFIG_LANE_0_7);
-				else {
-					if (dig_connector->linkb)
-						args.v1.ucConfig |= ATOM_TRANSMITTER_CONFIG_LINKB | ATOM_TRANSMITTER_CONFIG_LANE_0_3;
-					else
-						args.v1.ucConfig |= ATOM_TRANSMITTER_CONFIG_LINKA | ATOM_TRANSMITTER_CONFIG_LANE_0_3;
-				}
-			}
-			break;
-		case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_LVTMA:
-			args.v1.ucConfig |= ATOM_TRANSMITTER_CONFIG_DIG2_ENCODER;
-			if (radeon_encoder->pixel_clock > 165000)
-				args.v1.ucConfig |= (ATOM_TRANSMITTER_CONFIG_8LANE_LINK |
-						     ATOM_TRANSMITTER_CONFIG_LINKA_B |
-						     ATOM_TRANSMITTER_CONFIG_LANE_0_7);
-			else {
-				if (dig_connector->linkb)
-					args.v1.ucConfig |= ATOM_TRANSMITTER_CONFIG_LINKB | ATOM_TRANSMITTER_CONFIG_LANE_0_3;
-				else
-					args.v1.ucConfig |= ATOM_TRANSMITTER_CONFIG_LINKA | ATOM_TRANSMITTER_CONFIG_LANE_0_3;
 			}
 			break;
 		}
 
-		if (radeon_encoder->devices & (ATOM_DEVICE_DFP_SUPPORT)) {
+		if (radeon_encoder->pixel_clock > 165000)
+			args.v1.ucConfig |= ATOM_TRANSMITTER_CONFIG_8LANE_LINK;
+
+		if (dig_connector->linkb)
+			args.v1.ucConfig |= ATOM_TRANSMITTER_CONFIG_LINKB;
+		else
+			args.v1.ucConfig |= ATOM_TRANSMITTER_CONFIG_LINKA;
+
+		if (is_dp)
+			args.v1.ucConfig |= ATOM_TRANSMITTER_CONFIG_COHERENT;
+		else if (radeon_encoder->devices & (ATOM_DEVICE_DFP_SUPPORT)) {
 			if (dig->coherent_mode)
 				args.v1.ucConfig |= ATOM_TRANSMITTER_CONFIG_COHERENT;
 		}
 	}
 
 	atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args);
-
 }
 
 static void
@@ -874,16 +1000,9 @@ radeon_atom_encoder_dpms(struct drm_encoder *encoder, int mode)
 	DISPLAY_DEVICE_OUTPUT_CONTROL_PS_ALLOCATION args;
 	int index = 0;
 	bool is_dig = false;
-	int devices;
 
 	memset(&args, 0, sizeof(args));
 
-	/* on DPMS off we have no idea if active device is meaningful */
-	if (mode != DRM_MODE_DPMS_ON && !radeon_encoder->active_device)
-		devices = radeon_encoder->devices;
-	else
-		devices = radeon_encoder->active_device;
-
 	DRM_DEBUG("encoder dpms %d to mode %d, devices %08x, active_devices %08x\n",
 		  radeon_encoder->encoder_id, mode, radeon_encoder->devices,
 		  radeon_encoder->active_device);
@@ -914,18 +1033,18 @@ radeon_atom_encoder_dpms(struct drm_encoder *encoder, int mode)
 		break;
 	case ENCODER_OBJECT_ID_INTERNAL_DAC1:
 	case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DAC1:
-		if (devices & (ATOM_DEVICE_TV_SUPPORT))
+		if (radeon_encoder->active_device & (ATOM_DEVICE_TV_SUPPORT))
 			index = GetIndexIntoMasterTable(COMMAND, TV1OutputControl);
-		else if (devices & (ATOM_DEVICE_CV_SUPPORT))
+		else if (radeon_encoder->active_device & (ATOM_DEVICE_CV_SUPPORT))
 			index = GetIndexIntoMasterTable(COMMAND, CV1OutputControl);
 		else
 			index = GetIndexIntoMasterTable(COMMAND, DAC1OutputControl);
 		break;
 	case ENCODER_OBJECT_ID_INTERNAL_DAC2:
 	case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DAC2:
-		if (devices & (ATOM_DEVICE_TV_SUPPORT))
+		if (radeon_encoder->active_device & (ATOM_DEVICE_TV_SUPPORT))
 			index = GetIndexIntoMasterTable(COMMAND, TV1OutputControl);
-		else if (devices & (ATOM_DEVICE_CV_SUPPORT))
+		else if (radeon_encoder->active_device & (ATOM_DEVICE_CV_SUPPORT))
 			index = GetIndexIntoMasterTable(COMMAND, CV1OutputControl);
 		else
 			index = GetIndexIntoMasterTable(COMMAND, DAC2OutputControl);
@@ -935,12 +1054,16 @@ radeon_atom_encoder_dpms(struct drm_encoder *encoder, int mode)
 	if (is_dig) {
 		switch (mode) {
 		case DRM_MODE_DPMS_ON:
-			atombios_dig_transmitter_setup(encoder, ATOM_TRANSMITTER_ACTION_ENABLE);
+			atombios_dig_transmitter_setup(encoder, ATOM_TRANSMITTER_ACTION_ENABLE_OUTPUT, 0, 0);
+			{
+				struct drm_connector *connector = radeon_get_connector_for_encoder(encoder);
+				dp_link_train(encoder, connector);
+			}
 			break;
 		case DRM_MODE_DPMS_STANDBY:
 		case DRM_MODE_DPMS_SUSPEND:
 		case DRM_MODE_DPMS_OFF:
-			atombios_dig_transmitter_setup(encoder, ATOM_TRANSMITTER_ACTION_DISABLE);
+			atombios_dig_transmitter_setup(encoder, ATOM_TRANSMITTER_ACTION_DISABLE_OUTPUT, 0, 0);
 			break;
 		}
 	} else {
@@ -957,9 +1080,12 @@ radeon_atom_encoder_dpms(struct drm_encoder *encoder, int mode)
 		atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args);
 	}
 	radeon_atombios_encoder_dpms_scratch_regs(encoder, (mode == DRM_MODE_DPMS_ON) ? true : false);
+
+	/* adjust pm to dpms change */
+	radeon_pm_compute_clocks(rdev);
 }
 
-union crtc_sourc_param {
+union crtc_source_param {
 	SELECT_CRTC_SOURCE_PS_ALLOCATION v1;
 	SELECT_CRTC_SOURCE_PARAMETERS_V2 v2;
 };
@@ -971,9 +1097,10 @@ atombios_set_encoder_crtc_source(struct drm_encoder *encoder)
 	struct radeon_device *rdev = dev->dev_private;
 	struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
 	struct radeon_crtc *radeon_crtc = to_radeon_crtc(encoder->crtc);
-	union crtc_sourc_param args;
+	union crtc_source_param args;
 	int index = GetIndexIntoMasterTable(COMMAND, SelectCRTC_Source);
 	uint8_t frev, crev;
+	struct radeon_encoder_atom_dig *dig;
 
 	memset(&args, 0, sizeof(args));
 
@@ -1037,20 +1164,32 @@ atombios_set_encoder_crtc_source(struct drm_encoder *encoder)
 			case ENCODER_OBJECT_ID_INTERNAL_UNIPHY:
 			case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1:
 			case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2:
-				if (ASIC_IS_DCE32(rdev)) {
-					if (radeon_crtc->crtc_id)
-						args.v2.ucEncoderID = ASIC_INT_DIG2_ENCODER_ID;
-					else
-						args.v2.ucEncoderID = ASIC_INT_DIG1_ENCODER_ID;
-				} else
+			case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_LVTMA:
+				dig = radeon_encoder->enc_priv;
+				switch (dig->dig_encoder) {
+				case 0:
 					args.v2.ucEncoderID = ASIC_INT_DIG1_ENCODER_ID;
+					break;
+				case 1:
+					args.v2.ucEncoderID = ASIC_INT_DIG2_ENCODER_ID;
+					break;
+				case 2:
+					args.v2.ucEncoderID = ASIC_INT_DIG3_ENCODER_ID;
+					break;
+				case 3:
+					args.v2.ucEncoderID = ASIC_INT_DIG4_ENCODER_ID;
+					break;
+				case 4:
+					args.v2.ucEncoderID = ASIC_INT_DIG5_ENCODER_ID;
+					break;
+				case 5:
+					args.v2.ucEncoderID = ASIC_INT_DIG6_ENCODER_ID;
+					break;
+				}
 				break;
 			case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DVO1:
 				args.v2.ucEncoderID = ASIC_INT_DVO_ENCODER_ID;
 				break;
-			case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_LVTMA:
-				args.v2.ucEncoderID = ASIC_INT_DIG2_ENCODER_ID;
-				break;
 			case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DAC1:
 				if (radeon_encoder->active_device & (ATOM_DEVICE_TV_SUPPORT))
 					args.v2.ucEncoderID = ASIC_INT_TV_ENCODER_ID;
@@ -1077,7 +1216,6 @@ atombios_set_encoder_crtc_source(struct drm_encoder *encoder)
 	}
 
 	atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args);
-
 }
 
 static void
@@ -1104,8 +1242,80 @@ atombios_apply_encoder_quirks(struct drm_encoder *encoder,
 	}
 
 	/* set scaler clears this on some chips */
-	if (ASIC_IS_AVIVO(rdev) && (mode->flags & DRM_MODE_FLAG_INTERLACE))
-		WREG32(AVIVO_D1MODE_DATA_FORMAT + radeon_crtc->crtc_offset, AVIVO_D1MODE_INTERLEAVE_EN);
+	/* XXX check DCE4 */
+	if (!(radeon_encoder->active_device & (ATOM_DEVICE_TV_SUPPORT))) {
+		if (ASIC_IS_AVIVO(rdev) && (mode->flags & DRM_MODE_FLAG_INTERLACE))
+			WREG32(AVIVO_D1MODE_DATA_FORMAT + radeon_crtc->crtc_offset,
+			       AVIVO_D1MODE_INTERLEAVE_EN);
+	}
+}
+
+static int radeon_atom_pick_dig_encoder(struct drm_encoder *encoder)
+{
+	struct drm_device *dev = encoder->dev;
+	struct radeon_device *rdev = dev->dev_private;
+	struct radeon_crtc *radeon_crtc = to_radeon_crtc(encoder->crtc);
+	struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
+	struct drm_encoder *test_encoder;
+	struct radeon_encoder_atom_dig *dig;
+	uint32_t dig_enc_in_use = 0;
+
+	if (ASIC_IS_DCE4(rdev)) {
+		struct radeon_connector_atom_dig *dig_connector =
+			radeon_get_atom_connector_priv_from_encoder(encoder);
+
+		switch (radeon_encoder->encoder_id) {
+		case ENCODER_OBJECT_ID_INTERNAL_UNIPHY:
+			if (dig_connector->linkb)
+				return 1;
+			else
+				return 0;
+			break;
+		case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1:
+			if (dig_connector->linkb)
+				return 3;
+			else
+				return 2;
+			break;
+		case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2:
+			if (dig_connector->linkb)
+				return 5;
+			else
+				return 4;
+			break;
+		}
+	}
+
+	/* on DCE32 and encoder can driver any block so just crtc id */
+	if (ASIC_IS_DCE32(rdev)) {
+		return radeon_crtc->crtc_id;
+	}
+
+	/* on DCE3 - LVTMA can only be driven by DIGB */
+	list_for_each_entry(test_encoder, &dev->mode_config.encoder_list, head) {
+		struct radeon_encoder *radeon_test_encoder;
+
+		if (encoder == test_encoder)
+			continue;
+
+		if (!radeon_encoder_is_digital(test_encoder))
+			continue;
+
+		radeon_test_encoder = to_radeon_encoder(test_encoder);
+		dig = radeon_test_encoder->enc_priv;
+
+		if (dig->dig_encoder >= 0)
+			dig_enc_in_use |= (1 << dig->dig_encoder);
+	}
+
+	if (radeon_encoder->encoder_id == ENCODER_OBJECT_ID_INTERNAL_KLDSCP_LVTMA) {
+		if (dig_enc_in_use & 0x2)
+			DRM_ERROR("LVDS required digital encoder 2 but it was in use - stealing\n");
+		return 1;
+	}
+	if (!(dig_enc_in_use & 1))
+		return 0;
+	return 1;
 }
 
 static void
@@ -1118,11 +1328,11 @@ radeon_atom_encoder_mode_set(struct drm_encoder *encoder,
 	struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
 	struct radeon_crtc *radeon_crtc = to_radeon_crtc(encoder->crtc);
 
-	if (radeon_encoder->enc_priv) {
-		struct radeon_encoder_atom_dig *dig;
-
-		dig = radeon_encoder->enc_priv;
-		dig->dig_block = radeon_crtc->crtc_id;
+	if (radeon_encoder->active_device &
+	    (ATOM_DEVICE_DFP_SUPPORT | ATOM_DEVICE_LCD_SUPPORT)) {
+		struct radeon_encoder_atom_dig *dig = radeon_encoder->enc_priv;
+		if (dig)
+			dig->dig_encoder = radeon_atom_pick_dig_encoder(encoder);
 	}
 	radeon_encoder->pixel_clock = adjusted_mode->clock;
 
@@ -1147,14 +1357,26 @@ radeon_atom_encoder_mode_set(struct drm_encoder *encoder,
 	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1:
 	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2:
 	case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_LVTMA:
-		/* disable the encoder and transmitter */
-		atombios_dig_transmitter_setup(encoder, ATOM_TRANSMITTER_ACTION_DISABLE);
-		atombios_dig_encoder_setup(encoder, ATOM_DISABLE);
-
-		/* setup and enable the encoder and transmitter */
-		atombios_dig_encoder_setup(encoder, ATOM_ENABLE);
-		atombios_dig_transmitter_setup(encoder, ATOM_TRANSMITTER_ACTION_SETUP);
-		atombios_dig_transmitter_setup(encoder, ATOM_TRANSMITTER_ACTION_ENABLE);
+		if (ASIC_IS_DCE4(rdev)) {
+			/* disable the transmitter */
+			atombios_dig_transmitter_setup(encoder, ATOM_TRANSMITTER_ACTION_DISABLE, 0, 0);
+			/* setup and enable the encoder */
+			atombios_dig_encoder_setup(encoder, ATOM_ENCODER_CMD_SETUP);
+
+			/* init and enable the transmitter */
+			atombios_dig_transmitter_setup(encoder, ATOM_TRANSMITTER_ACTION_INIT, 0, 0);
+			atombios_dig_transmitter_setup(encoder, ATOM_TRANSMITTER_ACTION_ENABLE, 0, 0);
+		} else {
+			/* disable the encoder and transmitter */
+			atombios_dig_transmitter_setup(encoder, ATOM_TRANSMITTER_ACTION_DISABLE, 0, 0);
+			atombios_dig_encoder_setup(encoder, ATOM_DISABLE);
+
+			/* setup and enable the encoder and transmitter */
+			atombios_dig_encoder_setup(encoder, ATOM_ENABLE);
+			atombios_dig_transmitter_setup(encoder, ATOM_TRANSMITTER_ACTION_INIT, 0, 0);
+			atombios_dig_transmitter_setup(encoder, ATOM_TRANSMITTER_ACTION_SETUP, 0, 0);
+			atombios_dig_transmitter_setup(encoder, ATOM_TRANSMITTER_ACTION_ENABLE, 0, 0);
+		}
 		break;
 	case ENCODER_OBJECT_ID_INTERNAL_DDI:
 		atombios_ddia_setup(encoder, ATOM_ENABLE);
@@ -1173,6 +1395,10 @@ radeon_atom_encoder_mode_set(struct drm_encoder *encoder,
 		break;
 	}
 	atombios_apply_encoder_quirks(encoder, adjusted_mode);
+
+	/* XXX */
+	if (!ASIC_IS_DCE4(rdev))
+		r600_hdmi_setmode(encoder, adjusted_mode);
 }
 
 static bool
@@ -1268,8 +1494,6 @@ static void radeon_atom_encoder_prepare(struct drm_encoder *encoder)
 {
 	radeon_atom_output_lock(encoder, true);
 	radeon_atom_encoder_dpms(encoder, DRM_MODE_DPMS_OFF);
-
-	radeon_encoder_set_active_device(encoder);
 }
 
 static void radeon_atom_encoder_commit(struct drm_encoder *encoder)
@@ -1281,7 +1505,13 @@ static void radeon_atom_encoder_commit(struct drm_encoder *encoder)
 static void radeon_atom_encoder_disable(struct drm_encoder *encoder)
 {
 	struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
+	struct radeon_encoder_atom_dig *dig;
 	radeon_atom_encoder_dpms(encoder, DRM_MODE_DPMS_OFF);
+
+	if (radeon_encoder_is_digital(encoder)) {
+		dig = radeon_encoder->enc_priv;
+		dig->dig_encoder = -1;
+	}
 	radeon_encoder->active_device = 0;
 }
 
@@ -1338,6 +1568,7 @@ radeon_atombios_set_dig_info(struct radeon_encoder *radeon_encoder)
 
 	/* coherent mode by default */
 	dig->coherent_mode = true;
+	dig->dig_encoder = -1;
 
 	return dig;
 }
@@ -1345,6 +1576,7 @@ radeon_atombios_set_dig_info(struct radeon_encoder *radeon_encoder)
 void
 radeon_add_atom_encoder(struct drm_device *dev, uint32_t encoder_id, uint32_t supported_device)
 {
+	struct radeon_device *rdev = dev->dev_private;
 	struct drm_encoder *encoder;
 	struct radeon_encoder *radeon_encoder;
 
@@ -1364,8 +1596,18 @@ radeon_add_atom_encoder(struct drm_device *dev, uint32_t encoder_id, uint32_t su
 		return;
 
 	encoder = &radeon_encoder->base;
-	encoder->possible_crtcs = 0x3;
-	encoder->possible_clones = 0;
+	switch (rdev->num_crtc) {
+	case 1:
+		encoder->possible_crtcs = 0x1;
+		break;
+	case 2:
+	default:
+		encoder->possible_crtcs = 0x3;
+		break;
+	case 6:
+		encoder->possible_crtcs = 0x3f;
+		break;
+	}
 
 	radeon_encoder->enc_priv = NULL;
 
@@ -1417,4 +1659,6 @@ radeon_add_atom_encoder(struct drm_device *dev, uint32_t encoder_id, uint32_t su
 		drm_encoder_helper_add(encoder, &radeon_atom_dig_helper_funcs);
 		break;
 	}
+
+	r600_hdmi_init(encoder);
 }
diff --git a/drivers/gpu/drm/radeon/radeon_family.h b/drivers/gpu/drm/radeon/radeon_family.h
index 797972e344a6..93c7d5d41914 100644
--- a/drivers/gpu/drm/radeon/radeon_family.h
+++ b/drivers/gpu/drm/radeon/radeon_family.h
@@ -75,6 +75,11 @@ enum radeon_family {
 	CHIP_RV730,
 	CHIP_RV710,
 	CHIP_RV740,
+	CHIP_CEDAR,
+	CHIP_REDWOOD,
+	CHIP_JUNIPER,
+	CHIP_CYPRESS,
+	CHIP_HEMLOCK,
 	CHIP_LAST,
 };
 
diff --git a/drivers/gpu/drm/radeon/radeon_fb.c b/drivers/gpu/drm/radeon/radeon_fb.c
index 944e4fa78db5..8fccbf29235e 100644
--- a/drivers/gpu/drm/radeon/radeon_fb.c
+++ b/drivers/gpu/drm/radeon/radeon_fb.c
@@ -39,6 +39,8 @@
 
 #include "drm_fb_helper.h"
 
+#include <linux/vga_switcheroo.h>
+
 struct radeon_fb_device {
 	struct drm_fb_helper helper;
 	struct radeon_framebuffer	*rfb;
@@ -55,10 +57,11 @@ static struct fb_ops radeonfb_ops = {
 	.fb_imageblit = cfb_imageblit,
 	.fb_pan_display = drm_fb_helper_pan_display,
 	.fb_blank = drm_fb_helper_blank,
+	.fb_setcmap = drm_fb_helper_setcmap,
 };
 
 /**
- * Curretly it is assumed that the old framebuffer is reused.
+ * Currently it is assumed that the old framebuffer is reused.
  *
  * LOCKING
  * caller should hold the mode config lock.
@@ -123,11 +126,13 @@ static int radeon_align_pitch(struct radeon_device *rdev, int width, int bpp, bo
 
 static struct drm_fb_helper_funcs radeon_fb_helper_funcs = {
 	.gamma_set = radeon_crtc_fb_gamma_set,
+	.gamma_get = radeon_crtc_fb_gamma_get,
 };
 
 int radeonfb_create(struct drm_device *dev,
 		    uint32_t fb_width, uint32_t fb_height,
 		    uint32_t surface_width, uint32_t surface_height,
+		    uint32_t surface_depth, uint32_t surface_bpp,
 		    struct drm_framebuffer **fb_p)
 {
 	struct radeon_device *rdev = dev->dev_private;
@@ -137,7 +142,7 @@ int radeonfb_create(struct drm_device *dev,
 	struct radeon_framebuffer *rfb;
 	struct drm_mode_fb_cmd mode_cmd;
 	struct drm_gem_object *gobj = NULL;
-	struct radeon_object *robj = NULL;
+	struct radeon_bo *rbo = NULL;
 	struct device *device = &rdev->pdev->dev;
 	int size, aligned_size, ret;
 	u64 fb_gpuaddr;
@@ -148,10 +153,15 @@ int radeonfb_create(struct drm_device *dev,
 
 	mode_cmd.width = surface_width;
 	mode_cmd.height = surface_height;
-	mode_cmd.bpp = 32;
+
+	/* avivo can't scanout real 24bpp */
+	if ((surface_bpp == 24) && ASIC_IS_AVIVO(rdev))
+		surface_bpp = 32;
+
+	mode_cmd.bpp = surface_bpp;
 	/* need to align pitch with crtc limits */
 	mode_cmd.pitch = radeon_align_pitch(rdev, mode_cmd.width, mode_cmd.bpp, fb_tiled) * ((mode_cmd.bpp + 1) / 8);
-	mode_cmd.depth = 24;
+	mode_cmd.depth = surface_depth;
 
 	size = mode_cmd.pitch * mode_cmd.height;
 	aligned_size = ALIGN(size, PAGE_SIZE);
@@ -159,14 +169,14 @@ int radeonfb_create(struct drm_device *dev,
 	ret = radeon_gem_object_create(rdev, aligned_size, 0,
 			RADEON_GEM_DOMAIN_VRAM,
 			false, ttm_bo_type_kernel,
-			false, &gobj);
+			&gobj);
 	if (ret) {
 		printk(KERN_ERR "failed to allocate framebuffer (%d %d)\n",
 		       surface_width, surface_height);
 		ret = -ENOMEM;
 		goto out;
 	}
-	robj = gobj->driver_private;
+	rbo = gobj->driver_private;
 
 	if (fb_tiled)
 		tiling_flags = RADEON_TILING_MACRO;
@@ -183,8 +193,13 @@ int radeonfb_create(struct drm_device *dev,
 	}
 #endif
 
-	if (tiling_flags)
-		radeon_object_set_tiling_flags(robj, tiling_flags | RADEON_TILING_SURFACE, mode_cmd.pitch);
+	if (tiling_flags) {
+		ret = radeon_bo_set_tiling_flags(rbo,
+					tiling_flags | RADEON_TILING_SURFACE,
+					mode_cmd.pitch);
+		if (ret)
+			dev_err(rdev->dev, "FB failed to set tiling flags\n");
+	}
 	mutex_lock(&rdev->ddev->struct_mutex);
 	fb = radeon_framebuffer_create(rdev->ddev, &mode_cmd, gobj);
 	if (fb == NULL) {
@@ -192,10 +207,19 @@ int radeonfb_create(struct drm_device *dev,
 		ret = -ENOMEM;
 		goto out_unref;
 	}
-	ret = radeon_object_pin(robj, RADEON_GEM_DOMAIN_VRAM, &fb_gpuaddr);
+	ret = radeon_bo_reserve(rbo, false);
+	if (unlikely(ret != 0))
+		goto out_unref;
+	ret = radeon_bo_pin(rbo, RADEON_GEM_DOMAIN_VRAM, &fb_gpuaddr);
+	if (ret) {
+		radeon_bo_unreserve(rbo);
+		goto out_unref;
+	}
+	if (fb_tiled)
+		radeon_bo_check_tiling(rbo, 0, 0);
+	ret = radeon_bo_kmap(rbo, &fbptr);
+	radeon_bo_unreserve(rbo);
 	if (ret) {
-		printk(KERN_ERR "failed to pin framebuffer\n");
-		ret = -ENOMEM;
 		goto out_unref;
 	}
 
@@ -204,7 +228,7 @@ int radeonfb_create(struct drm_device *dev,
 	*fb_p = fb;
 	rfb = to_radeon_framebuffer(fb);
 	rdev->fbdev_rfb = rfb;
-	rdev->fbdev_robj = robj;
+	rdev->fbdev_rbo = rbo;
 
 	info = framebuffer_alloc(sizeof(struct radeon_fb_device), device);
 	if (info == NULL) {
@@ -216,29 +240,21 @@ int radeonfb_create(struct drm_device *dev,
 	rfbdev = info->par;
 	rfbdev->helper.funcs = &radeon_fb_helper_funcs;
 	rfbdev->helper.dev = dev;
-	ret = drm_fb_helper_init_crtc_count(&rfbdev->helper, 2,
+	ret = drm_fb_helper_init_crtc_count(&rfbdev->helper, rdev->num_crtc,
 					    RADEONFB_CONN_LIMIT);
 	if (ret)
 		goto out_unref;
 
-	if (fb_tiled)
-		radeon_object_check_tiling(robj, 0, 0);
-
-	ret = radeon_object_kmap(robj, &fbptr);
-	if (ret) {
-		goto out_unref;
-	}
-
-	memset_io(fbptr, 0, aligned_size);
+	memset_io(fbptr, 0x0, aligned_size);
 
 	strcpy(info->fix.id, "radeondrmfb");
 
-	drm_fb_helper_fill_fix(info, fb->pitch);
+	drm_fb_helper_fill_fix(info, fb->pitch, fb->depth);
 
 	info->flags = FBINFO_DEFAULT;
 	info->fbops = &radeonfb_ops;
 
-	tmp = fb_gpuaddr - rdev->mc.vram_location;
+	tmp = fb_gpuaddr - rdev->mc.vram_start;
 	info->fix.smem_start = rdev->mc.aper_base + tmp;
 	info->fix.smem_len = size;
 	info->screen_base = fbptr;
@@ -272,11 +288,16 @@ int radeonfb_create(struct drm_device *dev,
 	rfbdev->rdev = rdev;
 
 	mutex_unlock(&rdev->ddev->struct_mutex);
+	vga_switcheroo_client_fb_set(rdev->ddev->pdev, info);
 	return 0;
 
 out_unref:
-	if (robj) {
-		radeon_object_kunmap(robj);
+	if (rbo) {
+		ret = radeon_bo_reserve(rbo, false);
+		if (likely(ret == 0)) {
+			radeon_bo_kunmap(rbo);
+			radeon_bo_unreserve(rbo);
+		}
 	}
 	if (fb && ret) {
 		list_del(&fb->filp_head);
@@ -290,19 +311,40 @@ out:
 	return ret;
 }
 
+static char *mode_option;
+int radeon_parse_options(char *options)
+{
+	char *this_opt;
+
+	if (!options || !*options)
+		return 0;
+
+	while ((this_opt = strsep(&options, ",")) != NULL) {
+		if (!*this_opt)
+			continue;
+		mode_option = this_opt;
+	}
+	return 0;
+}
+
 int radeonfb_probe(struct drm_device *dev)
 {
-	int ret;
-	ret = drm_fb_helper_single_fb_probe(dev, &radeonfb_create);
-	return ret;
+	struct radeon_device *rdev = dev->dev_private;
+	int bpp_sel = 32;
+
+	/* select 8 bpp console on RN50 or 16MB cards */
+	if (ASIC_IS_RN50(rdev) || rdev->mc.real_vram_size <= (32*1024*1024))
+		bpp_sel = 8;
+
+	return drm_fb_helper_single_fb_probe(dev, bpp_sel, &radeonfb_create);
 }
-EXPORT_SYMBOL(radeonfb_probe);
 
 int radeonfb_remove(struct drm_device *dev, struct drm_framebuffer *fb)
 {
 	struct fb_info *info;
 	struct radeon_framebuffer *rfb = to_radeon_framebuffer(fb);
-	struct radeon_object *robj;
+	struct radeon_bo *rbo;
+	int r;
 
 	if (!fb) {
 		return -EINVAL;
@@ -310,10 +352,14 @@ int radeonfb_remove(struct drm_device *dev, struct drm_framebuffer *fb)
 	info = fb->fbdev;
 	if (info) {
 		struct radeon_fb_device *rfbdev = info->par;
-		robj = rfb->obj->driver_private;
+		rbo = rfb->obj->driver_private;
 		unregister_framebuffer(info);
-		radeon_object_kunmap(robj);
-		radeon_object_unpin(robj);
+		r = radeon_bo_reserve(rbo, false);
+		if (likely(r == 0)) {
+			radeon_bo_kunmap(rbo);
+			radeon_bo_unpin(rbo);
+			radeon_bo_unreserve(rbo);
+		}
 		drm_fb_helper_free(&rfbdev->helper);
 		framebuffer_release(info);
 	}
diff --git a/drivers/gpu/drm/radeon/radeon_fence.c b/drivers/gpu/drm/radeon/radeon_fence.c
index 3beb26d74719..8495d4e32e18 100644
--- a/drivers/gpu/drm/radeon/radeon_fence.c
+++ b/drivers/gpu/drm/radeon/radeon_fence.c
@@ -140,16 +140,15 @@ int radeon_fence_create(struct radeon_device *rdev, struct radeon_fence **fence)
 
 bool radeon_fence_signaled(struct radeon_fence *fence)
 {
-	struct radeon_device *rdev = fence->rdev;
 	unsigned long irq_flags;
 	bool signaled = false;
 
-	if (rdev->gpu_lockup) {
+	if (!fence)
 		return true;
-	}
-	if (fence == NULL) {
+
+	if (fence->rdev->gpu_lockup)
 		return true;
-	}
+
 	write_lock_irqsave(&fence->rdev->fence_drv.lock, irq_flags);
 	signaled = fence->signaled;
 	/* if we are shuting down report all fence as signaled */
@@ -168,37 +167,6 @@ bool radeon_fence_signaled(struct radeon_fence *fence)
 	return signaled;
 }
 
-int r600_fence_wait(struct radeon_fence *fence,  bool intr, bool lazy)
-{
-	struct radeon_device *rdev;
-	int ret = 0;
-
-	rdev = fence->rdev;
-
-	__set_current_state(intr ? TASK_INTERRUPTIBLE : TASK_UNINTERRUPTIBLE);
-
-	while (1) {
-		if (radeon_fence_signaled(fence))
-			break;
-
-		if (time_after_eq(jiffies, fence->timeout)) {
-			ret = -EBUSY;
-			break;
-		}
-
-		if (lazy)
-			schedule_timeout(1);
-
-		if (intr && signal_pending(current)) {
-			ret = -ERESTARTSYS;
-			break;
-		}
-	}
-	__set_current_state(TASK_RUNNING);
-	return ret;
-}
-
-
 int radeon_fence_wait(struct radeon_fence *fence, bool intr)
 {
 	struct radeon_device *rdev;
@@ -216,13 +184,6 @@ int radeon_fence_wait(struct radeon_fence *fence, bool intr)
 		return 0;
 	}
 
-	if (rdev->family >= CHIP_R600) {
-		r = r600_fence_wait(fence, intr, 0);
-		if (r == -ERESTARTSYS)
-			return -EBUSY;
-		return r;
-	}
-
 retry:
 	cur_jiffies = jiffies;
 	timeout = HZ / 100;
@@ -231,14 +192,17 @@ retry:
 	}
 
 	if (intr) {
+		radeon_irq_kms_sw_irq_get(rdev);
 		r = wait_event_interruptible_timeout(rdev->fence_drv.queue,
 				radeon_fence_signaled(fence), timeout);
-		if (unlikely(r == -ERESTARTSYS)) {
-			return -EBUSY;
-		}
+		radeon_irq_kms_sw_irq_put(rdev);
+		if (unlikely(r < 0))
+			return r;
 	} else {
+		radeon_irq_kms_sw_irq_get(rdev);
 		r = wait_event_timeout(rdev->fence_drv.queue,
 			 radeon_fence_signaled(fence), timeout);
+		radeon_irq_kms_sw_irq_put(rdev);
 	}
 	if (unlikely(!radeon_fence_signaled(fence))) {
 		if (unlikely(r == 0)) {
@@ -359,7 +323,7 @@ int radeon_fence_driver_init(struct radeon_device *rdev)
 	write_lock_irqsave(&rdev->fence_drv.lock, irq_flags);
 	r = radeon_scratch_get(rdev, &rdev->fence_drv.scratch_reg);
 	if (r) {
-		DRM_ERROR("Fence failed to get a scratch register.");
+		dev_err(rdev->dev, "fence failed to get scratch register\n");
 		write_unlock_irqrestore(&rdev->fence_drv.lock, irq_flags);
 		return r;
 	}
@@ -370,9 +334,10 @@ int radeon_fence_driver_init(struct radeon_device *rdev)
 	INIT_LIST_HEAD(&rdev->fence_drv.signaled);
 	rdev->fence_drv.count_timeout = 0;
 	init_waitqueue_head(&rdev->fence_drv.queue);
+	rdev->fence_drv.initialized = true;
 	write_unlock_irqrestore(&rdev->fence_drv.lock, irq_flags);
 	if (radeon_debugfs_fence_init(rdev)) {
-		DRM_ERROR("Failed to register debugfs file for fence !\n");
+		dev_err(rdev->dev, "fence debugfs file creation failed\n");
 	}
 	return 0;
 }
@@ -381,11 +346,13 @@ void radeon_fence_driver_fini(struct radeon_device *rdev)
 {
 	unsigned long irq_flags;
 
+	if (!rdev->fence_drv.initialized)
+		return;
 	wake_up_all(&rdev->fence_drv.queue);
 	write_lock_irqsave(&rdev->fence_drv.lock, irq_flags);
 	radeon_scratch_free(rdev, rdev->fence_drv.scratch_reg);
 	write_unlock_irqrestore(&rdev->fence_drv.lock, irq_flags);
-	DRM_INFO("radeon: fence finalized\n");
+	rdev->fence_drv.initialized = false;
 }
 
 
diff --git a/drivers/gpu/drm/radeon/radeon_fixed.h b/drivers/gpu/drm/radeon/radeon_fixed.h
index 90187d173847..3d4d84e078ac 100644
--- a/drivers/gpu/drm/radeon/radeon_fixed.h
+++ b/drivers/gpu/drm/radeon/radeon_fixed.h
@@ -38,6 +38,23 @@ typedef union rfixed {
 #define fixed_init_half(A) { .full = rfixed_const_half((A)) }
 #define rfixed_trunc(A) ((A).full >> 12)
 
+static inline u32 rfixed_floor(fixed20_12 A)
+{
+	u32 non_frac = rfixed_trunc(A);
+
+	return rfixed_const(non_frac);
+}
+
+static inline u32 rfixed_ceil(fixed20_12 A)
+{
+	u32 non_frac = rfixed_trunc(A);
+
+	if (A.full > rfixed_const(non_frac))
+		return rfixed_const(non_frac + 1);
+	else
+		return rfixed_const(non_frac);
+}
+
 static inline u32 rfixed_div(fixed20_12 A, fixed20_12 B)
 {
 	u64 tmp = ((u64)A.full << 13);
diff --git a/drivers/gpu/drm/radeon/radeon_gart.c b/drivers/gpu/drm/radeon/radeon_gart.c
index a931af065dd4..1770d3c07fd0 100644
--- a/drivers/gpu/drm/radeon/radeon_gart.c
+++ b/drivers/gpu/drm/radeon/radeon_gart.c
@@ -78,11 +78,9 @@ int radeon_gart_table_vram_alloc(struct radeon_device *rdev)
 	int r;
 
 	if (rdev->gart.table.vram.robj == NULL) {
-		r = radeon_object_create(rdev, NULL,
-					 rdev->gart.table_size,
-					 true,
-					 RADEON_GEM_DOMAIN_VRAM,
-					 false, &rdev->gart.table.vram.robj);
+		r = radeon_bo_create(rdev, NULL, rdev->gart.table_size,
+					true, RADEON_GEM_DOMAIN_VRAM,
+					&rdev->gart.table.vram.robj);
 		if (r) {
 			return r;
 		}
@@ -95,32 +93,38 @@ int radeon_gart_table_vram_pin(struct radeon_device *rdev)
 	uint64_t gpu_addr;
 	int r;
 
-	r = radeon_object_pin(rdev->gart.table.vram.robj,
-			      RADEON_GEM_DOMAIN_VRAM, &gpu_addr);
-	if (r) {
-		radeon_object_unref(&rdev->gart.table.vram.robj);
+	r = radeon_bo_reserve(rdev->gart.table.vram.robj, false);
+	if (unlikely(r != 0))
 		return r;
-	}
-	r = radeon_object_kmap(rdev->gart.table.vram.robj,
-			       (void **)&rdev->gart.table.vram.ptr);
+	r = radeon_bo_pin(rdev->gart.table.vram.robj,
+				RADEON_GEM_DOMAIN_VRAM, &gpu_addr);
 	if (r) {
-		radeon_object_unpin(rdev->gart.table.vram.robj);
-		radeon_object_unref(&rdev->gart.table.vram.robj);
-		DRM_ERROR("radeon: failed to map gart vram table.\n");
+		radeon_bo_unreserve(rdev->gart.table.vram.robj);
 		return r;
 	}
+	r = radeon_bo_kmap(rdev->gart.table.vram.robj,
+				(void **)&rdev->gart.table.vram.ptr);
+	if (r)
+		radeon_bo_unpin(rdev->gart.table.vram.robj);
+	radeon_bo_unreserve(rdev->gart.table.vram.robj);
 	rdev->gart.table_addr = gpu_addr;
-	return 0;
+	return r;
 }
 
 void radeon_gart_table_vram_free(struct radeon_device *rdev)
 {
+	int r;
+
 	if (rdev->gart.table.vram.robj == NULL) {
 		return;
 	}
-	radeon_object_kunmap(rdev->gart.table.vram.robj);
-	radeon_object_unpin(rdev->gart.table.vram.robj);
-	radeon_object_unref(&rdev->gart.table.vram.robj);
+	r = radeon_bo_reserve(rdev->gart.table.vram.robj, false);
+	if (likely(r == 0)) {
+		radeon_bo_kunmap(rdev->gart.table.vram.robj);
+		radeon_bo_unpin(rdev->gart.table.vram.robj);
+		radeon_bo_unreserve(rdev->gart.table.vram.robj);
+	}
+	radeon_bo_unref(&rdev->gart.table.vram.robj);
 }
 
 
@@ -135,21 +139,24 @@ void radeon_gart_unbind(struct radeon_device *rdev, unsigned offset,
 	unsigned t;
 	unsigned p;
 	int i, j;
+	u64 page_base;
 
 	if (!rdev->gart.ready) {
 		WARN(1, "trying to unbind memory to unitialized GART !\n");
 		return;
 	}
-	t = offset / 4096;
-	p = t / (PAGE_SIZE / 4096);
+	t = offset / RADEON_GPU_PAGE_SIZE;
+	p = t / (PAGE_SIZE / RADEON_GPU_PAGE_SIZE);
 	for (i = 0; i < pages; i++, p++) {
 		if (rdev->gart.pages[p]) {
 			pci_unmap_page(rdev->pdev, rdev->gart.pages_addr[p],
 				       PAGE_SIZE, PCI_DMA_BIDIRECTIONAL);
 			rdev->gart.pages[p] = NULL;
-			rdev->gart.pages_addr[p] = 0;
-			for (j = 0; j < (PAGE_SIZE / 4096); j++, t++) {
-				radeon_gart_set_page(rdev, t, 0);
+			rdev->gart.pages_addr[p] = rdev->dummy_page.addr;
+			page_base = rdev->gart.pages_addr[p];
+			for (j = 0; j < (PAGE_SIZE / RADEON_GPU_PAGE_SIZE); j++, t++) {
+				radeon_gart_set_page(rdev, t, page_base);
+				page_base += RADEON_GPU_PAGE_SIZE;
 			}
 		}
 	}
@@ -169,8 +176,8 @@ int radeon_gart_bind(struct radeon_device *rdev, unsigned offset,
 		DRM_ERROR("trying to bind memory to unitialized GART !\n");
 		return -EINVAL;
 	}
-	t = offset / 4096;
-	p = t / (PAGE_SIZE / 4096);
+	t = offset / RADEON_GPU_PAGE_SIZE;
+	p = t / (PAGE_SIZE / RADEON_GPU_PAGE_SIZE);
 
 	for (i = 0; i < pages; i++, p++) {
 		/* we need to support large memory configurations */
@@ -185,9 +192,9 @@ int radeon_gart_bind(struct radeon_device *rdev, unsigned offset,
 		}
 		rdev->gart.pages[p] = pagelist[i];
 		page_base = rdev->gart.pages_addr[p];
-		for (j = 0; j < (PAGE_SIZE / 4096); j++, t++) {
+		for (j = 0; j < (PAGE_SIZE / RADEON_GPU_PAGE_SIZE); j++, t++) {
 			radeon_gart_set_page(rdev, t, page_base);
-			page_base += 4096;
+			page_base += RADEON_GPU_PAGE_SIZE;
 		}
 	}
 	mb();
@@ -195,19 +202,40 @@ int radeon_gart_bind(struct radeon_device *rdev, unsigned offset,
 	return 0;
 }
 
+void radeon_gart_restore(struct radeon_device *rdev)
+{
+	int i, j, t;
+	u64 page_base;
+
+	for (i = 0, t = 0; i < rdev->gart.num_cpu_pages; i++) {
+		page_base = rdev->gart.pages_addr[i];
+		for (j = 0; j < (PAGE_SIZE / RADEON_GPU_PAGE_SIZE); j++, t++) {
+			radeon_gart_set_page(rdev, t, page_base);
+			page_base += RADEON_GPU_PAGE_SIZE;
+		}
+	}
+	mb();
+	radeon_gart_tlb_flush(rdev);
+}
+
 int radeon_gart_init(struct radeon_device *rdev)
 {
+	int r, i;
+
 	if (rdev->gart.pages) {
 		return 0;
 	}
-	/* We need PAGE_SIZE >= 4096 */
-	if (PAGE_SIZE < 4096) {
+	/* We need PAGE_SIZE >= RADEON_GPU_PAGE_SIZE */
+	if (PAGE_SIZE < RADEON_GPU_PAGE_SIZE) {
 		DRM_ERROR("Page size is smaller than GPU page size!\n");
 		return -EINVAL;
 	}
+	r = radeon_dummy_page_init(rdev);
+	if (r)
+		return r;
 	/* Compute table size */
 	rdev->gart.num_cpu_pages = rdev->mc.gtt_size / PAGE_SIZE;
-	rdev->gart.num_gpu_pages = rdev->mc.gtt_size / 4096;
+	rdev->gart.num_gpu_pages = rdev->mc.gtt_size / RADEON_GPU_PAGE_SIZE;
 	DRM_INFO("GART: num cpu pages %u, num gpu pages %u\n",
 		 rdev->gart.num_cpu_pages, rdev->gart.num_gpu_pages);
 	/* Allocate pages table */
@@ -223,6 +251,10 @@ int radeon_gart_init(struct radeon_device *rdev)
 		radeon_gart_fini(rdev);
 		return -ENOMEM;
 	}
+	/* set GART entry to point to the dummy page by default */
+	for (i = 0; i < rdev->gart.num_cpu_pages; i++) {
+		rdev->gart.pages_addr[i] = rdev->dummy_page.addr;
+	}
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/radeon/radeon_gem.c b/drivers/gpu/drm/radeon/radeon_gem.c
index d880edf254db..ef92d147d8f0 100644
--- a/drivers/gpu/drm/radeon/radeon_gem.c
+++ b/drivers/gpu/drm/radeon/radeon_gem.c
@@ -38,22 +38,21 @@ int radeon_gem_object_init(struct drm_gem_object *obj)
 
 void radeon_gem_object_free(struct drm_gem_object *gobj)
 {
-	struct radeon_object *robj = gobj->driver_private;
+	struct radeon_bo *robj = gobj->driver_private;
 
 	gobj->driver_private = NULL;
 	if (robj) {
-		radeon_object_unref(&robj);
+		radeon_bo_unref(&robj);
 	}
 }
 
 int radeon_gem_object_create(struct radeon_device *rdev, int size,
-			     int alignment, int initial_domain,
-			     bool discardable, bool kernel,
-			     bool interruptible,
-			     struct drm_gem_object **obj)
+				int alignment, int initial_domain,
+				bool discardable, bool kernel,
+				struct drm_gem_object **obj)
 {
 	struct drm_gem_object *gobj;
-	struct radeon_object *robj;
+	struct radeon_bo *robj;
 	int r;
 
 	*obj = NULL;
@@ -65,14 +64,12 @@ int radeon_gem_object_create(struct radeon_device *rdev, int size,
 	if (alignment < PAGE_SIZE) {
 		alignment = PAGE_SIZE;
 	}
-	r = radeon_object_create(rdev, gobj, size, kernel, initial_domain,
-				 interruptible, &robj);
+	r = radeon_bo_create(rdev, gobj, size, kernel, initial_domain, &robj);
 	if (r) {
-		DRM_ERROR("Failed to allocate GEM object (%d, %d, %u)\n",
-			  size, initial_domain, alignment);
-		mutex_lock(&rdev->ddev->struct_mutex);
-		drm_gem_object_unreference(gobj);
-		mutex_unlock(&rdev->ddev->struct_mutex);
+		if (r != -ERESTARTSYS)
+			DRM_ERROR("Failed to allocate GEM object (%d, %d, %u, %d)\n",
+				  size, initial_domain, alignment, r);
+		drm_gem_object_unreference_unlocked(gobj);
 		return r;
 	}
 	gobj->driver_private = robj;
@@ -83,33 +80,33 @@ int radeon_gem_object_create(struct radeon_device *rdev, int size,
 int radeon_gem_object_pin(struct drm_gem_object *obj, uint32_t pin_domain,
 			  uint64_t *gpu_addr)
 {
-	struct radeon_object *robj = obj->driver_private;
-	uint32_t flags;
+	struct radeon_bo *robj = obj->driver_private;
+	int r;
 
-	switch (pin_domain) {
-	case RADEON_GEM_DOMAIN_VRAM:
-		flags = TTM_PL_FLAG_VRAM;
-		break;
-	case RADEON_GEM_DOMAIN_GTT:
-		flags = TTM_PL_FLAG_TT;
-		break;
-	default:
-		flags = TTM_PL_FLAG_SYSTEM;
-		break;
-	}
-	return radeon_object_pin(robj, flags, gpu_addr);
+	r = radeon_bo_reserve(robj, false);
+	if (unlikely(r != 0))
+		return r;
+	r = radeon_bo_pin(robj, pin_domain, gpu_addr);
+	radeon_bo_unreserve(robj);
+	return r;
 }
 
 void radeon_gem_object_unpin(struct drm_gem_object *obj)
 {
-	struct radeon_object *robj = obj->driver_private;
-	radeon_object_unpin(robj);
+	struct radeon_bo *robj = obj->driver_private;
+	int r;
+
+	r = radeon_bo_reserve(robj, false);
+	if (likely(r == 0)) {
+		radeon_bo_unpin(robj);
+		radeon_bo_unreserve(robj);
+	}
 }
 
 int radeon_gem_set_domain(struct drm_gem_object *gobj,
 			  uint32_t rdomain, uint32_t wdomain)
 {
-	struct radeon_object *robj;
+	struct radeon_bo *robj;
 	uint32_t domain;
 	int r;
 
@@ -127,7 +124,7 @@ int radeon_gem_set_domain(struct drm_gem_object *gobj,
 	}
 	if (domain == RADEON_GEM_DOMAIN_CPU) {
 		/* Asking for cpu access wait for object idle */
-		r = radeon_object_wait(robj);
+		r = radeon_bo_wait(robj, NULL, false);
 		if (r) {
 			printk(KERN_ERR "Failed to wait for object !\n");
 			return r;
@@ -144,7 +141,7 @@ int radeon_gem_init(struct radeon_device *rdev)
 
 void radeon_gem_fini(struct radeon_device *rdev)
 {
-	radeon_object_force_delete(rdev);
+	radeon_bo_force_delete(rdev);
 }
 
 
@@ -158,9 +155,13 @@ int radeon_gem_info_ioctl(struct drm_device *dev, void *data,
 	struct drm_radeon_gem_info *args = data;
 
 	args->vram_size = rdev->mc.real_vram_size;
-	/* FIXME: report somethings that makes sense */
-	args->vram_visible = rdev->mc.real_vram_size - (4 * 1024 * 1024);
-	args->gart_size = rdev->mc.gtt_size;
+	args->vram_visible = rdev->mc.real_vram_size;
+	if (rdev->stollen_vga_memory)
+		args->vram_visible -= radeon_bo_size(rdev->stollen_vga_memory);
+	if (rdev->fbdev_rbo)
+		args->vram_visible -= radeon_bo_size(rdev->fbdev_rbo);
+	args->gart_size = rdev->mc.gtt_size - rdev->cp.ring_size - 4096 -
+		RADEON_IB_POOL_SIZE*64*1024;
 	return 0;
 }
 
@@ -192,21 +193,17 @@ int radeon_gem_create_ioctl(struct drm_device *dev, void *data,
 	/* create a gem object to contain this object in */
 	args->size = roundup(args->size, PAGE_SIZE);
 	r = radeon_gem_object_create(rdev, args->size, args->alignment,
-				     args->initial_domain, false,
-				     false, true, &gobj);
+					args->initial_domain, false,
+					false, &gobj);
 	if (r) {
 		return r;
 	}
 	r = drm_gem_handle_create(filp, gobj, &handle);
 	if (r) {
-		mutex_lock(&dev->struct_mutex);
-		drm_gem_object_unreference(gobj);
-		mutex_unlock(&dev->struct_mutex);
+		drm_gem_object_unreference_unlocked(gobj);
 		return r;
 	}
-	mutex_lock(&dev->struct_mutex);
-	drm_gem_object_handle_unreference(gobj);
-	mutex_unlock(&dev->struct_mutex);
+	drm_gem_object_handle_unreference_unlocked(gobj);
 	args->handle = handle;
 	return 0;
 }
@@ -218,7 +215,7 @@ int radeon_gem_set_domain_ioctl(struct drm_device *dev, void *data,
 	 * just validate the BO into a certain domain */
 	struct drm_radeon_gem_set_domain *args = data;
 	struct drm_gem_object *gobj;
-	struct radeon_object *robj;
+	struct radeon_bo *robj;
 	int r;
 
 	/* for now if someone requests domain CPU -
@@ -233,9 +230,7 @@ int radeon_gem_set_domain_ioctl(struct drm_device *dev, void *data,
 
 	r = radeon_gem_set_domain(gobj, args->read_domains, args->write_domain);
 
-	mutex_lock(&dev->struct_mutex);
-	drm_gem_object_unreference(gobj);
-	mutex_unlock(&dev->struct_mutex);
+	drm_gem_object_unreference_unlocked(gobj);
 	return r;
 }
 
@@ -244,19 +239,16 @@ int radeon_gem_mmap_ioctl(struct drm_device *dev, void *data,
 {
 	struct drm_radeon_gem_mmap *args = data;
 	struct drm_gem_object *gobj;
-	struct radeon_object *robj;
-	int r;
+	struct radeon_bo *robj;
 
 	gobj = drm_gem_object_lookup(dev, filp, args->handle);
 	if (gobj == NULL) {
 		return -EINVAL;
 	}
 	robj = gobj->driver_private;
-	r = radeon_object_mmap(robj, &args->addr_ptr);
-	mutex_lock(&dev->struct_mutex);
-	drm_gem_object_unreference(gobj);
-	mutex_unlock(&dev->struct_mutex);
-	return r;
+	args->addr_ptr = radeon_bo_mmap_offset(robj);
+	drm_gem_object_unreference_unlocked(gobj);
+	return 0;
 }
 
 int radeon_gem_busy_ioctl(struct drm_device *dev, void *data,
@@ -264,16 +256,16 @@ int radeon_gem_busy_ioctl(struct drm_device *dev, void *data,
 {
 	struct drm_radeon_gem_busy *args = data;
 	struct drm_gem_object *gobj;
-	struct radeon_object *robj;
+	struct radeon_bo *robj;
 	int r;
-	uint32_t cur_placement;
+	uint32_t cur_placement = 0;
 
 	gobj = drm_gem_object_lookup(dev, filp, args->handle);
 	if (gobj == NULL) {
 		return -EINVAL;
 	}
 	robj = gobj->driver_private;
-	r = radeon_object_busy_domain(robj, &cur_placement);
+	r = radeon_bo_wait(robj, &cur_placement, true);
 	switch (cur_placement) {
 	case TTM_PL_VRAM:
 		args->domain = RADEON_GEM_DOMAIN_VRAM;
@@ -286,9 +278,7 @@ int radeon_gem_busy_ioctl(struct drm_device *dev, void *data,
 	default:
 		break;
 	}
-	mutex_lock(&dev->struct_mutex);
-	drm_gem_object_unreference(gobj);
-	mutex_unlock(&dev->struct_mutex);
+	drm_gem_object_unreference_unlocked(gobj);
 	return r;
 }
 
@@ -297,7 +287,7 @@ int radeon_gem_wait_idle_ioctl(struct drm_device *dev, void *data,
 {
 	struct drm_radeon_gem_wait_idle *args = data;
 	struct drm_gem_object *gobj;
-	struct radeon_object *robj;
+	struct radeon_bo *robj;
 	int r;
 
 	gobj = drm_gem_object_lookup(dev, filp, args->handle);
@@ -305,10 +295,11 @@ int radeon_gem_wait_idle_ioctl(struct drm_device *dev, void *data,
 		return -EINVAL;
 	}
 	robj = gobj->driver_private;
-	r = radeon_object_wait(robj);
-	mutex_lock(&dev->struct_mutex);
-	drm_gem_object_unreference(gobj);
-	mutex_unlock(&dev->struct_mutex);
+	r = radeon_bo_wait(robj, NULL, false);
+	/* callback hw specific functions if any */
+	if (robj->rdev->asic->ioctl_wait_idle)
+		robj->rdev->asic->ioctl_wait_idle(robj->rdev, robj);
+	drm_gem_object_unreference_unlocked(gobj);
 	return r;
 }
 
@@ -317,7 +308,7 @@ int radeon_gem_set_tiling_ioctl(struct drm_device *dev, void *data,
 {
 	struct drm_radeon_gem_set_tiling *args = data;
 	struct drm_gem_object *gobj;
-	struct radeon_object *robj;
+	struct radeon_bo *robj;
 	int r = 0;
 
 	DRM_DEBUG("%d \n", args->handle);
@@ -325,10 +316,8 @@ int radeon_gem_set_tiling_ioctl(struct drm_device *dev, void *data,
 	if (gobj == NULL)
 		return -EINVAL;
 	robj = gobj->driver_private;
-	radeon_object_set_tiling_flags(robj, args->tiling_flags, args->pitch);
-	mutex_lock(&dev->struct_mutex);
-	drm_gem_object_unreference(gobj);
-	mutex_unlock(&dev->struct_mutex);
+	r = radeon_bo_set_tiling_flags(robj, args->tiling_flags, args->pitch);
+	drm_gem_object_unreference_unlocked(gobj);
 	return r;
 }
 
@@ -337,18 +326,20 @@ int radeon_gem_get_tiling_ioctl(struct drm_device *dev, void *data,
 {
 	struct drm_radeon_gem_get_tiling *args = data;
 	struct drm_gem_object *gobj;
-	struct radeon_object *robj;
+	struct radeon_bo *rbo;
 	int r = 0;
 
 	DRM_DEBUG("\n");
 	gobj = drm_gem_object_lookup(dev, filp, args->handle);
 	if (gobj == NULL)
 		return -EINVAL;
-	robj = gobj->driver_private;
-	radeon_object_get_tiling_flags(robj, &args->tiling_flags,
-				       &args->pitch);
-	mutex_lock(&dev->struct_mutex);
-	drm_gem_object_unreference(gobj);
-	mutex_unlock(&dev->struct_mutex);
+	rbo = gobj->driver_private;
+	r = radeon_bo_reserve(rbo, false);
+	if (unlikely(r != 0))
+		goto out;
+	radeon_bo_get_tiling_flags(rbo, &args->tiling_flags, &args->pitch);
+	radeon_bo_unreserve(rbo);
+out:
+	drm_gem_object_unreference_unlocked(gobj);
 	return r;
 }
diff --git a/drivers/gpu/drm/radeon/radeon_i2c.c b/drivers/gpu/drm/radeon/radeon_i2c.c
index dd438d32e5c0..4ae50c19589f 100644
--- a/drivers/gpu/drm/radeon/radeon_i2c.c
+++ b/drivers/gpu/drm/radeon/radeon_i2c.c
@@ -26,6 +26,7 @@
 #include "drmP.h"
 #include "radeon_drm.h"
 #include "radeon.h"
+#include "atom.h"
 
 /**
  * radeon_ddc_probe
@@ -59,35 +60,55 @@ bool radeon_ddc_probe(struct radeon_connector *radeon_connector)
 }
 
 
-void radeon_i2c_do_lock(struct radeon_connector *radeon_connector, int lock_state)
+static void radeon_i2c_do_lock(struct radeon_i2c_chan *i2c, int lock_state)
 {
-	struct radeon_device *rdev = radeon_connector->base.dev->dev_private;
+	struct radeon_device *rdev = i2c->dev->dev_private;
+	struct radeon_i2c_bus_rec *rec = &i2c->rec;
 	uint32_t temp;
-	struct radeon_i2c_bus_rec *rec = &radeon_connector->ddc_bus->rec;
 
 	/* RV410 appears to have a bug where the hw i2c in reset
 	 * holds the i2c port in a bad state - switch hw i2c away before
 	 * doing DDC - do this for all r200s/r300s/r400s for safety sake
 	 */
-	if ((rdev->family >= CHIP_R200) && !ASIC_IS_AVIVO(rdev)) {
-		if (rec->a_clk_reg == RADEON_GPIO_MONID) {
-			WREG32(RADEON_DVI_I2C_CNTL_0, (RADEON_I2C_SOFT_RST |
-						R200_DVI_I2C_PIN_SEL(R200_SEL_DDC1)));
-		} else {
-			WREG32(RADEON_DVI_I2C_CNTL_0, (RADEON_I2C_SOFT_RST |
-						R200_DVI_I2C_PIN_SEL(R200_SEL_DDC3)));
+	if (rec->hw_capable) {
+		if ((rdev->family >= CHIP_R200) && !ASIC_IS_AVIVO(rdev)) {
+			u32 reg;
+
+			if (rdev->family >= CHIP_RV350)
+				reg = RADEON_GPIO_MONID;
+			else if ((rdev->family == CHIP_R300) ||
+				 (rdev->family == CHIP_R350))
+				reg = RADEON_GPIO_DVI_DDC;
+			else
+				reg = RADEON_GPIO_CRT2_DDC;
+
+			mutex_lock(&rdev->dc_hw_i2c_mutex);
+			if (rec->a_clk_reg == reg) {
+				WREG32(RADEON_DVI_I2C_CNTL_0, (RADEON_I2C_SOFT_RST |
+							       R200_DVI_I2C_PIN_SEL(R200_SEL_DDC1)));
+			} else {
+				WREG32(RADEON_DVI_I2C_CNTL_0, (RADEON_I2C_SOFT_RST |
+							       R200_DVI_I2C_PIN_SEL(R200_SEL_DDC3)));
+			}
+			mutex_unlock(&rdev->dc_hw_i2c_mutex);
 		}
 	}
-	if (lock_state) {
-		temp = RREG32(rec->a_clk_reg);
-		temp &= ~(rec->a_clk_mask);
-		WREG32(rec->a_clk_reg, temp);
 
-		temp = RREG32(rec->a_data_reg);
-		temp &= ~(rec->a_data_mask);
-		WREG32(rec->a_data_reg, temp);
-	}
+	/* clear the output pin values */
+	temp = RREG32(rec->a_clk_reg) & ~rec->a_clk_mask;
+	WREG32(rec->a_clk_reg, temp);
+
+	temp = RREG32(rec->a_data_reg) & ~rec->a_data_mask;
+	WREG32(rec->a_data_reg, temp);
+
+	/* set the pins to input */
+	temp = RREG32(rec->en_clk_reg) & ~rec->en_clk_mask;
+	WREG32(rec->en_clk_reg, temp);
 
+	temp = RREG32(rec->en_data_reg) & ~rec->en_data_mask;
+	WREG32(rec->en_data_reg, temp);
+
+	/* mask the gpio pins for software use */
 	temp = RREG32(rec->mask_clk_reg);
 	if (lock_state)
 		temp |= rec->mask_clk_mask;
@@ -112,8 +133,9 @@ static int get_clock(void *i2c_priv)
 	struct radeon_i2c_bus_rec *rec = &i2c->rec;
 	uint32_t val;
 
-	val = RREG32(rec->get_clk_reg);
-	val &= rec->get_clk_mask;
+	/* read the value off the pin */
+	val = RREG32(rec->y_clk_reg);
+	val &= rec->y_clk_mask;
 
 	return (val != 0);
 }
@@ -126,8 +148,10 @@ static int get_data(void *i2c_priv)
 	struct radeon_i2c_bus_rec *rec = &i2c->rec;
 	uint32_t val;
 
-	val = RREG32(rec->get_data_reg);
-	val &= rec->get_data_mask;
+	/* read the value off the pin */
+	val = RREG32(rec->y_data_reg);
+	val &= rec->y_data_mask;
+
 	return (val != 0);
 }
 
@@ -138,9 +162,10 @@ static void set_clock(void *i2c_priv, int clock)
 	struct radeon_i2c_bus_rec *rec = &i2c->rec;
 	uint32_t val;
 
-	val = RREG32(rec->put_clk_reg) & (uint32_t)~(rec->put_clk_mask);
-	val |= clock ? 0 : rec->put_clk_mask;
-	WREG32(rec->put_clk_reg, val);
+	/* set pin direction */
+	val = RREG32(rec->en_clk_reg) & ~rec->en_clk_mask;
+	val |= clock ? 0 : rec->en_clk_mask;
+	WREG32(rec->en_clk_reg, val);
 }
 
 static void set_data(void *i2c_priv, int data)
@@ -150,14 +175,701 @@ static void set_data(void *i2c_priv, int data)
 	struct radeon_i2c_bus_rec *rec = &i2c->rec;
 	uint32_t val;
 
-	val = RREG32(rec->put_data_reg) & (uint32_t)~(rec->put_data_mask);
-	val |= data ? 0 : rec->put_data_mask;
-	WREG32(rec->put_data_reg, val);
+	/* set pin direction */
+	val = RREG32(rec->en_data_reg) & ~rec->en_data_mask;
+	val |= data ? 0 : rec->en_data_mask;
+	WREG32(rec->en_data_reg, val);
+}
+
+static u32 radeon_get_i2c_prescale(struct radeon_device *rdev)
+{
+	struct radeon_pll *spll = &rdev->clock.spll;
+	u32 sclk = radeon_get_engine_clock(rdev);
+	u32 prescale = 0;
+	u32 n, m;
+	u8 loop;
+	int i2c_clock;
+
+	switch (rdev->family) {
+	case CHIP_R100:
+	case CHIP_RV100:
+	case CHIP_RS100:
+	case CHIP_RV200:
+	case CHIP_RS200:
+	case CHIP_R200:
+	case CHIP_RV250:
+	case CHIP_RS300:
+	case CHIP_RV280:
+	case CHIP_R300:
+	case CHIP_R350:
+	case CHIP_RV350:
+		n = (spll->reference_freq) / (4 * 6);
+		for (loop = 1; loop < 255; loop++) {
+			if ((loop * (loop - 1)) > n)
+				break;
+		}
+		m = loop - 1;
+		prescale = m | (loop << 8);
+		break;
+	case CHIP_RV380:
+	case CHIP_RS400:
+	case CHIP_RS480:
+	case CHIP_R420:
+	case CHIP_R423:
+	case CHIP_RV410:
+		sclk = radeon_get_engine_clock(rdev);
+		prescale = (((sclk * 10)/(4 * 128 * 100) + 1) << 8) + 128;
+		break;
+	case CHIP_RS600:
+	case CHIP_RS690:
+	case CHIP_RS740:
+		/* todo */
+		break;
+	case CHIP_RV515:
+	case CHIP_R520:
+	case CHIP_RV530:
+	case CHIP_RV560:
+	case CHIP_RV570:
+	case CHIP_R580:
+		i2c_clock = 50;
+		sclk = radeon_get_engine_clock(rdev);
+		if (rdev->family == CHIP_R520)
+			prescale = (127 << 8) + ((sclk * 10) / (4 * 127 * i2c_clock));
+		else
+			prescale = (((sclk * 10)/(4 * 128 * 100) + 1) << 8) + 128;
+		break;
+	case CHIP_R600:
+	case CHIP_RV610:
+	case CHIP_RV630:
+	case CHIP_RV670:
+		/* todo */
+		break;
+	case CHIP_RV620:
+	case CHIP_RV635:
+	case CHIP_RS780:
+	case CHIP_RS880:
+	case CHIP_RV770:
+	case CHIP_RV730:
+	case CHIP_RV710:
+	case CHIP_RV740:
+		/* todo */
+		break;
+	case CHIP_CEDAR:
+	case CHIP_REDWOOD:
+	case CHIP_JUNIPER:
+	case CHIP_CYPRESS:
+	case CHIP_HEMLOCK:
+		/* todo */
+		break;
+	default:
+		DRM_ERROR("i2c: unhandled radeon chip\n");
+		break;
+	}
+	return prescale;
+}
+
+
+/* hw i2c engine for r1xx-4xx hardware
+ * hw can buffer up to 15 bytes
+ */
+static int r100_hw_i2c_xfer(struct i2c_adapter *i2c_adap,
+			    struct i2c_msg *msgs, int num)
+{
+	struct radeon_i2c_chan *i2c = i2c_get_adapdata(i2c_adap);
+	struct radeon_device *rdev = i2c->dev->dev_private;
+	struct radeon_i2c_bus_rec *rec = &i2c->rec;
+	struct i2c_msg *p;
+	int i, j, k, ret = num;
+	u32 prescale;
+	u32 i2c_cntl_0, i2c_cntl_1, i2c_data;
+	u32 tmp, reg;
+
+	mutex_lock(&rdev->dc_hw_i2c_mutex);
+	/* take the pm lock since we need a constant sclk */
+	mutex_lock(&rdev->pm.mutex);
+
+	prescale = radeon_get_i2c_prescale(rdev);
+
+	reg = ((prescale << RADEON_I2C_PRESCALE_SHIFT) |
+	       RADEON_I2C_START |
+	       RADEON_I2C_STOP |
+	       RADEON_I2C_GO);
+
+	if (rdev->is_atom_bios) {
+		tmp = RREG32(RADEON_BIOS_6_SCRATCH);
+		WREG32(RADEON_BIOS_6_SCRATCH, tmp | ATOM_S6_HW_I2C_BUSY_STATE);
+	}
+
+	if (rec->mm_i2c) {
+		i2c_cntl_0 = RADEON_I2C_CNTL_0;
+		i2c_cntl_1 = RADEON_I2C_CNTL_1;
+		i2c_data = RADEON_I2C_DATA;
+	} else {
+		i2c_cntl_0 = RADEON_DVI_I2C_CNTL_0;
+		i2c_cntl_1 = RADEON_DVI_I2C_CNTL_1;
+		i2c_data = RADEON_DVI_I2C_DATA;
+
+		switch (rdev->family) {
+		case CHIP_R100:
+		case CHIP_RV100:
+		case CHIP_RS100:
+		case CHIP_RV200:
+		case CHIP_RS200:
+		case CHIP_RS300:
+			switch (rec->mask_clk_reg) {
+			case RADEON_GPIO_DVI_DDC:
+				/* no gpio select bit */
+				break;
+			default:
+				DRM_ERROR("gpio not supported with hw i2c\n");
+				ret = -EINVAL;
+				goto done;
+			}
+			break;
+		case CHIP_R200:
+			/* only bit 4 on r200 */
+			switch (rec->mask_clk_reg) {
+			case RADEON_GPIO_DVI_DDC:
+				reg |= R200_DVI_I2C_PIN_SEL(R200_SEL_DDC1);
+				break;
+			case RADEON_GPIO_MONID:
+				reg |= R200_DVI_I2C_PIN_SEL(R200_SEL_DDC3);
+				break;
+			default:
+				DRM_ERROR("gpio not supported with hw i2c\n");
+				ret = -EINVAL;
+				goto done;
+			}
+			break;
+		case CHIP_RV250:
+		case CHIP_RV280:
+			/* bits 3 and 4 */
+			switch (rec->mask_clk_reg) {
+			case RADEON_GPIO_DVI_DDC:
+				reg |= R200_DVI_I2C_PIN_SEL(R200_SEL_DDC1);
+				break;
+			case RADEON_GPIO_VGA_DDC:
+				reg |= R200_DVI_I2C_PIN_SEL(R200_SEL_DDC2);
+				break;
+			case RADEON_GPIO_CRT2_DDC:
+				reg |= R200_DVI_I2C_PIN_SEL(R200_SEL_DDC3);
+				break;
+			default:
+				DRM_ERROR("gpio not supported with hw i2c\n");
+				ret = -EINVAL;
+				goto done;
+			}
+			break;
+		case CHIP_R300:
+		case CHIP_R350:
+			/* only bit 4 on r300/r350 */
+			switch (rec->mask_clk_reg) {
+			case RADEON_GPIO_VGA_DDC:
+				reg |= R200_DVI_I2C_PIN_SEL(R200_SEL_DDC1);
+				break;
+			case RADEON_GPIO_DVI_DDC:
+				reg |= R200_DVI_I2C_PIN_SEL(R200_SEL_DDC3);
+				break;
+			default:
+				DRM_ERROR("gpio not supported with hw i2c\n");
+				ret = -EINVAL;
+				goto done;
+			}
+			break;
+		case CHIP_RV350:
+		case CHIP_RV380:
+		case CHIP_R420:
+		case CHIP_R423:
+		case CHIP_RV410:
+		case CHIP_RS400:
+		case CHIP_RS480:
+			/* bits 3 and 4 */
+			switch (rec->mask_clk_reg) {
+			case RADEON_GPIO_VGA_DDC:
+				reg |= R200_DVI_I2C_PIN_SEL(R200_SEL_DDC1);
+				break;
+			case RADEON_GPIO_DVI_DDC:
+				reg |= R200_DVI_I2C_PIN_SEL(R200_SEL_DDC2);
+				break;
+			case RADEON_GPIO_MONID:
+				reg |= R200_DVI_I2C_PIN_SEL(R200_SEL_DDC3);
+				break;
+			default:
+				DRM_ERROR("gpio not supported with hw i2c\n");
+				ret = -EINVAL;
+				goto done;
+			}
+			break;
+		default:
+			DRM_ERROR("unsupported asic\n");
+			ret = -EINVAL;
+			goto done;
+			break;
+		}
+	}
+
+	/* check for bus probe */
+	p = &msgs[0];
+	if ((num == 1) && (p->len == 0)) {
+		WREG32(i2c_cntl_0, (RADEON_I2C_DONE |
+				    RADEON_I2C_NACK |
+				    RADEON_I2C_HALT |
+				    RADEON_I2C_SOFT_RST));
+		WREG32(i2c_data, (p->addr << 1) & 0xff);
+		WREG32(i2c_data, 0);
+		WREG32(i2c_cntl_1, ((1 << RADEON_I2C_DATA_COUNT_SHIFT) |
+				    (1 << RADEON_I2C_ADDR_COUNT_SHIFT) |
+				    RADEON_I2C_EN |
+				    (48 << RADEON_I2C_TIME_LIMIT_SHIFT)));
+		WREG32(i2c_cntl_0, reg);
+		for (k = 0; k < 32; k++) {
+			udelay(10);
+			tmp = RREG32(i2c_cntl_0);
+			if (tmp & RADEON_I2C_GO)
+				continue;
+			tmp = RREG32(i2c_cntl_0);
+			if (tmp & RADEON_I2C_DONE)
+				break;
+			else {
+				DRM_DEBUG("i2c write error 0x%08x\n", tmp);
+				WREG32(i2c_cntl_0, tmp | RADEON_I2C_ABORT);
+				ret = -EIO;
+				goto done;
+			}
+		}
+		goto done;
+	}
+
+	for (i = 0; i < num; i++) {
+		p = &msgs[i];
+		for (j = 0; j < p->len; j++) {
+			if (p->flags & I2C_M_RD) {
+				WREG32(i2c_cntl_0, (RADEON_I2C_DONE |
+						    RADEON_I2C_NACK |
+						    RADEON_I2C_HALT |
+						    RADEON_I2C_SOFT_RST));
+				WREG32(i2c_data, ((p->addr << 1) & 0xff) | 0x1);
+				WREG32(i2c_cntl_1, ((1 << RADEON_I2C_DATA_COUNT_SHIFT) |
+						    (1 << RADEON_I2C_ADDR_COUNT_SHIFT) |
+						    RADEON_I2C_EN |
+						    (48 << RADEON_I2C_TIME_LIMIT_SHIFT)));
+				WREG32(i2c_cntl_0, reg | RADEON_I2C_RECEIVE);
+				for (k = 0; k < 32; k++) {
+					udelay(10);
+					tmp = RREG32(i2c_cntl_0);
+					if (tmp & RADEON_I2C_GO)
+						continue;
+					tmp = RREG32(i2c_cntl_0);
+					if (tmp & RADEON_I2C_DONE)
+						break;
+					else {
+						DRM_DEBUG("i2c read error 0x%08x\n", tmp);
+						WREG32(i2c_cntl_0, tmp | RADEON_I2C_ABORT);
+						ret = -EIO;
+						goto done;
+					}
+				}
+				p->buf[j] = RREG32(i2c_data) & 0xff;
+			} else {
+				WREG32(i2c_cntl_0, (RADEON_I2C_DONE |
+						    RADEON_I2C_NACK |
+						    RADEON_I2C_HALT |
+						    RADEON_I2C_SOFT_RST));
+				WREG32(i2c_data, (p->addr << 1) & 0xff);
+				WREG32(i2c_data, p->buf[j]);
+				WREG32(i2c_cntl_1, ((1 << RADEON_I2C_DATA_COUNT_SHIFT) |
+						    (1 << RADEON_I2C_ADDR_COUNT_SHIFT) |
+						    RADEON_I2C_EN |
+						    (48 << RADEON_I2C_TIME_LIMIT_SHIFT)));
+				WREG32(i2c_cntl_0, reg);
+				for (k = 0; k < 32; k++) {
+					udelay(10);
+					tmp = RREG32(i2c_cntl_0);
+					if (tmp & RADEON_I2C_GO)
+						continue;
+					tmp = RREG32(i2c_cntl_0);
+					if (tmp & RADEON_I2C_DONE)
+						break;
+					else {
+						DRM_DEBUG("i2c write error 0x%08x\n", tmp);
+						WREG32(i2c_cntl_0, tmp | RADEON_I2C_ABORT);
+						ret = -EIO;
+						goto done;
+					}
+				}
+			}
+		}
+	}
+
+done:
+	WREG32(i2c_cntl_0, 0);
+	WREG32(i2c_cntl_1, 0);
+	WREG32(i2c_cntl_0, (RADEON_I2C_DONE |
+			    RADEON_I2C_NACK |
+			    RADEON_I2C_HALT |
+			    RADEON_I2C_SOFT_RST));
+
+	if (rdev->is_atom_bios) {
+		tmp = RREG32(RADEON_BIOS_6_SCRATCH);
+		tmp &= ~ATOM_S6_HW_I2C_BUSY_STATE;
+		WREG32(RADEON_BIOS_6_SCRATCH, tmp);
+	}
+
+	mutex_unlock(&rdev->pm.mutex);
+	mutex_unlock(&rdev->dc_hw_i2c_mutex);
+
+	return ret;
+}
+
+/* hw i2c engine for r5xx hardware
+ * hw can buffer up to 15 bytes
+ */
+static int r500_hw_i2c_xfer(struct i2c_adapter *i2c_adap,
+			    struct i2c_msg *msgs, int num)
+{
+	struct radeon_i2c_chan *i2c = i2c_get_adapdata(i2c_adap);
+	struct radeon_device *rdev = i2c->dev->dev_private;
+	struct radeon_i2c_bus_rec *rec = &i2c->rec;
+	struct i2c_msg *p;
+	int i, j, remaining, current_count, buffer_offset, ret = num;
+	u32 prescale;
+	u32 tmp, reg;
+	u32 saved1, saved2;
+
+	mutex_lock(&rdev->dc_hw_i2c_mutex);
+	/* take the pm lock since we need a constant sclk */
+	mutex_lock(&rdev->pm.mutex);
+
+	prescale = radeon_get_i2c_prescale(rdev);
+
+	/* clear gpio mask bits */
+	tmp = RREG32(rec->mask_clk_reg);
+	tmp &= ~rec->mask_clk_mask;
+	WREG32(rec->mask_clk_reg, tmp);
+	tmp = RREG32(rec->mask_clk_reg);
+
+	tmp = RREG32(rec->mask_data_reg);
+	tmp &= ~rec->mask_data_mask;
+	WREG32(rec->mask_data_reg, tmp);
+	tmp = RREG32(rec->mask_data_reg);
+
+	/* clear pin values */
+	tmp = RREG32(rec->a_clk_reg);
+	tmp &= ~rec->a_clk_mask;
+	WREG32(rec->a_clk_reg, tmp);
+	tmp = RREG32(rec->a_clk_reg);
+
+	tmp = RREG32(rec->a_data_reg);
+	tmp &= ~rec->a_data_mask;
+	WREG32(rec->a_data_reg, tmp);
+	tmp = RREG32(rec->a_data_reg);
+
+	/* set the pins to input */
+	tmp = RREG32(rec->en_clk_reg);
+	tmp &= ~rec->en_clk_mask;
+	WREG32(rec->en_clk_reg, tmp);
+	tmp = RREG32(rec->en_clk_reg);
+
+	tmp = RREG32(rec->en_data_reg);
+	tmp &= ~rec->en_data_mask;
+	WREG32(rec->en_data_reg, tmp);
+	tmp = RREG32(rec->en_data_reg);
+
+	/* */
+	tmp = RREG32(RADEON_BIOS_6_SCRATCH);
+	WREG32(RADEON_BIOS_6_SCRATCH, tmp | ATOM_S6_HW_I2C_BUSY_STATE);
+	saved1 = RREG32(AVIVO_DC_I2C_CONTROL1);
+	saved2 = RREG32(0x494);
+	WREG32(0x494, saved2 | 0x1);
+
+	WREG32(AVIVO_DC_I2C_ARBITRATION, AVIVO_DC_I2C_SW_WANTS_TO_USE_I2C);
+	for (i = 0; i < 50; i++) {
+		udelay(1);
+		if (RREG32(AVIVO_DC_I2C_ARBITRATION) & AVIVO_DC_I2C_SW_CAN_USE_I2C)
+			break;
+	}
+	if (i == 50) {
+		DRM_ERROR("failed to get i2c bus\n");
+		ret = -EBUSY;
+		goto done;
+	}
+
+	reg = AVIVO_DC_I2C_START | AVIVO_DC_I2C_STOP | AVIVO_DC_I2C_EN;
+	switch (rec->mask_clk_reg) {
+	case AVIVO_DC_GPIO_DDC1_MASK:
+		reg |= AVIVO_DC_I2C_PIN_SELECT(AVIVO_SEL_DDC1);
+		break;
+	case AVIVO_DC_GPIO_DDC2_MASK:
+		reg |= AVIVO_DC_I2C_PIN_SELECT(AVIVO_SEL_DDC2);
+		break;
+	case AVIVO_DC_GPIO_DDC3_MASK:
+		reg |= AVIVO_DC_I2C_PIN_SELECT(AVIVO_SEL_DDC3);
+		break;
+	default:
+		DRM_ERROR("gpio not supported with hw i2c\n");
+		ret = -EINVAL;
+		goto done;
+	}
+
+	/* check for bus probe */
+	p = &msgs[0];
+	if ((num == 1) && (p->len == 0)) {
+		WREG32(AVIVO_DC_I2C_STATUS1, (AVIVO_DC_I2C_DONE |
+					      AVIVO_DC_I2C_NACK |
+					      AVIVO_DC_I2C_HALT));
+		WREG32(AVIVO_DC_I2C_RESET, AVIVO_DC_I2C_SOFT_RESET);
+		udelay(1);
+		WREG32(AVIVO_DC_I2C_RESET, 0);
+
+		WREG32(AVIVO_DC_I2C_DATA, (p->addr << 1) & 0xff);
+		WREG32(AVIVO_DC_I2C_DATA, 0);
+
+		WREG32(AVIVO_DC_I2C_CONTROL3, AVIVO_DC_I2C_TIME_LIMIT(48));
+		WREG32(AVIVO_DC_I2C_CONTROL2, (AVIVO_DC_I2C_ADDR_COUNT(1) |
+					       AVIVO_DC_I2C_DATA_COUNT(1) |
+					       (prescale << 16)));
+		WREG32(AVIVO_DC_I2C_CONTROL1, reg);
+		WREG32(AVIVO_DC_I2C_STATUS1, AVIVO_DC_I2C_GO);
+		for (j = 0; j < 200; j++) {
+			udelay(50);
+			tmp = RREG32(AVIVO_DC_I2C_STATUS1);
+			if (tmp & AVIVO_DC_I2C_GO)
+				continue;
+			tmp = RREG32(AVIVO_DC_I2C_STATUS1);
+			if (tmp & AVIVO_DC_I2C_DONE)
+				break;
+			else {
+				DRM_DEBUG("i2c write error 0x%08x\n", tmp);
+				WREG32(AVIVO_DC_I2C_RESET, AVIVO_DC_I2C_ABORT);
+				ret = -EIO;
+				goto done;
+			}
+		}
+		goto done;
+	}
+
+	for (i = 0; i < num; i++) {
+		p = &msgs[i];
+		remaining = p->len;
+		buffer_offset = 0;
+		if (p->flags & I2C_M_RD) {
+			while (remaining) {
+				if (remaining > 15)
+					current_count = 15;
+				else
+					current_count = remaining;
+				WREG32(AVIVO_DC_I2C_STATUS1, (AVIVO_DC_I2C_DONE |
+							      AVIVO_DC_I2C_NACK |
+							      AVIVO_DC_I2C_HALT));
+				WREG32(AVIVO_DC_I2C_RESET, AVIVO_DC_I2C_SOFT_RESET);
+				udelay(1);
+				WREG32(AVIVO_DC_I2C_RESET, 0);
+
+				WREG32(AVIVO_DC_I2C_DATA, ((p->addr << 1) & 0xff) | 0x1);
+				WREG32(AVIVO_DC_I2C_CONTROL3, AVIVO_DC_I2C_TIME_LIMIT(48));
+				WREG32(AVIVO_DC_I2C_CONTROL2, (AVIVO_DC_I2C_ADDR_COUNT(1) |
+							       AVIVO_DC_I2C_DATA_COUNT(current_count) |
+							       (prescale << 16)));
+				WREG32(AVIVO_DC_I2C_CONTROL1, reg | AVIVO_DC_I2C_RECEIVE);
+				WREG32(AVIVO_DC_I2C_STATUS1, AVIVO_DC_I2C_GO);
+				for (j = 0; j < 200; j++) {
+					udelay(50);
+					tmp = RREG32(AVIVO_DC_I2C_STATUS1);
+					if (tmp & AVIVO_DC_I2C_GO)
+						continue;
+					tmp = RREG32(AVIVO_DC_I2C_STATUS1);
+					if (tmp & AVIVO_DC_I2C_DONE)
+						break;
+					else {
+						DRM_DEBUG("i2c read error 0x%08x\n", tmp);
+						WREG32(AVIVO_DC_I2C_RESET, AVIVO_DC_I2C_ABORT);
+						ret = -EIO;
+						goto done;
+					}
+				}
+				for (j = 0; j < current_count; j++)
+					p->buf[buffer_offset + j] = RREG32(AVIVO_DC_I2C_DATA) & 0xff;
+				remaining -= current_count;
+				buffer_offset += current_count;
+			}
+		} else {
+			while (remaining) {
+				if (remaining > 15)
+					current_count = 15;
+				else
+					current_count = remaining;
+				WREG32(AVIVO_DC_I2C_STATUS1, (AVIVO_DC_I2C_DONE |
+							      AVIVO_DC_I2C_NACK |
+							      AVIVO_DC_I2C_HALT));
+				WREG32(AVIVO_DC_I2C_RESET, AVIVO_DC_I2C_SOFT_RESET);
+				udelay(1);
+				WREG32(AVIVO_DC_I2C_RESET, 0);
+
+				WREG32(AVIVO_DC_I2C_DATA, (p->addr << 1) & 0xff);
+				for (j = 0; j < current_count; j++)
+					WREG32(AVIVO_DC_I2C_DATA, p->buf[buffer_offset + j]);
+
+				WREG32(AVIVO_DC_I2C_CONTROL3, AVIVO_DC_I2C_TIME_LIMIT(48));
+				WREG32(AVIVO_DC_I2C_CONTROL2, (AVIVO_DC_I2C_ADDR_COUNT(1) |
+							       AVIVO_DC_I2C_DATA_COUNT(current_count) |
+							       (prescale << 16)));
+				WREG32(AVIVO_DC_I2C_CONTROL1, reg);
+				WREG32(AVIVO_DC_I2C_STATUS1, AVIVO_DC_I2C_GO);
+				for (j = 0; j < 200; j++) {
+					udelay(50);
+					tmp = RREG32(AVIVO_DC_I2C_STATUS1);
+					if (tmp & AVIVO_DC_I2C_GO)
+						continue;
+					tmp = RREG32(AVIVO_DC_I2C_STATUS1);
+					if (tmp & AVIVO_DC_I2C_DONE)
+						break;
+					else {
+						DRM_DEBUG("i2c write error 0x%08x\n", tmp);
+						WREG32(AVIVO_DC_I2C_RESET, AVIVO_DC_I2C_ABORT);
+						ret = -EIO;
+						goto done;
+					}
+				}
+				remaining -= current_count;
+				buffer_offset += current_count;
+			}
+		}
+	}
+
+done:
+	WREG32(AVIVO_DC_I2C_STATUS1, (AVIVO_DC_I2C_DONE |
+				      AVIVO_DC_I2C_NACK |
+				      AVIVO_DC_I2C_HALT));
+	WREG32(AVIVO_DC_I2C_RESET, AVIVO_DC_I2C_SOFT_RESET);
+	udelay(1);
+	WREG32(AVIVO_DC_I2C_RESET, 0);
+
+	WREG32(AVIVO_DC_I2C_ARBITRATION, AVIVO_DC_I2C_SW_DONE_USING_I2C);
+	WREG32(AVIVO_DC_I2C_CONTROL1, saved1);
+	WREG32(0x494, saved2);
+	tmp = RREG32(RADEON_BIOS_6_SCRATCH);
+	tmp &= ~ATOM_S6_HW_I2C_BUSY_STATE;
+	WREG32(RADEON_BIOS_6_SCRATCH, tmp);
+
+	mutex_unlock(&rdev->pm.mutex);
+	mutex_unlock(&rdev->dc_hw_i2c_mutex);
+
+	return ret;
 }
 
+static int radeon_sw_i2c_xfer(struct i2c_adapter *i2c_adap,
+			      struct i2c_msg *msgs, int num)
+{
+	struct radeon_i2c_chan *i2c = i2c_get_adapdata(i2c_adap);
+	int ret;
+
+	radeon_i2c_do_lock(i2c, 1);
+	ret = i2c_transfer(&i2c->algo.radeon.bit_adapter, msgs, num);
+	radeon_i2c_do_lock(i2c, 0);
+
+	return ret;
+}
+
+static int radeon_i2c_xfer(struct i2c_adapter *i2c_adap,
+			   struct i2c_msg *msgs, int num)
+{
+	struct radeon_i2c_chan *i2c = i2c_get_adapdata(i2c_adap);
+	struct radeon_device *rdev = i2c->dev->dev_private;
+	struct radeon_i2c_bus_rec *rec = &i2c->rec;
+	int ret;
+
+	switch (rdev->family) {
+	case CHIP_R100:
+	case CHIP_RV100:
+	case CHIP_RS100:
+	case CHIP_RV200:
+	case CHIP_RS200:
+	case CHIP_R200:
+	case CHIP_RV250:
+	case CHIP_RS300:
+	case CHIP_RV280:
+	case CHIP_R300:
+	case CHIP_R350:
+	case CHIP_RV350:
+	case CHIP_RV380:
+	case CHIP_R420:
+	case CHIP_R423:
+	case CHIP_RV410:
+	case CHIP_RS400:
+	case CHIP_RS480:
+		if (rec->hw_capable)
+			ret = r100_hw_i2c_xfer(i2c_adap, msgs, num);
+		else
+			ret = radeon_sw_i2c_xfer(i2c_adap, msgs, num);
+		break;
+	case CHIP_RS600:
+	case CHIP_RS690:
+	case CHIP_RS740:
+		/* XXX fill in hw i2c implementation */
+		ret = radeon_sw_i2c_xfer(i2c_adap, msgs, num);
+		break;
+	case CHIP_RV515:
+	case CHIP_R520:
+	case CHIP_RV530:
+	case CHIP_RV560:
+	case CHIP_RV570:
+	case CHIP_R580:
+		if (rec->hw_capable) {
+			if (rec->mm_i2c)
+				ret = r100_hw_i2c_xfer(i2c_adap, msgs, num);
+			else
+				ret = r500_hw_i2c_xfer(i2c_adap, msgs, num);
+		} else
+			ret = radeon_sw_i2c_xfer(i2c_adap, msgs, num);
+		break;
+	case CHIP_R600:
+	case CHIP_RV610:
+	case CHIP_RV630:
+	case CHIP_RV670:
+		/* XXX fill in hw i2c implementation */
+		ret = radeon_sw_i2c_xfer(i2c_adap, msgs, num);
+		break;
+	case CHIP_RV620:
+	case CHIP_RV635:
+	case CHIP_RS780:
+	case CHIP_RS880:
+	case CHIP_RV770:
+	case CHIP_RV730:
+	case CHIP_RV710:
+	case CHIP_RV740:
+		/* XXX fill in hw i2c implementation */
+		ret = radeon_sw_i2c_xfer(i2c_adap, msgs, num);
+		break;
+	case CHIP_CEDAR:
+	case CHIP_REDWOOD:
+	case CHIP_JUNIPER:
+	case CHIP_CYPRESS:
+	case CHIP_HEMLOCK:
+		/* XXX fill in hw i2c implementation */
+		ret = radeon_sw_i2c_xfer(i2c_adap, msgs, num);
+		break;
+	default:
+		DRM_ERROR("i2c: unhandled radeon chip\n");
+		ret = -EIO;
+		break;
+	}
+
+	return ret;
+}
+
+static u32 radeon_i2c_func(struct i2c_adapter *adap)
+{
+	return I2C_FUNC_I2C | I2C_FUNC_SMBUS_EMUL;
+}
+
+static const struct i2c_algorithm radeon_i2c_algo = {
+	.master_xfer = radeon_i2c_xfer,
+	.functionality = radeon_i2c_func,
+};
+
 struct radeon_i2c_chan *radeon_i2c_create(struct drm_device *dev,
-		struct radeon_i2c_bus_rec *rec,
-		const char *name)
+					  struct radeon_i2c_bus_rec *rec,
+					  const char *name)
 {
 	struct radeon_i2c_chan *i2c;
 	int ret;
@@ -166,22 +878,65 @@ struct radeon_i2c_chan *radeon_i2c_create(struct drm_device *dev,
 	if (i2c == NULL)
 		return NULL;
 
-	i2c->adapter.owner = THIS_MODULE;
-	i2c->adapter.algo_data = &i2c->algo;
-	i2c->dev = dev;
-	i2c->algo.setsda = set_data;
-	i2c->algo.setscl = set_clock;
-	i2c->algo.getsda = get_data;
-	i2c->algo.getscl = get_clock;
-	i2c->algo.udelay = 20;
+	/* set the internal bit adapter */
+	i2c->algo.radeon.bit_adapter.owner = THIS_MODULE;
+	i2c_set_adapdata(&i2c->algo.radeon.bit_adapter, i2c);
+	sprintf(i2c->algo.radeon.bit_adapter.name, "Radeon internal i2c bit bus %s", name);
+	i2c->algo.radeon.bit_adapter.algo_data = &i2c->algo.radeon.bit_data;
+	i2c->algo.radeon.bit_data.setsda = set_data;
+	i2c->algo.radeon.bit_data.setscl = set_clock;
+	i2c->algo.radeon.bit_data.getsda = get_data;
+	i2c->algo.radeon.bit_data.getscl = get_clock;
+	i2c->algo.radeon.bit_data.udelay = 20;
 	/* vesa says 2.2 ms is enough, 1 jiffy doesn't seem to always
 	 * make this, 2 jiffies is a lot more reliable */
-	i2c->algo.timeout = 2;
-	i2c->algo.data = i2c;
+	i2c->algo.radeon.bit_data.timeout = 2;
+	i2c->algo.radeon.bit_data.data = i2c;
+	ret = i2c_bit_add_bus(&i2c->algo.radeon.bit_adapter);
+	if (ret) {
+		DRM_ERROR("Failed to register internal bit i2c %s\n", name);
+		goto out_free;
+	}
+	/* set the radeon i2c adapter */
+	i2c->dev = dev;
 	i2c->rec = *rec;
+	i2c->adapter.owner = THIS_MODULE;
 	i2c_set_adapdata(&i2c->adapter, i2c);
+	sprintf(i2c->adapter.name, "Radeon i2c %s", name);
+	i2c->adapter.algo_data = &i2c->algo.radeon;
+	i2c->adapter.algo = &radeon_i2c_algo;
+	ret = i2c_add_adapter(&i2c->adapter);
+	if (ret) {
+		DRM_ERROR("Failed to register i2c %s\n", name);
+		goto out_free;
+	}
+
+	return i2c;
+out_free:
+	kfree(i2c);
+	return NULL;
+
+}
+
+struct radeon_i2c_chan *radeon_i2c_create_dp(struct drm_device *dev,
+					     struct radeon_i2c_bus_rec *rec,
+					     const char *name)
+{
+	struct radeon_i2c_chan *i2c;
+	int ret;
 
-	ret = i2c_bit_add_bus(&i2c->adapter);
+	i2c = kzalloc(sizeof(struct radeon_i2c_chan), GFP_KERNEL);
+	if (i2c == NULL)
+		return NULL;
+
+	i2c->rec = *rec;
+	i2c->adapter.owner = THIS_MODULE;
+	i2c->dev = dev;
+	i2c_set_adapdata(&i2c->adapter, i2c);
+	i2c->adapter.algo_data = &i2c->algo.dp;
+	i2c->algo.dp.aux_ch = radeon_dp_i2c_aux_ch;
+	i2c->algo.dp.address = 0;
+	ret = i2c_dp_aux_add_bus(&i2c->adapter);
 	if (ret) {
 		DRM_INFO("Failed to register i2c %s\n", name);
 		goto out_free;
@@ -198,6 +953,15 @@ void radeon_i2c_destroy(struct radeon_i2c_chan *i2c)
 {
 	if (!i2c)
 		return;
+	i2c_del_adapter(&i2c->algo.radeon.bit_adapter);
+	i2c_del_adapter(&i2c->adapter);
+	kfree(i2c);
+}
+
+void radeon_i2c_destroy_dp(struct radeon_i2c_chan *i2c)
+{
+	if (!i2c)
+		return;
 
 	i2c_del_adapter(&i2c->adapter);
 	kfree(i2c);
@@ -207,3 +971,59 @@ struct drm_encoder *radeon_best_encoder(struct drm_connector *connector)
 {
 	return NULL;
 }
+
+void radeon_i2c_get_byte(struct radeon_i2c_chan *i2c_bus,
+			 u8 slave_addr,
+			 u8 addr,
+			 u8 *val)
+{
+	u8 out_buf[2];
+	u8 in_buf[2];
+	struct i2c_msg msgs[] = {
+		{
+			.addr = slave_addr,
+			.flags = 0,
+			.len = 1,
+			.buf = out_buf,
+		},
+		{
+			.addr = slave_addr,
+			.flags = I2C_M_RD,
+			.len = 1,
+			.buf = in_buf,
+		}
+	};
+
+	out_buf[0] = addr;
+	out_buf[1] = 0;
+
+	if (i2c_transfer(&i2c_bus->adapter, msgs, 2) == 2) {
+		*val = in_buf[0];
+		DRM_DEBUG("val = 0x%02x\n", *val);
+	} else {
+		DRM_ERROR("i2c 0x%02x 0x%02x read failed\n",
+			  addr, *val);
+	}
+}
+
+void radeon_i2c_put_byte(struct radeon_i2c_chan *i2c_bus,
+			 u8 slave_addr,
+			 u8 addr,
+			 u8 val)
+{
+	uint8_t out_buf[2];
+	struct i2c_msg msg = {
+		.addr = slave_addr,
+		.flags = 0,
+		.len = 2,
+		.buf = out_buf,
+	};
+
+	out_buf[0] = addr;
+	out_buf[1] = val;
+
+	if (i2c_transfer(&i2c_bus->adapter, &msg, 1) != 1)
+		DRM_ERROR("i2c 0x%02x 0x%02x write failed\n",
+			  addr, val);
+}
+
diff --git a/drivers/gpu/drm/radeon/radeon_ioc32.c b/drivers/gpu/drm/radeon/radeon_ioc32.c
index a1bf11de308a..48b7cea31e08 100644
--- a/drivers/gpu/drm/radeon/radeon_ioc32.c
+++ b/drivers/gpu/drm/radeon/radeon_ioc32.c
@@ -92,8 +92,7 @@ static int compat_radeon_cp_init(struct file *file, unsigned int cmd,
 			  &init->gart_textures_offset))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_RADEON_CP_INIT, (unsigned long)init);
+	return drm_ioctl(file, DRM_IOCTL_RADEON_CP_INIT, (unsigned long)init);
 }
 
 typedef struct drm_radeon_clear32 {
@@ -125,8 +124,7 @@ static int compat_radeon_cp_clear(struct file *file, unsigned int cmd,
 			  &clr->depth_boxes))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_RADEON_CLEAR, (unsigned long)clr);
+	return drm_ioctl(file, DRM_IOCTL_RADEON_CLEAR, (unsigned long)clr);
 }
 
 typedef struct drm_radeon_stipple32 {
@@ -149,8 +147,7 @@ static int compat_radeon_cp_stipple(struct file *file, unsigned int cmd,
 			  &request->mask))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_RADEON_STIPPLE, (unsigned long)request);
+	return drm_ioctl(file, DRM_IOCTL_RADEON_STIPPLE, (unsigned long)request);
 }
 
 typedef struct drm_radeon_tex_image32 {
@@ -204,8 +201,7 @@ static int compat_radeon_cp_texture(struct file *file, unsigned int cmd,
 			  &image->data))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_RADEON_TEXTURE, (unsigned long)request);
+	return drm_ioctl(file, DRM_IOCTL_RADEON_TEXTURE, (unsigned long)request);
 }
 
 typedef struct drm_radeon_vertex2_32 {
@@ -238,8 +234,7 @@ static int compat_radeon_cp_vertex2(struct file *file, unsigned int cmd,
 			  &request->prim))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_RADEON_VERTEX2, (unsigned long)request);
+	return drm_ioctl(file, DRM_IOCTL_RADEON_VERTEX2, (unsigned long)request);
 }
 
 typedef struct drm_radeon_cmd_buffer32 {
@@ -268,8 +263,7 @@ static int compat_radeon_cp_cmdbuf(struct file *file, unsigned int cmd,
 			  &request->boxes))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_RADEON_CMDBUF, (unsigned long)request);
+	return drm_ioctl(file, DRM_IOCTL_RADEON_CMDBUF, (unsigned long)request);
 }
 
 typedef struct drm_radeon_getparam32 {
@@ -293,8 +287,7 @@ static int compat_radeon_cp_getparam(struct file *file, unsigned int cmd,
 			  &request->value))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_RADEON_GETPARAM, (unsigned long)request);
+	return drm_ioctl(file, DRM_IOCTL_RADEON_GETPARAM, (unsigned long)request);
 }
 
 typedef struct drm_radeon_mem_alloc32 {
@@ -322,8 +315,7 @@ static int compat_radeon_mem_alloc(struct file *file, unsigned int cmd,
 			  &request->region_offset))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_RADEON_ALLOC, (unsigned long)request);
+	return drm_ioctl(file, DRM_IOCTL_RADEON_ALLOC, (unsigned long)request);
 }
 
 typedef struct drm_radeon_irq_emit32 {
@@ -345,8 +337,7 @@ static int compat_radeon_irq_emit(struct file *file, unsigned int cmd,
 			  &request->irq_seq))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_path.dentry->d_inode, file,
-			 DRM_IOCTL_RADEON_IRQ_EMIT, (unsigned long)request);
+	return drm_ioctl(file, DRM_IOCTL_RADEON_IRQ_EMIT, (unsigned long)request);
 }
 
 /* The two 64-bit arches where alignof(u64)==4 in 32-bit code */
@@ -372,8 +363,7 @@ static int compat_radeon_cp_setparam(struct file *file, unsigned int cmd,
 			  &request->value))
 		return -EFAULT;
 
-	return drm_ioctl(file->f_dentry->d_inode, file,
-			 DRM_IOCTL_RADEON_SETPARAM, (unsigned long) request);
+	return drm_ioctl(file, DRM_IOCTL_RADEON_SETPARAM, (unsigned long) request);
 }
 #else
 #define compat_radeon_cp_setparam NULL
@@ -413,12 +403,10 @@ long radeon_compat_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
 	if (nr < DRM_COMMAND_BASE + DRM_ARRAY_SIZE(radeon_compat_ioctls))
 		fn = radeon_compat_ioctls[nr - DRM_COMMAND_BASE];
 
-	lock_kernel();		/* XXX for now */
 	if (fn != NULL)
 		ret = (*fn) (filp, cmd, arg);
 	else
-		ret = drm_ioctl(filp->f_path.dentry->d_inode, filp, cmd, arg);
-	unlock_kernel();
+		ret = drm_ioctl(filp, cmd, arg);
 
 	return ret;
 }
@@ -431,9 +419,7 @@ long radeon_kms_compat_ioctl(struct file *filp, unsigned int cmd, unsigned long
 	if (nr < DRM_COMMAND_BASE)
 		return drm_compat_ioctl(filp, cmd, arg);
 
-	lock_kernel();		/* XXX for now */
-	ret = drm_ioctl(filp->f_path.dentry->d_inode, filp, cmd, arg);
-	unlock_kernel();
+	ret = drm_ioctl(filp, cmd, arg);
 
 	return ret;
 }
diff --git a/drivers/gpu/drm/radeon/radeon_irq.c b/drivers/gpu/drm/radeon/radeon_irq.c
index b79ecc4a7cc4..2f349a300195 100644
--- a/drivers/gpu/drm/radeon/radeon_irq.c
+++ b/drivers/gpu/drm/radeon/radeon_irq.c
@@ -289,16 +289,16 @@ int radeon_irq_emit(struct drm_device *dev, void *data, struct drm_file *file_pr
 	drm_radeon_irq_emit_t *emit = data;
 	int result;
 
-	if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_R600)
-		return -EINVAL;
-
-	LOCK_TEST_WITH_RETURN(dev, file_priv);
-
 	if (!dev_priv) {
 		DRM_ERROR("called with no initialization\n");
 		return -EINVAL;
 	}
 
+	if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_R600)
+		return -EINVAL;
+
+	LOCK_TEST_WITH_RETURN(dev, file_priv);
+
 	result = radeon_emit_irq(dev);
 
 	if (DRM_COPY_TO_USER(emit->irq_seq, &result, sizeof(int))) {
diff --git a/drivers/gpu/drm/radeon/radeon_irq_kms.c b/drivers/gpu/drm/radeon/radeon_irq_kms.c
index 1841145a7c4f..3cfd60fd0083 100644
--- a/drivers/gpu/drm/radeon/radeon_irq_kms.c
+++ b/drivers/gpu/drm/radeon/radeon_irq_kms.c
@@ -39,11 +39,32 @@ irqreturn_t radeon_driver_irq_handler_kms(DRM_IRQ_ARGS)
 	return radeon_irq_process(rdev);
 }
 
+/*
+ * Handle hotplug events outside the interrupt handler proper.
+ */
+static void radeon_hotplug_work_func(struct work_struct *work)
+{
+	struct radeon_device *rdev = container_of(work, struct radeon_device,
+						  hotplug_work);
+	struct drm_device *dev = rdev->ddev;
+	struct drm_mode_config *mode_config = &dev->mode_config;
+	struct drm_connector *connector;
+
+	if (mode_config->num_connector) {
+		list_for_each_entry(connector, &mode_config->connector_list, head)
+			radeon_connector_hotplug(connector);
+	}
+	/* Just fire off a uevent and let userspace tell us what to do */
+	drm_sysfs_hotplug_event(dev);
+}
+
 void radeon_driver_irq_preinstall_kms(struct drm_device *dev)
 {
 	struct radeon_device *rdev = dev->dev_private;
 	unsigned i;
 
+	INIT_WORK(&rdev->hotplug_work, radeon_hotplug_work_func);
+
 	/* Disable *all* interrupts */
 	rdev->irq.sw_int = false;
 	for (i = 0; i < 2; i++) {
@@ -76,6 +97,7 @@ void radeon_driver_irq_uninstall_kms(struct drm_device *dev)
 	rdev->irq.sw_int = false;
 	for (i = 0; i < 2; i++) {
 		rdev->irq.crtc_vblank_int[i] = false;
+		rdev->irq.hpd[i] = false;
 	}
 	radeon_irq_set(rdev);
 }
@@ -83,21 +105,73 @@ void radeon_driver_irq_uninstall_kms(struct drm_device *dev)
 int radeon_irq_kms_init(struct radeon_device *rdev)
 {
 	int r = 0;
+	int num_crtc = 2;
 
-	r = drm_vblank_init(rdev->ddev, 2);
+	if (rdev->flags & RADEON_SINGLE_CRTC)
+		num_crtc = 1;
+	spin_lock_init(&rdev->irq.sw_lock);
+	r = drm_vblank_init(rdev->ddev, num_crtc);
 	if (r) {
 		return r;
 	}
-	drm_irq_install(rdev->ddev);
+	/* enable msi */
+	rdev->msi_enabled = 0;
+	/* MSIs don't seem to work on my rs780;
+	 * not sure about rs880 or other rs780s.
+	 * Needs more investigation.
+	 */
+	if ((rdev->family >= CHIP_RV380) &&
+	    (rdev->family != CHIP_RS780) &&
+	    (rdev->family != CHIP_RS880)) {
+		int ret = pci_enable_msi(rdev->pdev);
+		if (!ret) {
+			rdev->msi_enabled = 1;
+			DRM_INFO("radeon: using MSI.\n");
+		}
+	}
 	rdev->irq.installed = true;
+	r = drm_irq_install(rdev->ddev);
+	if (r) {
+		rdev->irq.installed = false;
+		return r;
+	}
 	DRM_INFO("radeon: irq initialized.\n");
 	return 0;
 }
 
 void radeon_irq_kms_fini(struct radeon_device *rdev)
 {
+	drm_vblank_cleanup(rdev->ddev);
 	if (rdev->irq.installed) {
-		rdev->irq.installed = false;
 		drm_irq_uninstall(rdev->ddev);
+		rdev->irq.installed = false;
+		if (rdev->msi_enabled)
+			pci_disable_msi(rdev->pdev);
+	}
+}
+
+void radeon_irq_kms_sw_irq_get(struct radeon_device *rdev)
+{
+	unsigned long irqflags;
+
+	spin_lock_irqsave(&rdev->irq.sw_lock, irqflags);
+	if (rdev->ddev->irq_enabled && (++rdev->irq.sw_refcount == 1)) {
+		rdev->irq.sw_int = true;
+		radeon_irq_set(rdev);
 	}
+	spin_unlock_irqrestore(&rdev->irq.sw_lock, irqflags);
 }
+
+void radeon_irq_kms_sw_irq_put(struct radeon_device *rdev)
+{
+	unsigned long irqflags;
+
+	spin_lock_irqsave(&rdev->irq.sw_lock, irqflags);
+	BUG_ON(rdev->ddev->irq_enabled && rdev->irq.sw_refcount <= 0);
+	if (rdev->ddev->irq_enabled && (--rdev->irq.sw_refcount == 0)) {
+		rdev->irq.sw_int = false;
+		radeon_irq_set(rdev);
+	}
+	spin_unlock_irqrestore(&rdev->irq.sw_lock, irqflags);
+}
+
diff --git a/drivers/gpu/drm/radeon/radeon_kms.c b/drivers/gpu/drm/radeon/radeon_kms.c
index 709bd892b3a9..20ec276e7596 100644
--- a/drivers/gpu/drm/radeon/radeon_kms.c
+++ b/drivers/gpu/drm/radeon/radeon_kms.c
@@ -30,10 +30,21 @@
 #include "radeon.h"
 #include "radeon_drm.h"
 
+#include <linux/vga_switcheroo.h>
+
+int radeon_driver_unload_kms(struct drm_device *dev)
+{
+	struct radeon_device *rdev = dev->dev_private;
+
+	if (rdev == NULL)
+		return 0;
+	radeon_modeset_fini(rdev);
+	radeon_device_fini(rdev);
+	kfree(rdev);
+	dev->dev_private = NULL;
+	return 0;
+}
 
-/*
- * Driver load/unload
- */
 int radeon_driver_load_kms(struct drm_device *dev, unsigned long flags)
 {
 	struct radeon_device *rdev;
@@ -62,31 +73,20 @@ int radeon_driver_load_kms(struct drm_device *dev, unsigned long flags)
 	 */
 	r = radeon_device_init(rdev, dev, dev->pdev, flags);
 	if (r) {
-		DRM_ERROR("Fatal error while trying to initialize radeon.\n");
-		return r;
+		dev_err(&dev->pdev->dev, "Fatal error during GPU init\n");
+		goto out;
 	}
 	/* Again modeset_init should fail only on fatal error
 	 * otherwise it should provide enough functionalities
 	 * for shadowfb to run
 	 */
 	r = radeon_modeset_init(rdev);
-	if (r) {
-		return r;
-	}
-	return 0;
-}
-
-int radeon_driver_unload_kms(struct drm_device *dev)
-{
-	struct radeon_device *rdev = dev->dev_private;
-
-	if (rdev == NULL)
-		return 0;
-	radeon_modeset_fini(rdev);
-	radeon_device_fini(rdev);
-	kfree(rdev);
-	dev->dev_private = NULL;
-	return 0;
+	if (r)
+		dev_err(&dev->pdev->dev, "Fatal error during modeset init\n");
+out:
+	if (r)
+		radeon_driver_unload_kms(dev);
+	return r;
 }
 
 
@@ -138,6 +138,7 @@ int radeon_driver_firstopen_kms(struct drm_device *dev)
 
 void radeon_driver_lastclose_kms(struct drm_device *dev)
 {
+	vga_switcheroo_process_delayed_switch();
 }
 
 int radeon_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv)
@@ -201,55 +202,6 @@ void radeon_disable_vblank_kms(struct drm_device *dev, int crtc)
 
 
 /*
- * For multiple master (like multiple X).
- */
-struct drm_radeon_master_private {
-	drm_local_map_t *sarea;
-	drm_radeon_sarea_t *sarea_priv;
-};
-
-int radeon_master_create_kms(struct drm_device *dev, struct drm_master *master)
-{
-	struct drm_radeon_master_private *master_priv;
-	unsigned long sareapage;
-	int ret;
-
-	master_priv = kzalloc(sizeof(*master_priv), GFP_KERNEL);
-	if (master_priv == NULL) {
-		return -ENOMEM;
-	}
-	/* prebuild the SAREA */
-	sareapage = max_t(unsigned long, SAREA_MAX, PAGE_SIZE);
-	ret = drm_addmap(dev, 0, sareapage, _DRM_SHM,
-			 _DRM_CONTAINS_LOCK,
-			 &master_priv->sarea);
-	if (ret) {
-		DRM_ERROR("SAREA setup failed\n");
-		return ret;
-	}
-	master_priv->sarea_priv = master_priv->sarea->handle + sizeof(struct drm_sarea);
-	master_priv->sarea_priv->pfCurrentPage = 0;
-	master->driver_priv = master_priv;
-	return 0;
-}
-
-void radeon_master_destroy_kms(struct drm_device *dev,
-			       struct drm_master *master)
-{
-	struct drm_radeon_master_private *master_priv = master->driver_priv;
-
-	if (master_priv == NULL) {
-		return;
-	}
-	if (master_priv->sarea) {
-		drm_rmmap_locked(dev, master_priv->sarea);
-	}
-	kfree(master_priv);
-	master->driver_priv = NULL;
-}
-
-
-/*
  * IOCTL.
  */
 int radeon_dma_ioctl_kms(struct drm_device *dev, void *data,
@@ -327,17 +279,17 @@ struct drm_ioctl_desc radeon_ioctls_kms[] = {
 	DRM_IOCTL_DEF(DRM_RADEON_SURF_ALLOC, radeon_surface_alloc_kms, DRM_AUTH),
 	DRM_IOCTL_DEF(DRM_RADEON_SURF_FREE, radeon_surface_free_kms, DRM_AUTH),
 	/* KMS */
-	DRM_IOCTL_DEF(DRM_RADEON_GEM_INFO, radeon_gem_info_ioctl, DRM_AUTH),
-	DRM_IOCTL_DEF(DRM_RADEON_GEM_CREATE, radeon_gem_create_ioctl, DRM_AUTH),
-	DRM_IOCTL_DEF(DRM_RADEON_GEM_MMAP, radeon_gem_mmap_ioctl, DRM_AUTH),
-	DRM_IOCTL_DEF(DRM_RADEON_GEM_SET_DOMAIN, radeon_gem_set_domain_ioctl, DRM_AUTH),
-	DRM_IOCTL_DEF(DRM_RADEON_GEM_PREAD, radeon_gem_pread_ioctl, DRM_AUTH),
-	DRM_IOCTL_DEF(DRM_RADEON_GEM_PWRITE, radeon_gem_pwrite_ioctl, DRM_AUTH),
-	DRM_IOCTL_DEF(DRM_RADEON_GEM_WAIT_IDLE, radeon_gem_wait_idle_ioctl, DRM_AUTH),
-	DRM_IOCTL_DEF(DRM_RADEON_CS, radeon_cs_ioctl, DRM_AUTH),
-	DRM_IOCTL_DEF(DRM_RADEON_INFO, radeon_info_ioctl, DRM_AUTH),
-	DRM_IOCTL_DEF(DRM_RADEON_GEM_SET_TILING, radeon_gem_set_tiling_ioctl, DRM_AUTH),
-	DRM_IOCTL_DEF(DRM_RADEON_GEM_GET_TILING, radeon_gem_get_tiling_ioctl, DRM_AUTH),
-	DRM_IOCTL_DEF(DRM_RADEON_GEM_BUSY, radeon_gem_busy_ioctl, DRM_AUTH),
+	DRM_IOCTL_DEF(DRM_RADEON_GEM_INFO, radeon_gem_info_ioctl, DRM_AUTH|DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_RADEON_GEM_CREATE, radeon_gem_create_ioctl, DRM_AUTH|DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_RADEON_GEM_MMAP, radeon_gem_mmap_ioctl, DRM_AUTH|DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_RADEON_GEM_SET_DOMAIN, radeon_gem_set_domain_ioctl, DRM_AUTH|DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_RADEON_GEM_PREAD, radeon_gem_pread_ioctl, DRM_AUTH|DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_RADEON_GEM_PWRITE, radeon_gem_pwrite_ioctl, DRM_AUTH|DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_RADEON_GEM_WAIT_IDLE, radeon_gem_wait_idle_ioctl, DRM_AUTH|DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_RADEON_CS, radeon_cs_ioctl, DRM_AUTH|DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_RADEON_INFO, radeon_info_ioctl, DRM_AUTH|DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_RADEON_GEM_SET_TILING, radeon_gem_set_tiling_ioctl, DRM_AUTH|DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_RADEON_GEM_GET_TILING, radeon_gem_get_tiling_ioctl, DRM_AUTH|DRM_UNLOCKED),
+	DRM_IOCTL_DEF(DRM_RADEON_GEM_BUSY, radeon_gem_busy_ioctl, DRM_AUTH|DRM_UNLOCKED),
 };
 int radeon_max_kms_ioctl = DRM_ARRAY_SIZE(radeon_ioctls_kms);
diff --git a/drivers/gpu/drm/radeon/radeon_legacy_crtc.c b/drivers/gpu/drm/radeon/radeon_legacy_crtc.c
index 2b997a15fb1f..df23d6a01d02 100644
--- a/drivers/gpu/drm/radeon/radeon_legacy_crtc.c
+++ b/drivers/gpu/drm/radeon/radeon_legacy_crtc.c
@@ -30,9 +30,20 @@
 #include "radeon.h"
 #include "atom.h"
 
+static void radeon_overscan_setup(struct drm_crtc *crtc,
+				  struct drm_display_mode *mode)
+{
+	struct drm_device *dev = crtc->dev;
+	struct radeon_device *rdev = dev->dev_private;
+	struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc);
+
+	WREG32(RADEON_OVR_CLR + radeon_crtc->crtc_offset, 0);
+	WREG32(RADEON_OVR_WID_LEFT_RIGHT + radeon_crtc->crtc_offset, 0);
+	WREG32(RADEON_OVR_WID_TOP_BOTTOM + radeon_crtc->crtc_offset, 0);
+}
+
 static void radeon_legacy_rmx_mode_set(struct drm_crtc *crtc,
-				       struct drm_display_mode *mode,
-				       struct drm_display_mode *adjusted_mode)
+				       struct drm_display_mode *mode)
 {
 	struct drm_device *dev = crtc->dev;
 	struct radeon_device *rdev = dev->dev_private;
@@ -48,7 +59,7 @@ static void radeon_legacy_rmx_mode_set(struct drm_crtc *crtc,
 	u32 fp_horz_stretch, fp_vert_stretch, fp_horz_vert_active;
 	u32 fp_h_sync_strt_wid, fp_crtc_h_total_disp;
 	u32 fp_v_sync_strt_wid, fp_crtc_v_total_disp;
-	struct radeon_native_mode *native_mode = &radeon_crtc->native_mode;
+	struct drm_display_mode *native_mode = &radeon_crtc->native_mode;
 
 	fp_vert_stretch = RREG32(RADEON_FP_VERT_STRETCH) &
 		(RADEON_VERT_STRETCH_RESERVED |
@@ -95,19 +106,19 @@ static void radeon_legacy_rmx_mode_set(struct drm_crtc *crtc,
 
 	fp_horz_vert_active = 0;
 
-	if (native_mode->panel_xres == 0 ||
-	    native_mode->panel_yres == 0) {
+	if (native_mode->hdisplay == 0 ||
+	    native_mode->vdisplay == 0) {
 		hscale = false;
 		vscale = false;
 	} else {
-		if (xres > native_mode->panel_xres)
-			xres = native_mode->panel_xres;
-		if (yres > native_mode->panel_yres)
-			yres = native_mode->panel_yres;
+		if (xres > native_mode->hdisplay)
+			xres = native_mode->hdisplay;
+		if (yres > native_mode->vdisplay)
+			yres = native_mode->vdisplay;
 
-		if (xres == native_mode->panel_xres)
+		if (xres == native_mode->hdisplay)
 			hscale = false;
-		if (yres == native_mode->panel_yres)
+		if (yres == native_mode->vdisplay)
 			vscale = false;
 	}
 
@@ -119,11 +130,11 @@ static void radeon_legacy_rmx_mode_set(struct drm_crtc *crtc,
 		else {
 			inc = (fp_horz_stretch & RADEON_HORZ_AUTO_RATIO_INC) ? 1 : 0;
 			scale = ((xres + inc) * RADEON_HORZ_STRETCH_RATIO_MAX)
-				/ native_mode->panel_xres + 1;
+				/ native_mode->hdisplay + 1;
 			fp_horz_stretch |= (((scale) & RADEON_HORZ_STRETCH_RATIO_MASK) |
 					RADEON_HORZ_STRETCH_BLEND |
 					RADEON_HORZ_STRETCH_ENABLE |
-					((native_mode->panel_xres/8-1) << 16));
+					((native_mode->hdisplay/8-1) << 16));
 		}
 
 		if (!vscale)
@@ -131,11 +142,11 @@ static void radeon_legacy_rmx_mode_set(struct drm_crtc *crtc,
 		else {
 			inc = (fp_vert_stretch & RADEON_VERT_AUTO_RATIO_INC) ? 1 : 0;
 			scale = ((yres + inc) * RADEON_VERT_STRETCH_RATIO_MAX)
-				/ native_mode->panel_yres + 1;
+				/ native_mode->vdisplay + 1;
 			fp_vert_stretch |= (((scale) & RADEON_VERT_STRETCH_RATIO_MASK) |
 					RADEON_VERT_STRETCH_ENABLE |
 					RADEON_VERT_STRETCH_BLEND |
-					((native_mode->panel_yres-1) << 12));
+					((native_mode->vdisplay-1) << 12));
 		}
 		break;
 	case RMX_CENTER:
@@ -175,8 +186,8 @@ static void radeon_legacy_rmx_mode_set(struct drm_crtc *crtc,
 						? RADEON_CRTC_V_SYNC_POL
 						: 0)));
 
-		fp_horz_vert_active = (((native_mode->panel_yres) & 0xfff) |
-				(((native_mode->panel_xres / 8) & 0x1ff) << 16));
+		fp_horz_vert_active = (((native_mode->vdisplay) & 0xfff) |
+				(((native_mode->hdisplay / 8) & 0x1ff) << 16));
 		break;
 	case RMX_OFF:
 	default:
@@ -292,8 +303,7 @@ void radeon_crtc_dpms(struct drm_crtc *crtc, int mode)
 	uint32_t mask;
 
 	if (radeon_crtc->crtc_id)
-		mask = (RADEON_CRTC2_EN |
-			RADEON_CRTC2_DISP_DIS |
+		mask = (RADEON_CRTC2_DISP_DIS |
 			RADEON_CRTC2_VSYNC_DIS |
 			RADEON_CRTC2_HSYNC_DIS |
 			RADEON_CRTC2_DISP_REQ_EN_B);
@@ -305,7 +315,7 @@ void radeon_crtc_dpms(struct drm_crtc *crtc, int mode)
 	switch (mode) {
 	case DRM_MODE_DPMS_ON:
 		if (radeon_crtc->crtc_id)
-			WREG32_P(RADEON_CRTC2_GEN_CNTL, RADEON_CRTC2_EN, ~mask);
+			WREG32_P(RADEON_CRTC2_GEN_CNTL, RADEON_CRTC2_EN, ~(RADEON_CRTC2_EN | mask));
 		else {
 			WREG32_P(RADEON_CRTC_GEN_CNTL, RADEON_CRTC_EN, ~(RADEON_CRTC_EN |
 									 RADEON_CRTC_DISP_REQ_EN_B));
@@ -319,7 +329,7 @@ void radeon_crtc_dpms(struct drm_crtc *crtc, int mode)
 	case DRM_MODE_DPMS_OFF:
 		drm_vblank_pre_modeset(dev, radeon_crtc->crtc_id);
 		if (radeon_crtc->crtc_id)
-			WREG32_P(RADEON_CRTC2_GEN_CNTL, mask, ~mask);
+			WREG32_P(RADEON_CRTC2_GEN_CNTL, mask, ~(RADEON_CRTC2_EN | mask));
 		else {
 			WREG32_P(RADEON_CRTC_GEN_CNTL, RADEON_CRTC_DISP_REQ_EN_B, ~(RADEON_CRTC_EN |
 										    RADEON_CRTC_DISP_REQ_EN_B));
@@ -329,69 +339,6 @@ void radeon_crtc_dpms(struct drm_crtc *crtc, int mode)
 	}
 }
 
-/* properly set crtc bpp when using atombios */
-void radeon_legacy_atom_set_surface(struct drm_crtc *crtc)
-{
-	struct drm_device *dev = crtc->dev;
-	struct radeon_device *rdev = dev->dev_private;
-	struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc);
-	int format;
-	uint32_t crtc_gen_cntl;
-	uint32_t disp_merge_cntl;
-	uint32_t crtc_pitch;
-
-	switch (crtc->fb->bits_per_pixel) {
-	case 8:
-		format = 2;
-		break;
-	case 15:      /*  555 */
-		format = 3;
-		break;
-	case 16:      /*  565 */
-		format = 4;
-		break;
-	case 24:      /*  RGB */
-		format = 5;
-		break;
-	case 32:      /* xRGB */
-		format = 6;
-		break;
-	default:
-		return;
-	}
-
-	crtc_pitch  = ((((crtc->fb->pitch / (crtc->fb->bits_per_pixel / 8)) * crtc->fb->bits_per_pixel) +
-			((crtc->fb->bits_per_pixel * 8) - 1)) /
-		       (crtc->fb->bits_per_pixel * 8));
-	crtc_pitch |= crtc_pitch << 16;
-
-	WREG32(RADEON_CRTC_PITCH + radeon_crtc->crtc_offset, crtc_pitch);
-
-	switch (radeon_crtc->crtc_id) {
-	case 0:
-		disp_merge_cntl = RREG32(RADEON_DISP_MERGE_CNTL);
-		disp_merge_cntl &= ~RADEON_DISP_RGB_OFFSET_EN;
-		WREG32(RADEON_DISP_MERGE_CNTL, disp_merge_cntl);
-
-		crtc_gen_cntl = RREG32(RADEON_CRTC_GEN_CNTL) & 0xfffff0ff;
-		crtc_gen_cntl |= (format << 8);
-		crtc_gen_cntl |= RADEON_CRTC_EXT_DISP_EN;
-		WREG32(RADEON_CRTC_GEN_CNTL, crtc_gen_cntl);
-		break;
-	case 1:
-		disp_merge_cntl = RREG32(RADEON_DISP2_MERGE_CNTL);
-		disp_merge_cntl &= ~RADEON_DISP2_RGB_OFFSET_EN;
-		WREG32(RADEON_DISP2_MERGE_CNTL, disp_merge_cntl);
-
-		crtc_gen_cntl = RREG32(RADEON_CRTC2_GEN_CNTL) & 0xfffff0ff;
-		crtc_gen_cntl |= (format << 8);
-		WREG32(RADEON_CRTC2_GEN_CNTL, crtc_gen_cntl);
-		WREG32(RADEON_FP_H2_SYNC_STRT_WID,   RREG32(RADEON_CRTC2_H_SYNC_STRT_WID));
-		WREG32(RADEON_FP_V2_SYNC_STRT_WID,   RREG32(RADEON_CRTC2_V_SYNC_STRT_WID));
-		break;
-	}
-}
-
 int radeon_crtc_set_base(struct drm_crtc *crtc, int x, int y,
 			 struct drm_framebuffer *old_fb)
 {
@@ -400,14 +347,21 @@ int radeon_crtc_set_base(struct drm_crtc *crtc, int x, int y,
 	struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc);
 	struct radeon_framebuffer *radeon_fb;
 	struct drm_gem_object *obj;
+	struct radeon_bo *rbo;
 	uint64_t base;
 	uint32_t crtc_offset, crtc_offset_cntl, crtc_tile_x0_y0 = 0;
 	uint32_t crtc_pitch, pitch_pixels;
 	uint32_t tiling_flags;
 	int format;
 	uint32_t gen_cntl_reg, gen_cntl_val;
+	int r;
 
 	DRM_DEBUG("\n");
+	/* no fb bound */
+	if (!crtc->fb) {
+		DRM_DEBUG("No FB bound\n");
+		return 0;
+	}
 
 	radeon_fb = to_radeon_framebuffer(crtc->fb);
 
@@ -431,13 +385,25 @@ int radeon_crtc_set_base(struct drm_crtc *crtc, int x, int y,
 		return false;
 	}
 
+	/* Pin framebuffer & get tilling informations */
 	obj = radeon_fb->obj;
-	if (radeon_gem_object_pin(obj, RADEON_GEM_DOMAIN_VRAM, &base)) {
+	rbo = obj->driver_private;
+	r = radeon_bo_reserve(rbo, false);
+	if (unlikely(r != 0))
+		return r;
+	r = radeon_bo_pin(rbo, RADEON_GEM_DOMAIN_VRAM, &base);
+	if (unlikely(r != 0)) {
+		radeon_bo_unreserve(rbo);
 		return -EINVAL;
 	}
+	radeon_bo_get_tiling_flags(rbo, &tiling_flags, NULL);
+	radeon_bo_unreserve(rbo);
+	if (tiling_flags & RADEON_TILING_MICRO)
+		DRM_ERROR("trying to scanout microtiled buffer\n");
+
 	/* if scanout was in GTT this really wouldn't work */
 	/* crtc offset is from display base addr not FB location */
-	radeon_crtc->legacy_display_base_addr = rdev->mc.vram_location;
+	radeon_crtc->legacy_display_base_addr = rdev->mc.vram_start;
 
 	base -= radeon_crtc->legacy_display_base_addr;
 
@@ -449,10 +415,6 @@ int radeon_crtc_set_base(struct drm_crtc *crtc, int x, int y,
 		       (crtc->fb->bits_per_pixel * 8));
 	crtc_pitch |= crtc_pitch << 16;
 
-	radeon_object_get_tiling_flags(obj->driver_private,
-				       &tiling_flags, NULL);
-	if (tiling_flags & RADEON_TILING_MICRO)
-		DRM_ERROR("trying to scanout microtiled buffer\n");
 
 	if (tiling_flags & RADEON_TILING_MACRO) {
 		if (ASIC_IS_R300(rdev))
@@ -530,8 +492,17 @@ int radeon_crtc_set_base(struct drm_crtc *crtc, int x, int y,
 
 	if (old_fb && old_fb != crtc->fb) {
 		radeon_fb = to_radeon_framebuffer(old_fb);
-		radeon_gem_object_unpin(radeon_fb->obj);
+		rbo = radeon_fb->obj->driver_private;
+		r = radeon_bo_reserve(rbo, false);
+		if (unlikely(r != 0))
+			return r;
+		radeon_bo_unpin(rbo);
+		radeon_bo_unreserve(rbo);
 	}
+
+	/* Bytes per pixel may have changed */
+	radeon_bandwidth_update(rdev);
+
 	return 0;
 }
 
@@ -611,39 +582,12 @@ static bool radeon_set_crtc_timing(struct drm_crtc *crtc, struct drm_display_mod
 				   ? RADEON_CRTC_V_SYNC_POL
 				   : 0));
 
-	/* TODO -> Dell Server */
-	if (0) {
-		uint32_t disp_hw_debug = RREG32(RADEON_DISP_HW_DEBUG);
-		uint32_t tv_dac_cntl = RREG32(RADEON_TV_DAC_CNTL);
-		uint32_t dac2_cntl = RREG32(RADEON_DAC_CNTL2);
-		uint32_t crtc2_gen_cntl = RREG32(RADEON_CRTC2_GEN_CNTL);
-
-		dac2_cntl &= ~RADEON_DAC2_DAC_CLK_SEL;
-		dac2_cntl |= RADEON_DAC2_DAC2_CLK_SEL;
-
-		/* For CRT on DAC2, don't turn it on if BIOS didn't
-		   enable it, even it's detected.
-		*/
-		disp_hw_debug |= RADEON_CRT2_DISP1_SEL;
-		tv_dac_cntl &= ~((1<<2) | (3<<8) | (7<<24) | (0xff<<16));
-		tv_dac_cntl |= (0x03 | (2<<8) | (0x58<<16));
-
-		WREG32(RADEON_TV_DAC_CNTL, tv_dac_cntl);
-		WREG32(RADEON_DISP_HW_DEBUG, disp_hw_debug);
-		WREG32(RADEON_DAC_CNTL2, dac2_cntl);
-		WREG32(RADEON_CRTC2_GEN_CNTL, crtc2_gen_cntl);
-	}
-
 	if (radeon_crtc->crtc_id) {
 		uint32_t crtc2_gen_cntl;
 		uint32_t disp2_merge_cntl;
 
-		/* check to see if TV DAC is enabled for another crtc and keep it enabled */
-		if (RREG32(RADEON_CRTC2_GEN_CNTL) & RADEON_CRTC2_CRT2_ON)
-			crtc2_gen_cntl = RADEON_CRTC2_CRT2_ON;
-		else
-			crtc2_gen_cntl = 0;
-
+		/* if TV DAC is enabled for another crtc and keep it enabled */
+		crtc2_gen_cntl = RREG32(RADEON_CRTC2_GEN_CNTL) & 0x00718080;
 		crtc2_gen_cntl |= ((format << 8)
 				   | RADEON_CRTC2_VSYNC_DIS
 				   | RADEON_CRTC2_HSYNC_DIS
@@ -664,12 +608,16 @@ static bool radeon_set_crtc_timing(struct drm_crtc *crtc, struct drm_display_mod
 
 		WREG32(RADEON_DISP2_MERGE_CNTL, disp2_merge_cntl);
 		WREG32(RADEON_CRTC2_GEN_CNTL, crtc2_gen_cntl);
+
+		WREG32(RADEON_FP_H2_SYNC_STRT_WID, crtc_h_sync_strt_wid);
+		WREG32(RADEON_FP_V2_SYNC_STRT_WID, crtc_v_sync_strt_wid);
 	} else {
 		uint32_t crtc_gen_cntl;
 		uint32_t crtc_ext_cntl;
 		uint32_t disp_merge_cntl;
 
-		crtc_gen_cntl = (RADEON_CRTC_EXT_DISP_EN
+		crtc_gen_cntl = RREG32(RADEON_CRTC_GEN_CNTL) & 0x00718000;
+		crtc_gen_cntl |= (RADEON_CRTC_EXT_DISP_EN
 				 | (format << 8)
 				 | RADEON_CRTC_DISP_REQ_EN_B
 				 | ((mode->flags & DRM_MODE_FLAG_DBLSCAN)
@@ -721,7 +669,6 @@ static void radeon_set_pll(struct drm_crtc *crtc, struct drm_display_mode *mode)
 	uint32_t post_divider = 0;
 	uint32_t freq = 0;
 	uint8_t pll_gain;
-	int pll_flags = RADEON_PLL_LEGACY;
 	bool use_bios_divs = false;
 	/* PLL registers */
 	uint32_t pll_ref_div = 0;
@@ -755,10 +702,16 @@ static void radeon_set_pll(struct drm_crtc *crtc, struct drm_display_mode *mode)
 	else
 		pll = &rdev->clock.p1pll;
 
+	pll->flags = RADEON_PLL_LEGACY;
+	if (radeon_new_pll == 1)
+		pll->algo = PLL_ALGO_NEW;
+	else
+		pll->algo = PLL_ALGO_LEGACY;
+
 	if (mode->clock > 200000) /* range limits??? */
-		pll_flags |= RADEON_PLL_PREFER_HIGH_FB_DIV;
+		pll->flags |= RADEON_PLL_PREFER_HIGH_FB_DIV;
 	else
-		pll_flags |= RADEON_PLL_PREFER_LOW_REF_DIV;
+		pll->flags |= RADEON_PLL_PREFER_LOW_REF_DIV;
 
 	list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) {
 		if (encoder->crtc == crtc) {
@@ -770,20 +723,22 @@ static void radeon_set_pll(struct drm_crtc *crtc, struct drm_display_mode *mode)
 			}
 
 			if (encoder->encoder_type != DRM_MODE_ENCODER_DAC)
-				pll_flags |= RADEON_PLL_NO_ODD_POST_DIV;
+				pll->flags |= RADEON_PLL_NO_ODD_POST_DIV;
 			if (encoder->encoder_type == DRM_MODE_ENCODER_LVDS) {
-				struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
-				struct radeon_encoder_lvds *lvds = (struct radeon_encoder_lvds *)radeon_encoder->enc_priv;
-				if (lvds) {
-					if (lvds->use_bios_dividers) {
-						pll_ref_div = lvds->panel_ref_divider;
-						pll_fb_post_div   = (lvds->panel_fb_divider |
-								     (lvds->panel_post_divider << 16));
-						htotal_cntl  = 0;
-						use_bios_divs = true;
+				if (!rdev->is_atom_bios) {
+					struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
+					struct radeon_encoder_lvds *lvds = (struct radeon_encoder_lvds *)radeon_encoder->enc_priv;
+					if (lvds) {
+						if (lvds->use_bios_dividers) {
+							pll_ref_div = lvds->panel_ref_divider;
+							pll_fb_post_div   = (lvds->panel_fb_divider |
+									     (lvds->panel_post_divider << 16));
+							htotal_cntl  = 0;
+							use_bios_divs = true;
+						}
 					}
 				}
-				pll_flags |= RADEON_PLL_USE_REF_DIV;
+				pll->flags |= RADEON_PLL_USE_REF_DIV;
 			}
 		}
 	}
@@ -793,8 +748,7 @@ static void radeon_set_pll(struct drm_crtc *crtc, struct drm_display_mode *mode)
 	if (!use_bios_divs) {
 		radeon_compute_pll(pll, mode->clock,
 				   &freq, &feedback_div, &frac_fb_div,
-				   &reference_div, &post_divider,
-				   pll_flags);
+				   &reference_div, &post_divider);
 
 		for (post_div = &post_divs[0]; post_div->divider; ++post_div) {
 			if (post_div->divider == post_divider)
@@ -1015,16 +969,14 @@ static int radeon_crtc_mode_set(struct drm_crtc *crtc,
 				 int x, int y, struct drm_framebuffer *old_fb)
 {
 	struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc);
-	struct drm_device *dev = crtc->dev;
-	struct radeon_device *rdev = dev->dev_private;
 
 	/* TODO TV */
 	radeon_crtc_set_base(crtc, x, y, old_fb);
 	radeon_set_crtc_timing(crtc, adjusted_mode);
 	radeon_set_pll(crtc, adjusted_mode);
-	radeon_bandwidth_update(rdev);
+	radeon_overscan_setup(crtc, adjusted_mode);
 	if (radeon_crtc->crtc_id == 0) {
-		radeon_legacy_rmx_mode_set(crtc, mode, adjusted_mode);
+		radeon_legacy_rmx_mode_set(crtc, adjusted_mode);
 	} else {
 		if (radeon_crtc->rmx_type != RMX_OFF) {
 			/* FIXME: only first crtc has rmx what should we
@@ -1038,12 +990,29 @@ static int radeon_crtc_mode_set(struct drm_crtc *crtc,
 
 static void radeon_crtc_prepare(struct drm_crtc *crtc)
 {
-	radeon_crtc_dpms(crtc, DRM_MODE_DPMS_OFF);
+	struct drm_device *dev = crtc->dev;
+	struct drm_crtc *crtci;
+
+	/*
+	* The hardware wedges sometimes if you reconfigure one CRTC
+	* whilst another is running (see fdo bug #24611).
+	*/
+	list_for_each_entry(crtci, &dev->mode_config.crtc_list, head)
+		radeon_crtc_dpms(crtci, DRM_MODE_DPMS_OFF);
 }
 
 static void radeon_crtc_commit(struct drm_crtc *crtc)
 {
-	radeon_crtc_dpms(crtc, DRM_MODE_DPMS_ON);
+	struct drm_device *dev = crtc->dev;
+	struct drm_crtc *crtci;
+
+	/*
+	* Reenable the CRTCs that should be running.
+	*/
+	list_for_each_entry(crtci, &dev->mode_config.crtc_list, head) {
+		if (crtci->enabled)
+			radeon_crtc_dpms(crtci, DRM_MODE_DPMS_ON);
+	}
 }
 
 static const struct drm_crtc_helper_funcs legacy_helper_funcs = {
@@ -1053,6 +1022,7 @@ static const struct drm_crtc_helper_funcs legacy_helper_funcs = {
 	.mode_set_base = radeon_crtc_set_base,
 	.prepare = radeon_crtc_prepare,
 	.commit = radeon_crtc_commit,
+	.load_lut = radeon_crtc_load_lut,
 };
 
 
diff --git a/drivers/gpu/drm/radeon/radeon_legacy_encoders.c b/drivers/gpu/drm/radeon/radeon_legacy_encoders.c
index b1547f700d73..cf389ce50a8a 100644
--- a/drivers/gpu/drm/radeon/radeon_legacy_encoders.c
+++ b/drivers/gpu/drm/radeon/radeon_legacy_encoders.c
@@ -46,6 +46,7 @@ static void radeon_legacy_lvds_dpms(struct drm_encoder *encoder, int mode)
 	struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
 	uint32_t lvds_gen_cntl, lvds_pll_cntl, pixclks_cntl, disp_pwr_man;
 	int panel_pwr_delay = 2000;
+	bool is_mac = false;
 	DRM_DEBUG("\n");
 
 	if (radeon_encoder->enc_priv) {
@@ -58,6 +59,15 @@ static void radeon_legacy_lvds_dpms(struct drm_encoder *encoder, int mode)
 		}
 	}
 
+	/* macs (and possibly some x86 oem systems?) wire up LVDS strangely
+	 * Taken from radeonfb.
+	 */
+	if ((rdev->mode_info.connector_table == CT_IBOOK) ||
+	    (rdev->mode_info.connector_table == CT_POWERBOOK_EXTERNAL) ||
+	    (rdev->mode_info.connector_table == CT_POWERBOOK_INTERNAL) ||
+	    (rdev->mode_info.connector_table == CT_POWERBOOK_VGA))
+		is_mac = true;
+
 	switch (mode) {
 	case DRM_MODE_DPMS_ON:
 		disp_pwr_man = RREG32(RADEON_DISP_PWR_MAN);
@@ -74,6 +84,8 @@ static void radeon_legacy_lvds_dpms(struct drm_encoder *encoder, int mode)
 
 		lvds_gen_cntl = RREG32(RADEON_LVDS_GEN_CNTL);
 		lvds_gen_cntl |= (RADEON_LVDS_ON | RADEON_LVDS_EN | RADEON_LVDS_DIGON | RADEON_LVDS_BLON);
+		if (is_mac)
+			lvds_gen_cntl |= RADEON_LVDS_BL_MOD_EN;
 		lvds_gen_cntl &= ~(RADEON_LVDS_DISPLAY_DIS);
 		udelay(panel_pwr_delay * 1000);
 		WREG32(RADEON_LVDS_GEN_CNTL, lvds_gen_cntl);
@@ -85,7 +97,14 @@ static void radeon_legacy_lvds_dpms(struct drm_encoder *encoder, int mode)
 		WREG32_PLL_P(RADEON_PIXCLKS_CNTL, 0, ~RADEON_PIXCLK_LVDS_ALWAYS_ONb);
 		lvds_gen_cntl = RREG32(RADEON_LVDS_GEN_CNTL);
 		lvds_gen_cntl |= RADEON_LVDS_DISPLAY_DIS;
-		lvds_gen_cntl &= ~(RADEON_LVDS_ON | RADEON_LVDS_BLON | RADEON_LVDS_EN | RADEON_LVDS_DIGON);
+		if (is_mac) {
+			lvds_gen_cntl &= ~RADEON_LVDS_BL_MOD_EN;
+			WREG32(RADEON_LVDS_GEN_CNTL, lvds_gen_cntl);
+			lvds_gen_cntl &= ~(RADEON_LVDS_ON | RADEON_LVDS_EN);
+		} else {
+			WREG32(RADEON_LVDS_GEN_CNTL, lvds_gen_cntl);
+			lvds_gen_cntl &= ~(RADEON_LVDS_ON | RADEON_LVDS_BLON | RADEON_LVDS_EN | RADEON_LVDS_DIGON);
+		}
 		udelay(panel_pwr_delay * 1000);
 		WREG32(RADEON_LVDS_GEN_CNTL, lvds_gen_cntl);
 		WREG32_PLL(RADEON_PIXCLKS_CNTL, pixclks_cntl);
@@ -96,6 +115,9 @@ static void radeon_legacy_lvds_dpms(struct drm_encoder *encoder, int mode)
 		radeon_atombios_encoder_dpms_scratch_regs(encoder, (mode == DRM_MODE_DPMS_ON) ? true : false);
 	else
 		radeon_combios_encoder_dpms_scratch_regs(encoder, (mode == DRM_MODE_DPMS_ON) ? true : false);
+
+	/* adjust pm to dpms change */
+	radeon_pm_compute_clocks(rdev);
 }
 
 static void radeon_legacy_lvds_prepare(struct drm_encoder *encoder)
@@ -107,8 +129,6 @@ static void radeon_legacy_lvds_prepare(struct drm_encoder *encoder)
 	else
 		radeon_combios_output_lock(encoder, true);
 	radeon_legacy_lvds_dpms(encoder, DRM_MODE_DPMS_OFF);
-
-	radeon_encoder_set_active_device(encoder);
 }
 
 static void radeon_legacy_lvds_commit(struct drm_encoder *encoder)
@@ -138,7 +158,14 @@ static void radeon_legacy_lvds_mode_set(struct drm_encoder *encoder,
 	lvds_pll_cntl &= ~RADEON_LVDS_PLL_EN;
 
 	lvds_ss_gen_cntl = RREG32(RADEON_LVDS_SS_GEN_CNTL);
-	if ((!rdev->is_atom_bios)) {
+	if (rdev->is_atom_bios) {
+		/* LVDS_GEN_CNTL parameters are computed in LVDSEncoderControl
+		 * need to call that on resume to set up the reg properly.
+		 */
+		radeon_encoder->pixel_clock = adjusted_mode->clock;
+		atombios_digital_setup(encoder, PANEL_ENCODER_ACTION_ENABLE);
+		lvds_gen_cntl = RREG32(RADEON_LVDS_GEN_CNTL);
+	} else {
 		struct radeon_encoder_lvds *lvds = (struct radeon_encoder_lvds *)radeon_encoder->enc_priv;
 		if (lvds) {
 			DRM_DEBUG("bios LVDS_GEN_CNTL: 0x%x\n", lvds->lvds_gen_cntl);
@@ -149,8 +176,7 @@ static void radeon_legacy_lvds_mode_set(struct drm_encoder *encoder,
 					     (lvds->panel_blon_delay << RADEON_LVDS_PWRSEQ_DELAY2_SHIFT));
 		} else
 			lvds_gen_cntl = RREG32(RADEON_LVDS_GEN_CNTL);
-	} else
-		lvds_gen_cntl = RREG32(RADEON_LVDS_GEN_CNTL);
+	}
 	lvds_gen_cntl |= RADEON_LVDS_DISPLAY_DIS;
 	lvds_gen_cntl &= ~(RADEON_LVDS_ON |
 			   RADEON_LVDS_BLON |
@@ -186,23 +212,39 @@ static void radeon_legacy_lvds_mode_set(struct drm_encoder *encoder,
 		radeon_combios_encoder_crtc_scratch_regs(encoder, radeon_crtc->crtc_id);
 }
 
-static bool radeon_legacy_lvds_mode_fixup(struct drm_encoder *encoder,
-					  struct drm_display_mode *mode,
-					  struct drm_display_mode *adjusted_mode)
+static bool radeon_legacy_mode_fixup(struct drm_encoder *encoder,
+				     struct drm_display_mode *mode,
+				     struct drm_display_mode *adjusted_mode)
 {
 	struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
+	struct drm_device *dev = encoder->dev;
+	struct radeon_device *rdev = dev->dev_private;
+
+	/* adjust pm to upcoming mode change */
+	radeon_pm_compute_clocks(rdev);
 
+	/* set the active encoder to connector routing */
+	radeon_encoder_set_active_device(encoder);
 	drm_mode_set_crtcinfo(adjusted_mode, 0);
 
-	if (radeon_encoder->rmx_type != RMX_OFF)
-		radeon_rmx_mode_fixup(encoder, mode, adjusted_mode);
+	/* get the native mode for LVDS */
+	if (radeon_encoder->active_device & (ATOM_DEVICE_LCD_SUPPORT)) {
+		struct drm_display_mode *native_mode = &radeon_encoder->native_mode;
+		int mode_id = adjusted_mode->base.id;
+		*adjusted_mode = *native_mode;
+		adjusted_mode->hdisplay = mode->hdisplay;
+		adjusted_mode->vdisplay = mode->vdisplay;
+		adjusted_mode->crtc_hdisplay = mode->hdisplay;
+		adjusted_mode->crtc_vdisplay = mode->vdisplay;
+		adjusted_mode->base.id = mode_id;
+	}
 
 	return true;
 }
 
 static const struct drm_encoder_helper_funcs radeon_legacy_lvds_helper_funcs = {
 	.dpms = radeon_legacy_lvds_dpms,
-	.mode_fixup = radeon_legacy_lvds_mode_fixup,
+	.mode_fixup = radeon_legacy_mode_fixup,
 	.prepare = radeon_legacy_lvds_prepare,
 	.mode_set = radeon_legacy_lvds_mode_set,
 	.commit = radeon_legacy_lvds_commit,
@@ -214,16 +256,6 @@ static const struct drm_encoder_funcs radeon_legacy_lvds_enc_funcs = {
 	.destroy = radeon_enc_destroy,
 };
 
-static bool radeon_legacy_primary_dac_mode_fixup(struct drm_encoder *encoder,
-						 struct drm_display_mode *mode,
-						 struct drm_display_mode *adjusted_mode)
-{
-
-	drm_mode_set_crtcinfo(adjusted_mode, 0);
-
-	return true;
-}
-
 static void radeon_legacy_primary_dac_dpms(struct drm_encoder *encoder, int mode)
 {
 	struct drm_device *dev = encoder->dev;
@@ -261,6 +293,9 @@ static void radeon_legacy_primary_dac_dpms(struct drm_encoder *encoder, int mode
 		radeon_atombios_encoder_dpms_scratch_regs(encoder, (mode == DRM_MODE_DPMS_ON) ? true : false);
 	else
 		radeon_combios_encoder_dpms_scratch_regs(encoder, (mode == DRM_MODE_DPMS_ON) ? true : false);
+
+	/* adjust pm to dpms change */
+	radeon_pm_compute_clocks(rdev);
 }
 
 static void radeon_legacy_primary_dac_prepare(struct drm_encoder *encoder)
@@ -272,7 +307,6 @@ static void radeon_legacy_primary_dac_prepare(struct drm_encoder *encoder)
 	else
 		radeon_combios_output_lock(encoder, true);
 	radeon_legacy_primary_dac_dpms(encoder, DRM_MODE_DPMS_OFF);
-	radeon_encoder_set_active_device(encoder);
 }
 
 static void radeon_legacy_primary_dac_commit(struct drm_encoder *encoder)
@@ -410,7 +444,7 @@ static enum drm_connector_status radeon_legacy_primary_dac_detect(struct drm_enc
 
 static const struct drm_encoder_helper_funcs radeon_legacy_primary_dac_helper_funcs = {
 	.dpms = radeon_legacy_primary_dac_dpms,
-	.mode_fixup = radeon_legacy_primary_dac_mode_fixup,
+	.mode_fixup = radeon_legacy_mode_fixup,
 	.prepare = radeon_legacy_primary_dac_prepare,
 	.mode_set = radeon_legacy_primary_dac_mode_set,
 	.commit = radeon_legacy_primary_dac_commit,
@@ -423,16 +457,6 @@ static const struct drm_encoder_funcs radeon_legacy_primary_dac_enc_funcs = {
 	.destroy = radeon_enc_destroy,
 };
 
-static bool radeon_legacy_tmds_int_mode_fixup(struct drm_encoder *encoder,
-					      struct drm_display_mode *mode,
-					      struct drm_display_mode *adjusted_mode)
-{
-
-	drm_mode_set_crtcinfo(adjusted_mode, 0);
-
-	return true;
-}
-
 static void radeon_legacy_tmds_int_dpms(struct drm_encoder *encoder, int mode)
 {
 	struct drm_device *dev = encoder->dev;
@@ -457,6 +481,9 @@ static void radeon_legacy_tmds_int_dpms(struct drm_encoder *encoder, int mode)
 		radeon_atombios_encoder_dpms_scratch_regs(encoder, (mode == DRM_MODE_DPMS_ON) ? true : false);
 	else
 		radeon_combios_encoder_dpms_scratch_regs(encoder, (mode == DRM_MODE_DPMS_ON) ? true : false);
+
+	/* adjust pm to dpms change */
+	radeon_pm_compute_clocks(rdev);
 }
 
 static void radeon_legacy_tmds_int_prepare(struct drm_encoder *encoder)
@@ -468,7 +495,6 @@ static void radeon_legacy_tmds_int_prepare(struct drm_encoder *encoder)
 	else
 		radeon_combios_output_lock(encoder, true);
 	radeon_legacy_tmds_int_dpms(encoder, DRM_MODE_DPMS_OFF);
-	radeon_encoder_set_active_device(encoder);
 }
 
 static void radeon_legacy_tmds_int_commit(struct drm_encoder *encoder)
@@ -543,6 +569,14 @@ static void radeon_legacy_tmds_int_mode_set(struct drm_encoder *encoder,
 
     fp_gen_cntl &= ~(RADEON_FP_FPON | RADEON_FP_TMDS_EN);
 
+    fp_gen_cntl &= ~(RADEON_FP_RMX_HVSYNC_CONTROL_EN |
+		     RADEON_FP_DFP_SYNC_SEL |
+		     RADEON_FP_CRT_SYNC_SEL |
+		     RADEON_FP_CRTC_LOCK_8DOT |
+		     RADEON_FP_USE_SHADOW_EN |
+		     RADEON_FP_CRTC_USE_SHADOW_VEND |
+		     RADEON_FP_CRT_SYNC_ALT);
+
     if (1) /*  FIXME rgbBits == 8 */
 	    fp_gen_cntl |= RADEON_FP_PANEL_FORMAT;  /* 24 bit format */
     else
@@ -556,7 +590,7 @@ static void radeon_legacy_tmds_int_mode_set(struct drm_encoder *encoder,
 		    else
 			    fp_gen_cntl |= R200_FP_SOURCE_SEL_CRTC1;
 	    } else
-		    fp_gen_cntl |= RADEON_FP_SEL_CRTC1;
+		    fp_gen_cntl &= ~RADEON_FP_SEL_CRTC2;
     } else {
 	    if (ASIC_IS_R300(rdev) || rdev->family == CHIP_R200) {
 		    fp_gen_cntl &= ~R200_FP_SOURCE_SEL_MASK;
@@ -577,7 +611,7 @@ static void radeon_legacy_tmds_int_mode_set(struct drm_encoder *encoder,
 
 static const struct drm_encoder_helper_funcs radeon_legacy_tmds_int_helper_funcs = {
 	.dpms = radeon_legacy_tmds_int_dpms,
-	.mode_fixup = radeon_legacy_tmds_int_mode_fixup,
+	.mode_fixup = radeon_legacy_mode_fixup,
 	.prepare = radeon_legacy_tmds_int_prepare,
 	.mode_set = radeon_legacy_tmds_int_mode_set,
 	.commit = radeon_legacy_tmds_int_commit,
@@ -589,16 +623,6 @@ static const struct drm_encoder_funcs radeon_legacy_tmds_int_enc_funcs = {
 	.destroy = radeon_enc_destroy,
 };
 
-static bool radeon_legacy_tmds_ext_mode_fixup(struct drm_encoder *encoder,
-					      struct drm_display_mode *mode,
-					      struct drm_display_mode *adjusted_mode)
-{
-
-	drm_mode_set_crtcinfo(adjusted_mode, 0);
-
-	return true;
-}
-
 static void radeon_legacy_tmds_ext_dpms(struct drm_encoder *encoder, int mode)
 {
 	struct drm_device *dev = encoder->dev;
@@ -625,6 +649,9 @@ static void radeon_legacy_tmds_ext_dpms(struct drm_encoder *encoder, int mode)
 		radeon_atombios_encoder_dpms_scratch_regs(encoder, (mode == DRM_MODE_DPMS_ON) ? true : false);
 	else
 		radeon_combios_encoder_dpms_scratch_regs(encoder, (mode == DRM_MODE_DPMS_ON) ? true : false);
+
+	/* adjust pm to dpms change */
+	radeon_pm_compute_clocks(rdev);
 }
 
 static void radeon_legacy_tmds_ext_prepare(struct drm_encoder *encoder)
@@ -636,7 +663,6 @@ static void radeon_legacy_tmds_ext_prepare(struct drm_encoder *encoder)
 	else
 		radeon_combios_output_lock(encoder, true);
 	radeon_legacy_tmds_ext_dpms(encoder, DRM_MODE_DPMS_OFF);
-	radeon_encoder_set_active_device(encoder);
 }
 
 static void radeon_legacy_tmds_ext_commit(struct drm_encoder *encoder)
@@ -690,6 +716,8 @@ static void radeon_legacy_tmds_ext_mode_set(struct drm_encoder *encoder,
 			/*if (mode->clock > 165000)
 			  fp2_gen_cntl |= R300_FP2_DVO_DUAL_CHANNEL_EN;*/
 		}
+		if (!radeon_combios_external_tmds_setup(encoder))
+			radeon_external_tmds_setup(encoder);
 	}
 
 	if (radeon_crtc->crtc_id == 0) {
@@ -717,9 +745,22 @@ static void radeon_legacy_tmds_ext_mode_set(struct drm_encoder *encoder,
 		radeon_combios_encoder_crtc_scratch_regs(encoder, radeon_crtc->crtc_id);
 }
 
+static void radeon_ext_tmds_enc_destroy(struct drm_encoder *encoder)
+{
+	struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
+	struct radeon_encoder_ext_tmds *tmds = radeon_encoder->enc_priv;
+	if (tmds) {
+		if (tmds->i2c_bus)
+			radeon_i2c_destroy(tmds->i2c_bus);
+	}
+	kfree(radeon_encoder->enc_priv);
+	drm_encoder_cleanup(encoder);
+	kfree(radeon_encoder);
+}
+
 static const struct drm_encoder_helper_funcs radeon_legacy_tmds_ext_helper_funcs = {
 	.dpms = radeon_legacy_tmds_ext_dpms,
-	.mode_fixup = radeon_legacy_tmds_ext_mode_fixup,
+	.mode_fixup = radeon_legacy_mode_fixup,
 	.prepare = radeon_legacy_tmds_ext_prepare,
 	.mode_set = radeon_legacy_tmds_ext_mode_set,
 	.commit = radeon_legacy_tmds_ext_commit,
@@ -728,19 +769,9 @@ static const struct drm_encoder_helper_funcs radeon_legacy_tmds_ext_helper_funcs
 
 
 static const struct drm_encoder_funcs radeon_legacy_tmds_ext_enc_funcs = {
-	.destroy = radeon_enc_destroy,
+	.destroy = radeon_ext_tmds_enc_destroy,
 };
 
-static bool radeon_legacy_tv_dac_mode_fixup(struct drm_encoder *encoder,
-					    struct drm_display_mode *mode,
-					    struct drm_display_mode *adjusted_mode)
-{
-
-	drm_mode_set_crtcinfo(adjusted_mode, 0);
-
-	return true;
-}
-
 static void radeon_legacy_tv_dac_dpms(struct drm_encoder *encoder, int mode)
 {
 	struct drm_device *dev = encoder->dev;
@@ -828,6 +859,9 @@ static void radeon_legacy_tv_dac_dpms(struct drm_encoder *encoder, int mode)
 		radeon_atombios_encoder_dpms_scratch_regs(encoder, (mode == DRM_MODE_DPMS_ON) ? true : false);
 	else
 		radeon_combios_encoder_dpms_scratch_regs(encoder, (mode == DRM_MODE_DPMS_ON) ? true : false);
+
+	/* adjust pm to dpms change */
+	radeon_pm_compute_clocks(rdev);
 }
 
 static void radeon_legacy_tv_dac_prepare(struct drm_encoder *encoder)
@@ -839,7 +873,6 @@ static void radeon_legacy_tv_dac_prepare(struct drm_encoder *encoder)
 	else
 		radeon_combios_output_lock(encoder, true);
 	radeon_legacy_tv_dac_dpms(encoder, DRM_MODE_DPMS_OFF);
-	radeon_encoder_set_active_device(encoder);
 }
 
 static void radeon_legacy_tv_dac_commit(struct drm_encoder *encoder)
@@ -881,7 +914,7 @@ static void radeon_legacy_tv_dac_mode_set(struct drm_encoder *encoder,
 					R420_TV_DAC_DACADJ_MASK |
 					R420_TV_DAC_RDACPD |
 					R420_TV_DAC_GDACPD |
-					R420_TV_DAC_GDACPD |
+					R420_TV_DAC_BDACPD |
 					R420_TV_DAC_TVENABLE);
 		} else {
 			tv_dac_cntl &= ~(RADEON_TV_DAC_STD_MASK |
@@ -889,7 +922,7 @@ static void radeon_legacy_tv_dac_mode_set(struct drm_encoder *encoder,
 					RADEON_TV_DAC_DACADJ_MASK |
 					RADEON_TV_DAC_RDACPD |
 					RADEON_TV_DAC_GDACPD |
-					RADEON_TV_DAC_GDACPD);
+					RADEON_TV_DAC_BDACPD);
 		}
 
 		/*  FIXME TV */
@@ -1258,7 +1291,7 @@ static enum drm_connector_status radeon_legacy_tv_dac_detect(struct drm_encoder
 
 static const struct drm_encoder_helper_funcs radeon_legacy_tv_dac_helper_funcs = {
 	.dpms = radeon_legacy_tv_dac_dpms,
-	.mode_fixup = radeon_legacy_tv_dac_mode_fixup,
+	.mode_fixup = radeon_legacy_mode_fixup,
 	.prepare = radeon_legacy_tv_dac_prepare,
 	.mode_set = radeon_legacy_tv_dac_mode_set,
 	.commit = radeon_legacy_tv_dac_commit,
@@ -1295,6 +1328,29 @@ static struct radeon_encoder_int_tmds *radeon_legacy_get_tmds_info(struct radeon
 	return tmds;
 }
 
+static struct radeon_encoder_ext_tmds *radeon_legacy_get_ext_tmds_info(struct radeon_encoder *encoder)
+{
+	struct drm_device *dev = encoder->base.dev;
+	struct radeon_device *rdev = dev->dev_private;
+	struct radeon_encoder_ext_tmds *tmds = NULL;
+	bool ret;
+
+	if (rdev->is_atom_bios)
+		return NULL;
+
+	tmds = kzalloc(sizeof(struct radeon_encoder_ext_tmds), GFP_KERNEL);
+
+	if (!tmds)
+		return NULL;
+
+	ret = radeon_legacy_get_ext_tmds_info_from_combios(encoder, tmds);
+
+	if (ret == false)
+		radeon_legacy_get_ext_tmds_info_from_table(encoder, tmds);
+
+	return tmds;
+}
+
 void
 radeon_add_legacy_encoder(struct drm_device *dev, uint32_t encoder_id, uint32_t supported_device)
 {
@@ -1318,8 +1374,10 @@ radeon_add_legacy_encoder(struct drm_device *dev, uint32_t encoder_id, uint32_t
 		return;
 
 	encoder = &radeon_encoder->base;
-	encoder->possible_crtcs = 0x3;
-	encoder->possible_clones = 0;
+	if (rdev->flags & RADEON_SINGLE_CRTC)
+		encoder->possible_crtcs = 0x1;
+	else
+		encoder->possible_crtcs = 0x3;
 
 	radeon_encoder->enc_priv = NULL;
 
@@ -1363,7 +1421,7 @@ radeon_add_legacy_encoder(struct drm_device *dev, uint32_t encoder_id, uint32_t
 		drm_encoder_init(dev, encoder, &radeon_legacy_tmds_ext_enc_funcs, DRM_MODE_ENCODER_TMDS);
 		drm_encoder_helper_add(encoder, &radeon_legacy_tmds_ext_helper_funcs);
 		if (!rdev->is_atom_bios)
-			radeon_combios_get_ext_tmds_info(radeon_encoder);
+			radeon_encoder->enc_priv = radeon_legacy_get_ext_tmds_info(radeon_encoder);
 		break;
 	}
 }
diff --git a/drivers/gpu/drm/radeon/radeon_legacy_tv.c b/drivers/gpu/drm/radeon/radeon_legacy_tv.c
index 3a12bb0c0563..417684daef4c 100644
--- a/drivers/gpu/drm/radeon/radeon_legacy_tv.c
+++ b/drivers/gpu/drm/radeon/radeon_legacy_tv.c
@@ -77,7 +77,7 @@ struct radeon_tv_mode_constants {
 	unsigned pix_to_tv;
 };
 
-static const uint16_t hor_timing_NTSC[] = {
+static const uint16_t hor_timing_NTSC[MAX_H_CODE_TIMING_LEN] = {
 	0x0007,
 	0x003f,
 	0x0263,
@@ -98,7 +98,7 @@ static const uint16_t hor_timing_NTSC[] = {
 	0
 };
 
-static const uint16_t vert_timing_NTSC[] = {
+static const uint16_t vert_timing_NTSC[MAX_V_CODE_TIMING_LEN] = {
 	0x2001,
 	0x200d,
 	0x1006,
@@ -115,7 +115,7 @@ static const uint16_t vert_timing_NTSC[] = {
 	0
 };
 
-static const uint16_t hor_timing_PAL[] = {
+static const uint16_t hor_timing_PAL[MAX_H_CODE_TIMING_LEN] = {
 	0x0007,
 	0x0058,
 	0x027c,
@@ -136,7 +136,7 @@ static const uint16_t hor_timing_PAL[] = {
 	0
 };
 
-static const uint16_t vert_timing_PAL[] = 	{
+static const uint16_t vert_timing_PAL[MAX_V_CODE_TIMING_LEN] = {
 	0x2001,
 	0x200c,
 	0x1005,
@@ -623,9 +623,9 @@ void radeon_legacy_tv_mode_set(struct drm_encoder *encoder,
 	}
 	flicker_removal = (tmp + 500) / 1000;
 
-	if (flicker_removal < 3)
-		flicker_removal = 3;
-	for (i = 0; i < 6; ++i) {
+	if (flicker_removal < 2)
+		flicker_removal = 2;
+	for (i = 0; i < ARRAY_SIZE(SLOPE_limit); ++i) {
 		if (flicker_removal == SLOPE_limit[i])
 			break;
 	}
diff --git a/drivers/gpu/drm/radeon/radeon_mode.h b/drivers/gpu/drm/radeon/radeon_mode.h
index 570a58729daf..1702b820aa4d 100644
--- a/drivers/gpu/drm/radeon/radeon_mode.h
+++ b/drivers/gpu/drm/radeon/radeon_mode.h
@@ -33,6 +33,7 @@
 #include <drm_crtc.h>
 #include <drm_mode.h>
 #include <drm_edid.h>
+#include <drm_dp_helper.h>
 #include <linux/i2c.h>
 #include <linux/i2c-id.h>
 #include <linux/i2c-algo-bit.h>
@@ -45,32 +46,6 @@ struct radeon_device;
 #define to_radeon_encoder(x) container_of(x, struct radeon_encoder, base)
 #define to_radeon_framebuffer(x) container_of(x, struct radeon_framebuffer, base)
 
-enum radeon_connector_type {
-	CONNECTOR_NONE,
-	CONNECTOR_VGA,
-	CONNECTOR_DVI_I,
-	CONNECTOR_DVI_D,
-	CONNECTOR_DVI_A,
-	CONNECTOR_STV,
-	CONNECTOR_CTV,
-	CONNECTOR_LVDS,
-	CONNECTOR_DIGITAL,
-	CONNECTOR_SCART,
-	CONNECTOR_HDMI_TYPE_A,
-	CONNECTOR_HDMI_TYPE_B,
-	CONNECTOR_0XC,
-	CONNECTOR_0XD,
-	CONNECTOR_DIN,
-	CONNECTOR_DISPLAY_PORT,
-	CONNECTOR_UNSUPPORTED
-};
-
-enum radeon_dvi_type {
-	DVI_AUTO,
-	DVI_DIGITAL,
-	DVI_ANALOG
-};
-
 enum radeon_rmx_type {
 	RMX_OFF,
 	RMX_FULL,
@@ -87,26 +62,50 @@ enum radeon_tv_std {
 	TV_STD_SCART_PAL,
 	TV_STD_SECAM,
 	TV_STD_PAL_CN,
+	TV_STD_PAL_N,
 };
 
+/* radeon gpio-based i2c
+ * 1. "mask" reg and bits
+ *    grabs the gpio pins for software use
+ *    0=not held  1=held
+ * 2. "a" reg and bits
+ *    output pin value
+ *    0=low 1=high
+ * 3. "en" reg and bits
+ *    sets the pin direction
+ *    0=input 1=output
+ * 4. "y" reg and bits
+ *    input pin value
+ *    0=low 1=high
+ */
 struct radeon_i2c_bus_rec {
 	bool valid;
+	/* id used by atom */
+	uint8_t i2c_id;
+	/* id used by atom */
+	uint8_t hpd_id;
+	/* can be used with hw i2c engine */
+	bool hw_capable;
+	/* uses multi-media i2c engine */
+	bool mm_i2c;
+	/* regs and bits */
 	uint32_t mask_clk_reg;
 	uint32_t mask_data_reg;
 	uint32_t a_clk_reg;
 	uint32_t a_data_reg;
-	uint32_t put_clk_reg;
-	uint32_t put_data_reg;
-	uint32_t get_clk_reg;
-	uint32_t get_data_reg;
+	uint32_t en_clk_reg;
+	uint32_t en_data_reg;
+	uint32_t y_clk_reg;
+	uint32_t y_data_reg;
 	uint32_t mask_clk_mask;
 	uint32_t mask_data_mask;
-	uint32_t put_clk_mask;
-	uint32_t put_data_mask;
-	uint32_t get_clk_mask;
-	uint32_t get_data_mask;
 	uint32_t a_clk_mask;
 	uint32_t a_data_mask;
+	uint32_t en_clk_mask;
+	uint32_t en_data_mask;
+	uint32_t y_clk_mask;
+	uint32_t y_data_mask;
 };
 
 struct radeon_tmds_pll {
@@ -116,6 +115,7 @@ struct radeon_tmds_pll {
 
 #define RADEON_MAX_BIOS_CONNECTOR 16
 
+/* pll flags */
 #define RADEON_PLL_USE_BIOS_DIVS        (1 << 0)
 #define RADEON_PLL_NO_ODD_POST_DIV      (1 << 1)
 #define RADEON_PLL_USE_REF_DIV          (1 << 2)
@@ -128,16 +128,30 @@ struct radeon_tmds_pll {
 #define RADEON_PLL_PREFER_HIGH_POST_DIV (1 << 9)
 #define RADEON_PLL_USE_FRAC_FB_DIV      (1 << 10)
 #define RADEON_PLL_PREFER_CLOSEST_LOWER (1 << 11)
+#define RADEON_PLL_USE_POST_DIV         (1 << 12)
+
+/* pll algo */
+enum radeon_pll_algo {
+	PLL_ALGO_LEGACY,
+	PLL_ALGO_NEW
+};
 
 struct radeon_pll {
-	uint16_t reference_freq;
-	uint16_t reference_div;
+	/* reference frequency */
+	uint32_t reference_freq;
+
+	/* fixed dividers */
+	uint32_t reference_div;
+	uint32_t post_div;
+
+	/* pll in/out limits */
 	uint32_t pll_in_min;
 	uint32_t pll_in_max;
 	uint32_t pll_out_min;
 	uint32_t pll_out_max;
-	uint16_t xclk;
+	uint32_t best_vco;
 
+	/* divider limits */
 	uint32_t min_ref_div;
 	uint32_t max_ref_div;
 	uint32_t min_post_div;
@@ -146,13 +160,28 @@ struct radeon_pll {
 	uint32_t max_feedback_div;
 	uint32_t min_frac_feedback_div;
 	uint32_t max_frac_feedback_div;
-	uint32_t best_vco;
+
+	/* flags for the current clock */
+	uint32_t flags;
+
+	/* pll id */
+	uint32_t id;
+	/* pll algo */
+	enum radeon_pll_algo algo;
+};
+
+struct i2c_algo_radeon_data {
+	struct i2c_adapter bit_adapter;
+	struct i2c_algo_bit_data bit_data;
 };
 
 struct radeon_i2c_chan {
-	struct drm_device *dev;
 	struct i2c_adapter adapter;
-	struct i2c_algo_bit_data algo;
+	struct drm_device *dev;
+	union {
+		struct i2c_algo_dp_aux_data dp;
+		struct i2c_algo_radeon_data radeon;
+	} algo;
 	struct radeon_i2c_bus_rec rec;
 };
 
@@ -170,11 +199,17 @@ enum radeon_connector_table {
 	CT_EMAC,
 };
 
+enum radeon_dvo_chip {
+	DVO_SIL164,
+	DVO_SIL1178,
+};
+
 struct radeon_mode_info {
 	struct atom_context *atom_context;
+	struct card_info *atom_card_info;
 	enum radeon_connector_table connector_table;
 	bool mode_config_initialized;
-	struct radeon_crtc *crtcs[2];
+	struct radeon_crtc *crtcs[6];
 	/* DVI-I properties */
 	struct drm_property *coherent_mode_property;
 	/* DAC enable load detect */
@@ -183,18 +218,8 @@ struct radeon_mode_info {
 	struct drm_property *tv_std_property;
 	/* legacy TMDS PLL detect */
 	struct drm_property *tmds_pll_property;
-
-};
-
-struct radeon_native_mode {
-	/* preferred mode */
-	uint32_t panel_xres, panel_yres;
-	uint32_t hoverplus, hsync_width;
-	uint32_t hblank;
-	uint32_t voverplus, vsync_width;
-	uint32_t vblank;
-	uint32_t dotclock;
-	uint32_t flags;
+	/* hardcoded DFP edid from BIOS */
+	struct edid *bios_hardcoded_edid;
 };
 
 #define MAX_H_CODE_TIMING_LEN 32
@@ -228,7 +253,8 @@ struct radeon_crtc {
 	enum radeon_rmx_type rmx_type;
 	fixed20_12 vsc;
 	fixed20_12 hsc;
-	struct radeon_native_mode native_mode;
+	struct drm_display_mode native_mode;
+	int pll_id;
 };
 
 struct radeon_encoder_primary_dac {
@@ -248,7 +274,7 @@ struct radeon_encoder_lvds {
 	bool     use_bios_dividers;
 	uint32_t lvds_gen_cntl;
 	/* panel mode */
-	struct radeon_native_mode native_mode;
+	struct drm_display_mode native_mode;
 };
 
 struct radeon_encoder_tv_dac {
@@ -271,15 +297,34 @@ struct radeon_encoder_int_tmds {
 	struct radeon_tmds_pll tmds_pll[4];
 };
 
+struct radeon_encoder_ext_tmds {
+	/* tmds over dvo */
+	struct radeon_i2c_chan *i2c_bus;
+	uint8_t slave_addr;
+	enum radeon_dvo_chip dvo_chip;
+};
+
+/* spread spectrum */
+struct radeon_atom_ss {
+	uint16_t percentage;
+	uint8_t type;
+	uint8_t step;
+	uint8_t delay;
+	uint8_t range;
+	uint8_t refdiv;
+};
+
 struct radeon_encoder_atom_dig {
 	/* atom dig */
 	bool coherent_mode;
-	int dig_block;
+	int dig_encoder; /* -1 disabled, 0 DIGA, 1 DIGB */
 	/* atom lvds */
 	uint32_t lvds_misc;
 	uint16_t panel_pwr_delay;
+	enum radeon_pll_algo pll_algo;
+	struct radeon_atom_ss *ss;
 	/* panel mode */
-	struct radeon_native_mode native_mode;
+	struct drm_display_mode native_mode;
 };
 
 struct radeon_encoder_atom_dac {
@@ -294,13 +339,45 @@ struct radeon_encoder {
 	uint32_t flags;
 	uint32_t pixel_clock;
 	enum radeon_rmx_type rmx_type;
-	struct radeon_native_mode native_mode;
+	struct drm_display_mode native_mode;
 	void *enc_priv;
+	int hdmi_offset;
+	int hdmi_audio_workaround;
+	int hdmi_buffer_status;
 };
 
 struct radeon_connector_atom_dig {
 	uint32_t igp_lane_info;
 	bool linkb;
+	/* displayport */
+	struct radeon_i2c_chan *dp_i2c_bus;
+	u8 dpcd[8];
+	u8 dp_sink_type;
+	int dp_clock;
+	int dp_lane_count;
+};
+
+struct radeon_gpio_rec {
+	bool valid;
+	u8 id;
+	u32 reg;
+	u32 mask;
+};
+
+enum radeon_hpd_id {
+	RADEON_HPD_NONE = 0,
+	RADEON_HPD_1,
+	RADEON_HPD_2,
+	RADEON_HPD_3,
+	RADEON_HPD_4,
+	RADEON_HPD_5,
+	RADEON_HPD_6,
+};
+
+struct radeon_hpd {
+	enum radeon_hpd_id hpd;
+	u8 plugged_state;
+	struct radeon_gpio_rec gpio;
 };
 
 struct radeon_connector {
@@ -308,12 +385,16 @@ struct radeon_connector {
 	uint32_t connector_id;
 	uint32_t devices;
 	struct radeon_i2c_chan *ddc_bus;
+	/* some systems have a an hdmi and vga port with a shared ddc line */
+	bool shared_ddc;
 	bool use_digital;
 	/* we need to mind the EDID between detect
 	   and get modes due to analog/digital/tvencoder */
 	struct edid *edid;
 	void *con_priv;
 	bool dac_load_detect;
+	uint16_t connector_object_id;
+	struct radeon_hpd hpd;
 };
 
 struct radeon_framebuffer {
@@ -321,10 +402,44 @@ struct radeon_framebuffer {
 	struct drm_gem_object *obj;
 };
 
+extern enum radeon_tv_std
+radeon_combios_get_tv_info(struct radeon_device *rdev);
+extern enum radeon_tv_std
+radeon_atombios_get_tv_info(struct radeon_device *rdev);
+
+extern void radeon_connector_hotplug(struct drm_connector *connector);
+extern bool radeon_dp_needs_link_train(struct radeon_connector *radeon_connector);
+extern int radeon_dp_mode_valid_helper(struct radeon_connector *radeon_connector,
+				       struct drm_display_mode *mode);
+extern void radeon_dp_set_link_config(struct drm_connector *connector,
+				      struct drm_display_mode *mode);
+extern void dp_link_train(struct drm_encoder *encoder,
+			  struct drm_connector *connector);
+extern u8 radeon_dp_getsinktype(struct radeon_connector *radeon_connector);
+extern bool radeon_dp_getdpcd(struct radeon_connector *radeon_connector);
+extern void atombios_dig_encoder_setup(struct drm_encoder *encoder, int action);
+extern void atombios_dig_transmitter_setup(struct drm_encoder *encoder,
+					   int action, uint8_t lane_num,
+					   uint8_t lane_set);
+extern int radeon_dp_i2c_aux_ch(struct i2c_adapter *adapter, int mode,
+				uint8_t write_byte, uint8_t *read_byte);
+
+extern struct radeon_i2c_chan *radeon_i2c_create_dp(struct drm_device *dev,
+						    struct radeon_i2c_bus_rec *rec,
+						    const char *name);
 extern struct radeon_i2c_chan *radeon_i2c_create(struct drm_device *dev,
 						 struct radeon_i2c_bus_rec *rec,
 						 const char *name);
 extern void radeon_i2c_destroy(struct radeon_i2c_chan *i2c);
+extern void radeon_i2c_destroy_dp(struct radeon_i2c_chan *i2c);
+extern void radeon_i2c_get_byte(struct radeon_i2c_chan *i2c_bus,
+				u8 slave_addr,
+				u8 addr,
+				u8 *val);
+extern void radeon_i2c_put_byte(struct radeon_i2c_chan *i2c,
+				u8 slave_addr,
+				u8 addr,
+				u8 val);
 extern bool radeon_ddc_probe(struct radeon_connector *radeon_connector);
 extern int radeon_ddc_get_modes(struct radeon_connector *radeon_connector);
 
@@ -336,8 +451,9 @@ extern void radeon_compute_pll(struct radeon_pll *pll,
 			       uint32_t *fb_div_p,
 			       uint32_t *frac_fb_div_p,
 			       uint32_t *ref_div_p,
-			       uint32_t *post_div_p,
-			       int flags);
+			       uint32_t *post_div_p);
+
+extern void radeon_setup_encoder_clones(struct drm_device *dev);
 
 struct drm_encoder *radeon_encoder_legacy_lvds_add(struct drm_device *dev, int bios_index);
 struct drm_encoder *radeon_encoder_legacy_primary_dac_add(struct drm_device *dev, int bios_index, int with_tv);
@@ -345,6 +461,7 @@ struct drm_encoder *radeon_encoder_legacy_tv_dac_add(struct drm_device *dev, int
 struct drm_encoder *radeon_encoder_legacy_tmds_int_add(struct drm_device *dev, int bios_index);
 struct drm_encoder *radeon_encoder_legacy_tmds_ext_add(struct drm_device *dev, int bios_index);
 extern void atombios_external_tmds_setup(struct drm_encoder *encoder, int action);
+extern void atombios_digital_setup(struct drm_encoder *encoder, int action);
 extern int atombios_get_encoder_mode(struct drm_encoder *encoder);
 extern void radeon_encoder_set_active_device(struct drm_encoder *encoder);
 
@@ -360,7 +477,6 @@ extern void atombios_crtc_dpms(struct drm_crtc *crtc, int mode);
 
 extern int radeon_crtc_set_base(struct drm_crtc *crtc, int x, int y,
 				 struct drm_framebuffer *old_fb);
-extern void radeon_legacy_atom_set_surface(struct drm_crtc *crtc);
 
 extern int radeon_crtc_cursor_set(struct drm_crtc *crtc,
 				  struct drm_file *file_priv,
@@ -370,16 +486,23 @@ extern int radeon_crtc_cursor_set(struct drm_crtc *crtc,
 extern int radeon_crtc_cursor_move(struct drm_crtc *crtc,
 				   int x, int y);
 
+extern bool radeon_combios_check_hardcoded_edid(struct radeon_device *rdev);
+extern struct edid *
+radeon_combios_get_hardcoded_edid(struct radeon_device *rdev);
 extern bool radeon_atom_get_clock_info(struct drm_device *dev);
 extern bool radeon_combios_get_clock_info(struct drm_device *dev);
 extern struct radeon_encoder_atom_dig *
 radeon_atombios_get_lvds_info(struct radeon_encoder *encoder);
-bool radeon_atombios_get_tmds_info(struct radeon_encoder *encoder,
-				   struct radeon_encoder_int_tmds *tmds);
-bool radeon_legacy_get_tmds_info_from_combios(struct radeon_encoder *encoder,
-					   struct radeon_encoder_int_tmds *tmds);
-bool radeon_legacy_get_tmds_info_from_table(struct radeon_encoder *encoder,
-					    struct radeon_encoder_int_tmds *tmds);
+extern bool radeon_atombios_get_tmds_info(struct radeon_encoder *encoder,
+					  struct radeon_encoder_int_tmds *tmds);
+extern bool radeon_legacy_get_tmds_info_from_combios(struct radeon_encoder *encoder,
+						     struct radeon_encoder_int_tmds *tmds);
+extern bool radeon_legacy_get_tmds_info_from_table(struct radeon_encoder *encoder,
+						   struct radeon_encoder_int_tmds *tmds);
+extern bool radeon_legacy_get_ext_tmds_info_from_combios(struct radeon_encoder *encoder,
+							 struct radeon_encoder_ext_tmds *tmds);
+extern bool radeon_legacy_get_ext_tmds_info_from_table(struct radeon_encoder *encoder,
+						       struct radeon_encoder_ext_tmds *tmds);
 extern struct radeon_encoder_primary_dac *
 radeon_atombios_get_primary_dac_info(struct radeon_encoder *encoder);
 extern struct radeon_encoder_tv_dac *
@@ -391,6 +514,8 @@ extern struct radeon_encoder_tv_dac *
 radeon_combios_get_tv_dac_info(struct radeon_encoder *encoder);
 extern struct radeon_encoder_primary_dac *
 radeon_combios_get_primary_dac_info(struct radeon_encoder *encoder);
+extern bool radeon_combios_external_tmds_setup(struct drm_encoder *encoder);
+extern void radeon_external_tmds_setup(struct drm_encoder *encoder);
 extern void radeon_combios_output_lock(struct drm_encoder *encoder, bool lock);
 extern void radeon_combios_initialize_bios_scratch_regs(struct drm_device *dev);
 extern void radeon_atom_output_lock(struct drm_encoder *encoder, bool lock);
@@ -407,6 +532,8 @@ extern void
 radeon_combios_encoder_dpms_scratch_regs(struct drm_encoder *encoder, bool on);
 extern void radeon_crtc_fb_gamma_set(struct drm_crtc *crtc, u16 red, u16 green,
 				     u16 blue, int regno);
+extern void radeon_crtc_fb_gamma_get(struct drm_crtc *crtc, u16 *red, u16 *green,
+				     u16 *blue, int regno);
 struct drm_framebuffer *radeon_framebuffer_create(struct drm_device *dev,
 						  struct drm_mode_fb_cmd *mode_cmd,
 						  struct drm_gem_object *obj);
@@ -420,16 +547,12 @@ void radeon_atombios_init_crtc(struct drm_device *dev,
 			       struct radeon_crtc *radeon_crtc);
 void radeon_legacy_init_crtc(struct drm_device *dev,
 			     struct radeon_crtc *radeon_crtc);
-void radeon_i2c_do_lock(struct radeon_connector *radeon_connector, int lock_state);
 
 void radeon_get_clock_info(struct drm_device *dev);
 
 extern bool radeon_get_atom_connector_info_from_object_table(struct drm_device *dev);
 extern bool radeon_get_atom_connector_info_from_supported_devices_table(struct drm_device *dev);
 
-void radeon_rmx_mode_fixup(struct drm_encoder *encoder,
-			   struct drm_display_mode *mode,
-			   struct drm_display_mode *adjusted_mode);
 void radeon_enc_destroy(struct drm_encoder *encoder);
 void radeon_copy_fb(struct drm_device *dev, struct drm_gem_object *dst_obj);
 void radeon_combios_asic_init(struct drm_device *dev);
diff --git a/drivers/gpu/drm/radeon/radeon_object.c b/drivers/gpu/drm/radeon/radeon_object.c
index 73af463b7a59..fc9d00ac6b15 100644
--- a/drivers/gpu/drm/radeon/radeon_object.c
+++ b/drivers/gpu/drm/radeon/radeon_object.c
@@ -34,100 +34,62 @@
 #include "radeon_drm.h"
 #include "radeon.h"
 
-struct radeon_object {
-	struct ttm_buffer_object	tobj;
-	struct list_head		list;
-	struct radeon_device		*rdev;
-	struct drm_gem_object		*gobj;
-	struct ttm_bo_kmap_obj		kmap;
-	unsigned			pin_count;
-	uint64_t			gpu_addr;
-	void				*kptr;
-	bool				is_iomem;
-	uint32_t			tiling_flags;
-	uint32_t			pitch;
-	int				surface_reg;
-};
 
 int radeon_ttm_init(struct radeon_device *rdev);
 void radeon_ttm_fini(struct radeon_device *rdev);
+static void radeon_bo_clear_surface_reg(struct radeon_bo *bo);
 
 /*
  * To exclude mutual BO access we rely on bo_reserve exclusion, as all
  * function are calling it.
  */
 
-static int radeon_object_reserve(struct radeon_object *robj, bool interruptible)
+static void radeon_ttm_bo_destroy(struct ttm_buffer_object *tbo)
 {
-	return ttm_bo_reserve(&robj->tobj, interruptible, false, false, 0);
-}
+	struct radeon_bo *bo;
 
-static void radeon_object_unreserve(struct radeon_object *robj)
-{
-	ttm_bo_unreserve(&robj->tobj);
+	bo = container_of(tbo, struct radeon_bo, tbo);
+	mutex_lock(&bo->rdev->gem.mutex);
+	list_del_init(&bo->list);
+	mutex_unlock(&bo->rdev->gem.mutex);
+	radeon_bo_clear_surface_reg(bo);
+	kfree(bo);
 }
 
-static void radeon_ttm_object_object_destroy(struct ttm_buffer_object *tobj)
+bool radeon_ttm_bo_is_radeon_bo(struct ttm_buffer_object *bo)
 {
-	struct radeon_object *robj;
-
-	robj = container_of(tobj, struct radeon_object, tobj);
-	list_del_init(&robj->list);
-	radeon_object_clear_surface_reg(robj);
-	kfree(robj);
+	if (bo->destroy == &radeon_ttm_bo_destroy)
+		return true;
+	return false;
 }
 
-static inline void radeon_object_gpu_addr(struct radeon_object *robj)
+void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain)
 {
-	/* Default gpu address */
-	robj->gpu_addr = 0xFFFFFFFFFFFFFFFFULL;
-	if (robj->tobj.mem.mm_node == NULL) {
-		return;
-	}
-	robj->gpu_addr = ((u64)robj->tobj.mem.mm_node->start) << PAGE_SHIFT;
-	switch (robj->tobj.mem.mem_type) {
-	case TTM_PL_VRAM:
-		robj->gpu_addr += (u64)robj->rdev->mc.vram_location;
-		break;
-	case TTM_PL_TT:
-		robj->gpu_addr += (u64)robj->rdev->mc.gtt_location;
-		break;
-	default:
-		DRM_ERROR("Unknown placement %d\n", robj->tobj.mem.mem_type);
-		robj->gpu_addr = 0xFFFFFFFFFFFFFFFFULL;
-		return;
-	}
-}
+	u32 c = 0;
 
-static inline uint32_t radeon_object_flags_from_domain(uint32_t domain)
-{
-	uint32_t flags = 0;
-	if (domain & RADEON_GEM_DOMAIN_VRAM) {
-		flags |= TTM_PL_FLAG_VRAM | TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED;
-	}
-	if (domain & RADEON_GEM_DOMAIN_GTT) {
-		flags |= TTM_PL_FLAG_TT | TTM_PL_MASK_CACHING;
-	}
-	if (domain & RADEON_GEM_DOMAIN_CPU) {
-		flags |= TTM_PL_FLAG_SYSTEM | TTM_PL_MASK_CACHING;
-	}
-	if (!flags) {
-		flags |= TTM_PL_FLAG_SYSTEM | TTM_PL_MASK_CACHING;
-	}
-	return flags;
+	rbo->placement.fpfn = 0;
+	rbo->placement.lpfn = 0;
+	rbo->placement.placement = rbo->placements;
+	rbo->placement.busy_placement = rbo->placements;
+	if (domain & RADEON_GEM_DOMAIN_VRAM)
+		rbo->placements[c++] = TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED |
+					TTM_PL_FLAG_VRAM;
+	if (domain & RADEON_GEM_DOMAIN_GTT)
+		rbo->placements[c++] = TTM_PL_MASK_CACHING | TTM_PL_FLAG_TT;
+	if (domain & RADEON_GEM_DOMAIN_CPU)
+		rbo->placements[c++] = TTM_PL_MASK_CACHING | TTM_PL_FLAG_SYSTEM;
+	if (!c)
+		rbo->placements[c++] = TTM_PL_MASK_CACHING | TTM_PL_FLAG_SYSTEM;
+	rbo->placement.num_placement = c;
+	rbo->placement.num_busy_placement = c;
 }
 
-int radeon_object_create(struct radeon_device *rdev,
-			 struct drm_gem_object *gobj,
-			 unsigned long size,
-			 bool kernel,
-			 uint32_t domain,
-			 bool interruptible,
-			 struct radeon_object **robj_ptr)
+int radeon_bo_create(struct radeon_device *rdev, struct drm_gem_object *gobj,
+			unsigned long size, bool kernel, u32 domain,
+			struct radeon_bo **bo_ptr)
 {
-	struct radeon_object *robj;
+	struct radeon_bo *bo;
 	enum ttm_bo_type type;
-	uint32_t flags;
 	int r;
 
 	if (unlikely(rdev->mman.bdev.dev_mapping == NULL)) {
@@ -138,238 +100,162 @@ int radeon_object_create(struct radeon_device *rdev,
 	} else {
 		type = ttm_bo_type_device;
 	}
-	*robj_ptr = NULL;
-	robj = kzalloc(sizeof(struct radeon_object), GFP_KERNEL);
-	if (robj == NULL) {
+	*bo_ptr = NULL;
+	bo = kzalloc(sizeof(struct radeon_bo), GFP_KERNEL);
+	if (bo == NULL)
 		return -ENOMEM;
-	}
-	robj->rdev = rdev;
-	robj->gobj = gobj;
-	robj->surface_reg = -1;
-	INIT_LIST_HEAD(&robj->list);
-
-	flags = radeon_object_flags_from_domain(domain);
-	r = ttm_buffer_object_init(&rdev->mman.bdev, &robj->tobj, size, type, flags,
-				   0, 0, false, NULL, size,
-				   &radeon_ttm_object_object_destroy);
+	bo->rdev = rdev;
+	bo->gobj = gobj;
+	bo->surface_reg = -1;
+	INIT_LIST_HEAD(&bo->list);
+
+	radeon_ttm_placement_from_domain(bo, domain);
+	/* Kernel allocation are uninterruptible */
+	r = ttm_bo_init(&rdev->mman.bdev, &bo->tbo, size, type,
+			&bo->placement, 0, 0, !kernel, NULL, size,
+			&radeon_ttm_bo_destroy);
 	if (unlikely(r != 0)) {
-		/* ttm call radeon_ttm_object_object_destroy if error happen */
-		DRM_ERROR("Failed to allocate TTM object (%ld, 0x%08X, %u)\n",
-			  size, flags, 0);
+		if (r != -ERESTARTSYS)
+			dev_err(rdev->dev,
+				"object_init failed for (%lu, 0x%08X)\n",
+				size, domain);
 		return r;
 	}
-	*robj_ptr = robj;
+	*bo_ptr = bo;
 	if (gobj) {
-		list_add_tail(&robj->list, &rdev->gem.objects);
+		mutex_lock(&bo->rdev->gem.mutex);
+		list_add_tail(&bo->list, &rdev->gem.objects);
+		mutex_unlock(&bo->rdev->gem.mutex);
 	}
 	return 0;
 }
 
-int radeon_object_kmap(struct radeon_object *robj, void **ptr)
+int radeon_bo_kmap(struct radeon_bo *bo, void **ptr)
 {
+	bool is_iomem;
 	int r;
 
-	spin_lock(&robj->tobj.lock);
-	if (robj->kptr) {
+	if (bo->kptr) {
 		if (ptr) {
-			*ptr = robj->kptr;
+			*ptr = bo->kptr;
 		}
-		spin_unlock(&robj->tobj.lock);
 		return 0;
 	}
-	spin_unlock(&robj->tobj.lock);
-	r = ttm_bo_kmap(&robj->tobj, 0, robj->tobj.num_pages, &robj->kmap);
+	r = ttm_bo_kmap(&bo->tbo, 0, bo->tbo.num_pages, &bo->kmap);
 	if (r) {
 		return r;
 	}
-	spin_lock(&robj->tobj.lock);
-	robj->kptr = ttm_kmap_obj_virtual(&robj->kmap, &robj->is_iomem);
-	spin_unlock(&robj->tobj.lock);
+	bo->kptr = ttm_kmap_obj_virtual(&bo->kmap, &is_iomem);
 	if (ptr) {
-		*ptr = robj->kptr;
+		*ptr = bo->kptr;
 	}
-	radeon_object_check_tiling(robj, 0, 0);
+	radeon_bo_check_tiling(bo, 0, 0);
 	return 0;
 }
 
-void radeon_object_kunmap(struct radeon_object *robj)
+void radeon_bo_kunmap(struct radeon_bo *bo)
 {
-	spin_lock(&robj->tobj.lock);
-	if (robj->kptr == NULL) {
-		spin_unlock(&robj->tobj.lock);
+	if (bo->kptr == NULL)
 		return;
-	}
-	robj->kptr = NULL;
-	spin_unlock(&robj->tobj.lock);
-	radeon_object_check_tiling(robj, 0, 0);
-	ttm_bo_kunmap(&robj->kmap);
+	bo->kptr = NULL;
+	radeon_bo_check_tiling(bo, 0, 0);
+	ttm_bo_kunmap(&bo->kmap);
 }
 
-void radeon_object_unref(struct radeon_object **robj)
+void radeon_bo_unref(struct radeon_bo **bo)
 {
-	struct ttm_buffer_object *tobj;
+	struct ttm_buffer_object *tbo;
 
-	if ((*robj) == NULL) {
+	if ((*bo) == NULL)
 		return;
-	}
-	tobj = &((*robj)->tobj);
-	ttm_bo_unref(&tobj);
-	if (tobj == NULL) {
-		*robj = NULL;
-	}
-}
-
-int radeon_object_mmap(struct radeon_object *robj, uint64_t *offset)
-{
-	*offset = robj->tobj.addr_space_offset;
-	return 0;
+	tbo = &((*bo)->tbo);
+	ttm_bo_unref(&tbo);
+	if (tbo == NULL)
+		*bo = NULL;
 }
 
-int radeon_object_pin(struct radeon_object *robj, uint32_t domain,
-		      uint64_t *gpu_addr)
+int radeon_bo_pin(struct radeon_bo *bo, u32 domain, u64 *gpu_addr)
 {
-	uint32_t flags;
-	uint32_t tmp;
-	int r;
+	int r, i;
 
-	flags = radeon_object_flags_from_domain(domain);
-	spin_lock(&robj->tobj.lock);
-	if (robj->pin_count) {
-		robj->pin_count++;
-		if (gpu_addr != NULL) {
-			*gpu_addr = robj->gpu_addr;
-		}
-		spin_unlock(&robj->tobj.lock);
+	if (bo->pin_count) {
+		bo->pin_count++;
+		if (gpu_addr)
+			*gpu_addr = radeon_bo_gpu_offset(bo);
 		return 0;
 	}
-	spin_unlock(&robj->tobj.lock);
-	r = radeon_object_reserve(robj, false);
-	if (unlikely(r != 0)) {
-		DRM_ERROR("radeon: failed to reserve object for pinning it.\n");
-		return r;
-	}
-	tmp = robj->tobj.mem.placement;
-	ttm_flag_masked(&tmp, flags, TTM_PL_MASK_MEM);
-	robj->tobj.proposed_placement = tmp | TTM_PL_FLAG_NO_EVICT | TTM_PL_MASK_CACHING;
-	r = ttm_buffer_object_validate(&robj->tobj,
-				       robj->tobj.proposed_placement,
-				       false, false);
-	radeon_object_gpu_addr(robj);
-	if (gpu_addr != NULL) {
-		*gpu_addr = robj->gpu_addr;
-	}
-	robj->pin_count = 1;
-	if (unlikely(r != 0)) {
-		DRM_ERROR("radeon: failed to pin object.\n");
-	}
-	radeon_object_unreserve(robj);
-	return r;
-}
-
-void radeon_object_unpin(struct radeon_object *robj)
-{
-	uint32_t flags;
-	int r;
-
-	spin_lock(&robj->tobj.lock);
-	if (!robj->pin_count) {
-		spin_unlock(&robj->tobj.lock);
-		printk(KERN_WARNING "Unpin not necessary for %p !\n", robj);
-		return;
-	}
-	robj->pin_count--;
-	if (robj->pin_count) {
-		spin_unlock(&robj->tobj.lock);
-		return;
-	}
-	spin_unlock(&robj->tobj.lock);
-	r = radeon_object_reserve(robj, false);
-	if (unlikely(r != 0)) {
-		DRM_ERROR("radeon: failed to reserve object for unpinning it.\n");
-		return;
-	}
-	flags = robj->tobj.mem.placement;
-	robj->tobj.proposed_placement = flags & ~TTM_PL_FLAG_NO_EVICT;
-	r = ttm_buffer_object_validate(&robj->tobj,
-				       robj->tobj.proposed_placement,
-				       false, false);
-	if (unlikely(r != 0)) {
-		DRM_ERROR("radeon: failed to unpin buffer.\n");
-	}
-	radeon_object_unreserve(robj);
-}
-
-int radeon_object_wait(struct radeon_object *robj)
-{
-	int r = 0;
-
-	/* FIXME: should use block reservation instead */
-	r = radeon_object_reserve(robj, true);
-	if (unlikely(r != 0)) {
-		DRM_ERROR("radeon: failed to reserve object for waiting.\n");
-		return r;
-	}
-	spin_lock(&robj->tobj.lock);
-	if (robj->tobj.sync_obj) {
-		r = ttm_bo_wait(&robj->tobj, true, true, false);
-	}
-	spin_unlock(&robj->tobj.lock);
-	radeon_object_unreserve(robj);
+	radeon_ttm_placement_from_domain(bo, domain);
+	/* force to pin into visible video ram */
+	bo->placement.lpfn = bo->rdev->mc.visible_vram_size >> PAGE_SHIFT;
+	for (i = 0; i < bo->placement.num_placement; i++)
+		bo->placements[i] |= TTM_PL_FLAG_NO_EVICT;
+	r = ttm_bo_validate(&bo->tbo, &bo->placement, false, false);
+	if (likely(r == 0)) {
+		bo->pin_count = 1;
+		if (gpu_addr != NULL)
+			*gpu_addr = radeon_bo_gpu_offset(bo);
+	}
+	if (unlikely(r != 0))
+		dev_err(bo->rdev->dev, "%p pin failed\n", bo);
 	return r;
 }
 
-int radeon_object_busy_domain(struct radeon_object *robj, uint32_t *cur_placement)
+int radeon_bo_unpin(struct radeon_bo *bo)
 {
-	int r = 0;
+	int r, i;
 
-	r = radeon_object_reserve(robj, true);
-	if (unlikely(r != 0)) {
-		DRM_ERROR("radeon: failed to reserve object for waiting.\n");
-		return r;
-	}
-	spin_lock(&robj->tobj.lock);
-	*cur_placement = robj->tobj.mem.mem_type;
-	if (robj->tobj.sync_obj) {
-		r = ttm_bo_wait(&robj->tobj, true, true, true);
+	if (!bo->pin_count) {
+		dev_warn(bo->rdev->dev, "%p unpin not necessary\n", bo);
+		return 0;
 	}
-	spin_unlock(&robj->tobj.lock);
-	radeon_object_unreserve(robj);
+	bo->pin_count--;
+	if (bo->pin_count)
+		return 0;
+	for (i = 0; i < bo->placement.num_placement; i++)
+		bo->placements[i] &= ~TTM_PL_FLAG_NO_EVICT;
+	r = ttm_bo_validate(&bo->tbo, &bo->placement, false, false);
+	if (unlikely(r != 0))
+		dev_err(bo->rdev->dev, "%p validate failed for unpin\n", bo);
 	return r;
 }
 
-int radeon_object_evict_vram(struct radeon_device *rdev)
+int radeon_bo_evict_vram(struct radeon_device *rdev)
 {
-	if (rdev->flags & RADEON_IS_IGP) {
-		/* Useless to evict on IGP chips */
-		return 0;
+	/* late 2.6.33 fix IGP hibernate - we need pm ops to do this correct */
+	if (0 && (rdev->flags & RADEON_IS_IGP)) {
+		if (rdev->mc.igp_sideport_enabled == false)
+			/* Useless to evict on IGP chips */
+			return 0;
 	}
 	return ttm_bo_evict_mm(&rdev->mman.bdev, TTM_PL_VRAM);
 }
 
-void radeon_object_force_delete(struct radeon_device *rdev)
+void radeon_bo_force_delete(struct radeon_device *rdev)
 {
-	struct radeon_object *robj, *n;
+	struct radeon_bo *bo, *n;
 	struct drm_gem_object *gobj;
 
 	if (list_empty(&rdev->gem.objects)) {
 		return;
 	}
-	DRM_ERROR("Userspace still has active objects !\n");
-	list_for_each_entry_safe(robj, n, &rdev->gem.objects, list) {
+	dev_err(rdev->dev, "Userspace still has active objects !\n");
+	list_for_each_entry_safe(bo, n, &rdev->gem.objects, list) {
 		mutex_lock(&rdev->ddev->struct_mutex);
-		gobj = robj->gobj;
-		DRM_ERROR("Force free for (%p,%p,%lu,%lu)\n",
-			  gobj, robj, (unsigned long)gobj->size,
-			  *((unsigned long *)&gobj->refcount));
-		list_del_init(&robj->list);
-		radeon_object_unref(&robj);
+		gobj = bo->gobj;
+		dev_err(rdev->dev, "%p %p %lu %lu force free\n",
+			gobj, bo, (unsigned long)gobj->size,
+			*((unsigned long *)&gobj->refcount));
+		mutex_lock(&bo->rdev->gem.mutex);
+		list_del_init(&bo->list);
+		mutex_unlock(&bo->rdev->gem.mutex);
+		radeon_bo_unref(&bo);
 		gobj->driver_private = NULL;
 		drm_gem_object_unreference(gobj);
 		mutex_unlock(&rdev->ddev->struct_mutex);
 	}
 }
 
-int radeon_object_init(struct radeon_device *rdev)
+int radeon_bo_init(struct radeon_device *rdev)
 {
 	/* Add an MTRR for the VRAM */
 	rdev->mc.vram_mtrr = mtrr_add(rdev->mc.aper_base, rdev->mc.aper_size,
@@ -382,13 +268,13 @@ int radeon_object_init(struct radeon_device *rdev)
 	return radeon_ttm_init(rdev);
 }
 
-void radeon_object_fini(struct radeon_device *rdev)
+void radeon_bo_fini(struct radeon_device *rdev)
 {
 	radeon_ttm_fini(rdev);
 }
 
-void radeon_object_list_add_object(struct radeon_object_list *lobj,
-				   struct list_head *head)
+void radeon_bo_list_add_object(struct radeon_bo_list *lobj,
+				struct list_head *head)
 {
 	if (lobj->wdomain) {
 		list_add(&lobj->list, head);
@@ -397,134 +283,102 @@ void radeon_object_list_add_object(struct radeon_object_list *lobj,
 	}
 }
 
-int radeon_object_list_reserve(struct list_head *head)
+int radeon_bo_list_reserve(struct list_head *head)
 {
-	struct radeon_object_list *lobj;
-	struct list_head *i;
+	struct radeon_bo_list *lobj;
 	int r;
 
-	list_for_each(i, head) {
-		lobj = list_entry(i, struct radeon_object_list, list);
-		if (!lobj->robj->pin_count) {
-			r = radeon_object_reserve(lobj->robj, true);
-			if (unlikely(r != 0)) {
-				DRM_ERROR("radeon: failed to reserve object.\n");
-				return r;
-			}
-		} else {
-		}
+	list_for_each_entry(lobj, head, list){
+		r = radeon_bo_reserve(lobj->bo, false);
+		if (unlikely(r != 0))
+			return r;
 	}
 	return 0;
 }
 
-void radeon_object_list_unreserve(struct list_head *head)
+void radeon_bo_list_unreserve(struct list_head *head)
 {
-	struct radeon_object_list *lobj;
-	struct list_head *i;
-
-	list_for_each(i, head) {
-		lobj = list_entry(i, struct radeon_object_list, list);
-		if (!lobj->robj->pin_count) {
-			radeon_object_unreserve(lobj->robj);
-		} else {
-		}
+	struct radeon_bo_list *lobj;
+
+	list_for_each_entry(lobj, head, list) {
+		/* only unreserve object we successfully reserved */
+		if (radeon_bo_is_reserved(lobj->bo))
+			radeon_bo_unreserve(lobj->bo);
 	}
 }
 
-int radeon_object_list_validate(struct list_head *head, void *fence)
+int radeon_bo_list_validate(struct list_head *head)
 {
-	struct radeon_object_list *lobj;
-	struct radeon_object *robj;
-	struct radeon_fence *old_fence = NULL;
-	struct list_head *i;
+	struct radeon_bo_list *lobj;
+	struct radeon_bo *bo;
 	int r;
 
-	r = radeon_object_list_reserve(head);
+	r = radeon_bo_list_reserve(head);
 	if (unlikely(r != 0)) {
-		radeon_object_list_unreserve(head);
 		return r;
 	}
-	list_for_each(i, head) {
-		lobj = list_entry(i, struct radeon_object_list, list);
-		robj = lobj->robj;
-		if (!robj->pin_count) {
+	list_for_each_entry(lobj, head, list) {
+		bo = lobj->bo;
+		if (!bo->pin_count) {
 			if (lobj->wdomain) {
-				robj->tobj.proposed_placement =
-					radeon_object_flags_from_domain(lobj->wdomain);
+				radeon_ttm_placement_from_domain(bo,
+								lobj->wdomain);
 			} else {
-				robj->tobj.proposed_placement =
-					radeon_object_flags_from_domain(lobj->rdomain);
+				radeon_ttm_placement_from_domain(bo,
+								lobj->rdomain);
 			}
-			r = ttm_buffer_object_validate(&robj->tobj,
-						       robj->tobj.proposed_placement,
-						       true, false);
-			if (unlikely(r)) {
-				DRM_ERROR("radeon: failed to validate.\n");
+			r = ttm_bo_validate(&bo->tbo, &bo->placement,
+						true, false);
+			if (unlikely(r))
 				return r;
-			}
-			radeon_object_gpu_addr(robj);
-		}
-		lobj->gpu_offset = robj->gpu_addr;
-		lobj->tiling_flags = robj->tiling_flags;
-		if (fence) {
-			old_fence = (struct radeon_fence *)robj->tobj.sync_obj;
-			robj->tobj.sync_obj = radeon_fence_ref(fence);
-			robj->tobj.sync_obj_arg = NULL;
-		}
-		if (old_fence) {
-			radeon_fence_unref(&old_fence);
 		}
+		lobj->gpu_offset = radeon_bo_gpu_offset(bo);
+		lobj->tiling_flags = bo->tiling_flags;
 	}
 	return 0;
 }
 
-void radeon_object_list_unvalidate(struct list_head *head)
+void radeon_bo_list_fence(struct list_head *head, void *fence)
 {
-	struct radeon_object_list *lobj;
+	struct radeon_bo_list *lobj;
+	struct radeon_bo *bo;
 	struct radeon_fence *old_fence = NULL;
-	struct list_head *i;
 
-	list_for_each(i, head) {
-		lobj = list_entry(i, struct radeon_object_list, list);
-		old_fence = (struct radeon_fence *)lobj->robj->tobj.sync_obj;
-		lobj->robj->tobj.sync_obj = NULL;
+	list_for_each_entry(lobj, head, list) {
+		bo = lobj->bo;
+		spin_lock(&bo->tbo.lock);
+		old_fence = (struct radeon_fence *)bo->tbo.sync_obj;
+		bo->tbo.sync_obj = radeon_fence_ref(fence);
+		bo->tbo.sync_obj_arg = NULL;
+		spin_unlock(&bo->tbo.lock);
 		if (old_fence) {
 			radeon_fence_unref(&old_fence);
 		}
 	}
-	radeon_object_list_unreserve(head);
 }
 
-void radeon_object_list_clean(struct list_head *head)
-{
-	radeon_object_list_unreserve(head);
-}
-
-int radeon_object_fbdev_mmap(struct radeon_object *robj,
+int radeon_bo_fbdev_mmap(struct radeon_bo *bo,
 			     struct vm_area_struct *vma)
 {
-	return ttm_fbdev_mmap(vma, &robj->tobj);
+	return ttm_fbdev_mmap(vma, &bo->tbo);
 }
 
-unsigned long radeon_object_size(struct radeon_object *robj)
+int radeon_bo_get_surface_reg(struct radeon_bo *bo)
 {
-	return robj->tobj.num_pages << PAGE_SHIFT;
-}
-
-int radeon_object_get_surface_reg(struct radeon_object *robj)
-{
-	struct radeon_device *rdev = robj->rdev;
+	struct radeon_device *rdev = bo->rdev;
 	struct radeon_surface_reg *reg;
-	struct radeon_object *old_object;
+	struct radeon_bo *old_object;
 	int steal;
 	int i;
 
-	if (!robj->tiling_flags)
+	BUG_ON(!atomic_read(&bo->tbo.reserved));
+
+	if (!bo->tiling_flags)
 		return 0;
 
-	if (robj->surface_reg >= 0) {
-		reg = &rdev->surface_regs[robj->surface_reg];
-		i = robj->surface_reg;
+	if (bo->surface_reg >= 0) {
+		reg = &rdev->surface_regs[bo->surface_reg];
+		i = bo->surface_reg;
 		goto out;
 	}
 
@@ -532,10 +386,10 @@ int radeon_object_get_surface_reg(struct radeon_object *robj)
 	for (i = 0; i < RADEON_GEM_MAX_SURFACES; i++) {
 
 		reg = &rdev->surface_regs[i];
-		if (!reg->robj)
+		if (!reg->bo)
 			break;
 
-		old_object = reg->robj;
+		old_object = reg->bo;
 		if (old_object->pin_count == 0)
 			steal = i;
 	}
@@ -546,91 +400,107 @@ int radeon_object_get_surface_reg(struct radeon_object *robj)
 			return -ENOMEM;
 		/* find someone with a surface reg and nuke their BO */
 		reg = &rdev->surface_regs[steal];
-		old_object = reg->robj;
+		old_object = reg->bo;
 		/* blow away the mapping */
 		DRM_DEBUG("stealing surface reg %d from %p\n", steal, old_object);
-		ttm_bo_unmap_virtual(&old_object->tobj);
+		ttm_bo_unmap_virtual(&old_object->tbo);
 		old_object->surface_reg = -1;
 		i = steal;
 	}
 
-	robj->surface_reg = i;
-	reg->robj = robj;
+	bo->surface_reg = i;
+	reg->bo = bo;
 
 out:
-	radeon_set_surface_reg(rdev, i, robj->tiling_flags, robj->pitch,
-			       robj->tobj.mem.mm_node->start << PAGE_SHIFT,
-			       robj->tobj.num_pages << PAGE_SHIFT);
+	radeon_set_surface_reg(rdev, i, bo->tiling_flags, bo->pitch,
+			       bo->tbo.mem.mm_node->start << PAGE_SHIFT,
+			       bo->tbo.num_pages << PAGE_SHIFT);
 	return 0;
 }
 
-void radeon_object_clear_surface_reg(struct radeon_object *robj)
+static void radeon_bo_clear_surface_reg(struct radeon_bo *bo)
 {
-	struct radeon_device *rdev = robj->rdev;
+	struct radeon_device *rdev = bo->rdev;
 	struct radeon_surface_reg *reg;
 
-	if (robj->surface_reg == -1)
+	if (bo->surface_reg == -1)
 		return;
 
-	reg = &rdev->surface_regs[robj->surface_reg];
-	radeon_clear_surface_reg(rdev, robj->surface_reg);
+	reg = &rdev->surface_regs[bo->surface_reg];
+	radeon_clear_surface_reg(rdev, bo->surface_reg);
 
-	reg->robj = NULL;
-	robj->surface_reg = -1;
+	reg->bo = NULL;
+	bo->surface_reg = -1;
 }
 
-void radeon_object_set_tiling_flags(struct radeon_object *robj,
-				    uint32_t tiling_flags, uint32_t pitch)
+int radeon_bo_set_tiling_flags(struct radeon_bo *bo,
+				uint32_t tiling_flags, uint32_t pitch)
 {
-	robj->tiling_flags = tiling_flags;
-	robj->pitch = pitch;
+	int r;
+
+	r = radeon_bo_reserve(bo, false);
+	if (unlikely(r != 0))
+		return r;
+	bo->tiling_flags = tiling_flags;
+	bo->pitch = pitch;
+	radeon_bo_unreserve(bo);
+	return 0;
 }
 
-void radeon_object_get_tiling_flags(struct radeon_object *robj,
-				    uint32_t *tiling_flags,
-				    uint32_t *pitch)
+void radeon_bo_get_tiling_flags(struct radeon_bo *bo,
+				uint32_t *tiling_flags,
+				uint32_t *pitch)
 {
+	BUG_ON(!atomic_read(&bo->tbo.reserved));
 	if (tiling_flags)
-		*tiling_flags = robj->tiling_flags;
+		*tiling_flags = bo->tiling_flags;
 	if (pitch)
-		*pitch = robj->pitch;
+		*pitch = bo->pitch;
 }
 
-int radeon_object_check_tiling(struct radeon_object *robj, bool has_moved,
-			       bool force_drop)
+int radeon_bo_check_tiling(struct radeon_bo *bo, bool has_moved,
+				bool force_drop)
 {
-	if (!(robj->tiling_flags & RADEON_TILING_SURFACE))
+	BUG_ON(!atomic_read(&bo->tbo.reserved));
+
+	if (!(bo->tiling_flags & RADEON_TILING_SURFACE))
 		return 0;
 
 	if (force_drop) {
-		radeon_object_clear_surface_reg(robj);
+		radeon_bo_clear_surface_reg(bo);
 		return 0;
 	}
 
-	if (robj->tobj.mem.mem_type != TTM_PL_VRAM) {
+	if (bo->tbo.mem.mem_type != TTM_PL_VRAM) {
 		if (!has_moved)
 			return 0;
 
-		if (robj->surface_reg >= 0)
-			radeon_object_clear_surface_reg(robj);
+		if (bo->surface_reg >= 0)
+			radeon_bo_clear_surface_reg(bo);
 		return 0;
 	}
 
-	if ((robj->surface_reg >= 0) && !has_moved)
+	if ((bo->surface_reg >= 0) && !has_moved)
 		return 0;
 
-	return radeon_object_get_surface_reg(robj);
+	return radeon_bo_get_surface_reg(bo);
 }
 
 void radeon_bo_move_notify(struct ttm_buffer_object *bo,
-			  struct ttm_mem_reg *mem)
+			   struct ttm_mem_reg *mem)
 {
-	struct radeon_object *robj = container_of(bo, struct radeon_object, tobj);
-	radeon_object_check_tiling(robj, 0, 1);
+	struct radeon_bo *rbo;
+	if (!radeon_ttm_bo_is_radeon_bo(bo))
+		return;
+	rbo = container_of(bo, struct radeon_bo, tbo);
+	radeon_bo_check_tiling(rbo, 0, 1);
 }
 
 void radeon_bo_fault_reserve_notify(struct ttm_buffer_object *bo)
 {
-	struct radeon_object *robj = container_of(bo, struct radeon_object, tobj);
-	radeon_object_check_tiling(robj, 0, 0);
+	struct radeon_bo *rbo;
+	if (!radeon_ttm_bo_is_radeon_bo(bo))
+		return;
+	rbo = container_of(bo, struct radeon_bo, tbo);
+	radeon_bo_check_tiling(rbo, 0, 0);
 }
diff --git a/drivers/gpu/drm/radeon/radeon_object.h b/drivers/gpu/drm/radeon/radeon_object.h
index 10e8af6bb456..7ab43de1e244 100644
--- a/drivers/gpu/drm/radeon/radeon_object.h
+++ b/drivers/gpu/drm/radeon/radeon_object.h
@@ -28,19 +28,146 @@
 #ifndef __RADEON_OBJECT_H__
 #define __RADEON_OBJECT_H__
 
-#include <ttm/ttm_bo_api.h>
-#include <ttm/ttm_bo_driver.h>
-#include <ttm/ttm_placement.h>
-#include <ttm/ttm_module.h>
+#include <drm/radeon_drm.h>
+#include "radeon.h"
 
-/*
- * TTM.
+/**
+ * radeon_mem_type_to_domain - return domain corresponding to mem_type
+ * @mem_type:	ttm memory type
+ *
+ * Returns corresponding domain of the ttm mem_type
+ */
+static inline unsigned radeon_mem_type_to_domain(u32 mem_type)
+{
+	switch (mem_type) {
+	case TTM_PL_VRAM:
+		return RADEON_GEM_DOMAIN_VRAM;
+	case TTM_PL_TT:
+		return RADEON_GEM_DOMAIN_GTT;
+	case TTM_PL_SYSTEM:
+		return RADEON_GEM_DOMAIN_CPU;
+	default:
+		break;
+	}
+	return 0;
+}
+
+/**
+ * radeon_bo_reserve - reserve bo
+ * @bo:		bo structure
+ * @no_wait:		don't sleep while trying to reserve (return -EBUSY)
+ *
+ * Returns:
+ * -EBUSY: buffer is busy and @no_wait is true
+ * -ERESTARTSYS: A wait for the buffer to become unreserved was interrupted by
+ * a signal. Release all buffer reservations and return to user-space.
  */
-struct radeon_mman {
-	struct ttm_bo_global_ref        bo_global_ref;
-	struct ttm_global_reference	mem_global_ref;
-	bool				mem_global_referenced;
-	struct ttm_bo_device		bdev;
-};
+static inline int radeon_bo_reserve(struct radeon_bo *bo, bool no_wait)
+{
+	int r;
+
+	r = ttm_bo_reserve(&bo->tbo, true, no_wait, false, 0);
+	if (unlikely(r != 0)) {
+		if (r != -ERESTARTSYS)
+			dev_err(bo->rdev->dev, "%p reserve failed\n", bo);
+		return r;
+	}
+	return 0;
+}
+
+static inline void radeon_bo_unreserve(struct radeon_bo *bo)
+{
+	ttm_bo_unreserve(&bo->tbo);
+}
+
+/**
+ * radeon_bo_gpu_offset - return GPU offset of bo
+ * @bo:	radeon object for which we query the offset
+ *
+ * Returns current GPU offset of the object.
+ *
+ * Note: object should either be pinned or reserved when calling this
+ * function, it might be usefull to add check for this for debugging.
+ */
+static inline u64 radeon_bo_gpu_offset(struct radeon_bo *bo)
+{
+	return bo->tbo.offset;
+}
+
+static inline unsigned long radeon_bo_size(struct radeon_bo *bo)
+{
+	return bo->tbo.num_pages << PAGE_SHIFT;
+}
+
+static inline bool radeon_bo_is_reserved(struct radeon_bo *bo)
+{
+	return !!atomic_read(&bo->tbo.reserved);
+}
+
+/**
+ * radeon_bo_mmap_offset - return mmap offset of bo
+ * @bo:	radeon object for which we query the offset
+ *
+ * Returns mmap offset of the object.
+ *
+ * Note: addr_space_offset is constant after ttm bo init thus isn't protected
+ * by any lock.
+ */
+static inline u64 radeon_bo_mmap_offset(struct radeon_bo *bo)
+{
+	return bo->tbo.addr_space_offset;
+}
+
+static inline int radeon_bo_wait(struct radeon_bo *bo, u32 *mem_type,
+					bool no_wait)
+{
+	int r;
+
+	r = ttm_bo_reserve(&bo->tbo, true, no_wait, false, 0);
+	if (unlikely(r != 0)) {
+		if (r != -ERESTARTSYS)
+			dev_err(bo->rdev->dev, "%p reserve failed for wait\n", bo);
+		return r;
+	}
+	spin_lock(&bo->tbo.lock);
+	if (mem_type)
+		*mem_type = bo->tbo.mem.mem_type;
+	if (bo->tbo.sync_obj)
+		r = ttm_bo_wait(&bo->tbo, true, true, no_wait);
+	spin_unlock(&bo->tbo.lock);
+	ttm_bo_unreserve(&bo->tbo);
+	return r;
+}
 
+extern int radeon_bo_create(struct radeon_device *rdev,
+				struct drm_gem_object *gobj, unsigned long size,
+				bool kernel, u32 domain,
+				struct radeon_bo **bo_ptr);
+extern int radeon_bo_kmap(struct radeon_bo *bo, void **ptr);
+extern void radeon_bo_kunmap(struct radeon_bo *bo);
+extern void radeon_bo_unref(struct radeon_bo **bo);
+extern int radeon_bo_pin(struct radeon_bo *bo, u32 domain, u64 *gpu_addr);
+extern int radeon_bo_unpin(struct radeon_bo *bo);
+extern int radeon_bo_evict_vram(struct radeon_device *rdev);
+extern void radeon_bo_force_delete(struct radeon_device *rdev);
+extern int radeon_bo_init(struct radeon_device *rdev);
+extern void radeon_bo_fini(struct radeon_device *rdev);
+extern void radeon_bo_list_add_object(struct radeon_bo_list *lobj,
+				struct list_head *head);
+extern int radeon_bo_list_reserve(struct list_head *head);
+extern void radeon_bo_list_unreserve(struct list_head *head);
+extern int radeon_bo_list_validate(struct list_head *head);
+extern void radeon_bo_list_fence(struct list_head *head, void *fence);
+extern int radeon_bo_fbdev_mmap(struct radeon_bo *bo,
+				struct vm_area_struct *vma);
+extern int radeon_bo_set_tiling_flags(struct radeon_bo *bo,
+				u32 tiling_flags, u32 pitch);
+extern void radeon_bo_get_tiling_flags(struct radeon_bo *bo,
+				u32 *tiling_flags, u32 *pitch);
+extern int radeon_bo_check_tiling(struct radeon_bo *bo, bool has_moved,
+				bool force_drop);
+extern void radeon_bo_move_notify(struct ttm_buffer_object *bo,
+					struct ttm_mem_reg *mem);
+extern void radeon_bo_fault_reserve_notify(struct ttm_buffer_object *bo);
+extern int radeon_bo_get_surface_reg(struct radeon_bo *bo);
 #endif
diff --git a/drivers/gpu/drm/radeon/radeon_pm.c b/drivers/gpu/drm/radeon/radeon_pm.c
new file mode 100644
index 000000000000..d4d1c39a0e99
--- /dev/null
+++ b/drivers/gpu/drm/radeon/radeon_pm.c
@@ -0,0 +1,463 @@
+/*
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Rafał Miłecki <zajec5@gmail.com>
+ *          Alex Deucher <alexdeucher@gmail.com>
+ */
+#include "drmP.h"
+#include "radeon.h"
+#include "avivod.h"
+
+#define RADEON_IDLE_LOOP_MS 100
+#define RADEON_RECLOCK_DELAY_MS 200
+#define RADEON_WAIT_VBLANK_TIMEOUT 200
+
+static void radeon_pm_set_clocks_locked(struct radeon_device *rdev);
+static void radeon_pm_set_clocks(struct radeon_device *rdev);
+static void radeon_pm_idle_work_handler(struct work_struct *work);
+static int radeon_debugfs_pm_init(struct radeon_device *rdev);
+
+static const char *pm_state_names[4] = {
+	"PM_STATE_DISABLED",
+	"PM_STATE_MINIMUM",
+	"PM_STATE_PAUSED",
+	"PM_STATE_ACTIVE"
+};
+
+static const char *pm_state_types[5] = {
+	"Default",
+	"Powersave",
+	"Battery",
+	"Balanced",
+	"Performance",
+};
+
+static void radeon_print_power_mode_info(struct radeon_device *rdev)
+{
+	int i, j;
+	bool is_default;
+
+	DRM_INFO("%d Power State(s)\n", rdev->pm.num_power_states);
+	for (i = 0; i < rdev->pm.num_power_states; i++) {
+		if (rdev->pm.default_power_state == &rdev->pm.power_state[i])
+			is_default = true;
+		else
+			is_default = false;
+		DRM_INFO("State %d %s %s\n", i,
+			 pm_state_types[rdev->pm.power_state[i].type],
+			 is_default ? "(default)" : "");
+		if ((rdev->flags & RADEON_IS_PCIE) && !(rdev->flags & RADEON_IS_IGP))
+			DRM_INFO("\t%d PCIE Lanes\n", rdev->pm.power_state[i].non_clock_info.pcie_lanes);
+		DRM_INFO("\t%d Clock Mode(s)\n", rdev->pm.power_state[i].num_clock_modes);
+		for (j = 0; j < rdev->pm.power_state[i].num_clock_modes; j++) {
+			if (rdev->flags & RADEON_IS_IGP)
+				DRM_INFO("\t\t%d engine: %d\n",
+					 j,
+					 rdev->pm.power_state[i].clock_info[j].sclk * 10);
+			else
+				DRM_INFO("\t\t%d engine/memory: %d/%d\n",
+					 j,
+					 rdev->pm.power_state[i].clock_info[j].sclk * 10,
+					 rdev->pm.power_state[i].clock_info[j].mclk * 10);
+		}
+	}
+}
+
+static struct radeon_power_state * radeon_pick_power_state(struct radeon_device *rdev,
+							   enum radeon_pm_state_type type)
+{
+	int i, j;
+	enum radeon_pm_state_type wanted_types[2];
+	int wanted_count;
+
+	switch (type) {
+	case POWER_STATE_TYPE_DEFAULT:
+	default:
+		return rdev->pm.default_power_state;
+	case POWER_STATE_TYPE_POWERSAVE:
+		if (rdev->flags & RADEON_IS_MOBILITY) {
+			wanted_types[0] = POWER_STATE_TYPE_POWERSAVE;
+			wanted_types[1] = POWER_STATE_TYPE_BATTERY;
+			wanted_count = 2;
+		} else {
+			wanted_types[0] = POWER_STATE_TYPE_PERFORMANCE;
+			wanted_count = 1;
+		}
+		break;
+	case POWER_STATE_TYPE_BATTERY:
+		if (rdev->flags & RADEON_IS_MOBILITY) {
+			wanted_types[0] = POWER_STATE_TYPE_BATTERY;
+			wanted_types[1] = POWER_STATE_TYPE_POWERSAVE;
+			wanted_count = 2;
+		} else {
+			wanted_types[0] = POWER_STATE_TYPE_PERFORMANCE;
+			wanted_count = 1;
+		}
+		break;
+	case POWER_STATE_TYPE_BALANCED:
+	case POWER_STATE_TYPE_PERFORMANCE:
+		wanted_types[0] = type;
+		wanted_count = 1;
+		break;
+	}
+
+	for (i = 0; i < wanted_count; i++) {
+		for (j = 0; j < rdev->pm.num_power_states; j++) {
+			if (rdev->pm.power_state[j].type == wanted_types[i])
+				return &rdev->pm.power_state[j];
+		}
+	}
+
+	return rdev->pm.default_power_state;
+}
+
+static struct radeon_pm_clock_info * radeon_pick_clock_mode(struct radeon_device *rdev,
+							    struct radeon_power_state *power_state,
+							    enum radeon_pm_clock_mode_type type)
+{
+	switch (type) {
+	case POWER_MODE_TYPE_DEFAULT:
+	default:
+		return power_state->default_clock_mode;
+	case POWER_MODE_TYPE_LOW:
+		return &power_state->clock_info[0];
+	case POWER_MODE_TYPE_MID:
+		if (power_state->num_clock_modes > 2)
+			return &power_state->clock_info[1];
+		else
+			return &power_state->clock_info[0];
+		break;
+	case POWER_MODE_TYPE_HIGH:
+		return &power_state->clock_info[power_state->num_clock_modes - 1];
+	}
+
+}
+
+static void radeon_get_power_state(struct radeon_device *rdev,
+				   enum radeon_pm_action action)
+{
+	switch (action) {
+	case PM_ACTION_MINIMUM:
+		rdev->pm.requested_power_state = radeon_pick_power_state(rdev, POWER_STATE_TYPE_BATTERY);
+		rdev->pm.requested_clock_mode =
+			radeon_pick_clock_mode(rdev, rdev->pm.requested_power_state, POWER_MODE_TYPE_LOW);
+		break;
+	case PM_ACTION_DOWNCLOCK:
+		rdev->pm.requested_power_state = radeon_pick_power_state(rdev, POWER_STATE_TYPE_POWERSAVE);
+		rdev->pm.requested_clock_mode =
+			radeon_pick_clock_mode(rdev, rdev->pm.requested_power_state, POWER_MODE_TYPE_MID);
+		break;
+	case PM_ACTION_UPCLOCK:
+		rdev->pm.requested_power_state = radeon_pick_power_state(rdev, POWER_STATE_TYPE_DEFAULT);
+		rdev->pm.requested_clock_mode =
+			radeon_pick_clock_mode(rdev, rdev->pm.requested_power_state, POWER_MODE_TYPE_HIGH);
+		break;
+	case PM_ACTION_NONE:
+	default:
+		DRM_ERROR("Requested mode for not defined action\n");
+		return;
+	}
+	DRM_INFO("Requested: e: %d m: %d p: %d\n",
+		 rdev->pm.requested_clock_mode->sclk,
+		 rdev->pm.requested_clock_mode->mclk,
+		 rdev->pm.requested_power_state->non_clock_info.pcie_lanes);
+}
+
+static void radeon_set_power_state(struct radeon_device *rdev)
+{
+	/* if *_clock_mode are the same, *_power_state are as well */
+	if (rdev->pm.requested_clock_mode == rdev->pm.current_clock_mode)
+		return;
+
+	DRM_INFO("Setting: e: %d m: %d p: %d\n",
+		 rdev->pm.requested_clock_mode->sclk,
+		 rdev->pm.requested_clock_mode->mclk,
+		 rdev->pm.requested_power_state->non_clock_info.pcie_lanes);
+	/* set pcie lanes */
+	/* set voltage */
+	/* set engine clock */
+	radeon_set_engine_clock(rdev, rdev->pm.requested_clock_mode->sclk);
+	/* set memory clock */
+
+	rdev->pm.current_power_state = rdev->pm.requested_power_state;
+	rdev->pm.current_clock_mode = rdev->pm.requested_clock_mode;
+}
+
+int radeon_pm_init(struct radeon_device *rdev)
+{
+	rdev->pm.state = PM_STATE_DISABLED;
+	rdev->pm.planned_action = PM_ACTION_NONE;
+	rdev->pm.downclocked = false;
+
+	if (rdev->bios) {
+		if (rdev->is_atom_bios)
+			radeon_atombios_get_power_modes(rdev);
+		else
+			radeon_combios_get_power_modes(rdev);
+		radeon_print_power_mode_info(rdev);
+	}
+
+	if (radeon_debugfs_pm_init(rdev)) {
+		DRM_ERROR("Failed to register debugfs file for PM!\n");
+	}
+
+	INIT_DELAYED_WORK(&rdev->pm.idle_work, radeon_pm_idle_work_handler);
+
+	if (radeon_dynpm != -1 && radeon_dynpm) {
+		rdev->pm.state = PM_STATE_PAUSED;
+		DRM_INFO("radeon: dynamic power management enabled\n");
+	}
+
+	DRM_INFO("radeon: power management initialized\n");
+
+	return 0;
+}
+
+void radeon_pm_compute_clocks(struct radeon_device *rdev)
+{
+	struct drm_device *ddev = rdev->ddev;
+	struct drm_connector *connector;
+	struct radeon_crtc *radeon_crtc;
+	int count = 0;
+
+	if (rdev->pm.state == PM_STATE_DISABLED)
+		return;
+
+	mutex_lock(&rdev->pm.mutex);
+
+	rdev->pm.active_crtcs = 0;
+	list_for_each_entry(connector,
+		&ddev->mode_config.connector_list, head) {
+		if (connector->encoder &&
+			connector->dpms != DRM_MODE_DPMS_OFF) {
+			radeon_crtc = to_radeon_crtc(connector->encoder->crtc);
+			rdev->pm.active_crtcs |= (1 << radeon_crtc->crtc_id);
+			++count;
+		}
+	}
+
+	if (count > 1) {
+		if (rdev->pm.state == PM_STATE_ACTIVE) {
+			cancel_delayed_work(&rdev->pm.idle_work);
+
+			rdev->pm.state = PM_STATE_PAUSED;
+			rdev->pm.planned_action = PM_ACTION_UPCLOCK;
+			if (rdev->pm.downclocked)
+				radeon_pm_set_clocks(rdev);
+
+			DRM_DEBUG("radeon: dynamic power management deactivated\n");
+		}
+	} else if (count == 1) {
+		/* TODO: Increase clocks if needed for current mode */
+
+		if (rdev->pm.state == PM_STATE_MINIMUM) {
+			rdev->pm.state = PM_STATE_ACTIVE;
+			rdev->pm.planned_action = PM_ACTION_UPCLOCK;
+			radeon_pm_set_clocks(rdev);
+
+			queue_delayed_work(rdev->wq, &rdev->pm.idle_work,
+				msecs_to_jiffies(RADEON_IDLE_LOOP_MS));
+		}
+		else if (rdev->pm.state == PM_STATE_PAUSED) {
+			rdev->pm.state = PM_STATE_ACTIVE;
+			queue_delayed_work(rdev->wq, &rdev->pm.idle_work,
+				msecs_to_jiffies(RADEON_IDLE_LOOP_MS));
+			DRM_DEBUG("radeon: dynamic power management activated\n");
+		}
+	}
+	else { /* count == 0 */
+		if (rdev->pm.state != PM_STATE_MINIMUM) {
+			cancel_delayed_work(&rdev->pm.idle_work);
+
+			rdev->pm.state = PM_STATE_MINIMUM;
+			rdev->pm.planned_action = PM_ACTION_MINIMUM;
+			radeon_pm_set_clocks(rdev);
+		}
+	}
+
+	mutex_unlock(&rdev->pm.mutex);
+}
+
+static bool radeon_pm_debug_check_in_vbl(struct radeon_device *rdev, bool finish)
+{
+	u32 stat_crtc1 = 0, stat_crtc2 = 0;
+	bool in_vbl = true;
+
+	if (ASIC_IS_AVIVO(rdev)) {
+		if (rdev->pm.active_crtcs & (1 << 0)) {
+			stat_crtc1 = RREG32(D1CRTC_STATUS);
+			if (!(stat_crtc1 & 1))
+				in_vbl = false;
+		}
+		if (rdev->pm.active_crtcs & (1 << 1)) {
+			stat_crtc2 = RREG32(D2CRTC_STATUS);
+			if (!(stat_crtc2 & 1))
+				in_vbl = false;
+		}
+	}
+	if (in_vbl == false)
+		DRM_INFO("not in vbl for pm change %08x %08x at %s\n", stat_crtc1,
+			 stat_crtc2, finish ? "exit" : "entry");
+	return in_vbl;
+}
+static void radeon_pm_set_clocks_locked(struct radeon_device *rdev)
+{
+	/*radeon_fence_wait_last(rdev);*/
+	switch (rdev->pm.planned_action) {
+	case PM_ACTION_UPCLOCK:
+		rdev->pm.downclocked = false;
+		break;
+	case PM_ACTION_DOWNCLOCK:
+		rdev->pm.downclocked = true;
+		break;
+	case PM_ACTION_MINIMUM:
+		break;
+	case PM_ACTION_NONE:
+		DRM_ERROR("%s: PM_ACTION_NONE\n", __func__);
+		break;
+	}
+
+	/* check if we are in vblank */
+	radeon_pm_debug_check_in_vbl(rdev, false);
+	radeon_set_power_state(rdev);
+	radeon_pm_debug_check_in_vbl(rdev, true);
+	rdev->pm.planned_action = PM_ACTION_NONE;
+}
+
+static void radeon_pm_set_clocks(struct radeon_device *rdev)
+{
+	radeon_get_power_state(rdev, rdev->pm.planned_action);
+	mutex_lock(&rdev->cp.mutex);
+
+	if (rdev->pm.active_crtcs & (1 << 0)) {
+		rdev->pm.req_vblank |= (1 << 0);
+		drm_vblank_get(rdev->ddev, 0);
+	}
+	if (rdev->pm.active_crtcs & (1 << 1)) {
+		rdev->pm.req_vblank |= (1 << 1);
+		drm_vblank_get(rdev->ddev, 1);
+	}
+	if (rdev->pm.active_crtcs)
+		wait_event_interruptible_timeout(
+			rdev->irq.vblank_queue, 0,
+			msecs_to_jiffies(RADEON_WAIT_VBLANK_TIMEOUT));
+	if (rdev->pm.req_vblank & (1 << 0)) {
+		rdev->pm.req_vblank &= ~(1 << 0);
+		drm_vblank_put(rdev->ddev, 0);
+	}
+	if (rdev->pm.req_vblank & (1 << 1)) {
+		rdev->pm.req_vblank &= ~(1 << 1);
+		drm_vblank_put(rdev->ddev, 1);
+	}
+
+	radeon_pm_set_clocks_locked(rdev);
+	mutex_unlock(&rdev->cp.mutex);
+}
+
+static void radeon_pm_idle_work_handler(struct work_struct *work)
+{
+	struct radeon_device *rdev;
+	rdev = container_of(work, struct radeon_device,
+				pm.idle_work.work);
+
+	mutex_lock(&rdev->pm.mutex);
+	if (rdev->pm.state == PM_STATE_ACTIVE) {
+		unsigned long irq_flags;
+		int not_processed = 0;
+
+		read_lock_irqsave(&rdev->fence_drv.lock, irq_flags);
+		if (!list_empty(&rdev->fence_drv.emited)) {
+			struct list_head *ptr;
+			list_for_each(ptr, &rdev->fence_drv.emited) {
+				/* count up to 3, that's enought info */
+				if (++not_processed >= 3)
+					break;
+			}
+		}
+		read_unlock_irqrestore(&rdev->fence_drv.lock, irq_flags);
+
+		if (not_processed >= 3) { /* should upclock */
+			if (rdev->pm.planned_action == PM_ACTION_DOWNCLOCK) {
+				rdev->pm.planned_action = PM_ACTION_NONE;
+			} else if (rdev->pm.planned_action == PM_ACTION_NONE &&
+				rdev->pm.downclocked) {
+				rdev->pm.planned_action =
+					PM_ACTION_UPCLOCK;
+				rdev->pm.action_timeout = jiffies +
+				msecs_to_jiffies(RADEON_RECLOCK_DELAY_MS);
+			}
+		} else if (not_processed == 0) { /* should downclock */
+			if (rdev->pm.planned_action == PM_ACTION_UPCLOCK) {
+				rdev->pm.planned_action = PM_ACTION_NONE;
+			} else if (rdev->pm.planned_action == PM_ACTION_NONE &&
+				!rdev->pm.downclocked) {
+				rdev->pm.planned_action =
+					PM_ACTION_DOWNCLOCK;
+				rdev->pm.action_timeout = jiffies +
+				msecs_to_jiffies(RADEON_RECLOCK_DELAY_MS);
+			}
+		}
+
+		if (rdev->pm.planned_action != PM_ACTION_NONE &&
+		    jiffies > rdev->pm.action_timeout) {
+			radeon_pm_set_clocks(rdev);
+		}
+	}
+	mutex_unlock(&rdev->pm.mutex);
+
+	queue_delayed_work(rdev->wq, &rdev->pm.idle_work,
+					msecs_to_jiffies(RADEON_IDLE_LOOP_MS));
+}
+
+/*
+ * Debugfs info
+ */
+#if defined(CONFIG_DEBUG_FS)
+
+static int radeon_debugfs_pm_info(struct seq_file *m, void *data)
+{
+	struct drm_info_node *node = (struct drm_info_node *) m->private;
+	struct drm_device *dev = node->minor->dev;
+	struct radeon_device *rdev = dev->dev_private;
+
+	seq_printf(m, "state: %s\n", pm_state_names[rdev->pm.state]);
+	seq_printf(m, "default engine clock: %u0 kHz\n", rdev->clock.default_sclk);
+	seq_printf(m, "current engine clock: %u0 kHz\n", radeon_get_engine_clock(rdev));
+	seq_printf(m, "default memory clock: %u0 kHz\n", rdev->clock.default_mclk);
+	if (rdev->asic->get_memory_clock)
+		seq_printf(m, "current memory clock: %u0 kHz\n", radeon_get_memory_clock(rdev));
+	if (rdev->asic->get_pcie_lanes)
+		seq_printf(m, "PCIE lanes: %d\n", radeon_get_pcie_lanes(rdev));
+
+	return 0;
+}
+
+static struct drm_info_list radeon_pm_info_list[] = {
+	{"radeon_pm_info", radeon_debugfs_pm_info, 0, NULL},
+};
+#endif
+
+static int radeon_debugfs_pm_init(struct radeon_device *rdev)
+{
+#if defined(CONFIG_DEBUG_FS)
+	return radeon_debugfs_add_files(rdev, radeon_pm_info_list, ARRAY_SIZE(radeon_pm_info_list));
+#else
+	return 0;
+#endif
+}
diff --git a/drivers/gpu/drm/radeon/radeon_reg.h b/drivers/gpu/drm/radeon/radeon_reg.h
index 21da871a793c..5c0dc082d330 100644
--- a/drivers/gpu/drm/radeon/radeon_reg.h
+++ b/drivers/gpu/drm/radeon/radeon_reg.h
@@ -54,7 +54,7 @@
 #include "r300_reg.h"
 #include "r500_reg.h"
 #include "r600_reg.h"
-
+#include "evergreen_reg.h"
 
 #define RADEON_MC_AGP_LOCATION		0x014c
 #define		RADEON_MC_AGP_START_MASK	0x0000FFFF
@@ -290,6 +290,8 @@
 #define RADEON_BUS_CNTL                     0x0030
 #       define RADEON_BUS_MASTER_DIS         (1 << 6)
 #       define RADEON_BUS_BIOS_DIS_ROM       (1 << 12)
+#	define RS600_BUS_MASTER_DIS	     (1 << 14)
+#	define RS600_MSI_REARM		     (1 << 20) /* rs600/rs690/rs740 */
 #       define RADEON_BUS_RD_DISCARD_EN      (1 << 24)
 #       define RADEON_BUS_RD_ABORT_EN        (1 << 25)
 #       define RADEON_BUS_MSTR_DISCONNECT_EN (1 << 28)
@@ -297,6 +299,9 @@
 #       define RADEON_BUS_READ_BURST         (1 << 30)
 #define RADEON_BUS_CNTL1                    0x0034
 #       define RADEON_BUS_WAIT_ON_LOCK_EN    (1 << 4)
+/* rv370/rv380, rv410, r423/r430/r480, r5xx */
+#define RADEON_MSI_REARM_EN		    0x0160
+#	define RV370_MSI_REARM_EN	     (1 << 0)
 
 /* #define RADEON_PCIE_INDEX                   0x0030 */
 /* #define RADEON_PCIE_DATA                    0x0034 */
@@ -882,6 +887,7 @@
 #       define RADEON_FP_PANEL_FORMAT          (1 <<  3)
 #       define RADEON_FP_EN_TMDS               (1 <<  7)
 #       define RADEON_FP_DETECT_SENSE          (1 <<  8)
+#       define RADEON_FP_DETECT_INT_POL        (1 <<  9)
 #       define R200_FP_SOURCE_SEL_MASK         (3 <<  10)
 #       define R200_FP_SOURCE_SEL_CRTC1        (0 <<  10)
 #       define R200_FP_SOURCE_SEL_CRTC2        (1 <<  10)
@@ -889,6 +895,7 @@
 #       define R200_FP_SOURCE_SEL_TRANS        (3 <<  10)
 #       define RADEON_FP_SEL_CRTC1             (0 << 13)
 #       define RADEON_FP_SEL_CRTC2             (1 << 13)
+#       define R300_HPD_SEL(x)                 ((x) << 13)
 #       define RADEON_FP_CRTC_DONT_SHADOW_HPAR (1 << 15)
 #       define RADEON_FP_CRTC_DONT_SHADOW_VPAR (1 << 16)
 #       define RADEON_FP_CRTC_DONT_SHADOW_HEND (1 << 17)
@@ -904,6 +911,7 @@
 #       define RADEON_FP2_ON                   (1 <<  2)
 #       define RADEON_FP2_PANEL_FORMAT         (1 <<  3)
 #       define RADEON_FP2_DETECT_SENSE         (1 <<  8)
+#       define RADEON_FP2_DETECT_INT_POL       (1 <<  9)
 #       define R200_FP2_SOURCE_SEL_MASK        (3 << 10)
 #       define R200_FP2_SOURCE_SEL_CRTC1       (0 << 10)
 #       define R200_FP2_SOURCE_SEL_CRTC2       (1 << 10)
@@ -983,14 +991,20 @@
 
 #define RADEON_GEN_INT_CNTL                 0x0040
 #	define RADEON_CRTC_VBLANK_MASK		(1 << 0)
+#	define RADEON_FP_DETECT_MASK		(1 << 4)
 #	define RADEON_CRTC2_VBLANK_MASK		(1 << 9)
+#	define RADEON_FP2_DETECT_MASK		(1 << 10)
 #	define RADEON_SW_INT_ENABLE		(1 << 25)
 #define RADEON_GEN_INT_STATUS               0x0044
 #	define AVIVO_DISPLAY_INT_STATUS		(1 << 0)
 #	define RADEON_CRTC_VBLANK_STAT		(1 << 0)
 #	define RADEON_CRTC_VBLANK_STAT_ACK	(1 << 0)
+#	define RADEON_FP_DETECT_STAT		(1 << 4)
+#	define RADEON_FP_DETECT_STAT_ACK	(1 << 4)
 #	define RADEON_CRTC2_VBLANK_STAT		(1 << 9)
 #	define RADEON_CRTC2_VBLANK_STAT_ACK	(1 << 9)
+#	define RADEON_FP2_DETECT_STAT		(1 << 10)
+#	define RADEON_FP2_DETECT_STAT_ACK	(1 << 10)
 #	define RADEON_SW_INT_FIRE		(1 << 26)
 #	define RADEON_SW_INT_TEST		(1 << 25)
 #	define RADEON_SW_INT_TEST_ACK		(1 << 25)
@@ -1046,28 +1060,39 @@
 
        /* Multimedia I2C bus */
 #define RADEON_I2C_CNTL_0		    0x0090
-#define RADEON_I2C_DONE (1<<0)
-#define RADEON_I2C_NACK (1<<1)
-#define RADEON_I2C_HALT (1<<2)
-#define RADEON_I2C_SOFT_RST (1<<5)
-#define RADEON_I2C_DRIVE_EN (1<<6)
-#define RADEON_I2C_DRIVE_SEL (1<<7)
-#define RADEON_I2C_START (1<<8)
-#define RADEON_I2C_STOP (1<<9)
-#define RADEON_I2C_RECEIVE (1<<10)
-#define RADEON_I2C_ABORT (1<<11)
-#define RADEON_I2C_GO (1<<12)
+#       define RADEON_I2C_DONE              (1 << 0)
+#       define RADEON_I2C_NACK              (1 << 1)
+#       define RADEON_I2C_HALT              (1 << 2)
+#       define RADEON_I2C_SOFT_RST          (1 << 5)
+#       define RADEON_I2C_DRIVE_EN          (1 << 6)
+#       define RADEON_I2C_DRIVE_SEL         (1 << 7)
+#       define RADEON_I2C_START             (1 << 8)
+#       define RADEON_I2C_STOP              (1 << 9)
+#       define RADEON_I2C_RECEIVE           (1 << 10)
+#       define RADEON_I2C_ABORT             (1 << 11)
+#       define RADEON_I2C_GO                (1 << 12)
+#       define RADEON_I2C_PRESCALE_SHIFT    16
 #define RADEON_I2C_CNTL_1                   0x0094
-#define RADEON_I2C_SEL         (1<<16)
-#define RADEON_I2C_EN          (1<<17)
+#       define RADEON_I2C_DATA_COUNT_SHIFT  0
+#       define RADEON_I2C_ADDR_COUNT_SHIFT  4
+#       define RADEON_I2C_INTRA_BYTE_DELAY_SHIFT   8
+#       define RADEON_I2C_SEL               (1 << 16)
+#       define RADEON_I2C_EN                (1 << 17)
+#       define RADEON_I2C_TIME_LIMIT_SHIFT  24
 #define RADEON_I2C_DATA			    0x0098
 
 #define RADEON_DVI_I2C_CNTL_0		    0x02e0
 #       define R200_DVI_I2C_PIN_SEL(x)      ((x) << 3)
-#       define R200_SEL_DDC1                0 /* 0x60 - VGA_DDC */
-#       define R200_SEL_DDC2                1 /* 0x64 - DVI_DDC */
-#       define R200_SEL_DDC3                2 /* 0x68 - MONID_DDC */
-#define RADEON_DVI_I2C_CNTL_1               0x02e4 /* ? */
+#       define R200_SEL_DDC1                0 /* depends on asic */
+#       define R200_SEL_DDC2                1 /* depends on asic */
+#       define R200_SEL_DDC3                2 /* depends on asic */
+#	define RADEON_SW_WANTS_TO_USE_DVI_I2C (1 << 13)
+#	define RADEON_SW_CAN_USE_DVI_I2C      (1 << 13)
+#	define RADEON_SW_DONE_USING_DVI_I2C   (1 << 14)
+#	define RADEON_HW_NEEDS_DVI_I2C        (1 << 14)
+#	define RADEON_ABORT_HW_DVI_I2C        (1 << 15)
+#	define RADEON_HW_USING_DVI_I2C        (1 << 15)
+#define RADEON_DVI_I2C_CNTL_1               0x02e4
 #define RADEON_DVI_I2C_DATA		    0x02e8
 
 #define RADEON_INTERRUPT_LINE               0x0f3c /* PCI */
@@ -1138,15 +1163,16 @@
 #       define RADEON_IO_MCLK_MAX_DYN_STOP_LAT (1 << 13)
 #       define RADEON_MC_MCLK_DYN_ENABLE    (1 << 14)
 #       define RADEON_IO_MCLK_DYN_ENABLE    (1 << 15)
-#define RADEON_LCD_GPIO_MASK                0x01a0
-#define RADEON_GPIOPAD_EN                   0x01a0
-#define RADEON_LCD_GPIO_Y_REG               0x01a4
-#define RADEON_MDGPIO_A_REG                 0x01ac
-#define RADEON_MDGPIO_EN_REG                0x01b0
-#define RADEON_MDGPIO_MASK                  0x0198
+
 #define RADEON_GPIOPAD_MASK                 0x0198
 #define RADEON_GPIOPAD_A		    0x019c
-#define RADEON_MDGPIO_Y_REG                 0x01b4
+#define RADEON_GPIOPAD_EN                   0x01a0
+#define RADEON_GPIOPAD_Y                    0x01a4
+#define RADEON_MDGPIO_MASK                  0x01a8
+#define RADEON_MDGPIO_A                     0x01ac
+#define RADEON_MDGPIO_EN                    0x01b0
+#define RADEON_MDGPIO_Y                     0x01b4
+
 #define RADEON_MEM_ADDR_CONFIG              0x0148
 #define RADEON_MEM_BASE                     0x0f10 /* PCI */
 #define RADEON_MEM_CNTL                     0x0140
@@ -1355,6 +1381,9 @@
 #define RADEON_OVR_CLR                      0x0230
 #define RADEON_OVR_WID_LEFT_RIGHT           0x0234
 #define RADEON_OVR_WID_TOP_BOTTOM           0x0238
+#define RADEON_OVR2_CLR                     0x0330
+#define RADEON_OVR2_WID_LEFT_RIGHT          0x0334
+#define RADEON_OVR2_WID_TOP_BOTTOM          0x0338
 
 /* first capture unit */
 
@@ -3311,6 +3340,7 @@
 #define RADEON_AIC_CNTL                     0x01d0
 #       define RADEON_PCIGART_TRANSLATE_EN     (1 << 0)
 #       define RADEON_DIS_OUT_OF_PCI_GART_ACCESS     (1 << 1)
+#	define RS400_MSI_REARM	                (1 << 3) /* rs400/rs480 */
 #define RADEON_AIC_LO_ADDR                  0x01dc
 #define RADEON_AIC_PT_BASE		0x01d8
 #define RADEON_AIC_HI_ADDR		0x01e0
@@ -3333,6 +3363,7 @@
 #       define RADEON_CP_PACKET_MAX_DWORDS          (1 << 12)
 #       define RADEON_CP_PACKET0_REG_MASK           0x000007ff
 #       define R300_CP_PACKET0_REG_MASK             0x00001fff
+#       define R600_CP_PACKET0_REG_MASK             0x0000ffff
 #       define RADEON_CP_PACKET1_REG0_MASK          0x000007ff
 #       define RADEON_CP_PACKET1_REG1_MASK          0x003ff800
 
diff --git a/drivers/gpu/drm/radeon/radeon_ring.c b/drivers/gpu/drm/radeon/radeon_ring.c
index 747b4bffb84b..e50513a62735 100644
--- a/drivers/gpu/drm/radeon/radeon_ring.c
+++ b/drivers/gpu/drm/radeon/radeon_ring.c
@@ -34,6 +34,36 @@
 
 int radeon_debugfs_ib_init(struct radeon_device *rdev);
 
+void radeon_ib_bogus_cleanup(struct radeon_device *rdev)
+{
+	struct radeon_ib *ib, *n;
+
+	list_for_each_entry_safe(ib, n, &rdev->ib_pool.bogus_ib, list) {
+		list_del(&ib->list);
+		vfree(ib->ptr);
+		kfree(ib);
+	}
+}
+
+void radeon_ib_bogus_add(struct radeon_device *rdev, struct radeon_ib *ib)
+{
+	struct radeon_ib *bib;
+
+	bib = kmalloc(sizeof(*bib), GFP_KERNEL);
+	if (bib == NULL)
+		return;
+	bib->ptr = vmalloc(ib->length_dw * 4);
+	if (bib->ptr == NULL) {
+		kfree(bib);
+		return;
+	}
+	memcpy(bib->ptr, ib->ptr, ib->length_dw * 4);
+	bib->length_dw = ib->length_dw;
+	mutex_lock(&rdev->ib_pool.mutex);
+	list_add_tail(&bib->list, &rdev->ib_pool.bogus_ib);
+	mutex_unlock(&rdev->ib_pool.mutex);
+}
+
 /*
  * IB.
  */
@@ -41,68 +71,55 @@ int radeon_ib_get(struct radeon_device *rdev, struct radeon_ib **ib)
 {
 	struct radeon_fence *fence;
 	struct radeon_ib *nib;
-	unsigned long i;
-	int r = 0;
+	int r = 0, i, c;
 
 	*ib = NULL;
 	r = radeon_fence_create(rdev, &fence);
 	if (r) {
-		DRM_ERROR("failed to create fence for new IB\n");
+		dev_err(rdev->dev, "failed to create fence for new IB\n");
 		return r;
 	}
 	mutex_lock(&rdev->ib_pool.mutex);
-	i = find_first_zero_bit(rdev->ib_pool.alloc_bm, RADEON_IB_POOL_SIZE);
-	if (i < RADEON_IB_POOL_SIZE) {
-		set_bit(i, rdev->ib_pool.alloc_bm);
-		rdev->ib_pool.ibs[i].length_dw = 0;
-		*ib = &rdev->ib_pool.ibs[i];
-		mutex_unlock(&rdev->ib_pool.mutex);
-		goto out;
+	for (i = rdev->ib_pool.head_id, c = 0, nib = NULL; c < RADEON_IB_POOL_SIZE; c++, i++) {
+		i &= (RADEON_IB_POOL_SIZE - 1);
+		if (rdev->ib_pool.ibs[i].free) {
+			nib = &rdev->ib_pool.ibs[i];
+			break;
+		}
 	}
-	if (list_empty(&rdev->ib_pool.scheduled_ibs)) {
-		/* we go do nothings here */
+	if (nib == NULL) {
+		/* This should never happen, it means we allocated all
+		 * IB and haven't scheduled one yet, return EBUSY to
+		 * userspace hoping that on ioctl recall we get better
+		 * luck
+		 */
+		dev_err(rdev->dev, "no free indirect buffer !\n");
 		mutex_unlock(&rdev->ib_pool.mutex);
-		DRM_ERROR("all IB allocated none scheduled.\n");
-		r = -EINVAL;
-		goto out;
+		radeon_fence_unref(&fence);
+		return -EBUSY;
 	}
-	/* get the first ib on the scheduled list */
-	nib = list_entry(rdev->ib_pool.scheduled_ibs.next,
-			 struct radeon_ib, list);
-	if (nib->fence == NULL) {
-		/* we go do nothings here */
+	rdev->ib_pool.head_id = (nib->idx + 1) & (RADEON_IB_POOL_SIZE - 1);
+	nib->free = false;
+	if (nib->fence) {
 		mutex_unlock(&rdev->ib_pool.mutex);
-		DRM_ERROR("IB %lu scheduled without a fence.\n", nib->idx);
-		r = -EINVAL;
-		goto out;
-	}
-	mutex_unlock(&rdev->ib_pool.mutex);
-
-	r = radeon_fence_wait(nib->fence, false);
-	if (r) {
-		DRM_ERROR("radeon: IB(%lu:0x%016lX:%u)\n", nib->idx,
-			  (unsigned long)nib->gpu_addr, nib->length_dw);
-		DRM_ERROR("radeon: GPU lockup detected, fail to get a IB\n");
-		goto out;
+		r = radeon_fence_wait(nib->fence, false);
+		if (r) {
+			dev_err(rdev->dev, "error waiting fence of IB(%u:0x%016lX:%u)\n",
+				nib->idx, (unsigned long)nib->gpu_addr, nib->length_dw);
+			mutex_lock(&rdev->ib_pool.mutex);
+			nib->free = true;
+			mutex_unlock(&rdev->ib_pool.mutex);
+			radeon_fence_unref(&fence);
+			return r;
+		}
+		mutex_lock(&rdev->ib_pool.mutex);
 	}
 	radeon_fence_unref(&nib->fence);
-
+	nib->fence = fence;
 	nib->length_dw = 0;
-
-	/* scheduled list is accessed here */
-	mutex_lock(&rdev->ib_pool.mutex);
-	list_del(&nib->list);
-	INIT_LIST_HEAD(&nib->list);
 	mutex_unlock(&rdev->ib_pool.mutex);
-
 	*ib = nib;
-out:
-	if (r) {
-		radeon_fence_unref(&fence);
-	} else {
-		(*ib)->fence = fence;
-	}
-	return r;
+	return 0;
 }
 
 void radeon_ib_free(struct radeon_device *rdev, struct radeon_ib **ib)
@@ -113,19 +130,10 @@ void radeon_ib_free(struct radeon_device *rdev, struct radeon_ib **ib)
 	if (tmp == NULL) {
 		return;
 	}
-	mutex_lock(&rdev->ib_pool.mutex);
-	if (!list_empty(&tmp->list) && !radeon_fence_signaled(tmp->fence)) {
-		/* IB is scheduled & not signaled don't do anythings */
-		mutex_unlock(&rdev->ib_pool.mutex);
-		return;
-	}
-	list_del(&tmp->list);
-	INIT_LIST_HEAD(&tmp->list);
-	if (tmp->fence)
+	if (!tmp->fence->emited)
 		radeon_fence_unref(&tmp->fence);
-
-	tmp->length_dw = 0;
-	clear_bit(tmp->idx, rdev->ib_pool.alloc_bm);
+	mutex_lock(&rdev->ib_pool.mutex);
+	tmp->free = true;
 	mutex_unlock(&rdev->ib_pool.mutex);
 }
 
@@ -135,7 +143,7 @@ int radeon_ib_schedule(struct radeon_device *rdev, struct radeon_ib *ib)
 
 	if (!ib->length_dw || !rdev->cp.ready) {
 		/* TODO: Nothings in the ib we should report. */
-		DRM_ERROR("radeon: couldn't schedule IB(%lu).\n", ib->idx);
+		DRM_ERROR("radeon: couldn't schedule IB(%u).\n", ib->idx);
 		return -EINVAL;
 	}
 
@@ -148,7 +156,8 @@ int radeon_ib_schedule(struct radeon_device *rdev, struct radeon_ib *ib)
 	radeon_ring_ib_execute(rdev, ib);
 	radeon_fence_emit(rdev, ib->fence);
 	mutex_lock(&rdev->ib_pool.mutex);
-	list_add_tail(&ib->list, &rdev->ib_pool.scheduled_ibs);
+	/* once scheduled IB is considered free and protected by the fence */
+	ib->free = true;
 	mutex_unlock(&rdev->ib_pool.mutex);
 	radeon_ring_unlock_commit(rdev);
 	return 0;
@@ -163,21 +172,26 @@ int radeon_ib_pool_init(struct radeon_device *rdev)
 
 	if (rdev->ib_pool.robj)
 		return 0;
+	INIT_LIST_HEAD(&rdev->ib_pool.bogus_ib);
 	/* Allocate 1M object buffer */
-	INIT_LIST_HEAD(&rdev->ib_pool.scheduled_ibs);
-	r = radeon_object_create(rdev, NULL,  RADEON_IB_POOL_SIZE*64*1024,
-				 true, RADEON_GEM_DOMAIN_GTT,
-				 false, &rdev->ib_pool.robj);
+	r = radeon_bo_create(rdev, NULL,  RADEON_IB_POOL_SIZE*64*1024,
+				true, RADEON_GEM_DOMAIN_GTT,
+				&rdev->ib_pool.robj);
 	if (r) {
 		DRM_ERROR("radeon: failed to ib pool (%d).\n", r);
 		return r;
 	}
-	r = radeon_object_pin(rdev->ib_pool.robj, RADEON_GEM_DOMAIN_GTT, &gpu_addr);
+	r = radeon_bo_reserve(rdev->ib_pool.robj, false);
+	if (unlikely(r != 0))
+		return r;
+	r = radeon_bo_pin(rdev->ib_pool.robj, RADEON_GEM_DOMAIN_GTT, &gpu_addr);
 	if (r) {
+		radeon_bo_unreserve(rdev->ib_pool.robj);
 		DRM_ERROR("radeon: failed to pin ib pool (%d).\n", r);
 		return r;
 	}
-	r = radeon_object_kmap(rdev->ib_pool.robj, &ptr);
+	r = radeon_bo_kmap(rdev->ib_pool.robj, &ptr);
+	radeon_bo_unreserve(rdev->ib_pool.robj);
 	if (r) {
 		DRM_ERROR("radeon: failed to map ib poll (%d).\n", r);
 		return r;
@@ -190,9 +204,9 @@ int radeon_ib_pool_init(struct radeon_device *rdev)
 		rdev->ib_pool.ibs[i].ptr = ptr + offset;
 		rdev->ib_pool.ibs[i].idx = i;
 		rdev->ib_pool.ibs[i].length_dw = 0;
-		INIT_LIST_HEAD(&rdev->ib_pool.ibs[i].list);
+		rdev->ib_pool.ibs[i].free = true;
 	}
-	bitmap_zero(rdev->ib_pool.alloc_bm, RADEON_IB_POOL_SIZE);
+	rdev->ib_pool.head_id = 0;
 	rdev->ib_pool.ready = true;
 	DRM_INFO("radeon: ib pool ready.\n");
 	if (radeon_debugfs_ib_init(rdev)) {
@@ -203,14 +217,22 @@ int radeon_ib_pool_init(struct radeon_device *rdev)
 
 void radeon_ib_pool_fini(struct radeon_device *rdev)
 {
+	int r;
+
 	if (!rdev->ib_pool.ready) {
 		return;
 	}
 	mutex_lock(&rdev->ib_pool.mutex);
-	bitmap_zero(rdev->ib_pool.alloc_bm, RADEON_IB_POOL_SIZE);
+	radeon_ib_bogus_cleanup(rdev);
+
 	if (rdev->ib_pool.robj) {
-		radeon_object_kunmap(rdev->ib_pool.robj);
-		radeon_object_unref(&rdev->ib_pool.robj);
+		r = radeon_bo_reserve(rdev->ib_pool.robj, false);
+		if (likely(r == 0)) {
+			radeon_bo_kunmap(rdev->ib_pool.robj);
+			radeon_bo_unpin(rdev->ib_pool.robj);
+			radeon_bo_unreserve(rdev->ib_pool.robj);
+		}
+		radeon_bo_unref(&rdev->ib_pool.robj);
 		rdev->ib_pool.robj = NULL;
 	}
 	mutex_unlock(&rdev->ib_pool.mutex);
@@ -288,29 +310,28 @@ int radeon_ring_init(struct radeon_device *rdev, unsigned ring_size)
 	rdev->cp.ring_size = ring_size;
 	/* Allocate ring buffer */
 	if (rdev->cp.ring_obj == NULL) {
-		r = radeon_object_create(rdev, NULL, rdev->cp.ring_size,
-					 true,
-					 RADEON_GEM_DOMAIN_GTT,
-					 false,
-					 &rdev->cp.ring_obj);
+		r = radeon_bo_create(rdev, NULL, rdev->cp.ring_size, true,
+					RADEON_GEM_DOMAIN_GTT,
+					&rdev->cp.ring_obj);
 		if (r) {
-			DRM_ERROR("radeon: failed to create ring buffer (%d).\n", r);
-			mutex_unlock(&rdev->cp.mutex);
+			dev_err(rdev->dev, "(%d) ring create failed\n", r);
 			return r;
 		}
-		r = radeon_object_pin(rdev->cp.ring_obj,
-				      RADEON_GEM_DOMAIN_GTT,
-				      &rdev->cp.gpu_addr);
+		r = radeon_bo_reserve(rdev->cp.ring_obj, false);
+		if (unlikely(r != 0))
+			return r;
+		r = radeon_bo_pin(rdev->cp.ring_obj, RADEON_GEM_DOMAIN_GTT,
+					&rdev->cp.gpu_addr);
 		if (r) {
-			DRM_ERROR("radeon: failed to pin ring buffer (%d).\n", r);
-			mutex_unlock(&rdev->cp.mutex);
+			radeon_bo_unreserve(rdev->cp.ring_obj);
+			dev_err(rdev->dev, "(%d) ring pin failed\n", r);
 			return r;
 		}
-		r = radeon_object_kmap(rdev->cp.ring_obj,
+		r = radeon_bo_kmap(rdev->cp.ring_obj,
 				       (void **)&rdev->cp.ring);
+		radeon_bo_unreserve(rdev->cp.ring_obj);
 		if (r) {
-			DRM_ERROR("radeon: failed to map ring buffer (%d).\n", r);
-			mutex_unlock(&rdev->cp.mutex);
+			dev_err(rdev->dev, "(%d) ring map failed\n", r);
 			return r;
 		}
 	}
@@ -321,11 +342,17 @@ int radeon_ring_init(struct radeon_device *rdev, unsigned ring_size)
 
 void radeon_ring_fini(struct radeon_device *rdev)
 {
+	int r;
+
 	mutex_lock(&rdev->cp.mutex);
 	if (rdev->cp.ring_obj) {
-		radeon_object_kunmap(rdev->cp.ring_obj);
-		radeon_object_unpin(rdev->cp.ring_obj);
-		radeon_object_unref(&rdev->cp.ring_obj);
+		r = radeon_bo_reserve(rdev->cp.ring_obj, false);
+		if (likely(r == 0)) {
+			radeon_bo_kunmap(rdev->cp.ring_obj);
+			radeon_bo_unpin(rdev->cp.ring_obj);
+			radeon_bo_unreserve(rdev->cp.ring_obj);
+		}
+		radeon_bo_unref(&rdev->cp.ring_obj);
 		rdev->cp.ring = NULL;
 		rdev->cp.ring_obj = NULL;
 	}
@@ -346,7 +373,7 @@ static int radeon_debugfs_ib_info(struct seq_file *m, void *data)
 	if (ib == NULL) {
 		return 0;
 	}
-	seq_printf(m, "IB %04lu\n", ib->idx);
+	seq_printf(m, "IB %04u\n", ib->idx);
 	seq_printf(m, "IB fence %p\n", ib->fence);
 	seq_printf(m, "IB size %05u dwords\n", ib->length_dw);
 	for (i = 0; i < ib->length_dw; i++) {
@@ -355,15 +382,49 @@ static int radeon_debugfs_ib_info(struct seq_file *m, void *data)
 	return 0;
 }
 
+static int radeon_debugfs_ib_bogus_info(struct seq_file *m, void *data)
+{
+	struct drm_info_node *node = (struct drm_info_node *) m->private;
+	struct radeon_device *rdev = node->info_ent->data;
+	struct radeon_ib *ib;
+	unsigned i;
+
+	mutex_lock(&rdev->ib_pool.mutex);
+	if (list_empty(&rdev->ib_pool.bogus_ib)) {
+		mutex_unlock(&rdev->ib_pool.mutex);
+		seq_printf(m, "no bogus IB recorded\n");
+		return 0;
+	}
+	ib = list_first_entry(&rdev->ib_pool.bogus_ib, struct radeon_ib, list);
+	list_del_init(&ib->list);
+	mutex_unlock(&rdev->ib_pool.mutex);
+	seq_printf(m, "IB size %05u dwords\n", ib->length_dw);
+	for (i = 0; i < ib->length_dw; i++) {
+		seq_printf(m, "[%05u]=0x%08X\n", i, ib->ptr[i]);
+	}
+	vfree(ib->ptr);
+	kfree(ib);
+	return 0;
+}
+
 static struct drm_info_list radeon_debugfs_ib_list[RADEON_IB_POOL_SIZE];
 static char radeon_debugfs_ib_names[RADEON_IB_POOL_SIZE][32];
+
+static struct drm_info_list radeon_debugfs_ib_bogus_info_list[] = {
+	{"radeon_ib_bogus", radeon_debugfs_ib_bogus_info, 0, NULL},
+};
 #endif
 
 int radeon_debugfs_ib_init(struct radeon_device *rdev)
 {
 #if defined(CONFIG_DEBUG_FS)
 	unsigned i;
+	int r;
 
+	radeon_debugfs_ib_bogus_info_list[0].data = rdev;
+	r = radeon_debugfs_add_files(rdev, radeon_debugfs_ib_bogus_info_list, 1);
+	if (r)
+		return r;
 	for (i = 0; i < RADEON_IB_POOL_SIZE; i++) {
 		sprintf(radeon_debugfs_ib_names[i], "radeon_ib_%04u", i);
 		radeon_debugfs_ib_list[i].name = radeon_debugfs_ib_names[i];
diff --git a/drivers/gpu/drm/radeon/radeon_state.c b/drivers/gpu/drm/radeon/radeon_state.c
index 38537d971a3e..40ab6d9c3736 100644
--- a/drivers/gpu/drm/radeon/radeon_state.c
+++ b/drivers/gpu/drm/radeon/radeon_state.c
@@ -29,6 +29,7 @@
 
 #include "drmP.h"
 #include "drm.h"
+#include "drm_buffer.h"
 #include "drm_sarea.h"
 #include "radeon_drm.h"
 #include "radeon_drv.h"
@@ -91,21 +92,27 @@ static __inline__ int radeon_check_and_fixup_offset(drm_radeon_private_t *
 static __inline__ int radeon_check_and_fixup_packets(drm_radeon_private_t *
 						     dev_priv,
 						     struct drm_file *file_priv,
-						     int id, u32 *data)
+						     int id, struct drm_buffer *buf)
 {
+	u32 *data;
 	switch (id) {
 
 	case RADEON_EMIT_PP_MISC:
-		if (radeon_check_and_fixup_offset(dev_priv, file_priv,
-		    &data[(RADEON_RB3D_DEPTHOFFSET - RADEON_PP_MISC) / 4])) {
+		data = drm_buffer_pointer_to_dword(buf,
+			(RADEON_RB3D_DEPTHOFFSET - RADEON_PP_MISC) / 4);
+
+		if (radeon_check_and_fixup_offset(dev_priv, file_priv, data)) {
 			DRM_ERROR("Invalid depth buffer offset\n");
 			return -EINVAL;
 		}
+		dev_priv->have_z_offset = 1;
 		break;
 
 	case RADEON_EMIT_PP_CNTL:
-		if (radeon_check_and_fixup_offset(dev_priv, file_priv,
-		    &data[(RADEON_RB3D_COLOROFFSET - RADEON_PP_CNTL) / 4])) {
+		data = drm_buffer_pointer_to_dword(buf,
+			(RADEON_RB3D_COLOROFFSET - RADEON_PP_CNTL) / 4);
+
+		if (radeon_check_and_fixup_offset(dev_priv, file_priv, data)) {
 			DRM_ERROR("Invalid colour buffer offset\n");
 			return -EINVAL;
 		}
@@ -117,8 +124,8 @@ static __inline__ int radeon_check_and_fixup_packets(drm_radeon_private_t *
 	case R200_EMIT_PP_TXOFFSET_3:
 	case R200_EMIT_PP_TXOFFSET_4:
 	case R200_EMIT_PP_TXOFFSET_5:
-		if (radeon_check_and_fixup_offset(dev_priv, file_priv,
-						  &data[0])) {
+		data = drm_buffer_pointer_to_dword(buf, 0);
+		if (radeon_check_and_fixup_offset(dev_priv, file_priv, data)) {
 			DRM_ERROR("Invalid R200 texture offset\n");
 			return -EINVAL;
 		}
@@ -127,8 +134,9 @@ static __inline__ int radeon_check_and_fixup_packets(drm_radeon_private_t *
 	case RADEON_EMIT_PP_TXFILTER_0:
 	case RADEON_EMIT_PP_TXFILTER_1:
 	case RADEON_EMIT_PP_TXFILTER_2:
-		if (radeon_check_and_fixup_offset(dev_priv, file_priv,
-		    &data[(RADEON_PP_TXOFFSET_0 - RADEON_PP_TXFILTER_0) / 4])) {
+		data = drm_buffer_pointer_to_dword(buf,
+			(RADEON_PP_TXOFFSET_0 - RADEON_PP_TXFILTER_0) / 4);
+		if (radeon_check_and_fixup_offset(dev_priv, file_priv, data)) {
 			DRM_ERROR("Invalid R100 texture offset\n");
 			return -EINVAL;
 		}
@@ -142,9 +150,10 @@ static __inline__ int radeon_check_and_fixup_packets(drm_radeon_private_t *
 	case R200_EMIT_PP_CUBIC_OFFSETS_5:{
 			int i;
 			for (i = 0; i < 5; i++) {
+				data = drm_buffer_pointer_to_dword(buf, i);
 				if (radeon_check_and_fixup_offset(dev_priv,
 								  file_priv,
-								  &data[i])) {
+								  data)) {
 					DRM_ERROR
 					    ("Invalid R200 cubic texture offset\n");
 					return -EINVAL;
@@ -158,9 +167,10 @@ static __inline__ int radeon_check_and_fixup_packets(drm_radeon_private_t *
 	case RADEON_EMIT_PP_CUBIC_OFFSETS_T2:{
 			int i;
 			for (i = 0; i < 5; i++) {
+				data = drm_buffer_pointer_to_dword(buf, i);
 				if (radeon_check_and_fixup_offset(dev_priv,
 								  file_priv,
-								  &data[i])) {
+								  data)) {
 					DRM_ERROR
 					    ("Invalid R100 cubic texture offset\n");
 					return -EINVAL;
@@ -269,23 +279,24 @@ static __inline__ int radeon_check_and_fixup_packet3(drm_radeon_private_t *
 						     cmdbuf,
 						     unsigned int *cmdsz)
 {
-	u32 *cmd = (u32 *) cmdbuf->buf;
+	u32 *cmd = drm_buffer_pointer_to_dword(cmdbuf->buffer, 0);
 	u32 offset, narrays;
 	int count, i, k;
 
-	*cmdsz = 2 + ((cmd[0] & RADEON_CP_PACKET_COUNT_MASK) >> 16);
+	count = ((*cmd & RADEON_CP_PACKET_COUNT_MASK) >> 16);
+	*cmdsz = 2 + count;
 
-	if ((cmd[0] & 0xc0000000) != RADEON_CP_PACKET3) {
+	if ((*cmd & 0xc0000000) != RADEON_CP_PACKET3) {
 		DRM_ERROR("Not a type 3 packet\n");
 		return -EINVAL;
 	}
 
-	if (4 * *cmdsz > cmdbuf->bufsz) {
+	if (4 * *cmdsz > drm_buffer_unprocessed(cmdbuf->buffer)) {
 		DRM_ERROR("Packet size larger than size of data provided\n");
 		return -EINVAL;
 	}
 
-	switch(cmd[0] & 0xff00) {
+	switch (*cmd & 0xff00) {
 	/* XXX Are there old drivers needing other packets? */
 
 	case RADEON_3D_DRAW_IMMD:
@@ -312,7 +323,6 @@ static __inline__ int radeon_check_and_fixup_packet3(drm_radeon_private_t *
 		break;
 
 	case RADEON_3D_LOAD_VBPNTR:
-		count = (cmd[0] >> 16) & 0x3fff;
 
 		if (count > 18) { /* 12 arrays max */
 			DRM_ERROR("Too large payload in 3D_LOAD_VBPNTR (count=%d)\n",
@@ -321,13 +331,16 @@ static __inline__ int radeon_check_and_fixup_packet3(drm_radeon_private_t *
 		}
 
 		/* carefully check packet contents */
-		narrays = cmd[1] & ~0xc000;
+		cmd = drm_buffer_pointer_to_dword(cmdbuf->buffer, 1);
+
+		narrays = *cmd & ~0xc000;
 		k = 0;
 		i = 2;
 		while ((k < narrays) && (i < (count + 2))) {
 			i++;		/* skip attribute field */
+			cmd = drm_buffer_pointer_to_dword(cmdbuf->buffer, i);
 			if (radeon_check_and_fixup_offset(dev_priv, file_priv,
-							  &cmd[i])) {
+							  cmd)) {
 				DRM_ERROR
 				    ("Invalid offset (k=%d i=%d) in 3D_LOAD_VBPNTR packet.\n",
 				     k, i);
@@ -338,8 +351,10 @@ static __inline__ int radeon_check_and_fixup_packet3(drm_radeon_private_t *
 			if (k == narrays)
 				break;
 			/* have one more to process, they come in pairs */
+			cmd = drm_buffer_pointer_to_dword(cmdbuf->buffer, i);
+
 			if (radeon_check_and_fixup_offset(dev_priv,
-							  file_priv, &cmd[i]))
+							  file_priv, cmd))
 			{
 				DRM_ERROR
 				    ("Invalid offset (k=%d i=%d) in 3D_LOAD_VBPNTR packet.\n",
@@ -363,7 +378,9 @@ static __inline__ int radeon_check_and_fixup_packet3(drm_radeon_private_t *
 			DRM_ERROR("Invalid 3d packet for r200-class chip\n");
 			return -EINVAL;
 		}
-		if (radeon_check_and_fixup_offset(dev_priv, file_priv, &cmd[1])) {
+
+		cmd = drm_buffer_pointer_to_dword(cmdbuf->buffer, 1);
+		if (radeon_check_and_fixup_offset(dev_priv, file_priv, cmd)) {
 				DRM_ERROR("Invalid rndr_gen_indx offset\n");
 				return -EINVAL;
 		}
@@ -374,12 +391,15 @@ static __inline__ int radeon_check_and_fixup_packet3(drm_radeon_private_t *
 			DRM_ERROR("Invalid 3d packet for r100-class chip\n");
 			return -EINVAL;
 		}
-		if ((cmd[1] & 0x8000ffff) != 0x80000810) {
-			DRM_ERROR("Invalid indx_buffer reg address %08X\n", cmd[1]);
+
+		cmd = drm_buffer_pointer_to_dword(cmdbuf->buffer, 1);
+		if ((*cmd & 0x8000ffff) != 0x80000810) {
+			DRM_ERROR("Invalid indx_buffer reg address %08X\n", *cmd);
 			return -EINVAL;
 		}
-		if (radeon_check_and_fixup_offset(dev_priv, file_priv, &cmd[2])) {
-			DRM_ERROR("Invalid indx_buffer offset is %08X\n", cmd[2]);
+		cmd = drm_buffer_pointer_to_dword(cmdbuf->buffer, 2);
+		if (radeon_check_and_fixup_offset(dev_priv, file_priv, cmd)) {
+			DRM_ERROR("Invalid indx_buffer offset is %08X\n", *cmd);
 			return -EINVAL;
 		}
 		break;
@@ -388,31 +408,34 @@ static __inline__ int radeon_check_and_fixup_packet3(drm_radeon_private_t *
 	case RADEON_CNTL_PAINT_MULTI:
 	case RADEON_CNTL_BITBLT_MULTI:
 		/* MSB of opcode: next DWORD GUI_CNTL */
-		if (cmd[1] & (RADEON_GMC_SRC_PITCH_OFFSET_CNTL
+		cmd = drm_buffer_pointer_to_dword(cmdbuf->buffer, 1);
+		if (*cmd & (RADEON_GMC_SRC_PITCH_OFFSET_CNTL
 			      | RADEON_GMC_DST_PITCH_OFFSET_CNTL)) {
-			offset = cmd[2] << 10;
+			u32 *cmd2 = drm_buffer_pointer_to_dword(cmdbuf->buffer, 2);
+			offset = *cmd2 << 10;
 			if (radeon_check_and_fixup_offset
 			    (dev_priv, file_priv, &offset)) {
 				DRM_ERROR("Invalid first packet offset\n");
 				return -EINVAL;
 			}
-			cmd[2] = (cmd[2] & 0xffc00000) | offset >> 10;
+			*cmd2 = (*cmd2 & 0xffc00000) | offset >> 10;
 		}
 
-		if ((cmd[1] & RADEON_GMC_SRC_PITCH_OFFSET_CNTL) &&
-		    (cmd[1] & RADEON_GMC_DST_PITCH_OFFSET_CNTL)) {
-			offset = cmd[3] << 10;
+		if ((*cmd & RADEON_GMC_SRC_PITCH_OFFSET_CNTL) &&
+		    (*cmd & RADEON_GMC_DST_PITCH_OFFSET_CNTL)) {
+			u32 *cmd3 = drm_buffer_pointer_to_dword(cmdbuf->buffer, 3);
+			offset = *cmd << 10;
 			if (radeon_check_and_fixup_offset
 			    (dev_priv, file_priv, &offset)) {
 				DRM_ERROR("Invalid second packet offset\n");
 				return -EINVAL;
 			}
-			cmd[3] = (cmd[3] & 0xffc00000) | offset >> 10;
+			*cmd3 = (*cmd3 & 0xffc00000) | offset >> 10;
 		}
 		break;
 
 	default:
-		DRM_ERROR("Invalid packet type %x\n", cmd[0] & 0xff00);
+		DRM_ERROR("Invalid packet type %x\n", *cmd & 0xff00);
 		return -EINVAL;
 	}
 
@@ -876,6 +899,11 @@ static void radeon_cp_dispatch_clear(struct drm_device * dev,
 		if (tmp & RADEON_BACK)
 			flags |= RADEON_FRONT;
 	}
+	if (flags & (RADEON_DEPTH|RADEON_STENCIL)) {
+		if (!dev_priv->have_z_offset)
+			printk_once(KERN_ERR "radeon: illegal depth clear request. Buggy mesa detected - please update.\n");
+		flags &= ~(RADEON_DEPTH | RADEON_STENCIL);
+	}
 
 	if (flags & (RADEON_FRONT | RADEON_BACK)) {
 
@@ -1065,7 +1093,7 @@ static void radeon_cp_dispatch_clear(struct drm_device * dev,
 					/* judging by the first tile offset needed, could possibly
 					   directly address/clear 4x4 tiles instead of 8x2 * 4x4
 					   macro tiles, though would still need clear mask for
-					   right/bottom if truely 4x4 granularity is desired ? */
+					   right/bottom if truly 4x4 granularity is desired ? */
 					OUT_RING(tileoffset * 16);
 					/* the number of tiles to clear */
 					OUT_RING(nrtilesx + 1);
@@ -1950,7 +1978,7 @@ static void radeon_apply_surface_regs(int surf_index,
  * Note that refcount can be at most 2, since during a free refcount=3
  * might mean we have to allocate a new surface which might not always
  * be available.
- * For example : we allocate three contigous surfaces ABC. If B is
+ * For example : we allocate three contiguous surfaces ABC. If B is
  * freed, we suddenly need two surfaces to store A and C, which might
  * not always be available.
  */
@@ -2611,7 +2639,6 @@ static int radeon_emit_packets(drm_radeon_private_t * dev_priv,
 {
 	int id = (int)header.packet.packet_id;
 	int sz, reg;
-	int *data = (int *)cmdbuf->buf;
 	RING_LOCALS;
 
 	if (id >= RADEON_MAX_STATE_PACKETS)
@@ -2620,23 +2647,22 @@ static int radeon_emit_packets(drm_radeon_private_t * dev_priv,
 	sz = packet[id].len;
 	reg = packet[id].start;
 
-	if (sz * sizeof(int) > cmdbuf->bufsz) {
+	if (sz * sizeof(u32) > drm_buffer_unprocessed(cmdbuf->buffer)) {
 		DRM_ERROR("Packet size provided larger than data provided\n");
 		return -EINVAL;
 	}
 
-	if (radeon_check_and_fixup_packets(dev_priv, file_priv, id, data)) {
+	if (radeon_check_and_fixup_packets(dev_priv, file_priv, id,
+				cmdbuf->buffer)) {
 		DRM_ERROR("Packet verification failed\n");
 		return -EINVAL;
 	}
 
 	BEGIN_RING(sz + 1);
 	OUT_RING(CP_PACKET0(reg, (sz - 1)));
-	OUT_RING_TABLE(data, sz);
+	OUT_RING_DRM_BUFFER(cmdbuf->buffer, sz);
 	ADVANCE_RING();
 
-	cmdbuf->buf += sz * sizeof(int);
-	cmdbuf->bufsz -= sz * sizeof(int);
 	return 0;
 }
 
@@ -2653,10 +2679,8 @@ static __inline__ int radeon_emit_scalars(drm_radeon_private_t *dev_priv,
 	OUT_RING(CP_PACKET0(RADEON_SE_TCL_SCALAR_INDX_REG, 0));
 	OUT_RING(start | (stride << RADEON_SCAL_INDX_DWORD_STRIDE_SHIFT));
 	OUT_RING(CP_PACKET0_TABLE(RADEON_SE_TCL_SCALAR_DATA_REG, sz - 1));
-	OUT_RING_TABLE(cmdbuf->buf, sz);
+	OUT_RING_DRM_BUFFER(cmdbuf->buffer, sz);
 	ADVANCE_RING();
-	cmdbuf->buf += sz * sizeof(int);
-	cmdbuf->bufsz -= sz * sizeof(int);
 	return 0;
 }
 
@@ -2675,10 +2699,8 @@ static __inline__ int radeon_emit_scalars2(drm_radeon_private_t *dev_priv,
 	OUT_RING(CP_PACKET0(RADEON_SE_TCL_SCALAR_INDX_REG, 0));
 	OUT_RING(start | (stride << RADEON_SCAL_INDX_DWORD_STRIDE_SHIFT));
 	OUT_RING(CP_PACKET0_TABLE(RADEON_SE_TCL_SCALAR_DATA_REG, sz - 1));
-	OUT_RING_TABLE(cmdbuf->buf, sz);
+	OUT_RING_DRM_BUFFER(cmdbuf->buffer, sz);
 	ADVANCE_RING();
-	cmdbuf->buf += sz * sizeof(int);
-	cmdbuf->bufsz -= sz * sizeof(int);
 	return 0;
 }
 
@@ -2696,11 +2718,9 @@ static __inline__ int radeon_emit_vectors(drm_radeon_private_t *dev_priv,
 	OUT_RING(CP_PACKET0(RADEON_SE_TCL_VECTOR_INDX_REG, 0));
 	OUT_RING(start | (stride << RADEON_VEC_INDX_OCTWORD_STRIDE_SHIFT));
 	OUT_RING(CP_PACKET0_TABLE(RADEON_SE_TCL_VECTOR_DATA_REG, (sz - 1)));
-	OUT_RING_TABLE(cmdbuf->buf, sz);
+	OUT_RING_DRM_BUFFER(cmdbuf->buffer, sz);
 	ADVANCE_RING();
 
-	cmdbuf->buf += sz * sizeof(int);
-	cmdbuf->bufsz -= sz * sizeof(int);
 	return 0;
 }
 
@@ -2714,7 +2734,7 @@ static __inline__ int radeon_emit_veclinear(drm_radeon_private_t *dev_priv,
 
         if (!sz)
                 return 0;
-        if (sz * 4 > cmdbuf->bufsz)
+	if (sz * 4 > drm_buffer_unprocessed(cmdbuf->buffer))
                 return -EINVAL;
 
 	BEGIN_RING(5 + sz);
@@ -2722,11 +2742,9 @@ static __inline__ int radeon_emit_veclinear(drm_radeon_private_t *dev_priv,
 	OUT_RING(CP_PACKET0(RADEON_SE_TCL_VECTOR_INDX_REG, 0));
 	OUT_RING(start | (1 << RADEON_VEC_INDX_OCTWORD_STRIDE_SHIFT));
 	OUT_RING(CP_PACKET0_TABLE(RADEON_SE_TCL_VECTOR_DATA_REG, (sz - 1)));
-	OUT_RING_TABLE(cmdbuf->buf, sz);
+	OUT_RING_DRM_BUFFER(cmdbuf->buffer, sz);
 	ADVANCE_RING();
 
-	cmdbuf->buf += sz * sizeof(int);
-	cmdbuf->bufsz -= sz * sizeof(int);
 	return 0;
 }
 
@@ -2748,11 +2766,9 @@ static int radeon_emit_packet3(struct drm_device * dev,
 	}
 
 	BEGIN_RING(cmdsz);
-	OUT_RING_TABLE(cmdbuf->buf, cmdsz);
+	OUT_RING_DRM_BUFFER(cmdbuf->buffer, cmdsz);
 	ADVANCE_RING();
 
-	cmdbuf->buf += cmdsz * 4;
-	cmdbuf->bufsz -= cmdsz * 4;
 	return 0;
 }
 
@@ -2805,16 +2821,16 @@ static int radeon_emit_packet3_cliprect(struct drm_device *dev,
 		}
 
 		BEGIN_RING(cmdsz);
-		OUT_RING_TABLE(cmdbuf->buf, cmdsz);
+		OUT_RING_DRM_BUFFER(cmdbuf->buffer, cmdsz);
 		ADVANCE_RING();
 
 	} while (++i < cmdbuf->nbox);
 	if (cmdbuf->nbox == 1)
 		cmdbuf->nbox = 0;
 
+	return 0;
       out:
-	cmdbuf->buf += cmdsz * 4;
-	cmdbuf->bufsz -= cmdsz * 4;
+	drm_buffer_advance(cmdbuf->buffer, cmdsz * 4);
 	return 0;
 }
 
@@ -2847,16 +2863,16 @@ static int radeon_emit_wait(struct drm_device * dev, int flags)
 	return 0;
 }
 
-static int radeon_cp_cmdbuf(struct drm_device *dev, void *data, struct drm_file *file_priv)
+static int radeon_cp_cmdbuf(struct drm_device *dev, void *data,
+		struct drm_file *file_priv)
 {
 	drm_radeon_private_t *dev_priv = dev->dev_private;
 	struct drm_device_dma *dma = dev->dma;
 	struct drm_buf *buf = NULL;
+	drm_radeon_cmd_header_t stack_header;
 	int idx;
 	drm_radeon_kcmd_buffer_t *cmdbuf = data;
-	drm_radeon_cmd_header_t header;
-	int orig_nbox, orig_bufsz;
-	char *kbuf = NULL;
+	int orig_nbox;
 
 	LOCK_TEST_WITH_RETURN(dev, file_priv);
 
@@ -2871,17 +2887,16 @@ static int radeon_cp_cmdbuf(struct drm_device *dev, void *data, struct drm_file
 	 * races between checking values and using those values in other code,
 	 * and simply to avoid a lot of function calls to copy in data.
 	 */
-	orig_bufsz = cmdbuf->bufsz;
-	if (orig_bufsz != 0) {
-		kbuf = kmalloc(cmdbuf->bufsz, GFP_KERNEL);
-		if (kbuf == NULL)
-			return -ENOMEM;
-		if (DRM_COPY_FROM_USER(kbuf, (void __user *)cmdbuf->buf,
-				       cmdbuf->bufsz)) {
-			kfree(kbuf);
-			return -EFAULT;
-		}
-		cmdbuf->buf = kbuf;
+	if (cmdbuf->bufsz != 0) {
+		int rv;
+		void __user *buffer = cmdbuf->buffer;
+		rv = drm_buffer_alloc(&cmdbuf->buffer, cmdbuf->bufsz);
+		if (rv)
+			return rv;
+		rv = drm_buffer_copy_from_user(cmdbuf->buffer, buffer,
+						cmdbuf->bufsz);
+		if (rv)
+			return rv;
 	}
 
 	orig_nbox = cmdbuf->nbox;
@@ -2890,24 +2905,24 @@ static int radeon_cp_cmdbuf(struct drm_device *dev, void *data, struct drm_file
 		int temp;
 		temp = r300_do_cp_cmdbuf(dev, file_priv, cmdbuf);
 
-		if (orig_bufsz != 0)
-			kfree(kbuf);
+		if (cmdbuf->bufsz != 0)
+			drm_buffer_free(cmdbuf->buffer);
 
 		return temp;
 	}
 
 	/* microcode_version != r300 */
-	while (cmdbuf->bufsz >= sizeof(header)) {
+	while (drm_buffer_unprocessed(cmdbuf->buffer) >= sizeof(stack_header)) {
 
-		header.i = *(int *)cmdbuf->buf;
-		cmdbuf->buf += sizeof(header);
-		cmdbuf->bufsz -= sizeof(header);
+		drm_radeon_cmd_header_t *header;
+		header = drm_buffer_read_object(cmdbuf->buffer,
+				sizeof(stack_header), &stack_header);
 
-		switch (header.header.cmd_type) {
+		switch (header->header.cmd_type) {
 		case RADEON_CMD_PACKET:
 			DRM_DEBUG("RADEON_CMD_PACKET\n");
 			if (radeon_emit_packets
-			    (dev_priv, file_priv, header, cmdbuf)) {
+			    (dev_priv, file_priv, *header, cmdbuf)) {
 				DRM_ERROR("radeon_emit_packets failed\n");
 				goto err;
 			}
@@ -2915,7 +2930,7 @@ static int radeon_cp_cmdbuf(struct drm_device *dev, void *data, struct drm_file
 
 		case RADEON_CMD_SCALARS:
 			DRM_DEBUG("RADEON_CMD_SCALARS\n");
-			if (radeon_emit_scalars(dev_priv, header, cmdbuf)) {
+			if (radeon_emit_scalars(dev_priv, *header, cmdbuf)) {
 				DRM_ERROR("radeon_emit_scalars failed\n");
 				goto err;
 			}
@@ -2923,7 +2938,7 @@ static int radeon_cp_cmdbuf(struct drm_device *dev, void *data, struct drm_file
 
 		case RADEON_CMD_VECTORS:
 			DRM_DEBUG("RADEON_CMD_VECTORS\n");
-			if (radeon_emit_vectors(dev_priv, header, cmdbuf)) {
+			if (radeon_emit_vectors(dev_priv, *header, cmdbuf)) {
 				DRM_ERROR("radeon_emit_vectors failed\n");
 				goto err;
 			}
@@ -2931,7 +2946,7 @@ static int radeon_cp_cmdbuf(struct drm_device *dev, void *data, struct drm_file
 
 		case RADEON_CMD_DMA_DISCARD:
 			DRM_DEBUG("RADEON_CMD_DMA_DISCARD\n");
-			idx = header.dma.buf_idx;
+			idx = header->dma.buf_idx;
 			if (idx < 0 || idx >= dma->buf_count) {
 				DRM_ERROR("buffer index %d (of %d max)\n",
 					  idx, dma->buf_count - 1);
@@ -2968,7 +2983,7 @@ static int radeon_cp_cmdbuf(struct drm_device *dev, void *data, struct drm_file
 
 		case RADEON_CMD_SCALARS2:
 			DRM_DEBUG("RADEON_CMD_SCALARS2\n");
-			if (radeon_emit_scalars2(dev_priv, header, cmdbuf)) {
+			if (radeon_emit_scalars2(dev_priv, *header, cmdbuf)) {
 				DRM_ERROR("radeon_emit_scalars2 failed\n");
 				goto err;
 			}
@@ -2976,37 +2991,37 @@ static int radeon_cp_cmdbuf(struct drm_device *dev, void *data, struct drm_file
 
 		case RADEON_CMD_WAIT:
 			DRM_DEBUG("RADEON_CMD_WAIT\n");
-			if (radeon_emit_wait(dev, header.wait.flags)) {
+			if (radeon_emit_wait(dev, header->wait.flags)) {
 				DRM_ERROR("radeon_emit_wait failed\n");
 				goto err;
 			}
 			break;
 		case RADEON_CMD_VECLINEAR:
 			DRM_DEBUG("RADEON_CMD_VECLINEAR\n");
-			if (radeon_emit_veclinear(dev_priv, header, cmdbuf)) {
+			if (radeon_emit_veclinear(dev_priv, *header, cmdbuf)) {
 				DRM_ERROR("radeon_emit_veclinear failed\n");
 				goto err;
 			}
 			break;
 
 		default:
-			DRM_ERROR("bad cmd_type %d at %p\n",
-				  header.header.cmd_type,
-				  cmdbuf->buf - sizeof(header));
+			DRM_ERROR("bad cmd_type %d at byte %d\n",
+				  header->header.cmd_type,
+				  cmdbuf->buffer->iterator);
 			goto err;
 		}
 	}
 
-	if (orig_bufsz != 0)
-		kfree(kbuf);
+	if (cmdbuf->bufsz != 0)
+		drm_buffer_free(cmdbuf->buffer);
 
 	DRM_DEBUG("DONE\n");
 	COMMIT_RING();
 	return 0;
 
       err:
-	if (orig_bufsz != 0)
-		kfree(kbuf);
+	if (cmdbuf->bufsz != 0)
+		drm_buffer_free(cmdbuf->buffer);
 	return -EINVAL;
 }
 
diff --git a/drivers/gpu/drm/radeon/radeon_test.c b/drivers/gpu/drm/radeon/radeon_test.c
index 03c33cf4e14c..313c96bc09da 100644
--- a/drivers/gpu/drm/radeon/radeon_test.c
+++ b/drivers/gpu/drm/radeon/radeon_test.c
@@ -30,8 +30,8 @@
 /* Test BO GTT->VRAM and VRAM->GTT GPU copies across the whole GTT aperture */
 void radeon_test_moves(struct radeon_device *rdev)
 {
-	struct radeon_object *vram_obj = NULL;
-	struct radeon_object **gtt_obj = NULL;
+	struct radeon_bo *vram_obj = NULL;
+	struct radeon_bo **gtt_obj = NULL;
 	struct radeon_fence *fence = NULL;
 	uint64_t gtt_addr, vram_addr;
 	unsigned i, n, size;
@@ -42,8 +42,8 @@ void radeon_test_moves(struct radeon_device *rdev)
 	/* Number of tests =
 	 * (Total GTT - IB pool - writeback page - ring buffer) / test size
 	 */
-	n = (rdev->mc.gtt_size - RADEON_IB_POOL_SIZE*64*1024 - 4096 -
-	     rdev->cp.ring_size) / size;
+	n = ((u32)(rdev->mc.gtt_size - RADEON_IB_POOL_SIZE*64*1024 - RADEON_GPU_PAGE_SIZE -
+	     rdev->cp.ring_size)) / size;
 
 	gtt_obj = kzalloc(n * sizeof(*gtt_obj), GFP_KERNEL);
 	if (!gtt_obj) {
@@ -52,38 +52,42 @@ void radeon_test_moves(struct radeon_device *rdev)
 		goto out_cleanup;
 	}
 
-	r = radeon_object_create(rdev, NULL, size, true, RADEON_GEM_DOMAIN_VRAM,
-				 false, &vram_obj);
+	r = radeon_bo_create(rdev, NULL, size, true, RADEON_GEM_DOMAIN_VRAM,
+				&vram_obj);
 	if (r) {
 		DRM_ERROR("Failed to create VRAM object\n");
 		goto out_cleanup;
 	}
-
-	r = radeon_object_pin(vram_obj, RADEON_GEM_DOMAIN_VRAM, &vram_addr);
+	r = radeon_bo_reserve(vram_obj, false);
+	if (unlikely(r != 0))
+		goto out_cleanup;
+	r = radeon_bo_pin(vram_obj, RADEON_GEM_DOMAIN_VRAM, &vram_addr);
 	if (r) {
 		DRM_ERROR("Failed to pin VRAM object\n");
 		goto out_cleanup;
 	}
-
 	for (i = 0; i < n; i++) {
 		void *gtt_map, *vram_map;
 		void **gtt_start, **gtt_end;
 		void **vram_start, **vram_end;
 
-		r = radeon_object_create(rdev, NULL, size, true,
-					 RADEON_GEM_DOMAIN_GTT, false, gtt_obj + i);
+		r = radeon_bo_create(rdev, NULL, size, true,
+					 RADEON_GEM_DOMAIN_GTT, gtt_obj + i);
 		if (r) {
 			DRM_ERROR("Failed to create GTT object %d\n", i);
 			goto out_cleanup;
 		}
 
-		r = radeon_object_pin(gtt_obj[i], RADEON_GEM_DOMAIN_GTT, &gtt_addr);
+		r = radeon_bo_reserve(gtt_obj[i], false);
+		if (unlikely(r != 0))
+			goto out_cleanup;
+		r = radeon_bo_pin(gtt_obj[i], RADEON_GEM_DOMAIN_GTT, &gtt_addr);
 		if (r) {
 			DRM_ERROR("Failed to pin GTT object %d\n", i);
 			goto out_cleanup;
 		}
 
-		r = radeon_object_kmap(gtt_obj[i], &gtt_map);
+		r = radeon_bo_kmap(gtt_obj[i], &gtt_map);
 		if (r) {
 			DRM_ERROR("Failed to map GTT object %d\n", i);
 			goto out_cleanup;
@@ -94,7 +98,7 @@ void radeon_test_moves(struct radeon_device *rdev)
 		     gtt_start++)
 			*gtt_start = gtt_start;
 
-		radeon_object_kunmap(gtt_obj[i]);
+		radeon_bo_kunmap(gtt_obj[i]);
 
 		r = radeon_fence_create(rdev, &fence);
 		if (r) {
@@ -102,7 +106,7 @@ void radeon_test_moves(struct radeon_device *rdev)
 			goto out_cleanup;
 		}
 
-		r = radeon_copy(rdev, gtt_addr, vram_addr, size / 4096, fence);
+		r = radeon_copy(rdev, gtt_addr, vram_addr, size / RADEON_GPU_PAGE_SIZE, fence);
 		if (r) {
 			DRM_ERROR("Failed GTT->VRAM copy %d\n", i);
 			goto out_cleanup;
@@ -116,7 +120,7 @@ void radeon_test_moves(struct radeon_device *rdev)
 
 		radeon_fence_unref(&fence);
 
-		r = radeon_object_kmap(vram_obj, &vram_map);
+		r = radeon_bo_kmap(vram_obj, &vram_map);
 		if (r) {
 			DRM_ERROR("Failed to map VRAM object after copy %d\n", i);
 			goto out_cleanup;
@@ -131,13 +135,13 @@ void radeon_test_moves(struct radeon_device *rdev)
 					  "expected 0x%p (GTT map 0x%p-0x%p)\n",
 					  i, *vram_start, gtt_start, gtt_map,
 					  gtt_end);
-				radeon_object_kunmap(vram_obj);
+				radeon_bo_kunmap(vram_obj);
 				goto out_cleanup;
 			}
 			*vram_start = vram_start;
 		}
 
-		radeon_object_kunmap(vram_obj);
+		radeon_bo_kunmap(vram_obj);
 
 		r = radeon_fence_create(rdev, &fence);
 		if (r) {
@@ -145,7 +149,7 @@ void radeon_test_moves(struct radeon_device *rdev)
 			goto out_cleanup;
 		}
 
-		r = radeon_copy(rdev, vram_addr, gtt_addr, size / 4096, fence);
+		r = radeon_copy(rdev, vram_addr, gtt_addr, size / RADEON_GPU_PAGE_SIZE, fence);
 		if (r) {
 			DRM_ERROR("Failed VRAM->GTT copy %d\n", i);
 			goto out_cleanup;
@@ -159,7 +163,7 @@ void radeon_test_moves(struct radeon_device *rdev)
 
 		radeon_fence_unref(&fence);
 
-		r = radeon_object_kmap(gtt_obj[i], &gtt_map);
+		r = radeon_bo_kmap(gtt_obj[i], &gtt_map);
 		if (r) {
 			DRM_ERROR("Failed to map GTT object after copy %d\n", i);
 			goto out_cleanup;
@@ -174,27 +178,33 @@ void radeon_test_moves(struct radeon_device *rdev)
 					  "expected 0x%p (VRAM map 0x%p-0x%p)\n",
 					  i, *gtt_start, vram_start, vram_map,
 					  vram_end);
-				radeon_object_kunmap(gtt_obj[i]);
+				radeon_bo_kunmap(gtt_obj[i]);
 				goto out_cleanup;
 			}
 		}
 
-		radeon_object_kunmap(gtt_obj[i]);
+		radeon_bo_kunmap(gtt_obj[i]);
 
 		DRM_INFO("Tested GTT->VRAM and VRAM->GTT copy for GTT offset 0x%llx\n",
-			 gtt_addr - rdev->mc.gtt_location);
+			 gtt_addr - rdev->mc.gtt_start);
 	}
 
 out_cleanup:
 	if (vram_obj) {
-		radeon_object_unpin(vram_obj);
-		radeon_object_unref(&vram_obj);
+		if (radeon_bo_is_reserved(vram_obj)) {
+			radeon_bo_unpin(vram_obj);
+			radeon_bo_unreserve(vram_obj);
+		}
+		radeon_bo_unref(&vram_obj);
 	}
 	if (gtt_obj) {
 		for (i = 0; i < n; i++) {
 			if (gtt_obj[i]) {
-				radeon_object_unpin(gtt_obj[i]);
-				radeon_object_unref(&gtt_obj[i]);
+				if (radeon_bo_is_reserved(gtt_obj[i])) {
+					radeon_bo_unpin(gtt_obj[i]);
+					radeon_bo_unreserve(gtt_obj[i]);
+				}
+				radeon_bo_unref(&gtt_obj[i]);
 			}
 		}
 		kfree(gtt_obj);
@@ -206,4 +216,3 @@ out_cleanup:
 		printk(KERN_WARNING "Error while testing BO move.\n");
 	}
 }
-
diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon/radeon_ttm.c
index 5b1cf04a011a..43c5ab34b634 100644
--- a/drivers/gpu/drm/radeon/radeon_ttm.c
+++ b/drivers/gpu/drm/radeon/radeon_ttm.c
@@ -150,7 +150,7 @@ static int radeon_init_mem_type(struct ttm_bo_device *bdev, uint32_t type,
 		man->default_caching = TTM_PL_FLAG_CACHED;
 		break;
 	case TTM_PL_TT:
-		man->gpu_offset = 0;
+		man->gpu_offset = rdev->mc.gtt_start;
 		man->available_caching = TTM_PL_MASK_CACHING;
 		man->default_caching = TTM_PL_FLAG_CACHED;
 		man->flags = TTM_MEMTYPE_FLAG_MAPPABLE | TTM_MEMTYPE_FLAG_CMA;
@@ -180,7 +180,7 @@ static int radeon_init_mem_type(struct ttm_bo_device *bdev, uint32_t type,
 		break;
 	case TTM_PL_VRAM:
 		/* "On-card" video ram */
-		man->gpu_offset = 0;
+		man->gpu_offset = rdev->mc.vram_start;
 		man->flags = TTM_MEMTYPE_FLAG_FIXED |
 			     TTM_MEMTYPE_FLAG_NEEDS_IOREMAP |
 			     TTM_MEMTYPE_FLAG_MAPPABLE;
@@ -197,16 +197,34 @@ static int radeon_init_mem_type(struct ttm_bo_device *bdev, uint32_t type,
 	return 0;
 }
 
-static uint32_t radeon_evict_flags(struct ttm_buffer_object *bo)
+static void radeon_evict_flags(struct ttm_buffer_object *bo,
+				struct ttm_placement *placement)
 {
-	uint32_t cur_placement = bo->mem.placement & ~TTM_PL_MASK_MEMTYPE;
+	struct radeon_bo *rbo;
+	static u32 placements = TTM_PL_MASK_CACHING | TTM_PL_FLAG_SYSTEM;
 
+	if (!radeon_ttm_bo_is_radeon_bo(bo)) {
+		placement->fpfn = 0;
+		placement->lpfn = 0;
+		placement->placement = &placements;
+		placement->busy_placement = &placements;
+		placement->num_placement = 1;
+		placement->num_busy_placement = 1;
+		return;
+	}
+	rbo = container_of(bo, struct radeon_bo, tbo);
 	switch (bo->mem.mem_type) {
+	case TTM_PL_VRAM:
+		if (rbo->rdev->cp.ready == false)
+			radeon_ttm_placement_from_domain(rbo, RADEON_GEM_DOMAIN_CPU);
+		else
+			radeon_ttm_placement_from_domain(rbo, RADEON_GEM_DOMAIN_GTT);
+		break;
+	case TTM_PL_TT:
 	default:
-		return (cur_placement & ~TTM_PL_MASK_CACHING) |
-			TTM_PL_FLAG_SYSTEM |
-			TTM_PL_FLAG_CACHED;
+		radeon_ttm_placement_from_domain(rbo, RADEON_GEM_DOMAIN_CPU);
 	}
+	*placement = rbo->placement;
 }
 
 static int radeon_verify_access(struct ttm_buffer_object *bo, struct file *filp)
@@ -244,10 +262,10 @@ static int radeon_move_blit(struct ttm_buffer_object *bo,
 
 	switch (old_mem->mem_type) {
 	case TTM_PL_VRAM:
-		old_start += rdev->mc.vram_location;
+		old_start += rdev->mc.vram_start;
 		break;
 	case TTM_PL_TT:
-		old_start += rdev->mc.gtt_location;
+		old_start += rdev->mc.gtt_start;
 		break;
 	default:
 		DRM_ERROR("Unknown placement %d\n", old_mem->mem_type);
@@ -255,10 +273,10 @@ static int radeon_move_blit(struct ttm_buffer_object *bo,
 	}
 	switch (new_mem->mem_type) {
 	case TTM_PL_VRAM:
-		new_start += rdev->mc.vram_location;
+		new_start += rdev->mc.vram_start;
 		break;
 	case TTM_PL_TT:
-		new_start += rdev->mc.gtt_location;
+		new_start += rdev->mc.gtt_start;
 		break;
 	default:
 		DRM_ERROR("Unknown placement %d\n", old_mem->mem_type);
@@ -283,18 +301,31 @@ static int radeon_move_vram_ram(struct ttm_buffer_object *bo,
 	struct radeon_device *rdev;
 	struct ttm_mem_reg *old_mem = &bo->mem;
 	struct ttm_mem_reg tmp_mem;
-	uint32_t proposed_placement;
+	u32 placements;
+	struct ttm_placement placement;
 	int r;
 
 	rdev = radeon_get_rdev(bo->bdev);
 	tmp_mem = *new_mem;
 	tmp_mem.mm_node = NULL;
-	proposed_placement = TTM_PL_FLAG_TT | TTM_PL_MASK_CACHING;
-	r = ttm_bo_mem_space(bo, proposed_placement, &tmp_mem,
+	placement.fpfn = 0;
+	placement.lpfn = 0;
+	placement.num_placement = 1;
+	placement.placement = &placements;
+	placement.num_busy_placement = 1;
+	placement.busy_placement = &placements;
+	placements = TTM_PL_MASK_CACHING | TTM_PL_FLAG_TT;
+	r = ttm_bo_mem_space(bo, &placement, &tmp_mem,
 			     interruptible, no_wait);
 	if (unlikely(r)) {
 		return r;
 	}
+
+	r = ttm_tt_set_placement_caching(bo->ttm, tmp_mem.placement);
+	if (unlikely(r)) {
+		goto out_cleanup;
+	}
+
 	r = ttm_tt_bind(bo->ttm, &tmp_mem);
 	if (unlikely(r)) {
 		goto out_cleanup;
@@ -323,15 +354,21 @@ static int radeon_move_ram_vram(struct ttm_buffer_object *bo,
 	struct radeon_device *rdev;
 	struct ttm_mem_reg *old_mem = &bo->mem;
 	struct ttm_mem_reg tmp_mem;
-	uint32_t proposed_flags;
+	struct ttm_placement placement;
+	u32 placements;
 	int r;
 
 	rdev = radeon_get_rdev(bo->bdev);
 	tmp_mem = *new_mem;
 	tmp_mem.mm_node = NULL;
-	proposed_flags = TTM_PL_FLAG_TT | TTM_PL_MASK_CACHING;
-	r = ttm_bo_mem_space(bo, proposed_flags, &tmp_mem,
-			     interruptible, no_wait);
+	placement.fpfn = 0;
+	placement.lpfn = 0;
+	placement.num_placement = 1;
+	placement.placement = &placements;
+	placement.num_busy_placement = 1;
+	placement.busy_placement = &placements;
+	placements = TTM_PL_MASK_CACHING | TTM_PL_FLAG_TT;
+	r = ttm_bo_mem_space(bo, &placement, &tmp_mem, interruptible, no_wait);
 	if (unlikely(r)) {
 		return r;
 	}
@@ -372,7 +409,7 @@ static int radeon_bo_move(struct ttm_buffer_object *bo,
 	     new_mem->mem_type == TTM_PL_SYSTEM) ||
 	    (old_mem->mem_type == TTM_PL_SYSTEM &&
 	     new_mem->mem_type == TTM_PL_TT)) {
-		/* bind is enought */
+		/* bind is enough */
 		radeon_move_null(bo, new_mem);
 		return 0;
 	}
@@ -401,18 +438,6 @@ memcpy:
 	return r;
 }
 
-const uint32_t radeon_mem_prios[] = {
-	TTM_PL_VRAM,
-	TTM_PL_TT,
-	TTM_PL_SYSTEM,
-};
-
-const uint32_t radeon_busy_prios[] = {
-	TTM_PL_TT,
-	TTM_PL_VRAM,
-	TTM_PL_SYSTEM,
-};
-
 static int radeon_sync_obj_wait(void *sync_obj, void *sync_arg,
 				bool lazy, bool interruptible)
 {
@@ -440,10 +465,6 @@ static bool radeon_sync_obj_signaled(void *sync_obj, void *sync_arg)
 }
 
 static struct ttm_bo_driver radeon_bo_driver = {
-	.mem_type_prio = radeon_mem_prios,
-	.mem_busy_prio = radeon_busy_prios,
-	.num_mem_type_prio = ARRAY_SIZE(radeon_mem_prios),
-	.num_mem_busy_prio = ARRAY_SIZE(radeon_busy_prios),
 	.create_ttm_backend_entry = &radeon_create_ttm_backend_entry,
 	.invalidate_caches = &radeon_invalidate_caches,
 	.init_mem_type = &radeon_init_mem_type,
@@ -476,27 +497,32 @@ int radeon_ttm_init(struct radeon_device *rdev)
 		DRM_ERROR("failed initializing buffer object driver(%d).\n", r);
 		return r;
 	}
-	r = ttm_bo_init_mm(&rdev->mman.bdev, TTM_PL_VRAM, 0,
-			   ((rdev->mc.real_vram_size) >> PAGE_SHIFT));
+	rdev->mman.initialized = true;
+	r = ttm_bo_init_mm(&rdev->mman.bdev, TTM_PL_VRAM,
+				rdev->mc.real_vram_size >> PAGE_SHIFT);
 	if (r) {
 		DRM_ERROR("Failed initializing VRAM heap.\n");
 		return r;
 	}
-	r = radeon_object_create(rdev, NULL, 256 * 1024, true,
-				 RADEON_GEM_DOMAIN_VRAM, false,
-				 &rdev->stollen_vga_memory);
+	r = radeon_bo_create(rdev, NULL, 256 * 1024, true,
+				RADEON_GEM_DOMAIN_VRAM,
+				&rdev->stollen_vga_memory);
 	if (r) {
 		return r;
 	}
-	r = radeon_object_pin(rdev->stollen_vga_memory, RADEON_GEM_DOMAIN_VRAM, NULL);
+	r = radeon_bo_reserve(rdev->stollen_vga_memory, false);
+	if (r)
+		return r;
+	r = radeon_bo_pin(rdev->stollen_vga_memory, RADEON_GEM_DOMAIN_VRAM, NULL);
+	radeon_bo_unreserve(rdev->stollen_vga_memory);
 	if (r) {
-		radeon_object_unref(&rdev->stollen_vga_memory);
+		radeon_bo_unref(&rdev->stollen_vga_memory);
 		return r;
 	}
 	DRM_INFO("radeon: %uM of VRAM memory ready\n",
 		 (unsigned)rdev->mc.real_vram_size / (1024 * 1024));
-	r = ttm_bo_init_mm(&rdev->mman.bdev, TTM_PL_TT, 0,
-			   ((rdev->mc.gtt_size) >> PAGE_SHIFT));
+	r = ttm_bo_init_mm(&rdev->mman.bdev, TTM_PL_TT,
+				rdev->mc.gtt_size >> PAGE_SHIFT);
 	if (r) {
 		DRM_ERROR("Failed initializing GTT heap.\n");
 		return r;
@@ -517,15 +543,24 @@ int radeon_ttm_init(struct radeon_device *rdev)
 
 void radeon_ttm_fini(struct radeon_device *rdev)
 {
+	int r;
+
+	if (!rdev->mman.initialized)
+		return;
 	if (rdev->stollen_vga_memory) {
-		radeon_object_unpin(rdev->stollen_vga_memory);
-		radeon_object_unref(&rdev->stollen_vga_memory);
+		r = radeon_bo_reserve(rdev->stollen_vga_memory, false);
+		if (r == 0) {
+			radeon_bo_unpin(rdev->stollen_vga_memory);
+			radeon_bo_unreserve(rdev->stollen_vga_memory);
+		}
+		radeon_bo_unref(&rdev->stollen_vga_memory);
 	}
 	ttm_bo_clean_mm(&rdev->mman.bdev, TTM_PL_VRAM);
 	ttm_bo_clean_mm(&rdev->mman.bdev, TTM_PL_TT);
 	ttm_bo_device_release(&rdev->mman.bdev);
 	radeon_gart_fini(rdev);
 	radeon_ttm_global_fini(rdev);
+	rdev->mman.initialized = false;
 	DRM_INFO("radeon: ttm finalized\n");
 }
 
@@ -689,9 +724,6 @@ struct ttm_backend *radeon_ttm_backend_create(struct radeon_device *rdev)
 
 #define RADEON_DEBUGFS_MEM_TYPES 2
 
-static struct drm_info_list radeon_mem_types_list[RADEON_DEBUGFS_MEM_TYPES];
-static char radeon_mem_types_names[RADEON_DEBUGFS_MEM_TYPES][32];
-
 #if defined(CONFIG_DEBUG_FS)
 static int radeon_mm_dump_table(struct seq_file *m, void *data)
 {
@@ -711,9 +743,11 @@ static int radeon_mm_dump_table(struct seq_file *m, void *data)
 
 static int radeon_ttm_debugfs_init(struct radeon_device *rdev)
 {
+#if defined(CONFIG_DEBUG_FS)
+	static struct drm_info_list radeon_mem_types_list[RADEON_DEBUGFS_MEM_TYPES];
+	static char radeon_mem_types_names[RADEON_DEBUGFS_MEM_TYPES][32];
 	unsigned i;
 
-#if defined(CONFIG_DEBUG_FS)
 	for (i = 0; i < RADEON_DEBUGFS_MEM_TYPES; i++) {
 		if (i == 0)
 			sprintf(radeon_mem_types_names[i], "radeon_vram_mm");
diff --git a/drivers/gpu/drm/radeon/reg_srcs/r200 b/drivers/gpu/drm/radeon/reg_srcs/r200
index 6021c8849a16..c29ac434ac9c 100644
--- a/drivers/gpu/drm/radeon/reg_srcs/r200
+++ b/drivers/gpu/drm/radeon/reg_srcs/r200
@@ -91,6 +91,8 @@ r200 0x3294
 0x22b8 SE_TCL_TEX_CYL_WRAP_CTL
 0x22c0 SE_TCL_UCP_VERT_BLEND_CNTL
 0x22c4 SE_TCL_POINT_SPRITE_CNTL
+0x22d0 SE_PVS_CNTL
+0x22d4 SE_PVS_CONST_CNTL
 0x2648 RE_POINTSIZE
 0x26c0 RE_TOP_LEFT
 0x26c4 RE_MISC
diff --git a/drivers/gpu/drm/radeon/reg_srcs/r420 b/drivers/gpu/drm/radeon/reg_srcs/r420
new file mode 100644
index 000000000000..989f7a020832
--- /dev/null
+++ b/drivers/gpu/drm/radeon/reg_srcs/r420
@@ -0,0 +1,795 @@
+r420 0x4f60
+0x1434 SRC_Y_X
+0x1438 DST_Y_X
+0x143C DST_HEIGHT_WIDTH
+0x146C DP_GUI_MASTER_CNTL
+0x1474 BRUSH_Y_X
+0x1478 DP_BRUSH_BKGD_CLR
+0x147C DP_BRUSH_FRGD_CLR
+0x1480 BRUSH_DATA0
+0x1484 BRUSH_DATA1
+0x1598 DST_WIDTH_HEIGHT
+0x15C0 CLR_CMP_CNTL
+0x15C4 CLR_CMP_CLR_SRC
+0x15C8 CLR_CMP_CLR_DST
+0x15CC CLR_CMP_MSK
+0x15D8 DP_SRC_FRGD_CLR
+0x15DC DP_SRC_BKGD_CLR
+0x1600 DST_LINE_START
+0x1604 DST_LINE_END
+0x1608 DST_LINE_PATCOUNT
+0x16C0 DP_CNTL
+0x16CC DP_WRITE_MSK
+0x16D0 DP_CNTL_XDIR_YDIR_YMAJOR
+0x16E8 DEFAULT_SC_BOTTOM_RIGHT
+0x16EC SC_TOP_LEFT
+0x16F0 SC_BOTTOM_RIGHT
+0x16F4 SRC_SC_BOTTOM_RIGHT
+0x1714 DSTCACHE_CTLSTAT
+0x1720 WAIT_UNTIL
+0x172C RBBM_GUICNTL
+0x1D98 VAP_VPORT_XSCALE
+0x1D9C VAP_VPORT_XOFFSET
+0x1DA0 VAP_VPORT_YSCALE
+0x1DA4 VAP_VPORT_YOFFSET
+0x1DA8 VAP_VPORT_ZSCALE
+0x1DAC VAP_VPORT_ZOFFSET
+0x2080 VAP_CNTL
+0x2090 VAP_OUT_VTX_FMT_0
+0x2094 VAP_OUT_VTX_FMT_1
+0x20B0 VAP_VTE_CNTL
+0x2138 VAP_VF_MIN_VTX_INDX
+0x2140 VAP_CNTL_STATUS
+0x2150 VAP_PROG_STREAM_CNTL_0
+0x2154 VAP_PROG_STREAM_CNTL_1
+0x2158 VAP_PROG_STREAM_CNTL_2
+0x215C VAP_PROG_STREAM_CNTL_3
+0x2160 VAP_PROG_STREAM_CNTL_4
+0x2164 VAP_PROG_STREAM_CNTL_5
+0x2168 VAP_PROG_STREAM_CNTL_6
+0x216C VAP_PROG_STREAM_CNTL_7
+0x2180 VAP_VTX_STATE_CNTL
+0x2184 VAP_VSM_VTX_ASSM
+0x2188 VAP_VTX_STATE_IND_REG_0
+0x218C VAP_VTX_STATE_IND_REG_1
+0x2190 VAP_VTX_STATE_IND_REG_2
+0x2194 VAP_VTX_STATE_IND_REG_3
+0x2198 VAP_VTX_STATE_IND_REG_4
+0x219C VAP_VTX_STATE_IND_REG_5
+0x21A0 VAP_VTX_STATE_IND_REG_6
+0x21A4 VAP_VTX_STATE_IND_REG_7
+0x21A8 VAP_VTX_STATE_IND_REG_8
+0x21AC VAP_VTX_STATE_IND_REG_9
+0x21B0 VAP_VTX_STATE_IND_REG_10
+0x21B4 VAP_VTX_STATE_IND_REG_11
+0x21B8 VAP_VTX_STATE_IND_REG_12
+0x21BC VAP_VTX_STATE_IND_REG_13
+0x21C0 VAP_VTX_STATE_IND_REG_14
+0x21C4 VAP_VTX_STATE_IND_REG_15
+0x21DC VAP_PSC_SGN_NORM_CNTL
+0x21E0 VAP_PROG_STREAM_CNTL_EXT_0
+0x21E4 VAP_PROG_STREAM_CNTL_EXT_1
+0x21E8 VAP_PROG_STREAM_CNTL_EXT_2
+0x21EC VAP_PROG_STREAM_CNTL_EXT_3
+0x21F0 VAP_PROG_STREAM_CNTL_EXT_4
+0x21F4 VAP_PROG_STREAM_CNTL_EXT_5
+0x21F8 VAP_PROG_STREAM_CNTL_EXT_6
+0x21FC VAP_PROG_STREAM_CNTL_EXT_7
+0x2200 VAP_PVS_VECTOR_INDX_REG
+0x2204 VAP_PVS_VECTOR_DATA_REG
+0x2208 VAP_PVS_VECTOR_DATA_REG_128
+0x221C VAP_CLIP_CNTL
+0x2220 VAP_GB_VERT_CLIP_ADJ
+0x2224 VAP_GB_VERT_DISC_ADJ
+0x2228 VAP_GB_HORZ_CLIP_ADJ
+0x222C VAP_GB_HORZ_DISC_ADJ
+0x2230 VAP_PVS_FLOW_CNTL_ADDRS_0
+0x2234 VAP_PVS_FLOW_CNTL_ADDRS_1
+0x2238 VAP_PVS_FLOW_CNTL_ADDRS_2
+0x223C VAP_PVS_FLOW_CNTL_ADDRS_3
+0x2240 VAP_PVS_FLOW_CNTL_ADDRS_4
+0x2244 VAP_PVS_FLOW_CNTL_ADDRS_5
+0x2248 VAP_PVS_FLOW_CNTL_ADDRS_6
+0x224C VAP_PVS_FLOW_CNTL_ADDRS_7
+0x2250 VAP_PVS_FLOW_CNTL_ADDRS_8
+0x2254 VAP_PVS_FLOW_CNTL_ADDRS_9
+0x2258 VAP_PVS_FLOW_CNTL_ADDRS_10
+0x225C VAP_PVS_FLOW_CNTL_ADDRS_11
+0x2260 VAP_PVS_FLOW_CNTL_ADDRS_12
+0x2264 VAP_PVS_FLOW_CNTL_ADDRS_13
+0x2268 VAP_PVS_FLOW_CNTL_ADDRS_14
+0x226C VAP_PVS_FLOW_CNTL_ADDRS_15
+0x2284 VAP_PVS_STATE_FLUSH_REG
+0x2288 VAP_PVS_VTX_TIMEOUT_REG
+0x2290 VAP_PVS_FLOW_CNTL_LOOP_INDEX_0
+0x2294 VAP_PVS_FLOW_CNTL_LOOP_INDEX_1
+0x2298 VAP_PVS_FLOW_CNTL_LOOP_INDEX_2
+0x229C VAP_PVS_FLOW_CNTL_LOOP_INDEX_3
+0x22A0 VAP_PVS_FLOW_CNTL_LOOP_INDEX_4
+0x22A4 VAP_PVS_FLOW_CNTL_LOOP_INDEX_5
+0x22A8 VAP_PVS_FLOW_CNTL_LOOP_INDEX_6
+0x22AC VAP_PVS_FLOW_CNTL_LOOP_INDEX_7
+0x22B0 VAP_PVS_FLOW_CNTL_LOOP_INDEX_8
+0x22B4 VAP_PVS_FLOW_CNTL_LOOP_INDEX_9
+0x22B8 VAP_PVS_FLOW_CNTL_LOOP_INDEX_10
+0x22BC VAP_PVS_FLOW_CNTL_LOOP_INDEX_11
+0x22C0 VAP_PVS_FLOW_CNTL_LOOP_INDEX_12
+0x22C4 VAP_PVS_FLOW_CNTL_LOOP_INDEX_13
+0x22C8 VAP_PVS_FLOW_CNTL_LOOP_INDEX_14
+0x22CC VAP_PVS_FLOW_CNTL_LOOP_INDEX_15
+0x22D0 VAP_PVS_CODE_CNTL_0
+0x22D4 VAP_PVS_CONST_CNTL
+0x22D8 VAP_PVS_CODE_CNTL_1
+0x22DC VAP_PVS_FLOW_CNTL_OPC
+0x342C RB2D_DSTCACHE_CTLSTAT
+0x4000 GB_VAP_RASTER_VTX_FMT_0
+0x4004 GB_VAP_RASTER_VTX_FMT_1
+0x4008 GB_ENABLE
+0x401C GB_SELECT
+0x4020 GB_AA_CONFIG
+0x4024 GB_FIFO_SIZE
+0x4100 TX_INVALTAGS
+0x4200 GA_POINT_S0
+0x4204 GA_POINT_T0
+0x4208 GA_POINT_S1
+0x420C GA_POINT_T1
+0x4214 GA_TRIANGLE_STIPPLE
+0x421C GA_POINT_SIZE
+0x4230 GA_POINT_MINMAX
+0x4234 GA_LINE_CNTL
+0x4238 GA_LINE_STIPPLE_CONFIG
+0x4260 GA_LINE_STIPPLE_VALUE
+0x4264 GA_LINE_S0
+0x4268 GA_LINE_S1
+0x4278 GA_COLOR_CONTROL
+0x427C GA_SOLID_RG
+0x4280 GA_SOLID_BA
+0x4288 GA_POLY_MODE
+0x428C GA_ROUND_MODE
+0x4290 GA_OFFSET
+0x4294 GA_FOG_SCALE
+0x4298 GA_FOG_OFFSET
+0x42A0 SU_TEX_WRAP
+0x42A4 SU_POLY_OFFSET_FRONT_SCALE
+0x42A8 SU_POLY_OFFSET_FRONT_OFFSET
+0x42AC SU_POLY_OFFSET_BACK_SCALE
+0x42B0 SU_POLY_OFFSET_BACK_OFFSET
+0x42B4 SU_POLY_OFFSET_ENABLE
+0x42B8 SU_CULL_MODE
+0x42C0 SU_DEPTH_SCALE
+0x42C4 SU_DEPTH_OFFSET
+0x42C8 SU_REG_DEST
+0x4300 RS_COUNT
+0x4304 RS_INST_COUNT
+0x4310 RS_IP_0
+0x4314 RS_IP_1
+0x4318 RS_IP_2
+0x431C RS_IP_3
+0x4320 RS_IP_4
+0x4324 RS_IP_5
+0x4328 RS_IP_6
+0x432C RS_IP_7
+0x4330 RS_INST_0
+0x4334 RS_INST_1
+0x4338 RS_INST_2
+0x433C RS_INST_3
+0x4340 RS_INST_4
+0x4344 RS_INST_5
+0x4348 RS_INST_6
+0x434C RS_INST_7
+0x4350 RS_INST_8
+0x4354 RS_INST_9
+0x4358 RS_INST_10
+0x435C RS_INST_11
+0x4360 RS_INST_12
+0x4364 RS_INST_13
+0x4368 RS_INST_14
+0x436C RS_INST_15
+0x43A4 SC_HYPERZ_EN
+0x43A8 SC_EDGERULE
+0x43B0 SC_CLIP_0_A
+0x43B4 SC_CLIP_0_B
+0x43B8 SC_CLIP_1_A
+0x43BC SC_CLIP_1_B
+0x43C0 SC_CLIP_2_A
+0x43C4 SC_CLIP_2_B
+0x43C8 SC_CLIP_3_A
+0x43CC SC_CLIP_3_B
+0x43D0 SC_CLIP_RULE
+0x43E0 SC_SCISSOR0
+0x43E8 SC_SCREENDOOR
+0x4440 TX_FILTER1_0
+0x4444 TX_FILTER1_1
+0x4448 TX_FILTER1_2
+0x444C TX_FILTER1_3
+0x4450 TX_FILTER1_4
+0x4454 TX_FILTER1_5
+0x4458 TX_FILTER1_6
+0x445C TX_FILTER1_7
+0x4460 TX_FILTER1_8
+0x4464 TX_FILTER1_9
+0x4468 TX_FILTER1_10
+0x446C TX_FILTER1_11
+0x4470 TX_FILTER1_12
+0x4474 TX_FILTER1_13
+0x4478 TX_FILTER1_14
+0x447C TX_FILTER1_15
+0x4580 TX_CHROMA_KEY_0
+0x4584 TX_CHROMA_KEY_1
+0x4588 TX_CHROMA_KEY_2
+0x458C TX_CHROMA_KEY_3
+0x4590 TX_CHROMA_KEY_4
+0x4594 TX_CHROMA_KEY_5
+0x4598 TX_CHROMA_KEY_6
+0x459C TX_CHROMA_KEY_7
+0x45A0 TX_CHROMA_KEY_8
+0x45A4 TX_CHROMA_KEY_9
+0x45A8 TX_CHROMA_KEY_10
+0x45AC TX_CHROMA_KEY_11
+0x45B0 TX_CHROMA_KEY_12
+0x45B4 TX_CHROMA_KEY_13
+0x45B8 TX_CHROMA_KEY_14
+0x45BC TX_CHROMA_KEY_15
+0x45C0 TX_BORDER_COLOR_0
+0x45C4 TX_BORDER_COLOR_1
+0x45C8 TX_BORDER_COLOR_2
+0x45CC TX_BORDER_COLOR_3
+0x45D0 TX_BORDER_COLOR_4
+0x45D4 TX_BORDER_COLOR_5
+0x45D8 TX_BORDER_COLOR_6
+0x45DC TX_BORDER_COLOR_7
+0x45E0 TX_BORDER_COLOR_8
+0x45E4 TX_BORDER_COLOR_9
+0x45E8 TX_BORDER_COLOR_10
+0x45EC TX_BORDER_COLOR_11
+0x45F0 TX_BORDER_COLOR_12
+0x45F4 TX_BORDER_COLOR_13
+0x45F8 TX_BORDER_COLOR_14
+0x45FC TX_BORDER_COLOR_15
+0x4600 US_CONFIG
+0x4604 US_PIXSIZE
+0x4608 US_CODE_OFFSET
+0x460C US_RESET
+0x4610 US_CODE_ADDR_0
+0x4614 US_CODE_ADDR_1
+0x4618 US_CODE_ADDR_2
+0x461C US_CODE_ADDR_3
+0x4620 US_TEX_INST_0
+0x4624 US_TEX_INST_1
+0x4628 US_TEX_INST_2
+0x462C US_TEX_INST_3
+0x4630 US_TEX_INST_4
+0x4634 US_TEX_INST_5
+0x4638 US_TEX_INST_6
+0x463C US_TEX_INST_7
+0x4640 US_TEX_INST_8
+0x4644 US_TEX_INST_9
+0x4648 US_TEX_INST_10
+0x464C US_TEX_INST_11
+0x4650 US_TEX_INST_12
+0x4654 US_TEX_INST_13
+0x4658 US_TEX_INST_14
+0x465C US_TEX_INST_15
+0x4660 US_TEX_INST_16
+0x4664 US_TEX_INST_17
+0x4668 US_TEX_INST_18
+0x466C US_TEX_INST_19
+0x4670 US_TEX_INST_20
+0x4674 US_TEX_INST_21
+0x4678 US_TEX_INST_22
+0x467C US_TEX_INST_23
+0x4680 US_TEX_INST_24
+0x4684 US_TEX_INST_25
+0x4688 US_TEX_INST_26
+0x468C US_TEX_INST_27
+0x4690 US_TEX_INST_28
+0x4694 US_TEX_INST_29
+0x4698 US_TEX_INST_30
+0x469C US_TEX_INST_31
+0x46A4 US_OUT_FMT_0
+0x46A8 US_OUT_FMT_1
+0x46AC US_OUT_FMT_2
+0x46B0 US_OUT_FMT_3
+0x46B4 US_W_FMT
+0x46B8 US_CODE_BANK
+0x46BC US_CODE_EXT
+0x46C0 US_ALU_RGB_ADDR_0
+0x46C4 US_ALU_RGB_ADDR_1
+0x46C8 US_ALU_RGB_ADDR_2
+0x46CC US_ALU_RGB_ADDR_3
+0x46D0 US_ALU_RGB_ADDR_4
+0x46D4 US_ALU_RGB_ADDR_5
+0x46D8 US_ALU_RGB_ADDR_6
+0x46DC US_ALU_RGB_ADDR_7
+0x46E0 US_ALU_RGB_ADDR_8
+0x46E4 US_ALU_RGB_ADDR_9
+0x46E8 US_ALU_RGB_ADDR_10
+0x46EC US_ALU_RGB_ADDR_11
+0x46F0 US_ALU_RGB_ADDR_12
+0x46F4 US_ALU_RGB_ADDR_13
+0x46F8 US_ALU_RGB_ADDR_14
+0x46FC US_ALU_RGB_ADDR_15
+0x4700 US_ALU_RGB_ADDR_16
+0x4704 US_ALU_RGB_ADDR_17
+0x4708 US_ALU_RGB_ADDR_18
+0x470C US_ALU_RGB_ADDR_19
+0x4710 US_ALU_RGB_ADDR_20
+0x4714 US_ALU_RGB_ADDR_21
+0x4718 US_ALU_RGB_ADDR_22
+0x471C US_ALU_RGB_ADDR_23
+0x4720 US_ALU_RGB_ADDR_24
+0x4724 US_ALU_RGB_ADDR_25
+0x4728 US_ALU_RGB_ADDR_26
+0x472C US_ALU_RGB_ADDR_27
+0x4730 US_ALU_RGB_ADDR_28
+0x4734 US_ALU_RGB_ADDR_29
+0x4738 US_ALU_RGB_ADDR_30
+0x473C US_ALU_RGB_ADDR_31
+0x4740 US_ALU_RGB_ADDR_32
+0x4744 US_ALU_RGB_ADDR_33
+0x4748 US_ALU_RGB_ADDR_34
+0x474C US_ALU_RGB_ADDR_35
+0x4750 US_ALU_RGB_ADDR_36
+0x4754 US_ALU_RGB_ADDR_37
+0x4758 US_ALU_RGB_ADDR_38
+0x475C US_ALU_RGB_ADDR_39
+0x4760 US_ALU_RGB_ADDR_40
+0x4764 US_ALU_RGB_ADDR_41
+0x4768 US_ALU_RGB_ADDR_42
+0x476C US_ALU_RGB_ADDR_43
+0x4770 US_ALU_RGB_ADDR_44
+0x4774 US_ALU_RGB_ADDR_45
+0x4778 US_ALU_RGB_ADDR_46
+0x477C US_ALU_RGB_ADDR_47
+0x4780 US_ALU_RGB_ADDR_48
+0x4784 US_ALU_RGB_ADDR_49
+0x4788 US_ALU_RGB_ADDR_50
+0x478C US_ALU_RGB_ADDR_51
+0x4790 US_ALU_RGB_ADDR_52
+0x4794 US_ALU_RGB_ADDR_53
+0x4798 US_ALU_RGB_ADDR_54
+0x479C US_ALU_RGB_ADDR_55
+0x47A0 US_ALU_RGB_ADDR_56
+0x47A4 US_ALU_RGB_ADDR_57
+0x47A8 US_ALU_RGB_ADDR_58
+0x47AC US_ALU_RGB_ADDR_59
+0x47B0 US_ALU_RGB_ADDR_60
+0x47B4 US_ALU_RGB_ADDR_61
+0x47B8 US_ALU_RGB_ADDR_62
+0x47BC US_ALU_RGB_ADDR_63
+0x47C0 US_ALU_ALPHA_ADDR_0
+0x47C4 US_ALU_ALPHA_ADDR_1
+0x47C8 US_ALU_ALPHA_ADDR_2
+0x47CC US_ALU_ALPHA_ADDR_3
+0x47D0 US_ALU_ALPHA_ADDR_4
+0x47D4 US_ALU_ALPHA_ADDR_5
+0x47D8 US_ALU_ALPHA_ADDR_6
+0x47DC US_ALU_ALPHA_ADDR_7
+0x47E0 US_ALU_ALPHA_ADDR_8
+0x47E4 US_ALU_ALPHA_ADDR_9
+0x47E8 US_ALU_ALPHA_ADDR_10
+0x47EC US_ALU_ALPHA_ADDR_11
+0x47F0 US_ALU_ALPHA_ADDR_12
+0x47F4 US_ALU_ALPHA_ADDR_13
+0x47F8 US_ALU_ALPHA_ADDR_14
+0x47FC US_ALU_ALPHA_ADDR_15
+0x4800 US_ALU_ALPHA_ADDR_16
+0x4804 US_ALU_ALPHA_ADDR_17
+0x4808 US_ALU_ALPHA_ADDR_18
+0x480C US_ALU_ALPHA_ADDR_19
+0x4810 US_ALU_ALPHA_ADDR_20
+0x4814 US_ALU_ALPHA_ADDR_21
+0x4818 US_ALU_ALPHA_ADDR_22
+0x481C US_ALU_ALPHA_ADDR_23
+0x4820 US_ALU_ALPHA_ADDR_24
+0x4824 US_ALU_ALPHA_ADDR_25
+0x4828 US_ALU_ALPHA_ADDR_26
+0x482C US_ALU_ALPHA_ADDR_27
+0x4830 US_ALU_ALPHA_ADDR_28
+0x4834 US_ALU_ALPHA_ADDR_29
+0x4838 US_ALU_ALPHA_ADDR_30
+0x483C US_ALU_ALPHA_ADDR_31
+0x4840 US_ALU_ALPHA_ADDR_32
+0x4844 US_ALU_ALPHA_ADDR_33
+0x4848 US_ALU_ALPHA_ADDR_34
+0x484C US_ALU_ALPHA_ADDR_35
+0x4850 US_ALU_ALPHA_ADDR_36
+0x4854 US_ALU_ALPHA_ADDR_37
+0x4858 US_ALU_ALPHA_ADDR_38
+0x485C US_ALU_ALPHA_ADDR_39
+0x4860 US_ALU_ALPHA_ADDR_40
+0x4864 US_ALU_ALPHA_ADDR_41
+0x4868 US_ALU_ALPHA_ADDR_42
+0x486C US_ALU_ALPHA_ADDR_43
+0x4870 US_ALU_ALPHA_ADDR_44
+0x4874 US_ALU_ALPHA_ADDR_45
+0x4878 US_ALU_ALPHA_ADDR_46
+0x487C US_ALU_ALPHA_ADDR_47
+0x4880 US_ALU_ALPHA_ADDR_48
+0x4884 US_ALU_ALPHA_ADDR_49
+0x4888 US_ALU_ALPHA_ADDR_50
+0x488C US_ALU_ALPHA_ADDR_51
+0x4890 US_ALU_ALPHA_ADDR_52
+0x4894 US_ALU_ALPHA_ADDR_53
+0x4898 US_ALU_ALPHA_ADDR_54
+0x489C US_ALU_ALPHA_ADDR_55
+0x48A0 US_ALU_ALPHA_ADDR_56
+0x48A4 US_ALU_ALPHA_ADDR_57
+0x48A8 US_ALU_ALPHA_ADDR_58
+0x48AC US_ALU_ALPHA_ADDR_59
+0x48B0 US_ALU_ALPHA_ADDR_60
+0x48B4 US_ALU_ALPHA_ADDR_61
+0x48B8 US_ALU_ALPHA_ADDR_62
+0x48BC US_ALU_ALPHA_ADDR_63
+0x48C0 US_ALU_RGB_INST_0
+0x48C4 US_ALU_RGB_INST_1
+0x48C8 US_ALU_RGB_INST_2
+0x48CC US_ALU_RGB_INST_3
+0x48D0 US_ALU_RGB_INST_4
+0x48D4 US_ALU_RGB_INST_5
+0x48D8 US_ALU_RGB_INST_6
+0x48DC US_ALU_RGB_INST_7
+0x48E0 US_ALU_RGB_INST_8
+0x48E4 US_ALU_RGB_INST_9
+0x48E8 US_ALU_RGB_INST_10
+0x48EC US_ALU_RGB_INST_11
+0x48F0 US_ALU_RGB_INST_12
+0x48F4 US_ALU_RGB_INST_13
+0x48F8 US_ALU_RGB_INST_14
+0x48FC US_ALU_RGB_INST_15
+0x4900 US_ALU_RGB_INST_16
+0x4904 US_ALU_RGB_INST_17
+0x4908 US_ALU_RGB_INST_18
+0x490C US_ALU_RGB_INST_19
+0x4910 US_ALU_RGB_INST_20
+0x4914 US_ALU_RGB_INST_21
+0x4918 US_ALU_RGB_INST_22
+0x491C US_ALU_RGB_INST_23
+0x4920 US_ALU_RGB_INST_24
+0x4924 US_ALU_RGB_INST_25
+0x4928 US_ALU_RGB_INST_26
+0x492C US_ALU_RGB_INST_27
+0x4930 US_ALU_RGB_INST_28
+0x4934 US_ALU_RGB_INST_29
+0x4938 US_ALU_RGB_INST_30
+0x493C US_ALU_RGB_INST_31
+0x4940 US_ALU_RGB_INST_32
+0x4944 US_ALU_RGB_INST_33
+0x4948 US_ALU_RGB_INST_34
+0x494C US_ALU_RGB_INST_35
+0x4950 US_ALU_RGB_INST_36
+0x4954 US_ALU_RGB_INST_37
+0x4958 US_ALU_RGB_INST_38
+0x495C US_ALU_RGB_INST_39
+0x4960 US_ALU_RGB_INST_40
+0x4964 US_ALU_RGB_INST_41
+0x4968 US_ALU_RGB_INST_42
+0x496C US_ALU_RGB_INST_43
+0x4970 US_ALU_RGB_INST_44
+0x4974 US_ALU_RGB_INST_45
+0x4978 US_ALU_RGB_INST_46
+0x497C US_ALU_RGB_INST_47
+0x4980 US_ALU_RGB_INST_48
+0x4984 US_ALU_RGB_INST_49
+0x4988 US_ALU_RGB_INST_50
+0x498C US_ALU_RGB_INST_51
+0x4990 US_ALU_RGB_INST_52
+0x4994 US_ALU_RGB_INST_53
+0x4998 US_ALU_RGB_INST_54
+0x499C US_ALU_RGB_INST_55
+0x49A0 US_ALU_RGB_INST_56
+0x49A4 US_ALU_RGB_INST_57
+0x49A8 US_ALU_RGB_INST_58
+0x49AC US_ALU_RGB_INST_59
+0x49B0 US_ALU_RGB_INST_60
+0x49B4 US_ALU_RGB_INST_61
+0x49B8 US_ALU_RGB_INST_62
+0x49BC US_ALU_RGB_INST_63
+0x49C0 US_ALU_ALPHA_INST_0
+0x49C4 US_ALU_ALPHA_INST_1
+0x49C8 US_ALU_ALPHA_INST_2
+0x49CC US_ALU_ALPHA_INST_3
+0x49D0 US_ALU_ALPHA_INST_4
+0x49D4 US_ALU_ALPHA_INST_5
+0x49D8 US_ALU_ALPHA_INST_6
+0x49DC US_ALU_ALPHA_INST_7
+0x49E0 US_ALU_ALPHA_INST_8
+0x49E4 US_ALU_ALPHA_INST_9
+0x49E8 US_ALU_ALPHA_INST_10
+0x49EC US_ALU_ALPHA_INST_11
+0x49F0 US_ALU_ALPHA_INST_12
+0x49F4 US_ALU_ALPHA_INST_13
+0x49F8 US_ALU_ALPHA_INST_14
+0x49FC US_ALU_ALPHA_INST_15
+0x4A00 US_ALU_ALPHA_INST_16
+0x4A04 US_ALU_ALPHA_INST_17
+0x4A08 US_ALU_ALPHA_INST_18
+0x4A0C US_ALU_ALPHA_INST_19
+0x4A10 US_ALU_ALPHA_INST_20
+0x4A14 US_ALU_ALPHA_INST_21
+0x4A18 US_ALU_ALPHA_INST_22
+0x4A1C US_ALU_ALPHA_INST_23
+0x4A20 US_ALU_ALPHA_INST_24
+0x4A24 US_ALU_ALPHA_INST_25
+0x4A28 US_ALU_ALPHA_INST_26
+0x4A2C US_ALU_ALPHA_INST_27
+0x4A30 US_ALU_ALPHA_INST_28
+0x4A34 US_ALU_ALPHA_INST_29
+0x4A38 US_ALU_ALPHA_INST_30
+0x4A3C US_ALU_ALPHA_INST_31
+0x4A40 US_ALU_ALPHA_INST_32
+0x4A44 US_ALU_ALPHA_INST_33
+0x4A48 US_ALU_ALPHA_INST_34
+0x4A4C US_ALU_ALPHA_INST_35
+0x4A50 US_ALU_ALPHA_INST_36
+0x4A54 US_ALU_ALPHA_INST_37
+0x4A58 US_ALU_ALPHA_INST_38
+0x4A5C US_ALU_ALPHA_INST_39
+0x4A60 US_ALU_ALPHA_INST_40
+0x4A64 US_ALU_ALPHA_INST_41
+0x4A68 US_ALU_ALPHA_INST_42
+0x4A6C US_ALU_ALPHA_INST_43
+0x4A70 US_ALU_ALPHA_INST_44
+0x4A74 US_ALU_ALPHA_INST_45
+0x4A78 US_ALU_ALPHA_INST_46
+0x4A7C US_ALU_ALPHA_INST_47
+0x4A80 US_ALU_ALPHA_INST_48
+0x4A84 US_ALU_ALPHA_INST_49
+0x4A88 US_ALU_ALPHA_INST_50
+0x4A8C US_ALU_ALPHA_INST_51
+0x4A90 US_ALU_ALPHA_INST_52
+0x4A94 US_ALU_ALPHA_INST_53
+0x4A98 US_ALU_ALPHA_INST_54
+0x4A9C US_ALU_ALPHA_INST_55
+0x4AA0 US_ALU_ALPHA_INST_56
+0x4AA4 US_ALU_ALPHA_INST_57
+0x4AA8 US_ALU_ALPHA_INST_58
+0x4AAC US_ALU_ALPHA_INST_59
+0x4AB0 US_ALU_ALPHA_INST_60
+0x4AB4 US_ALU_ALPHA_INST_61
+0x4AB8 US_ALU_ALPHA_INST_62
+0x4ABC US_ALU_ALPHA_INST_63
+0x4AC0 US_ALU_EXT_ADDR_0
+0x4AC4 US_ALU_EXT_ADDR_1
+0x4AC8 US_ALU_EXT_ADDR_2
+0x4ACC US_ALU_EXT_ADDR_3
+0x4AD0 US_ALU_EXT_ADDR_4
+0x4AD4 US_ALU_EXT_ADDR_5
+0x4AD8 US_ALU_EXT_ADDR_6
+0x4ADC US_ALU_EXT_ADDR_7
+0x4AE0 US_ALU_EXT_ADDR_8
+0x4AE4 US_ALU_EXT_ADDR_9
+0x4AE8 US_ALU_EXT_ADDR_10
+0x4AEC US_ALU_EXT_ADDR_11
+0x4AF0 US_ALU_EXT_ADDR_12
+0x4AF4 US_ALU_EXT_ADDR_13
+0x4AF8 US_ALU_EXT_ADDR_14
+0x4AFC US_ALU_EXT_ADDR_15
+0x4B00 US_ALU_EXT_ADDR_16
+0x4B04 US_ALU_EXT_ADDR_17
+0x4B08 US_ALU_EXT_ADDR_18
+0x4B0C US_ALU_EXT_ADDR_19
+0x4B10 US_ALU_EXT_ADDR_20
+0x4B14 US_ALU_EXT_ADDR_21
+0x4B18 US_ALU_EXT_ADDR_22
+0x4B1C US_ALU_EXT_ADDR_23
+0x4B20 US_ALU_EXT_ADDR_24
+0x4B24 US_ALU_EXT_ADDR_25
+0x4B28 US_ALU_EXT_ADDR_26
+0x4B2C US_ALU_EXT_ADDR_27
+0x4B30 US_ALU_EXT_ADDR_28
+0x4B34 US_ALU_EXT_ADDR_29
+0x4B38 US_ALU_EXT_ADDR_30
+0x4B3C US_ALU_EXT_ADDR_31
+0x4B40 US_ALU_EXT_ADDR_32
+0x4B44 US_ALU_EXT_ADDR_33
+0x4B48 US_ALU_EXT_ADDR_34
+0x4B4C US_ALU_EXT_ADDR_35
+0x4B50 US_ALU_EXT_ADDR_36
+0x4B54 US_ALU_EXT_ADDR_37
+0x4B58 US_ALU_EXT_ADDR_38
+0x4B5C US_ALU_EXT_ADDR_39
+0x4B60 US_ALU_EXT_ADDR_40
+0x4B64 US_ALU_EXT_ADDR_41
+0x4B68 US_ALU_EXT_ADDR_42
+0x4B6C US_ALU_EXT_ADDR_43
+0x4B70 US_ALU_EXT_ADDR_44
+0x4B74 US_ALU_EXT_ADDR_45
+0x4B78 US_ALU_EXT_ADDR_46
+0x4B7C US_ALU_EXT_ADDR_47
+0x4B80 US_ALU_EXT_ADDR_48
+0x4B84 US_ALU_EXT_ADDR_49
+0x4B88 US_ALU_EXT_ADDR_50
+0x4B8C US_ALU_EXT_ADDR_51
+0x4B90 US_ALU_EXT_ADDR_52
+0x4B94 US_ALU_EXT_ADDR_53
+0x4B98 US_ALU_EXT_ADDR_54
+0x4B9C US_ALU_EXT_ADDR_55
+0x4BA0 US_ALU_EXT_ADDR_56
+0x4BA4 US_ALU_EXT_ADDR_57
+0x4BA8 US_ALU_EXT_ADDR_58
+0x4BAC US_ALU_EXT_ADDR_59
+0x4BB0 US_ALU_EXT_ADDR_60
+0x4BB4 US_ALU_EXT_ADDR_61
+0x4BB8 US_ALU_EXT_ADDR_62
+0x4BBC US_ALU_EXT_ADDR_63
+0x4BC0 FG_FOG_BLEND
+0x4BC4 FG_FOG_FACTOR
+0x4BC8 FG_FOG_COLOR_R
+0x4BCC FG_FOG_COLOR_G
+0x4BD0 FG_FOG_COLOR_B
+0x4BD4 FG_ALPHA_FUNC
+0x4BD8 FG_DEPTH_SRC
+0x4C00 US_ALU_CONST_R_0
+0x4C04 US_ALU_CONST_G_0
+0x4C08 US_ALU_CONST_B_0
+0x4C0C US_ALU_CONST_A_0
+0x4C10 US_ALU_CONST_R_1
+0x4C14 US_ALU_CONST_G_1
+0x4C18 US_ALU_CONST_B_1
+0x4C1C US_ALU_CONST_A_1
+0x4C20 US_ALU_CONST_R_2
+0x4C24 US_ALU_CONST_G_2
+0x4C28 US_ALU_CONST_B_2
+0x4C2C US_ALU_CONST_A_2
+0x4C30 US_ALU_CONST_R_3
+0x4C34 US_ALU_CONST_G_3
+0x4C38 US_ALU_CONST_B_3
+0x4C3C US_ALU_CONST_A_3
+0x4C40 US_ALU_CONST_R_4
+0x4C44 US_ALU_CONST_G_4
+0x4C48 US_ALU_CONST_B_4
+0x4C4C US_ALU_CONST_A_4
+0x4C50 US_ALU_CONST_R_5
+0x4C54 US_ALU_CONST_G_5
+0x4C58 US_ALU_CONST_B_5
+0x4C5C US_ALU_CONST_A_5
+0x4C60 US_ALU_CONST_R_6
+0x4C64 US_ALU_CONST_G_6
+0x4C68 US_ALU_CONST_B_6
+0x4C6C US_ALU_CONST_A_6
+0x4C70 US_ALU_CONST_R_7
+0x4C74 US_ALU_CONST_G_7
+0x4C78 US_ALU_CONST_B_7
+0x4C7C US_ALU_CONST_A_7
+0x4C80 US_ALU_CONST_R_8
+0x4C84 US_ALU_CONST_G_8
+0x4C88 US_ALU_CONST_B_8
+0x4C8C US_ALU_CONST_A_8
+0x4C90 US_ALU_CONST_R_9
+0x4C94 US_ALU_CONST_G_9
+0x4C98 US_ALU_CONST_B_9
+0x4C9C US_ALU_CONST_A_9
+0x4CA0 US_ALU_CONST_R_10
+0x4CA4 US_ALU_CONST_G_10
+0x4CA8 US_ALU_CONST_B_10
+0x4CAC US_ALU_CONST_A_10
+0x4CB0 US_ALU_CONST_R_11
+0x4CB4 US_ALU_CONST_G_11
+0x4CB8 US_ALU_CONST_B_11
+0x4CBC US_ALU_CONST_A_11
+0x4CC0 US_ALU_CONST_R_12
+0x4CC4 US_ALU_CONST_G_12
+0x4CC8 US_ALU_CONST_B_12
+0x4CCC US_ALU_CONST_A_12
+0x4CD0 US_ALU_CONST_R_13
+0x4CD4 US_ALU_CONST_G_13
+0x4CD8 US_ALU_CONST_B_13
+0x4CDC US_ALU_CONST_A_13
+0x4CE0 US_ALU_CONST_R_14
+0x4CE4 US_ALU_CONST_G_14
+0x4CE8 US_ALU_CONST_B_14
+0x4CEC US_ALU_CONST_A_14
+0x4CF0 US_ALU_CONST_R_15
+0x4CF4 US_ALU_CONST_G_15
+0x4CF8 US_ALU_CONST_B_15
+0x4CFC US_ALU_CONST_A_15
+0x4D00 US_ALU_CONST_R_16
+0x4D04 US_ALU_CONST_G_16
+0x4D08 US_ALU_CONST_B_16
+0x4D0C US_ALU_CONST_A_16
+0x4D10 US_ALU_CONST_R_17
+0x4D14 US_ALU_CONST_G_17
+0x4D18 US_ALU_CONST_B_17
+0x4D1C US_ALU_CONST_A_17
+0x4D20 US_ALU_CONST_R_18
+0x4D24 US_ALU_CONST_G_18
+0x4D28 US_ALU_CONST_B_18
+0x4D2C US_ALU_CONST_A_18
+0x4D30 US_ALU_CONST_R_19
+0x4D34 US_ALU_CONST_G_19
+0x4D38 US_ALU_CONST_B_19
+0x4D3C US_ALU_CONST_A_19
+0x4D40 US_ALU_CONST_R_20
+0x4D44 US_ALU_CONST_G_20
+0x4D48 US_ALU_CONST_B_20
+0x4D4C US_ALU_CONST_A_20
+0x4D50 US_ALU_CONST_R_21
+0x4D54 US_ALU_CONST_G_21
+0x4D58 US_ALU_CONST_B_21
+0x4D5C US_ALU_CONST_A_21
+0x4D60 US_ALU_CONST_R_22
+0x4D64 US_ALU_CONST_G_22
+0x4D68 US_ALU_CONST_B_22
+0x4D6C US_ALU_CONST_A_22
+0x4D70 US_ALU_CONST_R_23
+0x4D74 US_ALU_CONST_G_23
+0x4D78 US_ALU_CONST_B_23
+0x4D7C US_ALU_CONST_A_23
+0x4D80 US_ALU_CONST_R_24
+0x4D84 US_ALU_CONST_G_24
+0x4D88 US_ALU_CONST_B_24
+0x4D8C US_ALU_CONST_A_24
+0x4D90 US_ALU_CONST_R_25
+0x4D94 US_ALU_CONST_G_25
+0x4D98 US_ALU_CONST_B_25
+0x4D9C US_ALU_CONST_A_25
+0x4DA0 US_ALU_CONST_R_26
+0x4DA4 US_ALU_CONST_G_26
+0x4DA8 US_ALU_CONST_B_26
+0x4DAC US_ALU_CONST_A_26
+0x4DB0 US_ALU_CONST_R_27
+0x4DB4 US_ALU_CONST_G_27
+0x4DB8 US_ALU_CONST_B_27
+0x4DBC US_ALU_CONST_A_27
+0x4DC0 US_ALU_CONST_R_28
+0x4DC4 US_ALU_CONST_G_28
+0x4DC8 US_ALU_CONST_B_28
+0x4DCC US_ALU_CONST_A_28
+0x4DD0 US_ALU_CONST_R_29
+0x4DD4 US_ALU_CONST_G_29
+0x4DD8 US_ALU_CONST_B_29
+0x4DDC US_ALU_CONST_A_29
+0x4DE0 US_ALU_CONST_R_30
+0x4DE4 US_ALU_CONST_G_30
+0x4DE8 US_ALU_CONST_B_30
+0x4DEC US_ALU_CONST_A_30
+0x4DF0 US_ALU_CONST_R_31
+0x4DF4 US_ALU_CONST_G_31
+0x4DF8 US_ALU_CONST_B_31
+0x4DFC US_ALU_CONST_A_31
+0x4E04 RB3D_BLENDCNTL_R3
+0x4E08 RB3D_ABLENDCNTL_R3
+0x4E0C RB3D_COLOR_CHANNEL_MASK
+0x4E10 RB3D_CONSTANT_COLOR
+0x4E14 RB3D_COLOR_CLEAR_VALUE
+0x4E18 RB3D_ROPCNTL_R3
+0x4E1C RB3D_CLRCMP_FLIPE_R3
+0x4E20 RB3D_CLRCMP_CLR_R3
+0x4E24 RB3D_CLRCMP_MSK_R3
+0x4E48 RB3D_DEBUG_CTL
+0x4E4C RB3D_DSTCACHE_CTLSTAT_R3
+0x4E50 RB3D_DITHER_CTL
+0x4E54 RB3D_CMASK_OFFSET0
+0x4E58 RB3D_CMASK_OFFSET1
+0x4E5C RB3D_CMASK_OFFSET2
+0x4E60 RB3D_CMASK_OFFSET3
+0x4E64 RB3D_CMASK_PITCH0
+0x4E68 RB3D_CMASK_PITCH1
+0x4E6C RB3D_CMASK_PITCH2
+0x4E70 RB3D_CMASK_PITCH3
+0x4E74 RB3D_CMASK_WRINDEX
+0x4E78 RB3D_CMASK_DWORD
+0x4E7C RB3D_CMASK_RDINDEX
+0x4E80 RB3D_AARESOLVE_OFFSET
+0x4E84 RB3D_AARESOLVE_PITCH
+0x4E88 RB3D_AARESOLVE_CTL
+0x4EA0 RB3D_DISCARD_SRC_PIXEL_LTE_THRESHOLD
+0x4EA4 RB3D_DISCARD_SRC_PIXEL_GTE_THRESHOLD
+0x4F04 ZB_ZSTENCILCNTL
+0x4F08 ZB_STENCILREFMASK
+0x4F14 ZB_ZTOP
+0x4F18 ZB_ZCACHE_CTLSTAT
+0x4F1C ZB_BW_CNTL
+0x4F28 ZB_DEPTHCLEARVALUE
+0x4F30 ZB_ZMASK_OFFSET
+0x4F34 ZB_ZMASK_PITCH
+0x4F38 ZB_ZMASK_WRINDEX
+0x4F3C ZB_ZMASK_DWORD
+0x4F40 ZB_ZMASK_RDINDEX
+0x4F44 ZB_HIZ_OFFSET
+0x4F48 ZB_HIZ_WRINDEX
+0x4F4C ZB_HIZ_DWORD
+0x4F50 ZB_HIZ_RDINDEX
+0x4F54 ZB_HIZ_PITCH
+0x4F58 ZB_ZPASS_DATA
diff --git a/drivers/gpu/drm/radeon/reg_srcs/r600 b/drivers/gpu/drm/radeon/reg_srcs/r600
new file mode 100644
index 000000000000..8f414a5f520f
--- /dev/null
+++ b/drivers/gpu/drm/radeon/reg_srcs/r600
@@ -0,0 +1,837 @@
+r600 0x9400
+0x000287A0 R7xx_CB_SHADER_CONTROL
+0x00028230 R7xx_PA_SC_EDGERULE
+0x000286C8 R7xx_SPI_THREAD_GROUPING
+0x00008D8C R7xx_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ
+0x000088C4 VGT_CACHE_INVALIDATION
+0x00028A50 VGT_ENHANCE
+0x000088CC VGT_ES_PER_GS
+0x00028A2C VGT_GROUP_DECR
+0x00028A28 VGT_GROUP_FIRST_DECR
+0x00028A24 VGT_GROUP_PRIM_TYPE
+0x00028A30 VGT_GROUP_VECT_0_CNTL
+0x00028A38 VGT_GROUP_VECT_0_FMT_CNTL
+0x00028A34 VGT_GROUP_VECT_1_CNTL
+0x00028A3C VGT_GROUP_VECT_1_FMT_CNTL
+0x00028A40 VGT_GS_MODE
+0x00028A6C VGT_GS_OUT_PRIM_TYPE
+0x000088C8 VGT_GS_PER_ES
+0x000088E8 VGT_GS_PER_VS
+0x000088D4 VGT_GS_VERTEX_REUSE
+0x00028A14 VGT_HOS_CNTL
+0x00028A18 VGT_HOS_MAX_TESS_LEVEL
+0x00028A1C VGT_HOS_MIN_TESS_LEVEL
+0x00028A20 VGT_HOS_REUSE_DEPTH
+0x0000895C VGT_INDEX_TYPE
+0x00028408 VGT_INDX_OFFSET
+0x00028AA0 VGT_INSTANCE_STEP_RATE_0
+0x00028AA4 VGT_INSTANCE_STEP_RATE_1
+0x000088C0 VGT_LAST_COPY_STATE
+0x00028400 VGT_MAX_VTX_INDX
+0x000088D8 VGT_MC_LAT_CNTL
+0x00028404 VGT_MIN_VTX_INDX
+0x00028A94 VGT_MULTI_PRIM_IB_RESET_EN
+0x0002840C VGT_MULTI_PRIM_IB_RESET_INDX
+0x00008970 VGT_NUM_INDICES
+0x00008974 VGT_NUM_INSTANCES
+0x00028A10 VGT_OUTPUT_PATH_CNTL
+0x00028C5C VGT_OUT_DEALLOC_CNTL
+0x00028A84 VGT_PRIMITIVEID_EN
+0x00008958 VGT_PRIMITIVE_TYPE
+0x00028AB4 VGT_REUSE_OFF
+0x00028C58 VGT_VERTEX_REUSE_BLOCK_CNTL
+0x00028AB8 VGT_VTX_CNT_EN
+0x000088B0 VGT_VTX_VECT_EJECT_REG
+0x00028810 PA_CL_CLIP_CNTL
+0x00008A14 PA_CL_ENHANCE
+0x00028C14 PA_CL_GB_HORZ_CLIP_ADJ
+0x00028C18 PA_CL_GB_HORZ_DISC_ADJ
+0x00028C0C PA_CL_GB_VERT_CLIP_ADJ
+0x00028C10 PA_CL_GB_VERT_DISC_ADJ
+0x00028820 PA_CL_NANINF_CNTL
+0x00028E1C PA_CL_POINT_CULL_RAD
+0x00028E18 PA_CL_POINT_SIZE
+0x00028E10 PA_CL_POINT_X_RAD
+0x00028E14 PA_CL_POINT_Y_RAD
+0x00028E2C PA_CL_UCP_0_W
+0x00028E3C PA_CL_UCP_1_W
+0x00028E4C PA_CL_UCP_2_W
+0x00028E5C PA_CL_UCP_3_W
+0x00028E6C PA_CL_UCP_4_W
+0x00028E7C PA_CL_UCP_5_W
+0x00028E20 PA_CL_UCP_0_X
+0x00028E30 PA_CL_UCP_1_X
+0x00028E40 PA_CL_UCP_2_X
+0x00028E50 PA_CL_UCP_3_X
+0x00028E60 PA_CL_UCP_4_X
+0x00028E70 PA_CL_UCP_5_X
+0x00028E24 PA_CL_UCP_0_Y
+0x00028E34 PA_CL_UCP_1_Y
+0x00028E44 PA_CL_UCP_2_Y
+0x00028E54 PA_CL_UCP_3_Y
+0x00028E64 PA_CL_UCP_4_Y
+0x00028E74 PA_CL_UCP_5_Y
+0x00028E28 PA_CL_UCP_0_Z
+0x00028E38 PA_CL_UCP_1_Z
+0x00028E48 PA_CL_UCP_2_Z
+0x00028E58 PA_CL_UCP_3_Z
+0x00028E68 PA_CL_UCP_4_Z
+0x00028E78 PA_CL_UCP_5_Z
+0x00028440 PA_CL_VPORT_XOFFSET_0
+0x00028458 PA_CL_VPORT_XOFFSET_1
+0x00028470 PA_CL_VPORT_XOFFSET_2
+0x00028488 PA_CL_VPORT_XOFFSET_3
+0x000284A0 PA_CL_VPORT_XOFFSET_4
+0x000284B8 PA_CL_VPORT_XOFFSET_5
+0x000284D0 PA_CL_VPORT_XOFFSET_6
+0x000284E8 PA_CL_VPORT_XOFFSET_7
+0x00028500 PA_CL_VPORT_XOFFSET_8
+0x00028518 PA_CL_VPORT_XOFFSET_9
+0x00028530 PA_CL_VPORT_XOFFSET_10
+0x00028548 PA_CL_VPORT_XOFFSET_11
+0x00028560 PA_CL_VPORT_XOFFSET_12
+0x00028578 PA_CL_VPORT_XOFFSET_13
+0x00028590 PA_CL_VPORT_XOFFSET_14
+0x000285A8 PA_CL_VPORT_XOFFSET_15
+0x0002843C PA_CL_VPORT_XSCALE_0
+0x00028454 PA_CL_VPORT_XSCALE_1
+0x0002846C PA_CL_VPORT_XSCALE_2
+0x00028484 PA_CL_VPORT_XSCALE_3
+0x0002849C PA_CL_VPORT_XSCALE_4
+0x000284B4 PA_CL_VPORT_XSCALE_5
+0x000284CC PA_CL_VPORT_XSCALE_6
+0x000284E4 PA_CL_VPORT_XSCALE_7
+0x000284FC PA_CL_VPORT_XSCALE_8
+0x00028514 PA_CL_VPORT_XSCALE_9
+0x0002852C PA_CL_VPORT_XSCALE_10
+0x00028544 PA_CL_VPORT_XSCALE_11
+0x0002855C PA_CL_VPORT_XSCALE_12
+0x00028574 PA_CL_VPORT_XSCALE_13
+0x0002858C PA_CL_VPORT_XSCALE_14
+0x000285A4 PA_CL_VPORT_XSCALE_15
+0x00028448 PA_CL_VPORT_YOFFSET_0
+0x00028460 PA_CL_VPORT_YOFFSET_1
+0x00028478 PA_CL_VPORT_YOFFSET_2
+0x00028490 PA_CL_VPORT_YOFFSET_3
+0x000284A8 PA_CL_VPORT_YOFFSET_4
+0x000284C0 PA_CL_VPORT_YOFFSET_5
+0x000284D8 PA_CL_VPORT_YOFFSET_6
+0x000284F0 PA_CL_VPORT_YOFFSET_7
+0x00028508 PA_CL_VPORT_YOFFSET_8
+0x00028520 PA_CL_VPORT_YOFFSET_9
+0x00028538 PA_CL_VPORT_YOFFSET_10
+0x00028550 PA_CL_VPORT_YOFFSET_11
+0x00028568 PA_CL_VPORT_YOFFSET_12
+0x00028580 PA_CL_VPORT_YOFFSET_13
+0x00028598 PA_CL_VPORT_YOFFSET_14
+0x000285B0 PA_CL_VPORT_YOFFSET_15
+0x00028444 PA_CL_VPORT_YSCALE_0
+0x0002845C PA_CL_VPORT_YSCALE_1
+0x00028474 PA_CL_VPORT_YSCALE_2
+0x0002848C PA_CL_VPORT_YSCALE_3
+0x000284A4 PA_CL_VPORT_YSCALE_4
+0x000284BC PA_CL_VPORT_YSCALE_5
+0x000284D4 PA_CL_VPORT_YSCALE_6
+0x000284EC PA_CL_VPORT_YSCALE_7
+0x00028504 PA_CL_VPORT_YSCALE_8
+0x0002851C PA_CL_VPORT_YSCALE_9
+0x00028534 PA_CL_VPORT_YSCALE_10
+0x0002854C PA_CL_VPORT_YSCALE_11
+0x00028564 PA_CL_VPORT_YSCALE_12
+0x0002857C PA_CL_VPORT_YSCALE_13
+0x00028594 PA_CL_VPORT_YSCALE_14
+0x000285AC PA_CL_VPORT_YSCALE_15
+0x00028450 PA_CL_VPORT_ZOFFSET_0
+0x00028468 PA_CL_VPORT_ZOFFSET_1
+0x00028480 PA_CL_VPORT_ZOFFSET_2
+0x00028498 PA_CL_VPORT_ZOFFSET_3
+0x000284B0 PA_CL_VPORT_ZOFFSET_4
+0x000284C8 PA_CL_VPORT_ZOFFSET_5
+0x000284E0 PA_CL_VPORT_ZOFFSET_6
+0x000284F8 PA_CL_VPORT_ZOFFSET_7
+0x00028510 PA_CL_VPORT_ZOFFSET_8
+0x00028528 PA_CL_VPORT_ZOFFSET_9
+0x00028540 PA_CL_VPORT_ZOFFSET_10
+0x00028558 PA_CL_VPORT_ZOFFSET_11
+0x00028570 PA_CL_VPORT_ZOFFSET_12
+0x00028588 PA_CL_VPORT_ZOFFSET_13
+0x000285A0 PA_CL_VPORT_ZOFFSET_14
+0x000285B8 PA_CL_VPORT_ZOFFSET_15
+0x0002844C PA_CL_VPORT_ZSCALE_0
+0x00028464 PA_CL_VPORT_ZSCALE_1
+0x0002847C PA_CL_VPORT_ZSCALE_2
+0x00028494 PA_CL_VPORT_ZSCALE_3
+0x000284AC PA_CL_VPORT_ZSCALE_4
+0x000284C4 PA_CL_VPORT_ZSCALE_5
+0x000284DC PA_CL_VPORT_ZSCALE_6
+0x000284F4 PA_CL_VPORT_ZSCALE_7
+0x0002850C PA_CL_VPORT_ZSCALE_8
+0x00028524 PA_CL_VPORT_ZSCALE_9
+0x0002853C PA_CL_VPORT_ZSCALE_10
+0x00028554 PA_CL_VPORT_ZSCALE_11
+0x0002856C PA_CL_VPORT_ZSCALE_12
+0x00028584 PA_CL_VPORT_ZSCALE_13
+0x0002859C PA_CL_VPORT_ZSCALE_14
+0x000285B4 PA_CL_VPORT_ZSCALE_15
+0x0002881C PA_CL_VS_OUT_CNTL
+0x00028818 PA_CL_VTE_CNTL
+0x00028C48 PA_SC_AA_MASK
+0x00008B40 PA_SC_AA_SAMPLE_LOCS_2S
+0x00008B44 PA_SC_AA_SAMPLE_LOCS_4S
+0x00008B48 PA_SC_AA_SAMPLE_LOCS_8S_WD0
+0x00008B4C PA_SC_AA_SAMPLE_LOCS_8S_WD1
+0x00028C20 PA_SC_AA_SAMPLE_LOCS_8S_WD1_MCTX
+0x00028C1C PA_SC_AA_SAMPLE_LOCS_MCTX
+0x00028214 PA_SC_CLIPRECT_0_BR
+0x0002821C PA_SC_CLIPRECT_1_BR
+0x00028224 PA_SC_CLIPRECT_2_BR
+0x0002822C PA_SC_CLIPRECT_3_BR
+0x00028210 PA_SC_CLIPRECT_0_TL
+0x00028218 PA_SC_CLIPRECT_1_TL
+0x00028220 PA_SC_CLIPRECT_2_TL
+0x00028228 PA_SC_CLIPRECT_3_TL
+0x0002820C PA_SC_CLIPRECT_RULE
+0x00008BF0 PA_SC_ENHANCE
+0x00028244 PA_SC_GENERIC_SCISSOR_BR
+0x00028240 PA_SC_GENERIC_SCISSOR_TL
+0x00028C00 PA_SC_LINE_CNTL
+0x00028A0C PA_SC_LINE_STIPPLE
+0x00008B10 PA_SC_LINE_STIPPLE_STATE
+0x00028A4C PA_SC_MODE_CNTL
+0x00028A48 PA_SC_MPASS_PS_CNTL
+0x00008B20 PA_SC_MULTI_CHIP_CNTL
+0x00028034 PA_SC_SCREEN_SCISSOR_BR
+0x00028030 PA_SC_SCREEN_SCISSOR_TL
+0x00028254 PA_SC_VPORT_SCISSOR_0_BR
+0x0002825C PA_SC_VPORT_SCISSOR_1_BR
+0x00028264 PA_SC_VPORT_SCISSOR_2_BR
+0x0002826C PA_SC_VPORT_SCISSOR_3_BR
+0x00028274 PA_SC_VPORT_SCISSOR_4_BR
+0x0002827C PA_SC_VPORT_SCISSOR_5_BR
+0x00028284 PA_SC_VPORT_SCISSOR_6_BR
+0x0002828C PA_SC_VPORT_SCISSOR_7_BR
+0x00028294 PA_SC_VPORT_SCISSOR_8_BR
+0x0002829C PA_SC_VPORT_SCISSOR_9_BR
+0x000282A4 PA_SC_VPORT_SCISSOR_10_BR
+0x000282AC PA_SC_VPORT_SCISSOR_11_BR
+0x000282B4 PA_SC_VPORT_SCISSOR_12_BR
+0x000282BC PA_SC_VPORT_SCISSOR_13_BR
+0x000282C4 PA_SC_VPORT_SCISSOR_14_BR
+0x000282CC PA_SC_VPORT_SCISSOR_15_BR
+0x00028250 PA_SC_VPORT_SCISSOR_0_TL
+0x00028258 PA_SC_VPORT_SCISSOR_1_TL
+0x00028260 PA_SC_VPORT_SCISSOR_2_TL
+0x00028268 PA_SC_VPORT_SCISSOR_3_TL
+0x00028270 PA_SC_VPORT_SCISSOR_4_TL
+0x00028278 PA_SC_VPORT_SCISSOR_5_TL
+0x00028280 PA_SC_VPORT_SCISSOR_6_TL
+0x00028288 PA_SC_VPORT_SCISSOR_7_TL
+0x00028290 PA_SC_VPORT_SCISSOR_8_TL
+0x00028298 PA_SC_VPORT_SCISSOR_9_TL
+0x000282A0 PA_SC_VPORT_SCISSOR_10_TL
+0x000282A8 PA_SC_VPORT_SCISSOR_11_TL
+0x000282B0 PA_SC_VPORT_SCISSOR_12_TL
+0x000282B8 PA_SC_VPORT_SCISSOR_13_TL
+0x000282C0 PA_SC_VPORT_SCISSOR_14_TL
+0x000282C8 PA_SC_VPORT_SCISSOR_15_TL
+0x000282D4 PA_SC_VPORT_ZMAX_0
+0x000282DC PA_SC_VPORT_ZMAX_1
+0x000282E4 PA_SC_VPORT_ZMAX_2
+0x000282EC PA_SC_VPORT_ZMAX_3
+0x000282F4 PA_SC_VPORT_ZMAX_4
+0x000282FC PA_SC_VPORT_ZMAX_5
+0x00028304 PA_SC_VPORT_ZMAX_6
+0x0002830C PA_SC_VPORT_ZMAX_7
+0x00028314 PA_SC_VPORT_ZMAX_8
+0x0002831C PA_SC_VPORT_ZMAX_9
+0x00028324 PA_SC_VPORT_ZMAX_10
+0x0002832C PA_SC_VPORT_ZMAX_11
+0x00028334 PA_SC_VPORT_ZMAX_12
+0x0002833C PA_SC_VPORT_ZMAX_13
+0x00028344 PA_SC_VPORT_ZMAX_14
+0x0002834C PA_SC_VPORT_ZMAX_15
+0x000282D0 PA_SC_VPORT_ZMIN_0
+0x000282D8 PA_SC_VPORT_ZMIN_1
+0x000282E0 PA_SC_VPORT_ZMIN_2
+0x000282E8 PA_SC_VPORT_ZMIN_3
+0x000282F0 PA_SC_VPORT_ZMIN_4
+0x000282F8 PA_SC_VPORT_ZMIN_5
+0x00028300 PA_SC_VPORT_ZMIN_6
+0x00028308 PA_SC_VPORT_ZMIN_7
+0x00028310 PA_SC_VPORT_ZMIN_8
+0x00028318 PA_SC_VPORT_ZMIN_9
+0x00028320 PA_SC_VPORT_ZMIN_10
+0x00028328 PA_SC_VPORT_ZMIN_11
+0x00028330 PA_SC_VPORT_ZMIN_12
+0x00028338 PA_SC_VPORT_ZMIN_13
+0x00028340 PA_SC_VPORT_ZMIN_14
+0x00028348 PA_SC_VPORT_ZMIN_15
+0x00028200 PA_SC_WINDOW_OFFSET
+0x00028208 PA_SC_WINDOW_SCISSOR_BR
+0x00028204 PA_SC_WINDOW_SCISSOR_TL
+0x00028A08 PA_SU_LINE_CNTL
+0x00028A04 PA_SU_POINT_MINMAX
+0x00028A00 PA_SU_POINT_SIZE
+0x00028E0C PA_SU_POLY_OFFSET_BACK_OFFSET
+0x00028E08 PA_SU_POLY_OFFSET_BACK_SCALE
+0x00028DFC PA_SU_POLY_OFFSET_CLAMP
+0x00028DF8 PA_SU_POLY_OFFSET_DB_FMT_CNTL
+0x00028E04 PA_SU_POLY_OFFSET_FRONT_OFFSET
+0x00028E00 PA_SU_POLY_OFFSET_FRONT_SCALE
+0x00028814 PA_SU_SC_MODE_CNTL
+0x00028C08 PA_SU_VTX_CNTL
+0x00008C00 SQ_CONFIG
+0x00008C04 SQ_GPR_RESOURCE_MGMT_1
+0x00008C08 SQ_GPR_RESOURCE_MGMT_2
+0x00008C10 SQ_STACK_RESOURCE_MGMT_1
+0x00008C14 SQ_STACK_RESOURCE_MGMT_2
+0x00008C0C SQ_THREAD_RESOURCE_MGMT
+0x00028380 SQ_VTX_SEMANTIC_0
+0x00028384 SQ_VTX_SEMANTIC_1
+0x00028388 SQ_VTX_SEMANTIC_2
+0x0002838C SQ_VTX_SEMANTIC_3
+0x00028390 SQ_VTX_SEMANTIC_4
+0x00028394 SQ_VTX_SEMANTIC_5
+0x00028398 SQ_VTX_SEMANTIC_6
+0x0002839C SQ_VTX_SEMANTIC_7
+0x000283A0 SQ_VTX_SEMANTIC_8
+0x000283A4 SQ_VTX_SEMANTIC_9
+0x000283A8 SQ_VTX_SEMANTIC_10
+0x000283AC SQ_VTX_SEMANTIC_11
+0x000283B0 SQ_VTX_SEMANTIC_12
+0x000283B4 SQ_VTX_SEMANTIC_13
+0x000283B8 SQ_VTX_SEMANTIC_14
+0x000283BC SQ_VTX_SEMANTIC_15
+0x000283C0 SQ_VTX_SEMANTIC_16
+0x000283C4 SQ_VTX_SEMANTIC_17
+0x000283C8 SQ_VTX_SEMANTIC_18
+0x000283CC SQ_VTX_SEMANTIC_19
+0x000283D0 SQ_VTX_SEMANTIC_20
+0x000283D4 SQ_VTX_SEMANTIC_21
+0x000283D8 SQ_VTX_SEMANTIC_22
+0x000283DC SQ_VTX_SEMANTIC_23
+0x000283E0 SQ_VTX_SEMANTIC_24
+0x000283E4 SQ_VTX_SEMANTIC_25
+0x000283E8 SQ_VTX_SEMANTIC_26
+0x000283EC SQ_VTX_SEMANTIC_27
+0x000283F0 SQ_VTX_SEMANTIC_28
+0x000283F4 SQ_VTX_SEMANTIC_29
+0x000283F8 SQ_VTX_SEMANTIC_30
+0x000283FC SQ_VTX_SEMANTIC_31
+0x000288E0 SQ_VTX_SEMANTIC_CLEAR
+0x0003CFF4 SQ_VTX_START_INST_LOC
+0x0003C000 SQ_TEX_SAMPLER_WORD0_0
+0x0003C004 SQ_TEX_SAMPLER_WORD1_0
+0x0003C008 SQ_TEX_SAMPLER_WORD2_0
+0x00030000 SQ_ALU_CONSTANT0_0
+0x00030004 SQ_ALU_CONSTANT1_0
+0x00030008 SQ_ALU_CONSTANT2_0
+0x0003000C SQ_ALU_CONSTANT3_0
+0x0003E380 SQ_BOOL_CONST_0
+0x0003E384 SQ_BOOL_CONST_1
+0x0003E388 SQ_BOOL_CONST_2
+0x0003E200 SQ_LOOP_CONST_0
+0x0003E200 SQ_LOOP_CONST_DX10_0
+0x000281C0 SQ_ALU_CONST_BUFFER_SIZE_GS_0
+0x000281C4 SQ_ALU_CONST_BUFFER_SIZE_GS_1
+0x000281C8 SQ_ALU_CONST_BUFFER_SIZE_GS_2
+0x000281CC SQ_ALU_CONST_BUFFER_SIZE_GS_3
+0x000281D0 SQ_ALU_CONST_BUFFER_SIZE_GS_4
+0x000281D4 SQ_ALU_CONST_BUFFER_SIZE_GS_5
+0x000281D8 SQ_ALU_CONST_BUFFER_SIZE_GS_6
+0x000281DC SQ_ALU_CONST_BUFFER_SIZE_GS_7
+0x000281E0 SQ_ALU_CONST_BUFFER_SIZE_GS_8
+0x000281E4 SQ_ALU_CONST_BUFFER_SIZE_GS_9
+0x000281E8 SQ_ALU_CONST_BUFFER_SIZE_GS_10
+0x000281EC SQ_ALU_CONST_BUFFER_SIZE_GS_11
+0x000281F0 SQ_ALU_CONST_BUFFER_SIZE_GS_12
+0x000281F4 SQ_ALU_CONST_BUFFER_SIZE_GS_13
+0x000281F8 SQ_ALU_CONST_BUFFER_SIZE_GS_14
+0x000281FC SQ_ALU_CONST_BUFFER_SIZE_GS_15
+0x00028140 SQ_ALU_CONST_BUFFER_SIZE_PS_0
+0x00028144 SQ_ALU_CONST_BUFFER_SIZE_PS_1
+0x00028148 SQ_ALU_CONST_BUFFER_SIZE_PS_2
+0x0002814C SQ_ALU_CONST_BUFFER_SIZE_PS_3
+0x00028150 SQ_ALU_CONST_BUFFER_SIZE_PS_4
+0x00028154 SQ_ALU_CONST_BUFFER_SIZE_PS_5
+0x00028158 SQ_ALU_CONST_BUFFER_SIZE_PS_6
+0x0002815C SQ_ALU_CONST_BUFFER_SIZE_PS_7
+0x00028160 SQ_ALU_CONST_BUFFER_SIZE_PS_8
+0x00028164 SQ_ALU_CONST_BUFFER_SIZE_PS_9
+0x00028168 SQ_ALU_CONST_BUFFER_SIZE_PS_10
+0x0002816C SQ_ALU_CONST_BUFFER_SIZE_PS_11
+0x00028170 SQ_ALU_CONST_BUFFER_SIZE_PS_12
+0x00028174 SQ_ALU_CONST_BUFFER_SIZE_PS_13
+0x00028178 SQ_ALU_CONST_BUFFER_SIZE_PS_14
+0x0002817C SQ_ALU_CONST_BUFFER_SIZE_PS_15
+0x00028180 SQ_ALU_CONST_BUFFER_SIZE_VS_0
+0x00028184 SQ_ALU_CONST_BUFFER_SIZE_VS_1
+0x00028188 SQ_ALU_CONST_BUFFER_SIZE_VS_2
+0x0002818C SQ_ALU_CONST_BUFFER_SIZE_VS_3
+0x00028190 SQ_ALU_CONST_BUFFER_SIZE_VS_4
+0x00028194 SQ_ALU_CONST_BUFFER_SIZE_VS_5
+0x00028198 SQ_ALU_CONST_BUFFER_SIZE_VS_6
+0x0002819C SQ_ALU_CONST_BUFFER_SIZE_VS_7
+0x000281A0 SQ_ALU_CONST_BUFFER_SIZE_VS_8
+0x000281A4 SQ_ALU_CONST_BUFFER_SIZE_VS_9
+0x000281A8 SQ_ALU_CONST_BUFFER_SIZE_VS_10
+0x000281AC SQ_ALU_CONST_BUFFER_SIZE_VS_11
+0x000281B0 SQ_ALU_CONST_BUFFER_SIZE_VS_12
+0x000281B4 SQ_ALU_CONST_BUFFER_SIZE_VS_13
+0x000281B8 SQ_ALU_CONST_BUFFER_SIZE_VS_14
+0x000281BC SQ_ALU_CONST_BUFFER_SIZE_VS_15
+0x000289C0 SQ_ALU_CONST_CACHE_GS_0
+0x000289C4 SQ_ALU_CONST_CACHE_GS_1
+0x000289C8 SQ_ALU_CONST_CACHE_GS_2
+0x000289CC SQ_ALU_CONST_CACHE_GS_3
+0x000289D0 SQ_ALU_CONST_CACHE_GS_4
+0x000289D4 SQ_ALU_CONST_CACHE_GS_5
+0x000289D8 SQ_ALU_CONST_CACHE_GS_6
+0x000289DC SQ_ALU_CONST_CACHE_GS_7
+0x000289E0 SQ_ALU_CONST_CACHE_GS_8
+0x000289E4 SQ_ALU_CONST_CACHE_GS_9
+0x000289E8 SQ_ALU_CONST_CACHE_GS_10
+0x000289EC SQ_ALU_CONST_CACHE_GS_11
+0x000289F0 SQ_ALU_CONST_CACHE_GS_12
+0x000289F4 SQ_ALU_CONST_CACHE_GS_13
+0x000289F8 SQ_ALU_CONST_CACHE_GS_14
+0x000289FC SQ_ALU_CONST_CACHE_GS_15
+0x00028940 SQ_ALU_CONST_CACHE_PS_0
+0x00028944 SQ_ALU_CONST_CACHE_PS_1
+0x00028948 SQ_ALU_CONST_CACHE_PS_2
+0x0002894C SQ_ALU_CONST_CACHE_PS_3
+0x00028950 SQ_ALU_CONST_CACHE_PS_4
+0x00028954 SQ_ALU_CONST_CACHE_PS_5
+0x00028958 SQ_ALU_CONST_CACHE_PS_6
+0x0002895C SQ_ALU_CONST_CACHE_PS_7
+0x00028960 SQ_ALU_CONST_CACHE_PS_8
+0x00028964 SQ_ALU_CONST_CACHE_PS_9
+0x00028968 SQ_ALU_CONST_CACHE_PS_10
+0x0002896C SQ_ALU_CONST_CACHE_PS_11
+0x00028970 SQ_ALU_CONST_CACHE_PS_12
+0x00028974 SQ_ALU_CONST_CACHE_PS_13
+0x00028978 SQ_ALU_CONST_CACHE_PS_14
+0x0002897C SQ_ALU_CONST_CACHE_PS_15
+0x00028980 SQ_ALU_CONST_CACHE_VS_0
+0x00028984 SQ_ALU_CONST_CACHE_VS_1
+0x00028988 SQ_ALU_CONST_CACHE_VS_2
+0x0002898C SQ_ALU_CONST_CACHE_VS_3
+0x00028990 SQ_ALU_CONST_CACHE_VS_4
+0x00028994 SQ_ALU_CONST_CACHE_VS_5
+0x00028998 SQ_ALU_CONST_CACHE_VS_6
+0x0002899C SQ_ALU_CONST_CACHE_VS_7
+0x000289A0 SQ_ALU_CONST_CACHE_VS_8
+0x000289A4 SQ_ALU_CONST_CACHE_VS_9
+0x000289A8 SQ_ALU_CONST_CACHE_VS_10
+0x000289AC SQ_ALU_CONST_CACHE_VS_11
+0x000289B0 SQ_ALU_CONST_CACHE_VS_12
+0x000289B4 SQ_ALU_CONST_CACHE_VS_13
+0x000289B8 SQ_ALU_CONST_CACHE_VS_14
+0x000289BC SQ_ALU_CONST_CACHE_VS_15
+0x000288D8 SQ_PGM_CF_OFFSET_ES
+0x000288DC SQ_PGM_CF_OFFSET_FS
+0x000288D4 SQ_PGM_CF_OFFSET_GS
+0x000288CC SQ_PGM_CF_OFFSET_PS
+0x000288D0 SQ_PGM_CF_OFFSET_VS
+0x00028854 SQ_PGM_EXPORTS_PS
+0x00028890 SQ_PGM_RESOURCES_ES
+0x000288A4 SQ_PGM_RESOURCES_FS
+0x0002887C SQ_PGM_RESOURCES_GS
+0x00028850 SQ_PGM_RESOURCES_PS
+0x00028868 SQ_PGM_RESOURCES_VS
+0x00009100 SPI_CONFIG_CNTL
+0x0000913C SPI_CONFIG_CNTL_1
+0x000286DC SPI_FOG_CNTL
+0x000286E4 SPI_FOG_FUNC_BIAS
+0x000286E0 SPI_FOG_FUNC_SCALE
+0x000286D8 SPI_INPUT_Z
+0x000286D4 SPI_INTERP_CONTROL_0
+0x00028644 SPI_PS_INPUT_CNTL_0
+0x00028648 SPI_PS_INPUT_CNTL_1
+0x0002864C SPI_PS_INPUT_CNTL_2
+0x00028650 SPI_PS_INPUT_CNTL_3
+0x00028654 SPI_PS_INPUT_CNTL_4
+0x00028658 SPI_PS_INPUT_CNTL_5
+0x0002865C SPI_PS_INPUT_CNTL_6
+0x00028660 SPI_PS_INPUT_CNTL_7
+0x00028664 SPI_PS_INPUT_CNTL_8
+0x00028668 SPI_PS_INPUT_CNTL_9
+0x0002866C SPI_PS_INPUT_CNTL_10
+0x00028670 SPI_PS_INPUT_CNTL_11
+0x00028674 SPI_PS_INPUT_CNTL_12
+0x00028678 SPI_PS_INPUT_CNTL_13
+0x0002867C SPI_PS_INPUT_CNTL_14
+0x00028680 SPI_PS_INPUT_CNTL_15
+0x00028684 SPI_PS_INPUT_CNTL_16
+0x00028688 SPI_PS_INPUT_CNTL_17
+0x0002868C SPI_PS_INPUT_CNTL_18
+0x00028690 SPI_PS_INPUT_CNTL_19
+0x00028694 SPI_PS_INPUT_CNTL_20
+0x00028698 SPI_PS_INPUT_CNTL_21
+0x0002869C SPI_PS_INPUT_CNTL_22
+0x000286A0 SPI_PS_INPUT_CNTL_23
+0x000286A4 SPI_PS_INPUT_CNTL_24
+0x000286A8 SPI_PS_INPUT_CNTL_25
+0x000286AC SPI_PS_INPUT_CNTL_26
+0x000286B0 SPI_PS_INPUT_CNTL_27
+0x000286B4 SPI_PS_INPUT_CNTL_28
+0x000286B8 SPI_PS_INPUT_CNTL_29
+0x000286BC SPI_PS_INPUT_CNTL_30
+0x000286C0 SPI_PS_INPUT_CNTL_31
+0x000286CC SPI_PS_IN_CONTROL_0
+0x000286D0 SPI_PS_IN_CONTROL_1
+0x000286C4 SPI_VS_OUT_CONFIG
+0x00028614 SPI_VS_OUT_ID_0
+0x00028618 SPI_VS_OUT_ID_1
+0x0002861C SPI_VS_OUT_ID_2
+0x00028620 SPI_VS_OUT_ID_3
+0x00028624 SPI_VS_OUT_ID_4
+0x00028628 SPI_VS_OUT_ID_5
+0x0002862C SPI_VS_OUT_ID_6
+0x00028630 SPI_VS_OUT_ID_7
+0x00028634 SPI_VS_OUT_ID_8
+0x00028638 SPI_VS_OUT_ID_9
+0x00028438 SX_ALPHA_REF
+0x00028410 SX_ALPHA_TEST_CONTROL
+0x00028350 SX_MISC
+0x0000A020 SMX_DC_CTL0
+0x0000A024 SMX_DC_CTL1
+0x0000A028 SMX_DC_CTL2
+0x00009608 TC_CNTL
+0x00009604 TC_INVALIDATE
+0x00009490 TD_CNTL
+0x00009400 TD_FILTER4
+0x00009404 TD_FILTER4_1
+0x00009408 TD_FILTER4_2
+0x0000940C TD_FILTER4_3
+0x00009410 TD_FILTER4_4
+0x00009414 TD_FILTER4_5
+0x00009418 TD_FILTER4_6
+0x0000941C TD_FILTER4_7
+0x00009420 TD_FILTER4_8
+0x00009424 TD_FILTER4_9
+0x00009428 TD_FILTER4_10
+0x0000942C TD_FILTER4_11
+0x00009430 TD_FILTER4_12
+0x00009434 TD_FILTER4_13
+0x00009438 TD_FILTER4_14
+0x0000943C TD_FILTER4_15
+0x00009440 TD_FILTER4_16
+0x00009444 TD_FILTER4_17
+0x00009448 TD_FILTER4_18
+0x0000944C TD_FILTER4_19
+0x00009450 TD_FILTER4_20
+0x00009454 TD_FILTER4_21
+0x00009458 TD_FILTER4_22
+0x0000945C TD_FILTER4_23
+0x00009460 TD_FILTER4_24
+0x00009464 TD_FILTER4_25
+0x00009468 TD_FILTER4_26
+0x0000946C TD_FILTER4_27
+0x00009470 TD_FILTER4_28
+0x00009474 TD_FILTER4_29
+0x00009478 TD_FILTER4_30
+0x0000947C TD_FILTER4_31
+0x00009480 TD_FILTER4_32
+0x00009484 TD_FILTER4_33
+0x00009488 TD_FILTER4_34
+0x0000948C TD_FILTER4_35
+0x0000A80C TD_GS_SAMPLER0_BORDER_ALPHA
+0x0000A81C TD_GS_SAMPLER1_BORDER_ALPHA
+0x0000A82C TD_GS_SAMPLER2_BORDER_ALPHA
+0x0000A83C TD_GS_SAMPLER3_BORDER_ALPHA
+0x0000A84C TD_GS_SAMPLER4_BORDER_ALPHA
+0x0000A85C TD_GS_SAMPLER5_BORDER_ALPHA
+0x0000A86C TD_GS_SAMPLER6_BORDER_ALPHA
+0x0000A87C TD_GS_SAMPLER7_BORDER_ALPHA
+0x0000A88C TD_GS_SAMPLER8_BORDER_ALPHA
+0x0000A89C TD_GS_SAMPLER9_BORDER_ALPHA
+0x0000A8AC TD_GS_SAMPLER10_BORDER_ALPHA
+0x0000A8BC TD_GS_SAMPLER11_BORDER_ALPHA
+0x0000A8CC TD_GS_SAMPLER12_BORDER_ALPHA
+0x0000A8DC TD_GS_SAMPLER13_BORDER_ALPHA
+0x0000A8EC TD_GS_SAMPLER14_BORDER_ALPHA
+0x0000A8FC TD_GS_SAMPLER15_BORDER_ALPHA
+0x0000A90C TD_GS_SAMPLER16_BORDER_ALPHA
+0x0000A91C TD_GS_SAMPLER17_BORDER_ALPHA
+0x0000A808 TD_GS_SAMPLER0_BORDER_BLUE
+0x0000A818 TD_GS_SAMPLER1_BORDER_BLUE
+0x0000A828 TD_GS_SAMPLER2_BORDER_BLUE
+0x0000A838 TD_GS_SAMPLER3_BORDER_BLUE
+0x0000A848 TD_GS_SAMPLER4_BORDER_BLUE
+0x0000A858 TD_GS_SAMPLER5_BORDER_BLUE
+0x0000A868 TD_GS_SAMPLER6_BORDER_BLUE
+0x0000A878 TD_GS_SAMPLER7_BORDER_BLUE
+0x0000A888 TD_GS_SAMPLER8_BORDER_BLUE
+0x0000A898 TD_GS_SAMPLER9_BORDER_BLUE
+0x0000A8A8 TD_GS_SAMPLER10_BORDER_BLUE
+0x0000A8B8 TD_GS_SAMPLER11_BORDER_BLUE
+0x0000A8C8 TD_GS_SAMPLER12_BORDER_BLUE
+0x0000A8D8 TD_GS_SAMPLER13_BORDER_BLUE
+0x0000A8E8 TD_GS_SAMPLER14_BORDER_BLUE
+0x0000A8F8 TD_GS_SAMPLER15_BORDER_BLUE
+0x0000A908 TD_GS_SAMPLER16_BORDER_BLUE
+0x0000A918 TD_GS_SAMPLER17_BORDER_BLUE
+0x0000A804 TD_GS_SAMPLER0_BORDER_GREEN
+0x0000A814 TD_GS_SAMPLER1_BORDER_GREEN
+0x0000A824 TD_GS_SAMPLER2_BORDER_GREEN
+0x0000A834 TD_GS_SAMPLER3_BORDER_GREEN
+0x0000A844 TD_GS_SAMPLER4_BORDER_GREEN
+0x0000A854 TD_GS_SAMPLER5_BORDER_GREEN
+0x0000A864 TD_GS_SAMPLER6_BORDER_GREEN
+0x0000A874 TD_GS_SAMPLER7_BORDER_GREEN
+0x0000A884 TD_GS_SAMPLER8_BORDER_GREEN
+0x0000A894 TD_GS_SAMPLER9_BORDER_GREEN
+0x0000A8A4 TD_GS_SAMPLER10_BORDER_GREEN
+0x0000A8B4 TD_GS_SAMPLER11_BORDER_GREEN
+0x0000A8C4 TD_GS_SAMPLER12_BORDER_GREEN
+0x0000A8D4 TD_GS_SAMPLER13_BORDER_GREEN
+0x0000A8E4 TD_GS_SAMPLER14_BORDER_GREEN
+0x0000A8F4 TD_GS_SAMPLER15_BORDER_GREEN
+0x0000A904 TD_GS_SAMPLER16_BORDER_GREEN
+0x0000A914 TD_GS_SAMPLER17_BORDER_GREEN
+0x0000A800 TD_GS_SAMPLER0_BORDER_RED
+0x0000A810 TD_GS_SAMPLER1_BORDER_RED
+0x0000A820 TD_GS_SAMPLER2_BORDER_RED
+0x0000A830 TD_GS_SAMPLER3_BORDER_RED
+0x0000A840 TD_GS_SAMPLER4_BORDER_RED
+0x0000A850 TD_GS_SAMPLER5_BORDER_RED
+0x0000A860 TD_GS_SAMPLER6_BORDER_RED
+0x0000A870 TD_GS_SAMPLER7_BORDER_RED
+0x0000A880 TD_GS_SAMPLER8_BORDER_RED
+0x0000A890 TD_GS_SAMPLER9_BORDER_RED
+0x0000A8A0 TD_GS_SAMPLER10_BORDER_RED
+0x0000A8B0 TD_GS_SAMPLER11_BORDER_RED
+0x0000A8C0 TD_GS_SAMPLER12_BORDER_RED
+0x0000A8D0 TD_GS_SAMPLER13_BORDER_RED
+0x0000A8E0 TD_GS_SAMPLER14_BORDER_RED
+0x0000A8F0 TD_GS_SAMPLER15_BORDER_RED
+0x0000A900 TD_GS_SAMPLER16_BORDER_RED
+0x0000A910 TD_GS_SAMPLER17_BORDER_RED
+0x0000A40C TD_PS_SAMPLER0_BORDER_ALPHA
+0x0000A41C TD_PS_SAMPLER1_BORDER_ALPHA
+0x0000A42C TD_PS_SAMPLER2_BORDER_ALPHA
+0x0000A43C TD_PS_SAMPLER3_BORDER_ALPHA
+0x0000A44C TD_PS_SAMPLER4_BORDER_ALPHA
+0x0000A45C TD_PS_SAMPLER5_BORDER_ALPHA
+0x0000A46C TD_PS_SAMPLER6_BORDER_ALPHA
+0x0000A47C TD_PS_SAMPLER7_BORDER_ALPHA
+0x0000A48C TD_PS_SAMPLER8_BORDER_ALPHA
+0x0000A49C TD_PS_SAMPLER9_BORDER_ALPHA
+0x0000A4AC TD_PS_SAMPLER10_BORDER_ALPHA
+0x0000A4BC TD_PS_SAMPLER11_BORDER_ALPHA
+0x0000A4CC TD_PS_SAMPLER12_BORDER_ALPHA
+0x0000A4DC TD_PS_SAMPLER13_BORDER_ALPHA
+0x0000A4EC TD_PS_SAMPLER14_BORDER_ALPHA
+0x0000A4FC TD_PS_SAMPLER15_BORDER_ALPHA
+0x0000A50C TD_PS_SAMPLER16_BORDER_ALPHA
+0x0000A51C TD_PS_SAMPLER17_BORDER_ALPHA
+0x0000A408 TD_PS_SAMPLER0_BORDER_BLUE
+0x0000A418 TD_PS_SAMPLER1_BORDER_BLUE
+0x0000A428 TD_PS_SAMPLER2_BORDER_BLUE
+0x0000A438 TD_PS_SAMPLER3_BORDER_BLUE
+0x0000A448 TD_PS_SAMPLER4_BORDER_BLUE
+0x0000A458 TD_PS_SAMPLER5_BORDER_BLUE
+0x0000A468 TD_PS_SAMPLER6_BORDER_BLUE
+0x0000A478 TD_PS_SAMPLER7_BORDER_BLUE
+0x0000A488 TD_PS_SAMPLER8_BORDER_BLUE
+0x0000A498 TD_PS_SAMPLER9_BORDER_BLUE
+0x0000A4A8 TD_PS_SAMPLER10_BORDER_BLUE
+0x0000A4B8 TD_PS_SAMPLER11_BORDER_BLUE
+0x0000A4C8 TD_PS_SAMPLER12_BORDER_BLUE
+0x0000A4D8 TD_PS_SAMPLER13_BORDER_BLUE
+0x0000A4E8 TD_PS_SAMPLER14_BORDER_BLUE
+0x0000A4F8 TD_PS_SAMPLER15_BORDER_BLUE
+0x0000A508 TD_PS_SAMPLER16_BORDER_BLUE
+0x0000A518 TD_PS_SAMPLER17_BORDER_BLUE
+0x0000A404 TD_PS_SAMPLER0_BORDER_GREEN
+0x0000A414 TD_PS_SAMPLER1_BORDER_GREEN
+0x0000A424 TD_PS_SAMPLER2_BORDER_GREEN
+0x0000A434 TD_PS_SAMPLER3_BORDER_GREEN
+0x0000A444 TD_PS_SAMPLER4_BORDER_GREEN
+0x0000A454 TD_PS_SAMPLER5_BORDER_GREEN
+0x0000A464 TD_PS_SAMPLER6_BORDER_GREEN
+0x0000A474 TD_PS_SAMPLER7_BORDER_GREEN
+0x0000A484 TD_PS_SAMPLER8_BORDER_GREEN
+0x0000A494 TD_PS_SAMPLER9_BORDER_GREEN
+0x0000A4A4 TD_PS_SAMPLER10_BORDER_GREEN
+0x0000A4B4 TD_PS_SAMPLER11_BORDER_GREEN
+0x0000A4C4 TD_PS_SAMPLER12_BORDER_GREEN
+0x0000A4D4 TD_PS_SAMPLER13_BORDER_GREEN
+0x0000A4E4 TD_PS_SAMPLER14_BORDER_GREEN
+0x0000A4F4 TD_PS_SAMPLER15_BORDER_GREEN
+0x0000A504 TD_PS_SAMPLER16_BORDER_GREEN
+0x0000A514 TD_PS_SAMPLER17_BORDER_GREEN
+0x0000A400 TD_PS_SAMPLER0_BORDER_RED
+0x0000A410 TD_PS_SAMPLER1_BORDER_RED
+0x0000A420 TD_PS_SAMPLER2_BORDER_RED
+0x0000A430 TD_PS_SAMPLER3_BORDER_RED
+0x0000A440 TD_PS_SAMPLER4_BORDER_RED
+0x0000A450 TD_PS_SAMPLER5_BORDER_RED
+0x0000A460 TD_PS_SAMPLER6_BORDER_RED
+0x0000A470 TD_PS_SAMPLER7_BORDER_RED
+0x0000A480 TD_PS_SAMPLER8_BORDER_RED
+0x0000A490 TD_PS_SAMPLER9_BORDER_RED
+0x0000A4A0 TD_PS_SAMPLER10_BORDER_RED
+0x0000A4B0 TD_PS_SAMPLER11_BORDER_RED
+0x0000A4C0 TD_PS_SAMPLER12_BORDER_RED
+0x0000A4D0 TD_PS_SAMPLER13_BORDER_RED
+0x0000A4E0 TD_PS_SAMPLER14_BORDER_RED
+0x0000A4F0 TD_PS_SAMPLER15_BORDER_RED
+0x0000A500 TD_PS_SAMPLER16_BORDER_RED
+0x0000A510 TD_PS_SAMPLER17_BORDER_RED
+0x0000AA00 TD_PS_SAMPLER0_CLEARTYPE_KERNEL
+0x0000AA04 TD_PS_SAMPLER1_CLEARTYPE_KERNEL
+0x0000AA08 TD_PS_SAMPLER2_CLEARTYPE_KERNEL
+0x0000AA0C TD_PS_SAMPLER3_CLEARTYPE_KERNEL
+0x0000AA10 TD_PS_SAMPLER4_CLEARTYPE_KERNEL
+0x0000AA14 TD_PS_SAMPLER5_CLEARTYPE_KERNEL
+0x0000AA18 TD_PS_SAMPLER6_CLEARTYPE_KERNEL
+0x0000AA1C TD_PS_SAMPLER7_CLEARTYPE_KERNEL
+0x0000AA20 TD_PS_SAMPLER8_CLEARTYPE_KERNEL
+0x0000AA24 TD_PS_SAMPLER9_CLEARTYPE_KERNEL
+0x0000AA28 TD_PS_SAMPLER10_CLEARTYPE_KERNEL
+0x0000AA2C TD_PS_SAMPLER11_CLEARTYPE_KERNEL
+0x0000AA30 TD_PS_SAMPLER12_CLEARTYPE_KERNEL
+0x0000AA34 TD_PS_SAMPLER13_CLEARTYPE_KERNEL
+0x0000AA38 TD_PS_SAMPLER14_CLEARTYPE_KERNEL
+0x0000AA3C TD_PS_SAMPLER15_CLEARTYPE_KERNEL
+0x0000AA40 TD_PS_SAMPLER16_CLEARTYPE_KERNEL
+0x0000AA44 TD_PS_SAMPLER17_CLEARTYPE_KERNEL
+0x0000A60C TD_VS_SAMPLER0_BORDER_ALPHA
+0x0000A61C TD_VS_SAMPLER1_BORDER_ALPHA
+0x0000A62C TD_VS_SAMPLER2_BORDER_ALPHA
+0x0000A63C TD_VS_SAMPLER3_BORDER_ALPHA
+0x0000A64C TD_VS_SAMPLER4_BORDER_ALPHA
+0x0000A65C TD_VS_SAMPLER5_BORDER_ALPHA
+0x0000A66C TD_VS_SAMPLER6_BORDER_ALPHA
+0x0000A67C TD_VS_SAMPLER7_BORDER_ALPHA
+0x0000A68C TD_VS_SAMPLER8_BORDER_ALPHA
+0x0000A69C TD_VS_SAMPLER9_BORDER_ALPHA
+0x0000A6AC TD_VS_SAMPLER10_BORDER_ALPHA
+0x0000A6BC TD_VS_SAMPLER11_BORDER_ALPHA
+0x0000A6CC TD_VS_SAMPLER12_BORDER_ALPHA
+0x0000A6DC TD_VS_SAMPLER13_BORDER_ALPHA
+0x0000A6EC TD_VS_SAMPLER14_BORDER_ALPHA
+0x0000A6FC TD_VS_SAMPLER15_BORDER_ALPHA
+0x0000A70C TD_VS_SAMPLER16_BORDER_ALPHA
+0x0000A71C TD_VS_SAMPLER17_BORDER_ALPHA
+0x0000A608 TD_VS_SAMPLER0_BORDER_BLUE
+0x0000A618 TD_VS_SAMPLER1_BORDER_BLUE
+0x0000A628 TD_VS_SAMPLER2_BORDER_BLUE
+0x0000A638 TD_VS_SAMPLER3_BORDER_BLUE
+0x0000A648 TD_VS_SAMPLER4_BORDER_BLUE
+0x0000A658 TD_VS_SAMPLER5_BORDER_BLUE
+0x0000A668 TD_VS_SAMPLER6_BORDER_BLUE
+0x0000A678 TD_VS_SAMPLER7_BORDER_BLUE
+0x0000A688 TD_VS_SAMPLER8_BORDER_BLUE
+0x0000A698 TD_VS_SAMPLER9_BORDER_BLUE
+0x0000A6A8 TD_VS_SAMPLER10_BORDER_BLUE
+0x0000A6B8 TD_VS_SAMPLER11_BORDER_BLUE
+0x0000A6C8 TD_VS_SAMPLER12_BORDER_BLUE
+0x0000A6D8 TD_VS_SAMPLER13_BORDER_BLUE
+0x0000A6E8 TD_VS_SAMPLER14_BORDER_BLUE
+0x0000A6F8 TD_VS_SAMPLER15_BORDER_BLUE
+0x0000A708 TD_VS_SAMPLER16_BORDER_BLUE
+0x0000A718 TD_VS_SAMPLER17_BORDER_BLUE
+0x0000A604 TD_VS_SAMPLER0_BORDER_GREEN
+0x0000A614 TD_VS_SAMPLER1_BORDER_GREEN
+0x0000A624 TD_VS_SAMPLER2_BORDER_GREEN
+0x0000A634 TD_VS_SAMPLER3_BORDER_GREEN
+0x0000A644 TD_VS_SAMPLER4_BORDER_GREEN
+0x0000A654 TD_VS_SAMPLER5_BORDER_GREEN
+0x0000A664 TD_VS_SAMPLER6_BORDER_GREEN
+0x0000A674 TD_VS_SAMPLER7_BORDER_GREEN
+0x0000A684 TD_VS_SAMPLER8_BORDER_GREEN
+0x0000A694 TD_VS_SAMPLER9_BORDER_GREEN
+0x0000A6A4 TD_VS_SAMPLER10_BORDER_GREEN
+0x0000A6B4 TD_VS_SAMPLER11_BORDER_GREEN
+0x0000A6C4 TD_VS_SAMPLER12_BORDER_GREEN
+0x0000A6D4 TD_VS_SAMPLER13_BORDER_GREEN
+0x0000A6E4 TD_VS_SAMPLER14_BORDER_GREEN
+0x0000A6F4 TD_VS_SAMPLER15_BORDER_GREEN
+0x0000A704 TD_VS_SAMPLER16_BORDER_GREEN
+0x0000A714 TD_VS_SAMPLER17_BORDER_GREEN
+0x0000A600 TD_VS_SAMPLER0_BORDER_RED
+0x0000A610 TD_VS_SAMPLER1_BORDER_RED
+0x0000A620 TD_VS_SAMPLER2_BORDER_RED
+0x0000A630 TD_VS_SAMPLER3_BORDER_RED
+0x0000A640 TD_VS_SAMPLER4_BORDER_RED
+0x0000A650 TD_VS_SAMPLER5_BORDER_RED
+0x0000A660 TD_VS_SAMPLER6_BORDER_RED
+0x0000A670 TD_VS_SAMPLER7_BORDER_RED
+0x0000A680 TD_VS_SAMPLER8_BORDER_RED
+0x0000A690 TD_VS_SAMPLER9_BORDER_RED
+0x0000A6A0 TD_VS_SAMPLER10_BORDER_RED
+0x0000A6B0 TD_VS_SAMPLER11_BORDER_RED
+0x0000A6C0 TD_VS_SAMPLER12_BORDER_RED
+0x0000A6D0 TD_VS_SAMPLER13_BORDER_RED
+0x0000A6E0 TD_VS_SAMPLER14_BORDER_RED
+0x0000A6F0 TD_VS_SAMPLER15_BORDER_RED
+0x0000A700 TD_VS_SAMPLER16_BORDER_RED
+0x0000A710 TD_VS_SAMPLER17_BORDER_RED
+0x00009508 TA_CNTL_AUX
+0x0002802C DB_DEPTH_CLEAR
+0x00028D24 DB_HTILE_SURFACE
+0x00028D34 DB_PREFETCH_LIMIT
+0x00028D30 DB_PRELOAD_CONTROL
+0x00028D0C DB_RENDER_CONTROL
+0x00028D10 DB_RENDER_OVERRIDE
+0x0002880C DB_SHADER_CONTROL
+0x00028D2C DB_SRESULTS_COMPARE_STATE1
+0x00028430 DB_STENCILREFMASK
+0x00028434 DB_STENCILREFMASK_BF
+0x00028028 DB_STENCIL_CLEAR
+0x00028780 CB_BLEND0_CONTROL
+0x00028784 CB_BLEND1_CONTROL
+0x00028788 CB_BLEND2_CONTROL
+0x0002878C CB_BLEND3_CONTROL
+0x00028790 CB_BLEND4_CONTROL
+0x00028794 CB_BLEND5_CONTROL
+0x00028798 CB_BLEND6_CONTROL
+0x0002879C CB_BLEND7_CONTROL
+0x00028804 CB_BLEND_CONTROL
+0x00028420 CB_BLEND_ALPHA
+0x0002841C CB_BLEND_BLUE
+0x00028418 CB_BLEND_GREEN
+0x00028414 CB_BLEND_RED
+0x0002812C CB_CLEAR_ALPHA
+0x00028128 CB_CLEAR_BLUE
+0x00028124 CB_CLEAR_GREEN
+0x00028120 CB_CLEAR_RED
+0x00028C30 CB_CLRCMP_CONTROL
+0x00028C38 CB_CLRCMP_DST
+0x00028C3C CB_CLRCMP_MSK
+0x00028C34 CB_CLRCMP_SRC
+0x00028100 CB_COLOR0_MASK
+0x00028104 CB_COLOR1_MASK
+0x00028108 CB_COLOR2_MASK
+0x0002810C CB_COLOR3_MASK
+0x00028110 CB_COLOR4_MASK
+0x00028114 CB_COLOR5_MASK
+0x00028118 CB_COLOR6_MASK
+0x0002811C CB_COLOR7_MASK
+0x00028080 CB_COLOR0_VIEW
+0x00028084 CB_COLOR1_VIEW
+0x00028088 CB_COLOR2_VIEW
+0x0002808C CB_COLOR3_VIEW
+0x00028090 CB_COLOR4_VIEW
+0x00028094 CB_COLOR5_VIEW
+0x00028098 CB_COLOR6_VIEW
+0x0002809C CB_COLOR7_VIEW
+0x00028808 CB_COLOR_CONTROL
+0x0002842C CB_FOG_BLUE
+0x00028428 CB_FOG_GREEN
+0x00028424 CB_FOG_RED
+0x00008040 WAIT_UNTIL
+0x00008950 CC_GC_SHADER_PIPE_CONFIG
+0x00008954 GC_USER_SHADER_PIPE_CONFIG
+0x00009714 VC_ENHANCE
+0x00009830 DB_DEBUG
+0x00009838 DB_WATERMARKS
+0x00028D28 DB_SRESULTS_COMPARE_STATE0
+0x00028D44 DB_ALPHA_TO_MASK
+0x00009504 TA_CNTL
+0x00009700 VC_CNTL
+0x00009718 VC_CONFIG
+0x0000A02C SMX_DC_MC_INTF_CTL
diff --git a/drivers/gpu/drm/radeon/reg_srcs/rs600 b/drivers/gpu/drm/radeon/reg_srcs/rs600
index 8e3c0b807add..6801b865d1c4 100644
--- a/drivers/gpu/drm/radeon/reg_srcs/rs600
+++ b/drivers/gpu/drm/radeon/reg_srcs/rs600
@@ -153,7 +153,7 @@ rs600 0x6d40
 0x42A4 SU_POLY_OFFSET_FRONT_SCALE
 0x42A8 SU_POLY_OFFSET_FRONT_OFFSET
 0x42AC SU_POLY_OFFSET_BACK_SCALE
-0x42B0 SU_POLY_OFFSET_BACK_OFFSET 
+0x42B0 SU_POLY_OFFSET_BACK_OFFSET
 0x42B4 SU_POLY_OFFSET_ENABLE
 0x42B8 SU_CULL_MODE
 0x42C0 SU_DEPTH_SCALE
@@ -291,6 +291,8 @@ rs600 0x6d40
 0x46AC US_OUT_FMT_2
 0x46B0 US_OUT_FMT_3
 0x46B4 US_W_FMT
+0x46B8 US_CODE_BANK
+0x46BC US_CODE_EXT
 0x46C0 US_ALU_RGB_ADDR_0
 0x46C4 US_ALU_RGB_ADDR_1
 0x46C8 US_ALU_RGB_ADDR_2
@@ -547,6 +549,70 @@ rs600 0x6d40
 0x4AB4 US_ALU_ALPHA_INST_61
 0x4AB8 US_ALU_ALPHA_INST_62
 0x4ABC US_ALU_ALPHA_INST_63
+0x4AC0 US_ALU_EXT_ADDR_0
+0x4AC4 US_ALU_EXT_ADDR_1
+0x4AC8 US_ALU_EXT_ADDR_2
+0x4ACC US_ALU_EXT_ADDR_3
+0x4AD0 US_ALU_EXT_ADDR_4
+0x4AD4 US_ALU_EXT_ADDR_5
+0x4AD8 US_ALU_EXT_ADDR_6
+0x4ADC US_ALU_EXT_ADDR_7
+0x4AE0 US_ALU_EXT_ADDR_8
+0x4AE4 US_ALU_EXT_ADDR_9
+0x4AE8 US_ALU_EXT_ADDR_10
+0x4AEC US_ALU_EXT_ADDR_11
+0x4AF0 US_ALU_EXT_ADDR_12
+0x4AF4 US_ALU_EXT_ADDR_13
+0x4AF8 US_ALU_EXT_ADDR_14
+0x4AFC US_ALU_EXT_ADDR_15
+0x4B00 US_ALU_EXT_ADDR_16
+0x4B04 US_ALU_EXT_ADDR_17
+0x4B08 US_ALU_EXT_ADDR_18
+0x4B0C US_ALU_EXT_ADDR_19
+0x4B10 US_ALU_EXT_ADDR_20
+0x4B14 US_ALU_EXT_ADDR_21
+0x4B18 US_ALU_EXT_ADDR_22
+0x4B1C US_ALU_EXT_ADDR_23
+0x4B20 US_ALU_EXT_ADDR_24
+0x4B24 US_ALU_EXT_ADDR_25
+0x4B28 US_ALU_EXT_ADDR_26
+0x4B2C US_ALU_EXT_ADDR_27
+0x4B30 US_ALU_EXT_ADDR_28
+0x4B34 US_ALU_EXT_ADDR_29
+0x4B38 US_ALU_EXT_ADDR_30
+0x4B3C US_ALU_EXT_ADDR_31
+0x4B40 US_ALU_EXT_ADDR_32
+0x4B44 US_ALU_EXT_ADDR_33
+0x4B48 US_ALU_EXT_ADDR_34
+0x4B4C US_ALU_EXT_ADDR_35
+0x4B50 US_ALU_EXT_ADDR_36
+0x4B54 US_ALU_EXT_ADDR_37
+0x4B58 US_ALU_EXT_ADDR_38
+0x4B5C US_ALU_EXT_ADDR_39
+0x4B60 US_ALU_EXT_ADDR_40
+0x4B64 US_ALU_EXT_ADDR_41
+0x4B68 US_ALU_EXT_ADDR_42
+0x4B6C US_ALU_EXT_ADDR_43
+0x4B70 US_ALU_EXT_ADDR_44
+0x4B74 US_ALU_EXT_ADDR_45
+0x4B78 US_ALU_EXT_ADDR_46
+0x4B7C US_ALU_EXT_ADDR_47
+0x4B80 US_ALU_EXT_ADDR_48
+0x4B84 US_ALU_EXT_ADDR_49
+0x4B88 US_ALU_EXT_ADDR_50
+0x4B8C US_ALU_EXT_ADDR_51
+0x4B90 US_ALU_EXT_ADDR_52
+0x4B94 US_ALU_EXT_ADDR_53
+0x4B98 US_ALU_EXT_ADDR_54
+0x4B9C US_ALU_EXT_ADDR_55
+0x4BA0 US_ALU_EXT_ADDR_56
+0x4BA4 US_ALU_EXT_ADDR_57
+0x4BA8 US_ALU_EXT_ADDR_58
+0x4BAC US_ALU_EXT_ADDR_59
+0x4BB0 US_ALU_EXT_ADDR_60
+0x4BB4 US_ALU_EXT_ADDR_61
+0x4BB8 US_ALU_EXT_ADDR_62
+0x4BBC US_ALU_EXT_ADDR_63
 0x4BC0 FG_FOG_BLEND
 0x4BC4 FG_FOG_FACTOR
 0x4BC8 FG_FOG_COLOR_R
diff --git a/drivers/gpu/drm/radeon/reg_srcs/rv515 b/drivers/gpu/drm/radeon/reg_srcs/rv515
index 0102a0d5735c..38abf63bf2cd 100644
--- a/drivers/gpu/drm/radeon/reg_srcs/rv515
+++ b/drivers/gpu/drm/radeon/reg_srcs/rv515
@@ -161,7 +161,12 @@ rv515 0x6d40
 0x401C GB_SELECT
 0x4020 GB_AA_CONFIG
 0x4024 GB_FIFO_SIZE
+0x4028 GB_Z_PEQ_CONFIG
 0x4100 TX_INVALTAGS
+0x4114 SU_TEX_WRAP_PS3
+0x4118 PS3_ENABLE
+0x411c PS3_VTX_FMT
+0x4120 PS3_TEX_SOURCE
 0x4200 GA_POINT_S0
 0x4204 GA_POINT_T0
 0x4208 GA_POINT_S1
@@ -171,6 +176,7 @@ rv515 0x6d40
 0x4230 GA_POINT_MINMAX
 0x4234 GA_LINE_CNTL
 0x4238 GA_LINE_STIPPLE_CONFIG
+0x4258 GA_COLOR_CONTROL_PS3
 0x4260 GA_LINE_STIPPLE_VALUE
 0x4264 GA_LINE_S0
 0x4268 GA_LINE_S1
diff --git a/drivers/gpu/drm/radeon/rs100d.h b/drivers/gpu/drm/radeon/rs100d.h
new file mode 100644
index 000000000000..48a913a06cfd
--- /dev/null
+++ b/drivers/gpu/drm/radeon/rs100d.h
@@ -0,0 +1,40 @@
+/*
+ * Copyright 2008 Advanced Micro Devices, Inc.
+ * Copyright 2008 Red Hat Inc.
+ * Copyright 2009 Jerome Glisse.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Dave Airlie
+ *          Alex Deucher
+ *          Jerome Glisse
+ */
+#ifndef __RS100D_H__
+#define __RS100D_H__
+
+/* Registers */
+#define R_00015C_NB_TOM                              0x00015C
+#define   S_00015C_MC_FB_START(x)                      (((x) & 0xFFFF) << 0)
+#define   G_00015C_MC_FB_START(x)                      (((x) >> 0) & 0xFFFF)
+#define   C_00015C_MC_FB_START                         0xFFFF0000
+#define   S_00015C_MC_FB_TOP(x)                        (((x) & 0xFFFF) << 16)
+#define   G_00015C_MC_FB_TOP(x)                        (((x) >> 16) & 0xFFFF)
+#define   C_00015C_MC_FB_TOP                           0x0000FFFF
+
+#endif
diff --git a/drivers/gpu/drm/radeon/rs400.c b/drivers/gpu/drm/radeon/rs400.c
index a3fbdad938c7..626d51891ee9 100644
--- a/drivers/gpu/drm/radeon/rs400.c
+++ b/drivers/gpu/drm/radeon/rs400.c
@@ -27,27 +27,12 @@
  */
 #include <linux/seq_file.h>
 #include <drm/drmP.h>
-#include "radeon_reg.h"
 #include "radeon.h"
+#include "rs400d.h"
 
-/* rs400,rs480 depends on : */
-void r100_hdp_reset(struct radeon_device *rdev);
-void r100_mc_disable_clients(struct radeon_device *rdev);
-int r300_mc_wait_for_idle(struct radeon_device *rdev);
-void r420_pipes_init(struct radeon_device *rdev);
+/* This files gather functions specifics to : rs400,rs480 */
+static int rs400_debugfs_pcie_gart_info_init(struct radeon_device *rdev);
 
-/* This files gather functions specifics to :
- * rs400,rs480
- *
- * Some of these functions might be used by newer ASICs.
- */
-void rs400_gpu_init(struct radeon_device *rdev);
-int rs400_debugfs_pcie_gart_info_init(struct radeon_device *rdev);
-
-
-/*
- * GART functions.
- */
 void rs400_gart_adjust_size(struct radeon_device *rdev)
 {
 	/* Check gart size */
@@ -128,6 +113,7 @@ int rs400_gart_enable(struct radeon_device *rdev)
 	uint32_t size_reg;
 	uint32_t tmp;
 
+	radeon_gart_restore(rdev);
 	tmp = RREG32_MC(RS690_AIC_CTRL_SCRATCH);
 	tmp |= RS690_DIS_OUT_OF_PCI_GART_ACCESS;
 	WREG32_MC(RS690_AIC_CTRL_SCRATCH, tmp);
@@ -165,9 +151,8 @@ int rs400_gart_enable(struct radeon_device *rdev)
 		WREG32(RADEON_AGP_BASE, 0xFFFFFFFF);
 		WREG32(RS480_AGP_BASE_2, 0);
 	}
-	tmp = rdev->mc.gtt_location + rdev->mc.gtt_size - 1;
-	tmp = REG_SET(RS690_MC_AGP_TOP, tmp >> 16);
-	tmp |= REG_SET(RS690_MC_AGP_START, rdev->mc.gtt_location >> 16);
+	tmp = REG_SET(RS690_MC_AGP_TOP, rdev->mc.gtt_end >> 16);
+	tmp |= REG_SET(RS690_MC_AGP_START, rdev->mc.gtt_start >> 16);
 	if ((rdev->family == CHIP_RS690) || (rdev->family == CHIP_RS740)) {
 		WREG32_MC(RS690_MCCFG_AGP_LOCATION, tmp);
 		tmp = RREG32(RADEON_BUS_CNTL) & ~RS600_BUS_MASTER_DIS;
@@ -238,59 +223,20 @@ int rs400_gart_set_page(struct radeon_device *rdev, int i, uint64_t addr)
 	return 0;
 }
 
-
-/*
- * MC functions.
- */
-int rs400_mc_init(struct radeon_device *rdev)
+int rs400_mc_wait_for_idle(struct radeon_device *rdev)
 {
+	unsigned i;
 	uint32_t tmp;
-	int r;
 
-	if (r100_debugfs_rbbm_init(rdev)) {
-		DRM_ERROR("Failed to register debugfs file for RBBM !\n");
-	}
-
-	rs400_gpu_init(rdev);
-	rs400_gart_disable(rdev);
-	rdev->mc.gtt_location = rdev->mc.mc_vram_size;
-	rdev->mc.gtt_location += (rdev->mc.gtt_size - 1);
-	rdev->mc.gtt_location &= ~(rdev->mc.gtt_size - 1);
-	r = radeon_mc_setup(rdev);
-	if (r) {
-		return r;
-	}
-
-	r100_mc_disable_clients(rdev);
-	if (r300_mc_wait_for_idle(rdev)) {
-		printk(KERN_WARNING "Failed to wait MC idle while "
-		       "programming pipes. Bad things might happen.\n");
+	for (i = 0; i < rdev->usec_timeout; i++) {
+		/* read MC_STATUS */
+		tmp = RREG32(0x0150);
+		if (tmp & (1 << 2)) {
+			return 0;
+		}
+		DRM_UDELAY(1);
 	}
-
-	tmp = rdev->mc.vram_location + rdev->mc.mc_vram_size - 1;
-	tmp = REG_SET(RADEON_MC_FB_TOP, tmp >> 16);
-	tmp |= REG_SET(RADEON_MC_FB_START, rdev->mc.vram_location >> 16);
-	WREG32(RADEON_MC_FB_LOCATION, tmp);
-	tmp = RREG32(RADEON_HOST_PATH_CNTL) | RADEON_HP_LIN_RD_CACHE_DIS;
-	WREG32(RADEON_HOST_PATH_CNTL, tmp | RADEON_HDP_SOFT_RESET | RADEON_HDP_READ_BUFFER_INVALIDATE);
-	(void)RREG32(RADEON_HOST_PATH_CNTL);
-	WREG32(RADEON_HOST_PATH_CNTL, tmp);
-	(void)RREG32(RADEON_HOST_PATH_CNTL);
-
-	return 0;
-}
-
-void rs400_mc_fini(struct radeon_device *rdev)
-{
-}
-
-
-/*
- * Global GPU functions
- */
-void rs400_errata(struct radeon_device *rdev)
-{
-	rdev->pll_errata = 0;
+	return -1;
 }
 
 void rs400_gpu_init(struct radeon_device *rdev)
@@ -299,30 +245,27 @@ void rs400_gpu_init(struct radeon_device *rdev)
 	r100_hdp_reset(rdev);
 	/* FIXME: is this correct ? */
 	r420_pipes_init(rdev);
-	if (r300_mc_wait_for_idle(rdev)) {
-		printk(KERN_WARNING "Failed to wait MC idle while "
-		       "programming pipes. Bad things might happen.\n");
+	if (rs400_mc_wait_for_idle(rdev)) {
+		printk(KERN_WARNING "rs400: Failed to wait MC idle while "
+		       "programming pipes. Bad things might happen. %08x\n", RREG32(0x150));
 	}
 }
 
-
-/*
- * VRAM info.
- */
-void rs400_vram_info(struct radeon_device *rdev)
+void rs400_mc_init(struct radeon_device *rdev)
 {
+	u64 base;
+
 	rs400_gart_adjust_size(rdev);
+	rdev->mc.igp_sideport_enabled = radeon_combios_sideport_present(rdev);
 	/* DDR for all card after R300 & IGP */
 	rdev->mc.vram_is_ddr = true;
 	rdev->mc.vram_width = 128;
-
 	r100_vram_init_sizes(rdev);
+	base = (RREG32(RADEON_NB_TOM) & 0xffff) << 16;
+	radeon_vram_location(rdev, &rdev->mc, base);
+	radeon_gtt_location(rdev, &rdev->mc);
 }
 
-
-/*
- * Indirect registers accessor
- */
 uint32_t rs400_mc_rreg(struct radeon_device *rdev, uint32_t reg)
 {
 	uint32_t r;
@@ -340,10 +283,6 @@ void rs400_mc_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v)
 	WREG32(RS480_NB_MC_INDEX, 0xff);
 }
 
-
-/*
- * Debugfs info
- */
 #if defined(CONFIG_DEBUG_FS)
 static int rs400_debugfs_gart_info(struct seq_file *m, void *data)
 {
@@ -419,7 +358,7 @@ static struct drm_info_list rs400_gart_info_list[] = {
 };
 #endif
 
-int rs400_debugfs_pcie_gart_info_init(struct radeon_device *rdev)
+static int rs400_debugfs_pcie_gart_info_init(struct radeon_device *rdev)
 {
 #if defined(CONFIG_DEBUG_FS)
 	return radeon_debugfs_add_files(rdev, rs400_gart_info_list, 1);
@@ -427,3 +366,173 @@ int rs400_debugfs_pcie_gart_info_init(struct radeon_device *rdev)
 	return 0;
 #endif
 }
+
+void rs400_mc_program(struct radeon_device *rdev)
+{
+	struct r100_mc_save save;
+
+	/* Stops all mc clients */
+	r100_mc_stop(rdev, &save);
+
+	/* Wait for mc idle */
+	if (rs400_mc_wait_for_idle(rdev))
+		dev_warn(rdev->dev, "rs400: Wait MC idle timeout before updating MC.\n");
+	WREG32(R_000148_MC_FB_LOCATION,
+		S_000148_MC_FB_START(rdev->mc.vram_start >> 16) |
+		S_000148_MC_FB_TOP(rdev->mc.vram_end >> 16));
+
+	r100_mc_resume(rdev, &save);
+}
+
+static int rs400_startup(struct radeon_device *rdev)
+{
+	int r;
+
+	rs400_mc_program(rdev);
+	/* Resume clock */
+	r300_clock_startup(rdev);
+	/* Initialize GPU configuration (# pipes, ...) */
+	rs400_gpu_init(rdev);
+	r100_enable_bm(rdev);
+	/* Initialize GART (initialize after TTM so we can allocate
+	 * memory through TTM but finalize after TTM) */
+	r = rs400_gart_enable(rdev);
+	if (r)
+		return r;
+	/* Enable IRQ */
+	r100_irq_set(rdev);
+	rdev->config.r300.hdp_cntl = RREG32(RADEON_HOST_PATH_CNTL);
+	/* 1M ring buffer */
+	r = r100_cp_init(rdev, 1024 * 1024);
+	if (r) {
+		dev_err(rdev->dev, "failled initializing CP (%d).\n", r);
+		return r;
+	}
+	r = r100_wb_init(rdev);
+	if (r)
+		dev_err(rdev->dev, "failled initializing WB (%d).\n", r);
+	r = r100_ib_init(rdev);
+	if (r) {
+		dev_err(rdev->dev, "failled initializing IB (%d).\n", r);
+		return r;
+	}
+	return 0;
+}
+
+int rs400_resume(struct radeon_device *rdev)
+{
+	/* Make sur GART are not working */
+	rs400_gart_disable(rdev);
+	/* Resume clock before doing reset */
+	r300_clock_startup(rdev);
+	/* setup MC before calling post tables */
+	rs400_mc_program(rdev);
+	/* Reset gpu before posting otherwise ATOM will enter infinite loop */
+	if (radeon_gpu_reset(rdev)) {
+		dev_warn(rdev->dev, "GPU reset failed ! (0xE40=0x%08X, 0x7C0=0x%08X)\n",
+			RREG32(R_000E40_RBBM_STATUS),
+			RREG32(R_0007C0_CP_STAT));
+	}
+	/* post */
+	radeon_combios_asic_init(rdev->ddev);
+	/* Resume clock after posting */
+	r300_clock_startup(rdev);
+	/* Initialize surface registers */
+	radeon_surface_init(rdev);
+	return rs400_startup(rdev);
+}
+
+int rs400_suspend(struct radeon_device *rdev)
+{
+	r100_cp_disable(rdev);
+	r100_wb_disable(rdev);
+	r100_irq_disable(rdev);
+	rs400_gart_disable(rdev);
+	return 0;
+}
+
+void rs400_fini(struct radeon_device *rdev)
+{
+	r100_cp_fini(rdev);
+	r100_wb_fini(rdev);
+	r100_ib_fini(rdev);
+	radeon_gem_fini(rdev);
+	rs400_gart_fini(rdev);
+	radeon_irq_kms_fini(rdev);
+	radeon_fence_driver_fini(rdev);
+	radeon_bo_fini(rdev);
+	radeon_atombios_fini(rdev);
+	kfree(rdev->bios);
+	rdev->bios = NULL;
+}
+
+int rs400_init(struct radeon_device *rdev)
+{
+	int r;
+
+	/* Disable VGA */
+	r100_vga_render_disable(rdev);
+	/* Initialize scratch registers */
+	radeon_scratch_init(rdev);
+	/* Initialize surface registers */
+	radeon_surface_init(rdev);
+	/* TODO: disable VGA need to use VGA request */
+	/* BIOS*/
+	if (!radeon_get_bios(rdev)) {
+		if (ASIC_IS_AVIVO(rdev))
+			return -EINVAL;
+	}
+	if (rdev->is_atom_bios) {
+		dev_err(rdev->dev, "Expecting combios for RS400/RS480 GPU\n");
+		return -EINVAL;
+	} else {
+		r = radeon_combios_init(rdev);
+		if (r)
+			return r;
+	}
+	/* Reset gpu before posting otherwise ATOM will enter infinite loop */
+	if (radeon_gpu_reset(rdev)) {
+		dev_warn(rdev->dev,
+			"GPU reset failed ! (0xE40=0x%08X, 0x7C0=0x%08X)\n",
+			RREG32(R_000E40_RBBM_STATUS),
+			RREG32(R_0007C0_CP_STAT));
+	}
+	/* check if cards are posted or not */
+	if (radeon_boot_test_post_card(rdev) == false)
+		return -EINVAL;
+
+	/* Initialize clocks */
+	radeon_get_clock_info(rdev->ddev);
+	/* Initialize power management */
+	radeon_pm_init(rdev);
+	/* initialize memory controller */
+	rs400_mc_init(rdev);
+	/* Fence driver */
+	r = radeon_fence_driver_init(rdev);
+	if (r)
+		return r;
+	r = radeon_irq_kms_init(rdev);
+	if (r)
+		return r;
+	/* Memory manager */
+	r = radeon_bo_init(rdev);
+	if (r)
+		return r;
+	r = rs400_gart_init(rdev);
+	if (r)
+		return r;
+	r300_set_reg_safe(rdev);
+	rdev->accel_working = true;
+	r = rs400_startup(rdev);
+	if (r) {
+		/* Somethings want wront with the accel init stop accel */
+		dev_err(rdev->dev, "Disabling GPU acceleration\n");
+		r100_cp_fini(rdev);
+		r100_wb_fini(rdev);
+		r100_ib_fini(rdev);
+		rs400_gart_fini(rdev);
+		radeon_irq_kms_fini(rdev);
+		rdev->accel_working = false;
+	}
+	return 0;
+}
diff --git a/drivers/gpu/drm/radeon/rs400d.h b/drivers/gpu/drm/radeon/rs400d.h
new file mode 100644
index 000000000000..6d8bac58ced9
--- /dev/null
+++ b/drivers/gpu/drm/radeon/rs400d.h
@@ -0,0 +1,160 @@
+/*
+ * Copyright 2008 Advanced Micro Devices, Inc.
+ * Copyright 2008 Red Hat Inc.
+ * Copyright 2009 Jerome Glisse.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Dave Airlie
+ *          Alex Deucher
+ *          Jerome Glisse
+ */
+#ifndef __RS400D_H__
+#define __RS400D_H__
+
+/* Registers */
+#define R_000148_MC_FB_LOCATION                      0x000148
+#define   S_000148_MC_FB_START(x)                      (((x) & 0xFFFF) << 0)
+#define   G_000148_MC_FB_START(x)                      (((x) >> 0) & 0xFFFF)
+#define   C_000148_MC_FB_START                         0xFFFF0000
+#define   S_000148_MC_FB_TOP(x)                        (((x) & 0xFFFF) << 16)
+#define   G_000148_MC_FB_TOP(x)                        (((x) >> 16) & 0xFFFF)
+#define   C_000148_MC_FB_TOP                           0x0000FFFF
+#define R_00015C_NB_TOM                              0x00015C
+#define   S_00015C_MC_FB_START(x)                      (((x) & 0xFFFF) << 0)
+#define   G_00015C_MC_FB_START(x)                      (((x) >> 0) & 0xFFFF)
+#define   C_00015C_MC_FB_START                         0xFFFF0000
+#define   S_00015C_MC_FB_TOP(x)                        (((x) & 0xFFFF) << 16)
+#define   G_00015C_MC_FB_TOP(x)                        (((x) >> 16) & 0xFFFF)
+#define   C_00015C_MC_FB_TOP                           0x0000FFFF
+#define R_0007C0_CP_STAT                             0x0007C0
+#define   S_0007C0_MRU_BUSY(x)                         (((x) & 0x1) << 0)
+#define   G_0007C0_MRU_BUSY(x)                         (((x) >> 0) & 0x1)
+#define   C_0007C0_MRU_BUSY                            0xFFFFFFFE
+#define   S_0007C0_MWU_BUSY(x)                         (((x) & 0x1) << 1)
+#define   G_0007C0_MWU_BUSY(x)                         (((x) >> 1) & 0x1)
+#define   C_0007C0_MWU_BUSY                            0xFFFFFFFD
+#define   S_0007C0_RSIU_BUSY(x)                        (((x) & 0x1) << 2)
+#define   G_0007C0_RSIU_BUSY(x)                        (((x) >> 2) & 0x1)
+#define   C_0007C0_RSIU_BUSY                           0xFFFFFFFB
+#define   S_0007C0_RCIU_BUSY(x)                        (((x) & 0x1) << 3)
+#define   G_0007C0_RCIU_BUSY(x)                        (((x) >> 3) & 0x1)
+#define   C_0007C0_RCIU_BUSY                           0xFFFFFFF7
+#define   S_0007C0_CSF_PRIMARY_BUSY(x)                 (((x) & 0x1) << 9)
+#define   G_0007C0_CSF_PRIMARY_BUSY(x)                 (((x) >> 9) & 0x1)
+#define   C_0007C0_CSF_PRIMARY_BUSY                    0xFFFFFDFF
+#define   S_0007C0_CSF_INDIRECT_BUSY(x)                (((x) & 0x1) << 10)
+#define   G_0007C0_CSF_INDIRECT_BUSY(x)                (((x) >> 10) & 0x1)
+#define   C_0007C0_CSF_INDIRECT_BUSY                   0xFFFFFBFF
+#define   S_0007C0_CSQ_PRIMARY_BUSY(x)                 (((x) & 0x1) << 11)
+#define   G_0007C0_CSQ_PRIMARY_BUSY(x)                 (((x) >> 11) & 0x1)
+#define   C_0007C0_CSQ_PRIMARY_BUSY                    0xFFFFF7FF
+#define   S_0007C0_CSQ_INDIRECT_BUSY(x)                (((x) & 0x1) << 12)
+#define   G_0007C0_CSQ_INDIRECT_BUSY(x)                (((x) >> 12) & 0x1)
+#define   C_0007C0_CSQ_INDIRECT_BUSY                   0xFFFFEFFF
+#define   S_0007C0_CSI_BUSY(x)                         (((x) & 0x1) << 13)
+#define   G_0007C0_CSI_BUSY(x)                         (((x) >> 13) & 0x1)
+#define   C_0007C0_CSI_BUSY                            0xFFFFDFFF
+#define   S_0007C0_CSF_INDIRECT2_BUSY(x)               (((x) & 0x1) << 14)
+#define   G_0007C0_CSF_INDIRECT2_BUSY(x)               (((x) >> 14) & 0x1)
+#define   C_0007C0_CSF_INDIRECT2_BUSY                  0xFFFFBFFF
+#define   S_0007C0_CSQ_INDIRECT2_BUSY(x)               (((x) & 0x1) << 15)
+#define   G_0007C0_CSQ_INDIRECT2_BUSY(x)               (((x) >> 15) & 0x1)
+#define   C_0007C0_CSQ_INDIRECT2_BUSY                  0xFFFF7FFF
+#define   S_0007C0_GUIDMA_BUSY(x)                      (((x) & 0x1) << 28)
+#define   G_0007C0_GUIDMA_BUSY(x)                      (((x) >> 28) & 0x1)
+#define   C_0007C0_GUIDMA_BUSY                         0xEFFFFFFF
+#define   S_0007C0_VIDDMA_BUSY(x)                      (((x) & 0x1) << 29)
+#define   G_0007C0_VIDDMA_BUSY(x)                      (((x) >> 29) & 0x1)
+#define   C_0007C0_VIDDMA_BUSY                         0xDFFFFFFF
+#define   S_0007C0_CMDSTRM_BUSY(x)                     (((x) & 0x1) << 30)
+#define   G_0007C0_CMDSTRM_BUSY(x)                     (((x) >> 30) & 0x1)
+#define   C_0007C0_CMDSTRM_BUSY                        0xBFFFFFFF
+#define   S_0007C0_CP_BUSY(x)                          (((x) & 0x1) << 31)
+#define   G_0007C0_CP_BUSY(x)                          (((x) >> 31) & 0x1)
+#define   C_0007C0_CP_BUSY                             0x7FFFFFFF
+#define R_000E40_RBBM_STATUS                         0x000E40
+#define   S_000E40_CMDFIFO_AVAIL(x)                    (((x) & 0x7F) << 0)
+#define   G_000E40_CMDFIFO_AVAIL(x)                    (((x) >> 0) & 0x7F)
+#define   C_000E40_CMDFIFO_AVAIL                       0xFFFFFF80
+#define   S_000E40_HIRQ_ON_RBB(x)                      (((x) & 0x1) << 8)
+#define   G_000E40_HIRQ_ON_RBB(x)                      (((x) >> 8) & 0x1)
+#define   C_000E40_HIRQ_ON_RBB                         0xFFFFFEFF
+#define   S_000E40_CPRQ_ON_RBB(x)                      (((x) & 0x1) << 9)
+#define   G_000E40_CPRQ_ON_RBB(x)                      (((x) >> 9) & 0x1)
+#define   C_000E40_CPRQ_ON_RBB                         0xFFFFFDFF
+#define   S_000E40_CFRQ_ON_RBB(x)                      (((x) & 0x1) << 10)
+#define   G_000E40_CFRQ_ON_RBB(x)                      (((x) >> 10) & 0x1)
+#define   C_000E40_CFRQ_ON_RBB                         0xFFFFFBFF
+#define   S_000E40_HIRQ_IN_RTBUF(x)                    (((x) & 0x1) << 11)
+#define   G_000E40_HIRQ_IN_RTBUF(x)                    (((x) >> 11) & 0x1)
+#define   C_000E40_HIRQ_IN_RTBUF                       0xFFFFF7FF
+#define   S_000E40_CPRQ_IN_RTBUF(x)                    (((x) & 0x1) << 12)
+#define   G_000E40_CPRQ_IN_RTBUF(x)                    (((x) >> 12) & 0x1)
+#define   C_000E40_CPRQ_IN_RTBUF                       0xFFFFEFFF
+#define   S_000E40_CFRQ_IN_RTBUF(x)                    (((x) & 0x1) << 13)
+#define   G_000E40_CFRQ_IN_RTBUF(x)                    (((x) >> 13) & 0x1)
+#define   C_000E40_CFRQ_IN_RTBUF                       0xFFFFDFFF
+#define   S_000E40_CF_PIPE_BUSY(x)                     (((x) & 0x1) << 14)
+#define   G_000E40_CF_PIPE_BUSY(x)                     (((x) >> 14) & 0x1)
+#define   C_000E40_CF_PIPE_BUSY                        0xFFFFBFFF
+#define   S_000E40_ENG_EV_BUSY(x)                      (((x) & 0x1) << 15)
+#define   G_000E40_ENG_EV_BUSY(x)                      (((x) >> 15) & 0x1)
+#define   C_000E40_ENG_EV_BUSY                         0xFFFF7FFF
+#define   S_000E40_CP_CMDSTRM_BUSY(x)                  (((x) & 0x1) << 16)
+#define   G_000E40_CP_CMDSTRM_BUSY(x)                  (((x) >> 16) & 0x1)
+#define   C_000E40_CP_CMDSTRM_BUSY                     0xFFFEFFFF
+#define   S_000E40_E2_BUSY(x)                          (((x) & 0x1) << 17)
+#define   G_000E40_E2_BUSY(x)                          (((x) >> 17) & 0x1)
+#define   C_000E40_E2_BUSY                             0xFFFDFFFF
+#define   S_000E40_RB2D_BUSY(x)                        (((x) & 0x1) << 18)
+#define   G_000E40_RB2D_BUSY(x)                        (((x) >> 18) & 0x1)
+#define   C_000E40_RB2D_BUSY                           0xFFFBFFFF
+#define   S_000E40_RB3D_BUSY(x)                        (((x) & 0x1) << 19)
+#define   G_000E40_RB3D_BUSY(x)                        (((x) >> 19) & 0x1)
+#define   C_000E40_RB3D_BUSY                           0xFFF7FFFF
+#define   S_000E40_VAP_BUSY(x)                         (((x) & 0x1) << 20)
+#define   G_000E40_VAP_BUSY(x)                         (((x) >> 20) & 0x1)
+#define   C_000E40_VAP_BUSY                            0xFFEFFFFF
+#define   S_000E40_RE_BUSY(x)                          (((x) & 0x1) << 21)
+#define   G_000E40_RE_BUSY(x)                          (((x) >> 21) & 0x1)
+#define   C_000E40_RE_BUSY                             0xFFDFFFFF
+#define   S_000E40_TAM_BUSY(x)                         (((x) & 0x1) << 22)
+#define   G_000E40_TAM_BUSY(x)                         (((x) >> 22) & 0x1)
+#define   C_000E40_TAM_BUSY                            0xFFBFFFFF
+#define   S_000E40_TDM_BUSY(x)                         (((x) & 0x1) << 23)
+#define   G_000E40_TDM_BUSY(x)                         (((x) >> 23) & 0x1)
+#define   C_000E40_TDM_BUSY                            0xFF7FFFFF
+#define   S_000E40_PB_BUSY(x)                          (((x) & 0x1) << 24)
+#define   G_000E40_PB_BUSY(x)                          (((x) >> 24) & 0x1)
+#define   C_000E40_PB_BUSY                             0xFEFFFFFF
+#define   S_000E40_TIM_BUSY(x)                         (((x) & 0x1) << 25)
+#define   G_000E40_TIM_BUSY(x)                         (((x) >> 25) & 0x1)
+#define   C_000E40_TIM_BUSY                            0xFDFFFFFF
+#define   S_000E40_GA_BUSY(x)                          (((x) & 0x1) << 26)
+#define   G_000E40_GA_BUSY(x)                          (((x) >> 26) & 0x1)
+#define   C_000E40_GA_BUSY                             0xFBFFFFFF
+#define   S_000E40_CBA2D_BUSY(x)                       (((x) & 0x1) << 27)
+#define   G_000E40_CBA2D_BUSY(x)                       (((x) >> 27) & 0x1)
+#define   C_000E40_CBA2D_BUSY                          0xF7FFFFFF
+#define   S_000E40_GUI_ACTIVE(x)                       (((x) & 0x1) << 31)
+#define   G_000E40_GUI_ACTIVE(x)                       (((x) >> 31) & 0x1)
+#define   C_000E40_GUI_ACTIVE                          0x7FFFFFFF
+
+#endif
diff --git a/drivers/gpu/drm/radeon/rs600.c b/drivers/gpu/drm/radeon/rs600.c
index 0e791e26def3..47f046b78c6b 100644
--- a/drivers/gpu/drm/radeon/rs600.c
+++ b/drivers/gpu/drm/radeon/rs600.c
@@ -25,28 +25,126 @@
  *          Alex Deucher
  *          Jerome Glisse
  */
+/* RS600 / Radeon X1250/X1270 integrated GPU
+ *
+ * This file gather function specific to RS600 which is the IGP of
+ * the X1250/X1270 family supporting intel CPU (while RS690/RS740
+ * is the X1250/X1270 supporting AMD CPU). The display engine are
+ * the avivo one, bios is an atombios, 3D block are the one of the
+ * R4XX family. The GART is different from the RS400 one and is very
+ * close to the one of the R600 family (R600 likely being an evolution
+ * of the RS600 GART block).
+ */
 #include "drmP.h"
-#include "radeon_reg.h"
 #include "radeon.h"
-#include "avivod.h"
+#include "atom.h"
+#include "rs600d.h"
 
 #include "rs600_reg_safe.h"
 
-/* rs600 depends on : */
-void r100_hdp_reset(struct radeon_device *rdev);
-int r100_gui_wait_for_idle(struct radeon_device *rdev);
-int r300_mc_wait_for_idle(struct radeon_device *rdev);
-void r420_pipes_init(struct radeon_device *rdev);
-
-/* This files gather functions specifics to :
- * rs600
- *
- * Some of these functions might be used by newer ASICs.
- */
 void rs600_gpu_init(struct radeon_device *rdev);
 int rs600_mc_wait_for_idle(struct radeon_device *rdev);
-void rs600_disable_vga(struct radeon_device *rdev);
 
+/* hpd for digital panel detect/disconnect */
+bool rs600_hpd_sense(struct radeon_device *rdev, enum radeon_hpd_id hpd)
+{
+	u32 tmp;
+	bool connected = false;
+
+	switch (hpd) {
+	case RADEON_HPD_1:
+		tmp = RREG32(R_007D04_DC_HOT_PLUG_DETECT1_INT_STATUS);
+		if (G_007D04_DC_HOT_PLUG_DETECT1_SENSE(tmp))
+			connected = true;
+		break;
+	case RADEON_HPD_2:
+		tmp = RREG32(R_007D14_DC_HOT_PLUG_DETECT2_INT_STATUS);
+		if (G_007D14_DC_HOT_PLUG_DETECT2_SENSE(tmp))
+			connected = true;
+		break;
+	default:
+		break;
+	}
+	return connected;
+}
+
+void rs600_hpd_set_polarity(struct radeon_device *rdev,
+			    enum radeon_hpd_id hpd)
+{
+	u32 tmp;
+	bool connected = rs600_hpd_sense(rdev, hpd);
+
+	switch (hpd) {
+	case RADEON_HPD_1:
+		tmp = RREG32(R_007D08_DC_HOT_PLUG_DETECT1_INT_CONTROL);
+		if (connected)
+			tmp &= ~S_007D08_DC_HOT_PLUG_DETECT1_INT_POLARITY(1);
+		else
+			tmp |= S_007D08_DC_HOT_PLUG_DETECT1_INT_POLARITY(1);
+		WREG32(R_007D08_DC_HOT_PLUG_DETECT1_INT_CONTROL, tmp);
+		break;
+	case RADEON_HPD_2:
+		tmp = RREG32(R_007D18_DC_HOT_PLUG_DETECT2_INT_CONTROL);
+		if (connected)
+			tmp &= ~S_007D18_DC_HOT_PLUG_DETECT2_INT_POLARITY(1);
+		else
+			tmp |= S_007D18_DC_HOT_PLUG_DETECT2_INT_POLARITY(1);
+		WREG32(R_007D18_DC_HOT_PLUG_DETECT2_INT_CONTROL, tmp);
+		break;
+	default:
+		break;
+	}
+}
+
+void rs600_hpd_init(struct radeon_device *rdev)
+{
+	struct drm_device *dev = rdev->ddev;
+	struct drm_connector *connector;
+
+	list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
+		struct radeon_connector *radeon_connector = to_radeon_connector(connector);
+		switch (radeon_connector->hpd.hpd) {
+		case RADEON_HPD_1:
+			WREG32(R_007D00_DC_HOT_PLUG_DETECT1_CONTROL,
+			       S_007D00_DC_HOT_PLUG_DETECT1_EN(1));
+			rdev->irq.hpd[0] = true;
+			break;
+		case RADEON_HPD_2:
+			WREG32(R_007D10_DC_HOT_PLUG_DETECT2_CONTROL,
+			       S_007D10_DC_HOT_PLUG_DETECT2_EN(1));
+			rdev->irq.hpd[1] = true;
+			break;
+		default:
+			break;
+		}
+	}
+	if (rdev->irq.installed)
+		rs600_irq_set(rdev);
+}
+
+void rs600_hpd_fini(struct radeon_device *rdev)
+{
+	struct drm_device *dev = rdev->ddev;
+	struct drm_connector *connector;
+
+	list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
+		struct radeon_connector *radeon_connector = to_radeon_connector(connector);
+		switch (radeon_connector->hpd.hpd) {
+		case RADEON_HPD_1:
+			WREG32(R_007D00_DC_HOT_PLUG_DETECT1_CONTROL,
+			       S_007D00_DC_HOT_PLUG_DETECT1_EN(0));
+			rdev->irq.hpd[0] = false;
+			break;
+		case RADEON_HPD_2:
+			WREG32(R_007D10_DC_HOT_PLUG_DETECT2_CONTROL,
+			       S_007D10_DC_HOT_PLUG_DETECT2_EN(0));
+			rdev->irq.hpd[1] = false;
+			break;
+		default:
+			break;
+		}
+	}
+}
 
 /*
  * GART.
@@ -55,18 +153,18 @@ void rs600_gart_tlb_flush(struct radeon_device *rdev)
 {
 	uint32_t tmp;
 
-	tmp = RREG32_MC(RS600_MC_PT0_CNTL);
-	tmp &= ~(RS600_INVALIDATE_ALL_L1_TLBS | RS600_INVALIDATE_L2_CACHE);
-	WREG32_MC(RS600_MC_PT0_CNTL, tmp);
+	tmp = RREG32_MC(R_000100_MC_PT0_CNTL);
+	tmp &= C_000100_INVALIDATE_ALL_L1_TLBS & C_000100_INVALIDATE_L2_CACHE;
+	WREG32_MC(R_000100_MC_PT0_CNTL, tmp);
 
-	tmp = RREG32_MC(RS600_MC_PT0_CNTL);
-	tmp |= RS600_INVALIDATE_ALL_L1_TLBS | RS600_INVALIDATE_L2_CACHE;
-	WREG32_MC(RS600_MC_PT0_CNTL, tmp);
+	tmp = RREG32_MC(R_000100_MC_PT0_CNTL);
+	tmp |= S_000100_INVALIDATE_ALL_L1_TLBS(1) & S_000100_INVALIDATE_L2_CACHE(1);
+	WREG32_MC(R_000100_MC_PT0_CNTL, tmp);
 
-	tmp = RREG32_MC(RS600_MC_PT0_CNTL);
-	tmp &= ~(RS600_INVALIDATE_ALL_L1_TLBS | RS600_INVALIDATE_L2_CACHE);
-	WREG32_MC(RS600_MC_PT0_CNTL, tmp);
-	tmp = RREG32_MC(RS600_MC_PT0_CNTL);
+	tmp = RREG32_MC(R_000100_MC_PT0_CNTL);
+	tmp &= C_000100_INVALIDATE_ALL_L1_TLBS & C_000100_INVALIDATE_L2_CACHE;
+	WREG32_MC(R_000100_MC_PT0_CNTL, tmp);
+	tmp = RREG32_MC(R_000100_MC_PT0_CNTL);
 }
 
 int rs600_gart_init(struct radeon_device *rdev)
@@ -88,7 +186,7 @@ int rs600_gart_init(struct radeon_device *rdev)
 
 int rs600_gart_enable(struct radeon_device *rdev)
 {
-	uint32_t tmp;
+	u32 tmp;
 	int r, i;
 
 	if (rdev->gart.table.vram.robj == NULL) {
@@ -98,46 +196,51 @@ int rs600_gart_enable(struct radeon_device *rdev)
 	r = radeon_gart_table_vram_pin(rdev);
 	if (r)
 		return r;
+	radeon_gart_restore(rdev);
+	/* Enable bus master */
+	tmp = RREG32(R_00004C_BUS_CNTL) & C_00004C_BUS_MASTER_DIS;
+	WREG32(R_00004C_BUS_CNTL, tmp);
 	/* FIXME: setup default page */
-	WREG32_MC(RS600_MC_PT0_CNTL,
-		 (RS600_EFFECTIVE_L2_CACHE_SIZE(6) |
-		  RS600_EFFECTIVE_L2_QUEUE_SIZE(6)));
+	WREG32_MC(R_000100_MC_PT0_CNTL,
+		  (S_000100_EFFECTIVE_L2_CACHE_SIZE(6) |
+		   S_000100_EFFECTIVE_L2_QUEUE_SIZE(6)));
+
 	for (i = 0; i < 19; i++) {
-		WREG32_MC(RS600_MC_PT0_CLIENT0_CNTL + i,
-			 (RS600_ENABLE_TRANSLATION_MODE_OVERRIDE |
-			  RS600_SYSTEM_ACCESS_MODE_IN_SYS |
-			  RS600_SYSTEM_APERTURE_UNMAPPED_ACCESS_DEFAULT_PAGE |
-			  RS600_EFFECTIVE_L1_CACHE_SIZE(3) |
-			  RS600_ENABLE_FRAGMENT_PROCESSING |
-			  RS600_EFFECTIVE_L1_QUEUE_SIZE(3)));
+		WREG32_MC(R_00016C_MC_PT0_CLIENT0_CNTL + i,
+			  S_00016C_ENABLE_TRANSLATION_MODE_OVERRIDE(1) |
+			  S_00016C_SYSTEM_ACCESS_MODE_MASK(
+				  V_00016C_SYSTEM_ACCESS_MODE_NOT_IN_SYS) |
+			  S_00016C_SYSTEM_APERTURE_UNMAPPED_ACCESS(
+				  V_00016C_SYSTEM_APERTURE_UNMAPPED_PASSTHROUGH) |
+			  S_00016C_EFFECTIVE_L1_CACHE_SIZE(3) |
+			  S_00016C_ENABLE_FRAGMENT_PROCESSING(1) |
+			  S_00016C_EFFECTIVE_L1_QUEUE_SIZE(3));
 	}
-
-	/* System context map to GART space */
-	WREG32_MC(RS600_MC_PT0_SYSTEM_APERTURE_LOW_ADDR, rdev->mc.gtt_location);
-	tmp = rdev->mc.gtt_location + rdev->mc.gtt_size - 1;
-	WREG32_MC(RS600_MC_PT0_SYSTEM_APERTURE_HIGH_ADDR, tmp);
-
 	/* enable first context */
-	WREG32_MC(RS600_MC_PT0_CONTEXT0_FLAT_START_ADDR, rdev->mc.gtt_location);
-	tmp = rdev->mc.gtt_location + rdev->mc.gtt_size - 1;
-	WREG32_MC(RS600_MC_PT0_CONTEXT0_FLAT_END_ADDR, tmp);
-	WREG32_MC(RS600_MC_PT0_CONTEXT0_CNTL,
-		 (RS600_ENABLE_PAGE_TABLE | RS600_PAGE_TABLE_TYPE_FLAT));
+	WREG32_MC(R_000102_MC_PT0_CONTEXT0_CNTL,
+		  S_000102_ENABLE_PAGE_TABLE(1) |
+		  S_000102_PAGE_TABLE_DEPTH(V_000102_PAGE_TABLE_FLAT));
+
 	/* disable all other contexts */
-	for (i = 1; i < 8; i++) {
-		WREG32_MC(RS600_MC_PT0_CONTEXT0_CNTL + i, 0);
-	}
+	for (i = 1; i < 8; i++)
+		WREG32_MC(R_000102_MC_PT0_CONTEXT0_CNTL + i, 0);
 
 	/* setup the page table */
-	WREG32_MC(RS600_MC_PT0_CONTEXT0_FLAT_BASE_ADDR,
-		 rdev->gart.table_addr);
-	WREG32_MC(RS600_MC_PT0_CONTEXT0_DEFAULT_READ_ADDR, 0);
+	WREG32_MC(R_00012C_MC_PT0_CONTEXT0_FLAT_BASE_ADDR,
+		  rdev->gart.table_addr);
+	WREG32_MC(R_00013C_MC_PT0_CONTEXT0_FLAT_START_ADDR, rdev->mc.gtt_start);
+	WREG32_MC(R_00014C_MC_PT0_CONTEXT0_FLAT_END_ADDR, rdev->mc.gtt_end);
+	WREG32_MC(R_00011C_MC_PT0_CONTEXT0_DEFAULT_READ_ADDR, 0);
+
+	/* System context maps to VRAM space */
+	WREG32_MC(R_000112_MC_PT0_SYSTEM_APERTURE_LOW_ADDR, rdev->mc.vram_start);
+	WREG32_MC(R_000114_MC_PT0_SYSTEM_APERTURE_HIGH_ADDR, rdev->mc.vram_end);
 
 	/* enable page tables */
-	tmp = RREG32_MC(RS600_MC_PT0_CNTL);
-	WREG32_MC(RS600_MC_PT0_CNTL, (tmp | RS600_ENABLE_PT));
-	tmp = RREG32_MC(RS600_MC_CNTL1);
-	WREG32_MC(RS600_MC_CNTL1, (tmp | RS600_ENABLE_PAGE_TABLES));
+	tmp = RREG32_MC(R_000100_MC_PT0_CNTL);
+	WREG32_MC(R_000100_MC_PT0_CNTL, (tmp | S_000100_ENABLE_PT(1)));
+	tmp = RREG32_MC(R_000009_MC_CNTL1);
+	WREG32_MC(R_000009_MC_CNTL1, (tmp | S_000009_ENABLE_PAGE_TABLES(1)));
 	rs600_gart_tlb_flush(rdev);
 	rdev->gart.ready = true;
 	return 0;
@@ -145,16 +248,20 @@ int rs600_gart_enable(struct radeon_device *rdev)
 
 void rs600_gart_disable(struct radeon_device *rdev)
 {
-	uint32_t tmp;
+	u32 tmp;
+	int r;
 
 	/* FIXME: disable out of gart access */
-	WREG32_MC(RS600_MC_PT0_CNTL, 0);
-	tmp = RREG32_MC(RS600_MC_CNTL1);
-	tmp &= ~RS600_ENABLE_PAGE_TABLES;
-	WREG32_MC(RS600_MC_CNTL1, tmp);
+	WREG32_MC(R_000100_MC_PT0_CNTL, 0);
+	tmp = RREG32_MC(R_000009_MC_CNTL1);
+	WREG32_MC(R_000009_MC_CNTL1, tmp & C_000009_ENABLE_PAGE_TABLES);
 	if (rdev->gart.table.vram.robj) {
-		radeon_object_kunmap(rdev->gart.table.vram.robj);
-		radeon_object_unpin(rdev->gart.table.vram.robj);
+		r = radeon_bo_reserve(rdev->gart.table.vram.robj, false);
+		if (r == 0) {
+			radeon_bo_kunmap(rdev->gart.table.vram.robj);
+			radeon_bo_unpin(rdev->gart.table.vram.robj);
+			radeon_bo_unreserve(rdev->gart.table.vram.robj);
+		}
 	}
 }
 
@@ -185,133 +292,94 @@ int rs600_gart_set_page(struct radeon_device *rdev, int i, uint64_t addr)
 	return 0;
 }
 
-
-/*
- * MC.
- */
-void rs600_mc_disable_clients(struct radeon_device *rdev)
-{
-	unsigned tmp;
-
-	if (r100_gui_wait_for_idle(rdev)) {
-		printk(KERN_WARNING "Failed to wait GUI idle while "
-		       "programming pipes. Bad things might happen.\n");
-	}
-
-	radeon_avivo_vga_render_disable(rdev);
-
-	tmp = RREG32(AVIVO_D1VGA_CONTROL);
-	WREG32(AVIVO_D1VGA_CONTROL, tmp & ~AVIVO_DVGA_CONTROL_MODE_ENABLE);
-	tmp = RREG32(AVIVO_D2VGA_CONTROL);
-	WREG32(AVIVO_D2VGA_CONTROL, tmp & ~AVIVO_DVGA_CONTROL_MODE_ENABLE);
-
-	tmp = RREG32(AVIVO_D1CRTC_CONTROL);
-	WREG32(AVIVO_D1CRTC_CONTROL, tmp & ~AVIVO_CRTC_EN);
-	tmp = RREG32(AVIVO_D2CRTC_CONTROL);
-	WREG32(AVIVO_D2CRTC_CONTROL, tmp & ~AVIVO_CRTC_EN);
-
-	/* make sure all previous write got through */
-	tmp = RREG32(AVIVO_D2CRTC_CONTROL);
-
-	mdelay(1);
-}
-
-int rs600_mc_init(struct radeon_device *rdev)
-{
-	uint32_t tmp;
-	int r;
-
-	if (r100_debugfs_rbbm_init(rdev)) {
-		DRM_ERROR("Failed to register debugfs file for RBBM !\n");
-	}
-
-	rs600_gpu_init(rdev);
-	rs600_gart_disable(rdev);
-
-	/* Setup GPU memory space */
-	rdev->mc.vram_location = 0xFFFFFFFFUL;
-	rdev->mc.gtt_location = 0xFFFFFFFFUL;
-	r = radeon_mc_setup(rdev);
-	if (r) {
-		return r;
-	}
-
-	/* Program GPU memory space */
-	/* Enable bus master */
-	tmp = RREG32(RADEON_BUS_CNTL) & ~RS600_BUS_MASTER_DIS;
-	WREG32(RADEON_BUS_CNTL, tmp);
-	/* FIXME: What does AGP means for such chipset ? */
-	WREG32_MC(RS600_MC_AGP_LOCATION, 0x0FFFFFFF);
-	/* FIXME: are this AGP reg in indirect MC range ? */
-	WREG32_MC(RS600_MC_AGP_BASE, 0);
-	WREG32_MC(RS600_MC_AGP_BASE_2, 0);
-	rs600_mc_disable_clients(rdev);
-	if (rs600_mc_wait_for_idle(rdev)) {
-		printk(KERN_WARNING "Failed to wait MC idle while "
-		       "programming pipes. Bad things might happen.\n");
-	}
-	tmp = rdev->mc.vram_location + rdev->mc.mc_vram_size - 1;
-	tmp = REG_SET(RS600_MC_FB_TOP, tmp >> 16);
-	tmp |= REG_SET(RS600_MC_FB_START, rdev->mc.vram_location >> 16);
-	WREG32_MC(RS600_MC_FB_LOCATION, tmp);
-	WREG32(RS690_HDP_FB_LOCATION, rdev->mc.vram_location >> 16);
-	return 0;
-}
-
-void rs600_mc_fini(struct radeon_device *rdev)
-{
-}
-
-
-/*
- * Interrupts
- */
 int rs600_irq_set(struct radeon_device *rdev)
 {
 	uint32_t tmp = 0;
 	uint32_t mode_int = 0;
-
+	u32 hpd1 = RREG32(R_007D08_DC_HOT_PLUG_DETECT1_INT_CONTROL) &
+		~S_007D08_DC_HOT_PLUG_DETECT1_INT_EN(1);
+	u32 hpd2 = RREG32(R_007D18_DC_HOT_PLUG_DETECT2_INT_CONTROL) &
+		~S_007D18_DC_HOT_PLUG_DETECT2_INT_EN(1);
+
+	if (!rdev->irq.installed) {
+		WARN(1, "Can't enable IRQ/MSI because no handler is installed.\n");
+		WREG32(R_000040_GEN_INT_CNTL, 0);
+		return -EINVAL;
+	}
 	if (rdev->irq.sw_int) {
-		tmp |= RADEON_SW_INT_ENABLE;
+		tmp |= S_000040_SW_INT_EN(1);
 	}
 	if (rdev->irq.crtc_vblank_int[0]) {
-		mode_int |= AVIVO_D1MODE_INT_MASK;
+		mode_int |= S_006540_D1MODE_VBLANK_INT_MASK(1);
 	}
 	if (rdev->irq.crtc_vblank_int[1]) {
-		mode_int |= AVIVO_D2MODE_INT_MASK;
+		mode_int |= S_006540_D2MODE_VBLANK_INT_MASK(1);
+	}
+	if (rdev->irq.hpd[0]) {
+		hpd1 |= S_007D08_DC_HOT_PLUG_DETECT1_INT_EN(1);
 	}
-	WREG32(RADEON_GEN_INT_CNTL, tmp);
-	WREG32(AVIVO_DxMODE_INT_MASK, mode_int);
+	if (rdev->irq.hpd[1]) {
+		hpd2 |= S_007D18_DC_HOT_PLUG_DETECT2_INT_EN(1);
+	}
+	WREG32(R_000040_GEN_INT_CNTL, tmp);
+	WREG32(R_006540_DxMODE_INT_MASK, mode_int);
+	WREG32(R_007D08_DC_HOT_PLUG_DETECT1_INT_CONTROL, hpd1);
+	WREG32(R_007D18_DC_HOT_PLUG_DETECT2_INT_CONTROL, hpd2);
 	return 0;
 }
 
 static inline uint32_t rs600_irq_ack(struct radeon_device *rdev, u32 *r500_disp_int)
 {
-	uint32_t irqs = RREG32(RADEON_GEN_INT_STATUS);
-	uint32_t irq_mask = RADEON_SW_INT_TEST;
-
-	if (irqs & AVIVO_DISPLAY_INT_STATUS) {
-		*r500_disp_int = RREG32(AVIVO_DISP_INTERRUPT_STATUS);
-		if (*r500_disp_int & AVIVO_D1_VBLANK_INTERRUPT) {
-			WREG32(AVIVO_D1MODE_VBLANK_STATUS, AVIVO_VBLANK_ACK);
+	uint32_t irqs = RREG32(R_000044_GEN_INT_STATUS);
+	uint32_t irq_mask = ~C_000044_SW_INT;
+	u32 tmp;
+
+	if (G_000044_DISPLAY_INT_STAT(irqs)) {
+		*r500_disp_int = RREG32(R_007EDC_DISP_INTERRUPT_STATUS);
+		if (G_007EDC_LB_D1_VBLANK_INTERRUPT(*r500_disp_int)) {
+			WREG32(R_006534_D1MODE_VBLANK_STATUS,
+				S_006534_D1MODE_VBLANK_ACK(1));
+		}
+		if (G_007EDC_LB_D2_VBLANK_INTERRUPT(*r500_disp_int)) {
+			WREG32(R_006D34_D2MODE_VBLANK_STATUS,
+				S_006D34_D2MODE_VBLANK_ACK(1));
 		}
-		if (*r500_disp_int & AVIVO_D2_VBLANK_INTERRUPT) {
-			WREG32(AVIVO_D2MODE_VBLANK_STATUS, AVIVO_VBLANK_ACK);
+		if (G_007EDC_DC_HOT_PLUG_DETECT1_INTERRUPT(*r500_disp_int)) {
+			tmp = RREG32(R_007D08_DC_HOT_PLUG_DETECT1_INT_CONTROL);
+			tmp |= S_007D08_DC_HOT_PLUG_DETECT1_INT_ACK(1);
+			WREG32(R_007D08_DC_HOT_PLUG_DETECT1_INT_CONTROL, tmp);
+		}
+		if (G_007EDC_DC_HOT_PLUG_DETECT2_INTERRUPT(*r500_disp_int)) {
+			tmp = RREG32(R_007D18_DC_HOT_PLUG_DETECT2_INT_CONTROL);
+			tmp |= S_007D18_DC_HOT_PLUG_DETECT2_INT_ACK(1);
+			WREG32(R_007D18_DC_HOT_PLUG_DETECT2_INT_CONTROL, tmp);
 		}
 	} else {
 		*r500_disp_int = 0;
 	}
 
 	if (irqs) {
-		WREG32(RADEON_GEN_INT_STATUS, irqs);
+		WREG32(R_000044_GEN_INT_STATUS, irqs);
 	}
 	return irqs & irq_mask;
 }
 
+void rs600_irq_disable(struct radeon_device *rdev)
+{
+	u32 tmp;
+
+	WREG32(R_000040_GEN_INT_CNTL, 0);
+	WREG32(R_006540_DxMODE_INT_MASK, 0);
+	/* Wait and acknowledge irq */
+	mdelay(1);
+	rs600_irq_ack(rdev, &tmp);
+}
+
 int rs600_irq_process(struct radeon_device *rdev)
 {
-	uint32_t status;
+	uint32_t status, msi_rearm;
 	uint32_t r500_disp_int;
+	bool queue_hotplug = false;
 
 	status = rs600_irq_ack(rdev, &r500_disp_int);
 	if (!status && !r500_disp_int) {
@@ -319,90 +387,93 @@ int rs600_irq_process(struct radeon_device *rdev)
 	}
 	while (status || r500_disp_int) {
 		/* SW interrupt */
-		if (status & RADEON_SW_INT_TEST) {
+		if (G_000044_SW_INT(status))
 			radeon_fence_process(rdev);
-		}
 		/* Vertical blank interrupts */
-		if (r500_disp_int & AVIVO_D1_VBLANK_INTERRUPT) {
+		if (G_007EDC_LB_D1_VBLANK_INTERRUPT(r500_disp_int)) {
 			drm_handle_vblank(rdev->ddev, 0);
+			wake_up(&rdev->irq.vblank_queue);
 		}
-		if (r500_disp_int & AVIVO_D2_VBLANK_INTERRUPT) {
+		if (G_007EDC_LB_D2_VBLANK_INTERRUPT(r500_disp_int)) {
 			drm_handle_vblank(rdev->ddev, 1);
+			wake_up(&rdev->irq.vblank_queue);
+		}
+		if (G_007EDC_DC_HOT_PLUG_DETECT1_INTERRUPT(r500_disp_int)) {
+			queue_hotplug = true;
+			DRM_DEBUG("HPD1\n");
+		}
+		if (G_007EDC_DC_HOT_PLUG_DETECT2_INTERRUPT(r500_disp_int)) {
+			queue_hotplug = true;
+			DRM_DEBUG("HPD2\n");
 		}
 		status = rs600_irq_ack(rdev, &r500_disp_int);
 	}
+	if (queue_hotplug)
+		queue_work(rdev->wq, &rdev->hotplug_work);
+	if (rdev->msi_enabled) {
+		switch (rdev->family) {
+		case CHIP_RS600:
+		case CHIP_RS690:
+		case CHIP_RS740:
+			msi_rearm = RREG32(RADEON_BUS_CNTL) & ~RS600_MSI_REARM;
+			WREG32(RADEON_BUS_CNTL, msi_rearm);
+			WREG32(RADEON_BUS_CNTL, msi_rearm | RS600_MSI_REARM);
+			break;
+		default:
+			msi_rearm = RREG32(RADEON_MSI_REARM_EN) & ~RV370_MSI_REARM_EN;
+			WREG32(RADEON_MSI_REARM_EN, msi_rearm);
+			WREG32(RADEON_MSI_REARM_EN, msi_rearm | RV370_MSI_REARM_EN);
+			break;
+		}
+	}
 	return IRQ_HANDLED;
 }
 
 u32 rs600_get_vblank_counter(struct radeon_device *rdev, int crtc)
 {
 	if (crtc == 0)
-		return RREG32(AVIVO_D1CRTC_FRAME_COUNT);
+		return RREG32(R_0060A4_D1CRTC_STATUS_FRAME_COUNT);
 	else
-		return RREG32(AVIVO_D2CRTC_FRAME_COUNT);
-}
-
-
-/*
- * Global GPU functions
- */
-void rs600_disable_vga(struct radeon_device *rdev)
-{
-	unsigned tmp;
-
-	WREG32(0x330, 0);
-	WREG32(0x338, 0);
-	tmp = RREG32(0x300);
-	tmp &= ~(3 << 16);
-	WREG32(0x300, tmp);
-	WREG32(0x308, (1 << 8));
-	WREG32(0x310, rdev->mc.vram_location);
-	WREG32(0x594, 0);
+		return RREG32(R_0068A4_D2CRTC_STATUS_FRAME_COUNT);
 }
 
 int rs600_mc_wait_for_idle(struct radeon_device *rdev)
 {
 	unsigned i;
-	uint32_t tmp;
 
 	for (i = 0; i < rdev->usec_timeout; i++) {
-		/* read MC_STATUS */
-		tmp = RREG32_MC(RS600_MC_STATUS);
-		if (tmp & RS600_MC_STATUS_IDLE) {
+		if (G_000000_MC_IDLE(RREG32_MC(R_000000_MC_STATUS)))
 			return 0;
-		}
-		DRM_UDELAY(1);
+		udelay(1);
 	}
 	return -1;
 }
 
-void rs600_errata(struct radeon_device *rdev)
-{
-	rdev->pll_errata = 0;
-}
-
 void rs600_gpu_init(struct radeon_device *rdev)
 {
-	/* FIXME: HDP same place on rs600 ? */
 	r100_hdp_reset(rdev);
-	rs600_disable_vga(rdev);
-	/* FIXME: is this correct ? */
 	r420_pipes_init(rdev);
-	if (rs600_mc_wait_for_idle(rdev)) {
-		printk(KERN_WARNING "Failed to wait MC idle while "
-		       "programming pipes. Bad things might happen.\n");
-	}
+	/* Wait for mc idle */
+	if (rs600_mc_wait_for_idle(rdev))
+		dev_warn(rdev->dev, "Wait MC idle timeout before updating MC.\n");
 }
 
-
-/*
- * VRAM info.
- */
-void rs600_vram_info(struct radeon_device *rdev)
+void rs600_mc_init(struct radeon_device *rdev)
 {
-	/* FIXME: to do or is these values sane ? */
+	u64 base;
+
+	rdev->mc.aper_base = drm_get_resource_start(rdev->ddev, 0);
+	rdev->mc.aper_size = drm_get_resource_len(rdev->ddev, 0);
 	rdev->mc.vram_is_ddr = true;
 	rdev->mc.vram_width = 128;
+	rdev->mc.real_vram_size = RREG32(RADEON_CONFIG_MEMSIZE);
+	rdev->mc.mc_vram_size = rdev->mc.real_vram_size;
+	rdev->mc.visible_vram_size = rdev->mc.aper_size;
+	rdev->mc.igp_sideport_enabled = radeon_atombios_sideport_present(rdev);
+	base = RREG32_MC(R_000004_MC_FB_LOCATION);
+	base = G_000004_MC_FB_START(base) << 16;
+	radeon_vram_location(rdev, &rdev->mc, base);
+	radeon_gtt_location(rdev, &rdev->mc);
 }
 
 void rs600_bandwidth_update(struct radeon_device *rdev)
@@ -410,31 +481,203 @@ void rs600_bandwidth_update(struct radeon_device *rdev)
 	/* FIXME: implement, should this be like rs690 ? */
 }
 
-
-/*
- * Indirect registers accessor
- */
 uint32_t rs600_mc_rreg(struct radeon_device *rdev, uint32_t reg)
 {
-	uint32_t r;
-
-	WREG32(RS600_MC_INDEX,
-	       ((reg & RS600_MC_ADDR_MASK) | RS600_MC_IND_CITF_ARB0));
-	r = RREG32(RS600_MC_DATA);
-	return r;
+	WREG32(R_000070_MC_IND_INDEX, S_000070_MC_IND_ADDR(reg) |
+		S_000070_MC_IND_CITF_ARB0(1));
+	return RREG32(R_000074_MC_IND_DATA);
 }
 
 void rs600_mc_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v)
 {
-	WREG32(RS600_MC_INDEX,
-		RS600_MC_IND_WR_EN | RS600_MC_IND_CITF_ARB0 |
-		((reg) & RS600_MC_ADDR_MASK));
-	WREG32(RS600_MC_DATA, v);
+	WREG32(R_000070_MC_IND_INDEX, S_000070_MC_IND_ADDR(reg) |
+		S_000070_MC_IND_CITF_ARB0(1) | S_000070_MC_IND_WR_EN(1));
+	WREG32(R_000074_MC_IND_DATA, v);
 }
 
-int rs600_init(struct radeon_device *rdev)
+void rs600_debugfs(struct radeon_device *rdev)
+{
+	if (r100_debugfs_rbbm_init(rdev))
+		DRM_ERROR("Failed to register debugfs file for RBBM !\n");
+}
+
+void rs600_set_safe_registers(struct radeon_device *rdev)
 {
 	rdev->config.r300.reg_safe_bm = rs600_reg_safe_bm;
 	rdev->config.r300.reg_safe_bm_size = ARRAY_SIZE(rs600_reg_safe_bm);
+}
+
+static void rs600_mc_program(struct radeon_device *rdev)
+{
+	struct rv515_mc_save save;
+
+	/* Stops all mc clients */
+	rv515_mc_stop(rdev, &save);
+
+	/* Wait for mc idle */
+	if (rs600_mc_wait_for_idle(rdev))
+		dev_warn(rdev->dev, "Wait MC idle timeout before updating MC.\n");
+
+	/* FIXME: What does AGP means for such chipset ? */
+	WREG32_MC(R_000005_MC_AGP_LOCATION, 0x0FFFFFFF);
+	WREG32_MC(R_000006_AGP_BASE, 0);
+	WREG32_MC(R_000007_AGP_BASE_2, 0);
+	/* Program MC */
+	WREG32_MC(R_000004_MC_FB_LOCATION,
+			S_000004_MC_FB_START(rdev->mc.vram_start >> 16) |
+			S_000004_MC_FB_TOP(rdev->mc.vram_end >> 16));
+	WREG32(R_000134_HDP_FB_LOCATION,
+		S_000134_HDP_FB_START(rdev->mc.vram_start >> 16));
+
+	rv515_mc_resume(rdev, &save);
+}
+
+static int rs600_startup(struct radeon_device *rdev)
+{
+	int r;
+
+	rs600_mc_program(rdev);
+	/* Resume clock */
+	rv515_clock_startup(rdev);
+	/* Initialize GPU configuration (# pipes, ...) */
+	rs600_gpu_init(rdev);
+	/* Initialize GART (initialize after TTM so we can allocate
+	 * memory through TTM but finalize after TTM) */
+	r = rs600_gart_enable(rdev);
+	if (r)
+		return r;
+	/* Enable IRQ */
+	rs600_irq_set(rdev);
+	rdev->config.r300.hdp_cntl = RREG32(RADEON_HOST_PATH_CNTL);
+	/* 1M ring buffer */
+	r = r100_cp_init(rdev, 1024 * 1024);
+	if (r) {
+		dev_err(rdev->dev, "failled initializing CP (%d).\n", r);
+		return r;
+	}
+	r = r100_wb_init(rdev);
+	if (r)
+		dev_err(rdev->dev, "failled initializing WB (%d).\n", r);
+	r = r100_ib_init(rdev);
+	if (r) {
+		dev_err(rdev->dev, "failled initializing IB (%d).\n", r);
+		return r;
+	}
+	return 0;
+}
+
+int rs600_resume(struct radeon_device *rdev)
+{
+	/* Make sur GART are not working */
+	rs600_gart_disable(rdev);
+	/* Resume clock before doing reset */
+	rv515_clock_startup(rdev);
+	/* Reset gpu before posting otherwise ATOM will enter infinite loop */
+	if (radeon_gpu_reset(rdev)) {
+		dev_warn(rdev->dev, "GPU reset failed ! (0xE40=0x%08X, 0x7C0=0x%08X)\n",
+			RREG32(R_000E40_RBBM_STATUS),
+			RREG32(R_0007C0_CP_STAT));
+	}
+	/* post */
+	atom_asic_init(rdev->mode_info.atom_context);
+	/* Resume clock after posting */
+	rv515_clock_startup(rdev);
+	/* Initialize surface registers */
+	radeon_surface_init(rdev);
+	return rs600_startup(rdev);
+}
+
+int rs600_suspend(struct radeon_device *rdev)
+{
+	r100_cp_disable(rdev);
+	r100_wb_disable(rdev);
+	rs600_irq_disable(rdev);
+	rs600_gart_disable(rdev);
+	return 0;
+}
+
+void rs600_fini(struct radeon_device *rdev)
+{
+	r100_cp_fini(rdev);
+	r100_wb_fini(rdev);
+	r100_ib_fini(rdev);
+	radeon_gem_fini(rdev);
+	rs600_gart_fini(rdev);
+	radeon_irq_kms_fini(rdev);
+	radeon_fence_driver_fini(rdev);
+	radeon_bo_fini(rdev);
+	radeon_atombios_fini(rdev);
+	kfree(rdev->bios);
+	rdev->bios = NULL;
+}
+
+int rs600_init(struct radeon_device *rdev)
+{
+	int r;
+
+	/* Disable VGA */
+	rv515_vga_render_disable(rdev);
+	/* Initialize scratch registers */
+	radeon_scratch_init(rdev);
+	/* Initialize surface registers */
+	radeon_surface_init(rdev);
+	/* BIOS */
+	if (!radeon_get_bios(rdev)) {
+		if (ASIC_IS_AVIVO(rdev))
+			return -EINVAL;
+	}
+	if (rdev->is_atom_bios) {
+		r = radeon_atombios_init(rdev);
+		if (r)
+			return r;
+	} else {
+		dev_err(rdev->dev, "Expecting atombios for RS600 GPU\n");
+		return -EINVAL;
+	}
+	/* Reset gpu before posting otherwise ATOM will enter infinite loop */
+	if (radeon_gpu_reset(rdev)) {
+		dev_warn(rdev->dev,
+			"GPU reset failed ! (0xE40=0x%08X, 0x7C0=0x%08X)\n",
+			RREG32(R_000E40_RBBM_STATUS),
+			RREG32(R_0007C0_CP_STAT));
+	}
+	/* check if cards are posted or not */
+	if (radeon_boot_test_post_card(rdev) == false)
+		return -EINVAL;
+
+	/* Initialize clocks */
+	radeon_get_clock_info(rdev->ddev);
+	/* Initialize power management */
+	radeon_pm_init(rdev);
+	/* initialize memory controller */
+	rs600_mc_init(rdev);
+	rs600_debugfs(rdev);
+	/* Fence driver */
+	r = radeon_fence_driver_init(rdev);
+	if (r)
+		return r;
+	r = radeon_irq_kms_init(rdev);
+	if (r)
+		return r;
+	/* Memory manager */
+	r = radeon_bo_init(rdev);
+	if (r)
+		return r;
+	r = rs600_gart_init(rdev);
+	if (r)
+		return r;
+	rs600_set_safe_registers(rdev);
+	rdev->accel_working = true;
+	r = rs600_startup(rdev);
+	if (r) {
+		/* Somethings want wront with the accel init stop accel */
+		dev_err(rdev->dev, "Disabling GPU acceleration\n");
+		r100_cp_fini(rdev);
+		r100_wb_fini(rdev);
+		r100_ib_fini(rdev);
+		rs600_gart_fini(rdev);
+		radeon_irq_kms_fini(rdev);
+		rdev->accel_working = false;
+	}
 	return 0;
 }
diff --git a/drivers/gpu/drm/radeon/rs600d.h b/drivers/gpu/drm/radeon/rs600d.h
new file mode 100644
index 000000000000..c1c8f5885cbb
--- /dev/null
+++ b/drivers/gpu/drm/radeon/rs600d.h
@@ -0,0 +1,538 @@
+/*
+ * Copyright 2008 Advanced Micro Devices, Inc.
+ * Copyright 2008 Red Hat Inc.
+ * Copyright 2009 Jerome Glisse.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Dave Airlie
+ *          Alex Deucher
+ *          Jerome Glisse
+ */
+#ifndef __RS600D_H__
+#define __RS600D_H__
+
+/* Registers */
+#define R_000040_GEN_INT_CNTL                        0x000040
+#define   S_000040_SCRATCH_INT_MASK(x)                 (((x) & 0x1) << 18)
+#define   G_000040_SCRATCH_INT_MASK(x)                 (((x) >> 18) & 0x1)
+#define   C_000040_SCRATCH_INT_MASK                    0xFFFBFFFF
+#define   S_000040_GUI_IDLE_MASK(x)                    (((x) & 0x1) << 19)
+#define   G_000040_GUI_IDLE_MASK(x)                    (((x) >> 19) & 0x1)
+#define   C_000040_GUI_IDLE_MASK                       0xFFF7FFFF
+#define   S_000040_DMA_VIPH1_INT_EN(x)                 (((x) & 0x1) << 13)
+#define   G_000040_DMA_VIPH1_INT_EN(x)                 (((x) >> 13) & 0x1)
+#define   C_000040_DMA_VIPH1_INT_EN                    0xFFFFDFFF
+#define   S_000040_DMA_VIPH2_INT_EN(x)                 (((x) & 0x1) << 14)
+#define   G_000040_DMA_VIPH2_INT_EN(x)                 (((x) >> 14) & 0x1)
+#define   C_000040_DMA_VIPH2_INT_EN                    0xFFFFBFFF
+#define   S_000040_DMA_VIPH3_INT_EN(x)                 (((x) & 0x1) << 15)
+#define   G_000040_DMA_VIPH3_INT_EN(x)                 (((x) >> 15) & 0x1)
+#define   C_000040_DMA_VIPH3_INT_EN                    0xFFFF7FFF
+#define   S_000040_I2C_INT_EN(x)                       (((x) & 0x1) << 17)
+#define   G_000040_I2C_INT_EN(x)                       (((x) >> 17) & 0x1)
+#define   C_000040_I2C_INT_EN                          0xFFFDFFFF
+#define   S_000040_GUI_IDLE(x)                         (((x) & 0x1) << 19)
+#define   G_000040_GUI_IDLE(x)                         (((x) >> 19) & 0x1)
+#define   C_000040_GUI_IDLE                            0xFFF7FFFF
+#define   S_000040_VIPH_INT_EN(x)                      (((x) & 0x1) << 24)
+#define   G_000040_VIPH_INT_EN(x)                      (((x) >> 24) & 0x1)
+#define   C_000040_VIPH_INT_EN                         0xFEFFFFFF
+#define   S_000040_SW_INT_EN(x)                        (((x) & 0x1) << 25)
+#define   G_000040_SW_INT_EN(x)                        (((x) >> 25) & 0x1)
+#define   C_000040_SW_INT_EN                           0xFDFFFFFF
+#define   S_000040_GEYSERVILLE(x)                      (((x) & 0x1) << 27)
+#define   G_000040_GEYSERVILLE(x)                      (((x) >> 27) & 0x1)
+#define   C_000040_GEYSERVILLE                         0xF7FFFFFF
+#define   S_000040_HDCP_AUTHORIZED_INT(x)              (((x) & 0x1) << 28)
+#define   G_000040_HDCP_AUTHORIZED_INT(x)              (((x) >> 28) & 0x1)
+#define   C_000040_HDCP_AUTHORIZED_INT                 0xEFFFFFFF
+#define   S_000040_DVI_I2C_INT(x)                      (((x) & 0x1) << 29)
+#define   G_000040_DVI_I2C_INT(x)                      (((x) >> 29) & 0x1)
+#define   C_000040_DVI_I2C_INT                         0xDFFFFFFF
+#define   S_000040_GUIDMA(x)                           (((x) & 0x1) << 30)
+#define   G_000040_GUIDMA(x)                           (((x) >> 30) & 0x1)
+#define   C_000040_GUIDMA                              0xBFFFFFFF
+#define   S_000040_VIDDMA(x)                           (((x) & 0x1) << 31)
+#define   G_000040_VIDDMA(x)                           (((x) >> 31) & 0x1)
+#define   C_000040_VIDDMA                              0x7FFFFFFF
+#define R_000044_GEN_INT_STATUS                      0x000044
+#define   S_000044_DISPLAY_INT_STAT(x)                 (((x) & 0x1) << 0)
+#define   G_000044_DISPLAY_INT_STAT(x)                 (((x) >> 0) & 0x1)
+#define   C_000044_DISPLAY_INT_STAT                    0xFFFFFFFE
+#define   S_000044_VGA_INT_STAT(x)                     (((x) & 0x1) << 1)
+#define   G_000044_VGA_INT_STAT(x)                     (((x) >> 1) & 0x1)
+#define   C_000044_VGA_INT_STAT                        0xFFFFFFFD
+#define   S_000044_CAP0_INT_ACTIVE(x)                  (((x) & 0x1) << 8)
+#define   G_000044_CAP0_INT_ACTIVE(x)                  (((x) >> 8) & 0x1)
+#define   C_000044_CAP0_INT_ACTIVE                     0xFFFFFEFF
+#define   S_000044_DMA_VIPH0_INT(x)                    (((x) & 0x1) << 12)
+#define   G_000044_DMA_VIPH0_INT(x)                    (((x) >> 12) & 0x1)
+#define   C_000044_DMA_VIPH0_INT                       0xFFFFEFFF
+#define   S_000044_DMA_VIPH1_INT(x)                    (((x) & 0x1) << 13)
+#define   G_000044_DMA_VIPH1_INT(x)                    (((x) >> 13) & 0x1)
+#define   C_000044_DMA_VIPH1_INT                       0xFFFFDFFF
+#define   S_000044_DMA_VIPH2_INT(x)                    (((x) & 0x1) << 14)
+#define   G_000044_DMA_VIPH2_INT(x)                    (((x) >> 14) & 0x1)
+#define   C_000044_DMA_VIPH2_INT                       0xFFFFBFFF
+#define   S_000044_DMA_VIPH3_INT(x)                    (((x) & 0x1) << 15)
+#define   G_000044_DMA_VIPH3_INT(x)                    (((x) >> 15) & 0x1)
+#define   C_000044_DMA_VIPH3_INT                       0xFFFF7FFF
+#define   S_000044_MC_PROBE_FAULT_STAT(x)              (((x) & 0x1) << 16)
+#define   G_000044_MC_PROBE_FAULT_STAT(x)              (((x) >> 16) & 0x1)
+#define   C_000044_MC_PROBE_FAULT_STAT                 0xFFFEFFFF
+#define   S_000044_I2C_INT(x)                          (((x) & 0x1) << 17)
+#define   G_000044_I2C_INT(x)                          (((x) >> 17) & 0x1)
+#define   C_000044_I2C_INT                             0xFFFDFFFF
+#define   S_000044_SCRATCH_INT_STAT(x)                 (((x) & 0x1) << 18)
+#define   G_000044_SCRATCH_INT_STAT(x)                 (((x) >> 18) & 0x1)
+#define   C_000044_SCRATCH_INT_STAT                    0xFFFBFFFF
+#define   S_000044_GUI_IDLE_STAT(x)                    (((x) & 0x1) << 19)
+#define   G_000044_GUI_IDLE_STAT(x)                    (((x) >> 19) & 0x1)
+#define   C_000044_GUI_IDLE_STAT                       0xFFF7FFFF
+#define   S_000044_ATI_OVERDRIVE_INT_STAT(x)           (((x) & 0x1) << 20)
+#define   G_000044_ATI_OVERDRIVE_INT_STAT(x)           (((x) >> 20) & 0x1)
+#define   C_000044_ATI_OVERDRIVE_INT_STAT              0xFFEFFFFF
+#define   S_000044_MC_PROTECTION_FAULT_STAT(x)         (((x) & 0x1) << 21)
+#define   G_000044_MC_PROTECTION_FAULT_STAT(x)         (((x) >> 21) & 0x1)
+#define   C_000044_MC_PROTECTION_FAULT_STAT            0xFFDFFFFF
+#define   S_000044_RBBM_READ_INT_STAT(x)               (((x) & 0x1) << 22)
+#define   G_000044_RBBM_READ_INT_STAT(x)               (((x) >> 22) & 0x1)
+#define   C_000044_RBBM_READ_INT_STAT                  0xFFBFFFFF
+#define   S_000044_CB_CONTEXT_SWITCH_STAT(x)           (((x) & 0x1) << 23)
+#define   G_000044_CB_CONTEXT_SWITCH_STAT(x)           (((x) >> 23) & 0x1)
+#define   C_000044_CB_CONTEXT_SWITCH_STAT              0xFF7FFFFF
+#define   S_000044_VIPH_INT(x)                         (((x) & 0x1) << 24)
+#define   G_000044_VIPH_INT(x)                         (((x) >> 24) & 0x1)
+#define   C_000044_VIPH_INT                            0xFEFFFFFF
+#define   S_000044_SW_INT(x)                           (((x) & 0x1) << 25)
+#define   G_000044_SW_INT(x)                           (((x) >> 25) & 0x1)
+#define   C_000044_SW_INT                              0xFDFFFFFF
+#define   S_000044_SW_INT_SET(x)                       (((x) & 0x1) << 26)
+#define   G_000044_SW_INT_SET(x)                       (((x) >> 26) & 0x1)
+#define   C_000044_SW_INT_SET                          0xFBFFFFFF
+#define   S_000044_IDCT_INT_STAT(x)                    (((x) & 0x1) << 27)
+#define   G_000044_IDCT_INT_STAT(x)                    (((x) >> 27) & 0x1)
+#define   C_000044_IDCT_INT_STAT                       0xF7FFFFFF
+#define   S_000044_GUIDMA_STAT(x)                      (((x) & 0x1) << 30)
+#define   G_000044_GUIDMA_STAT(x)                      (((x) >> 30) & 0x1)
+#define   C_000044_GUIDMA_STAT                         0xBFFFFFFF
+#define   S_000044_VIDDMA_STAT(x)                      (((x) & 0x1) << 31)
+#define   G_000044_VIDDMA_STAT(x)                      (((x) >> 31) & 0x1)
+#define   C_000044_VIDDMA_STAT                         0x7FFFFFFF
+#define R_00004C_BUS_CNTL                            0x00004C
+#define   S_00004C_BUS_MASTER_DIS(x)                   (((x) & 0x1) << 14)
+#define   G_00004C_BUS_MASTER_DIS(x)                   (((x) >> 14) & 0x1)
+#define   C_00004C_BUS_MASTER_DIS                      0xFFFFBFFF
+#define   S_00004C_BUS_MSI_REARM(x)                    (((x) & 0x1) << 20)
+#define   G_00004C_BUS_MSI_REARM(x)                    (((x) >> 20) & 0x1)
+#define   C_00004C_BUS_MSI_REARM                       0xFFEFFFFF
+#define R_000070_MC_IND_INDEX                        0x000070
+#define   S_000070_MC_IND_ADDR(x)                      (((x) & 0xFFFF) << 0)
+#define   G_000070_MC_IND_ADDR(x)                      (((x) >> 0) & 0xFFFF)
+#define   C_000070_MC_IND_ADDR                         0xFFFF0000
+#define   S_000070_MC_IND_SEQ_RBS_0(x)                 (((x) & 0x1) << 16)
+#define   G_000070_MC_IND_SEQ_RBS_0(x)                 (((x) >> 16) & 0x1)
+#define   C_000070_MC_IND_SEQ_RBS_0                    0xFFFEFFFF
+#define   S_000070_MC_IND_SEQ_RBS_1(x)                 (((x) & 0x1) << 17)
+#define   G_000070_MC_IND_SEQ_RBS_1(x)                 (((x) >> 17) & 0x1)
+#define   C_000070_MC_IND_SEQ_RBS_1                    0xFFFDFFFF
+#define   S_000070_MC_IND_SEQ_RBS_2(x)                 (((x) & 0x1) << 18)
+#define   G_000070_MC_IND_SEQ_RBS_2(x)                 (((x) >> 18) & 0x1)
+#define   C_000070_MC_IND_SEQ_RBS_2                    0xFFFBFFFF
+#define   S_000070_MC_IND_SEQ_RBS_3(x)                 (((x) & 0x1) << 19)
+#define   G_000070_MC_IND_SEQ_RBS_3(x)                 (((x) >> 19) & 0x1)
+#define   C_000070_MC_IND_SEQ_RBS_3                    0xFFF7FFFF
+#define   S_000070_MC_IND_AIC_RBS(x)                   (((x) & 0x1) << 20)
+#define   G_000070_MC_IND_AIC_RBS(x)                   (((x) >> 20) & 0x1)
+#define   C_000070_MC_IND_AIC_RBS                      0xFFEFFFFF
+#define   S_000070_MC_IND_CITF_ARB0(x)                 (((x) & 0x1) << 21)
+#define   G_000070_MC_IND_CITF_ARB0(x)                 (((x) >> 21) & 0x1)
+#define   C_000070_MC_IND_CITF_ARB0                    0xFFDFFFFF
+#define   S_000070_MC_IND_CITF_ARB1(x)                 (((x) & 0x1) << 22)
+#define   G_000070_MC_IND_CITF_ARB1(x)                 (((x) >> 22) & 0x1)
+#define   C_000070_MC_IND_CITF_ARB1                    0xFFBFFFFF
+#define   S_000070_MC_IND_WR_EN(x)                     (((x) & 0x1) << 23)
+#define   G_000070_MC_IND_WR_EN(x)                     (((x) >> 23) & 0x1)
+#define   C_000070_MC_IND_WR_EN                        0xFF7FFFFF
+#define   S_000070_MC_IND_RD_INV(x)                    (((x) & 0x1) << 24)
+#define   G_000070_MC_IND_RD_INV(x)                    (((x) >> 24) & 0x1)
+#define   C_000070_MC_IND_RD_INV                       0xFEFFFFFF
+#define R_000074_MC_IND_DATA                         0x000074
+#define   S_000074_MC_IND_DATA(x)                      (((x) & 0xFFFFFFFF) << 0)
+#define   G_000074_MC_IND_DATA(x)                      (((x) >> 0) & 0xFFFFFFFF)
+#define   C_000074_MC_IND_DATA                         0x00000000
+#define R_000134_HDP_FB_LOCATION                     0x000134
+#define   S_000134_HDP_FB_START(x)                     (((x) & 0xFFFF) << 0)
+#define   G_000134_HDP_FB_START(x)                     (((x) >> 0) & 0xFFFF)
+#define   C_000134_HDP_FB_START                        0xFFFF0000
+#define R_0007C0_CP_STAT                             0x0007C0
+#define   S_0007C0_MRU_BUSY(x)                         (((x) & 0x1) << 0)
+#define   G_0007C0_MRU_BUSY(x)                         (((x) >> 0) & 0x1)
+#define   C_0007C0_MRU_BUSY                            0xFFFFFFFE
+#define   S_0007C0_MWU_BUSY(x)                         (((x) & 0x1) << 1)
+#define   G_0007C0_MWU_BUSY(x)                         (((x) >> 1) & 0x1)
+#define   C_0007C0_MWU_BUSY                            0xFFFFFFFD
+#define   S_0007C0_RSIU_BUSY(x)                        (((x) & 0x1) << 2)
+#define   G_0007C0_RSIU_BUSY(x)                        (((x) >> 2) & 0x1)
+#define   C_0007C0_RSIU_BUSY                           0xFFFFFFFB
+#define   S_0007C0_RCIU_BUSY(x)                        (((x) & 0x1) << 3)
+#define   G_0007C0_RCIU_BUSY(x)                        (((x) >> 3) & 0x1)
+#define   C_0007C0_RCIU_BUSY                           0xFFFFFFF7
+#define   S_0007C0_CSF_PRIMARY_BUSY(x)                 (((x) & 0x1) << 9)
+#define   G_0007C0_CSF_PRIMARY_BUSY(x)                 (((x) >> 9) & 0x1)
+#define   C_0007C0_CSF_PRIMARY_BUSY                    0xFFFFFDFF
+#define   S_0007C0_CSF_INDIRECT_BUSY(x)                (((x) & 0x1) << 10)
+#define   G_0007C0_CSF_INDIRECT_BUSY(x)                (((x) >> 10) & 0x1)
+#define   C_0007C0_CSF_INDIRECT_BUSY                   0xFFFFFBFF
+#define   S_0007C0_CSQ_PRIMARY_BUSY(x)                 (((x) & 0x1) << 11)
+#define   G_0007C0_CSQ_PRIMARY_BUSY(x)                 (((x) >> 11) & 0x1)
+#define   C_0007C0_CSQ_PRIMARY_BUSY                    0xFFFFF7FF
+#define   S_0007C0_CSQ_INDIRECT_BUSY(x)                (((x) & 0x1) << 12)
+#define   G_0007C0_CSQ_INDIRECT_BUSY(x)                (((x) >> 12) & 0x1)
+#define   C_0007C0_CSQ_INDIRECT_BUSY                   0xFFFFEFFF
+#define   S_0007C0_CSI_BUSY(x)                         (((x) & 0x1) << 13)
+#define   G_0007C0_CSI_BUSY(x)                         (((x) >> 13) & 0x1)
+#define   C_0007C0_CSI_BUSY                            0xFFFFDFFF
+#define   S_0007C0_CSF_INDIRECT2_BUSY(x)               (((x) & 0x1) << 14)
+#define   G_0007C0_CSF_INDIRECT2_BUSY(x)               (((x) >> 14) & 0x1)
+#define   C_0007C0_CSF_INDIRECT2_BUSY                  0xFFFFBFFF
+#define   S_0007C0_CSQ_INDIRECT2_BUSY(x)               (((x) & 0x1) << 15)
+#define   G_0007C0_CSQ_INDIRECT2_BUSY(x)               (((x) >> 15) & 0x1)
+#define   C_0007C0_CSQ_INDIRECT2_BUSY                  0xFFFF7FFF
+#define   S_0007C0_GUIDMA_BUSY(x)                      (((x) & 0x1) << 28)
+#define   G_0007C0_GUIDMA_BUSY(x)                      (((x) >> 28) & 0x1)
+#define   C_0007C0_GUIDMA_BUSY                         0xEFFFFFFF
+#define   S_0007C0_VIDDMA_BUSY(x)                      (((x) & 0x1) << 29)
+#define   G_0007C0_VIDDMA_BUSY(x)                      (((x) >> 29) & 0x1)
+#define   C_0007C0_VIDDMA_BUSY                         0xDFFFFFFF
+#define   S_0007C0_CMDSTRM_BUSY(x)                     (((x) & 0x1) << 30)
+#define   G_0007C0_CMDSTRM_BUSY(x)                     (((x) >> 30) & 0x1)
+#define   C_0007C0_CMDSTRM_BUSY                        0xBFFFFFFF
+#define   S_0007C0_CP_BUSY(x)                          (((x) & 0x1) << 31)
+#define   G_0007C0_CP_BUSY(x)                          (((x) >> 31) & 0x1)
+#define   C_0007C0_CP_BUSY                             0x7FFFFFFF
+#define R_000E40_RBBM_STATUS                         0x000E40
+#define   S_000E40_CMDFIFO_AVAIL(x)                    (((x) & 0x7F) << 0)
+#define   G_000E40_CMDFIFO_AVAIL(x)                    (((x) >> 0) & 0x7F)
+#define   C_000E40_CMDFIFO_AVAIL                       0xFFFFFF80
+#define   S_000E40_HIRQ_ON_RBB(x)                      (((x) & 0x1) << 8)
+#define   G_000E40_HIRQ_ON_RBB(x)                      (((x) >> 8) & 0x1)
+#define   C_000E40_HIRQ_ON_RBB                         0xFFFFFEFF
+#define   S_000E40_CPRQ_ON_RBB(x)                      (((x) & 0x1) << 9)
+#define   G_000E40_CPRQ_ON_RBB(x)                      (((x) >> 9) & 0x1)
+#define   C_000E40_CPRQ_ON_RBB                         0xFFFFFDFF
+#define   S_000E40_CFRQ_ON_RBB(x)                      (((x) & 0x1) << 10)
+#define   G_000E40_CFRQ_ON_RBB(x)                      (((x) >> 10) & 0x1)
+#define   C_000E40_CFRQ_ON_RBB                         0xFFFFFBFF
+#define   S_000E40_HIRQ_IN_RTBUF(x)                    (((x) & 0x1) << 11)
+#define   G_000E40_HIRQ_IN_RTBUF(x)                    (((x) >> 11) & 0x1)
+#define   C_000E40_HIRQ_IN_RTBUF                       0xFFFFF7FF
+#define   S_000E40_CPRQ_IN_RTBUF(x)                    (((x) & 0x1) << 12)
+#define   G_000E40_CPRQ_IN_RTBUF(x)                    (((x) >> 12) & 0x1)
+#define   C_000E40_CPRQ_IN_RTBUF                       0xFFFFEFFF
+#define   S_000E40_CFRQ_IN_RTBUF(x)                    (((x) & 0x1) << 13)
+#define   G_000E40_CFRQ_IN_RTBUF(x)                    (((x) >> 13) & 0x1)
+#define   C_000E40_CFRQ_IN_RTBUF                       0xFFFFDFFF
+#define   S_000E40_CF_PIPE_BUSY(x)                     (((x) & 0x1) << 14)
+#define   G_000E40_CF_PIPE_BUSY(x)                     (((x) >> 14) & 0x1)
+#define   C_000E40_CF_PIPE_BUSY                        0xFFFFBFFF
+#define   S_000E40_ENG_EV_BUSY(x)                      (((x) & 0x1) << 15)
+#define   G_000E40_ENG_EV_BUSY(x)                      (((x) >> 15) & 0x1)
+#define   C_000E40_ENG_EV_BUSY                         0xFFFF7FFF
+#define   S_000E40_CP_CMDSTRM_BUSY(x)                  (((x) & 0x1) << 16)
+#define   G_000E40_CP_CMDSTRM_BUSY(x)                  (((x) >> 16) & 0x1)
+#define   C_000E40_CP_CMDSTRM_BUSY                     0xFFFEFFFF
+#define   S_000E40_E2_BUSY(x)                          (((x) & 0x1) << 17)
+#define   G_000E40_E2_BUSY(x)                          (((x) >> 17) & 0x1)
+#define   C_000E40_E2_BUSY                             0xFFFDFFFF
+#define   S_000E40_RB2D_BUSY(x)                        (((x) & 0x1) << 18)
+#define   G_000E40_RB2D_BUSY(x)                        (((x) >> 18) & 0x1)
+#define   C_000E40_RB2D_BUSY                           0xFFFBFFFF
+#define   S_000E40_RB3D_BUSY(x)                        (((x) & 0x1) << 19)
+#define   G_000E40_RB3D_BUSY(x)                        (((x) >> 19) & 0x1)
+#define   C_000E40_RB3D_BUSY                           0xFFF7FFFF
+#define   S_000E40_VAP_BUSY(x)                         (((x) & 0x1) << 20)
+#define   G_000E40_VAP_BUSY(x)                         (((x) >> 20) & 0x1)
+#define   C_000E40_VAP_BUSY                            0xFFEFFFFF
+#define   S_000E40_RE_BUSY(x)                          (((x) & 0x1) << 21)
+#define   G_000E40_RE_BUSY(x)                          (((x) >> 21) & 0x1)
+#define   C_000E40_RE_BUSY                             0xFFDFFFFF
+#define   S_000E40_TAM_BUSY(x)                         (((x) & 0x1) << 22)
+#define   G_000E40_TAM_BUSY(x)                         (((x) >> 22) & 0x1)
+#define   C_000E40_TAM_BUSY                            0xFFBFFFFF
+#define   S_000E40_TDM_BUSY(x)                         (((x) & 0x1) << 23)
+#define   G_000E40_TDM_BUSY(x)                         (((x) >> 23) & 0x1)
+#define   C_000E40_TDM_BUSY                            0xFF7FFFFF
+#define   S_000E40_PB_BUSY(x)                          (((x) & 0x1) << 24)
+#define   G_000E40_PB_BUSY(x)                          (((x) >> 24) & 0x1)
+#define   C_000E40_PB_BUSY                             0xFEFFFFFF
+#define   S_000E40_TIM_BUSY(x)                         (((x) & 0x1) << 25)
+#define   G_000E40_TIM_BUSY(x)                         (((x) >> 25) & 0x1)
+#define   C_000E40_TIM_BUSY                            0xFDFFFFFF
+#define   S_000E40_GA_BUSY(x)                          (((x) & 0x1) << 26)
+#define   G_000E40_GA_BUSY(x)                          (((x) >> 26) & 0x1)
+#define   C_000E40_GA_BUSY                             0xFBFFFFFF
+#define   S_000E40_CBA2D_BUSY(x)                       (((x) & 0x1) << 27)
+#define   G_000E40_CBA2D_BUSY(x)                       (((x) >> 27) & 0x1)
+#define   C_000E40_CBA2D_BUSY                          0xF7FFFFFF
+#define   S_000E40_GUI_ACTIVE(x)                       (((x) & 0x1) << 31)
+#define   G_000E40_GUI_ACTIVE(x)                       (((x) >> 31) & 0x1)
+#define   C_000E40_GUI_ACTIVE                          0x7FFFFFFF
+#define R_0060A4_D1CRTC_STATUS_FRAME_COUNT           0x0060A4
+#define   S_0060A4_D1CRTC_FRAME_COUNT(x)               (((x) & 0xFFFFFF) << 0)
+#define   G_0060A4_D1CRTC_FRAME_COUNT(x)               (((x) >> 0) & 0xFFFFFF)
+#define   C_0060A4_D1CRTC_FRAME_COUNT                  0xFF000000
+#define R_006534_D1MODE_VBLANK_STATUS                0x006534
+#define   S_006534_D1MODE_VBLANK_OCCURRED(x)           (((x) & 0x1) << 0)
+#define   G_006534_D1MODE_VBLANK_OCCURRED(x)           (((x) >> 0) & 0x1)
+#define   C_006534_D1MODE_VBLANK_OCCURRED              0xFFFFFFFE
+#define   S_006534_D1MODE_VBLANK_ACK(x)                (((x) & 0x1) << 4)
+#define   G_006534_D1MODE_VBLANK_ACK(x)                (((x) >> 4) & 0x1)
+#define   C_006534_D1MODE_VBLANK_ACK                   0xFFFFFFEF
+#define   S_006534_D1MODE_VBLANK_STAT(x)               (((x) & 0x1) << 12)
+#define   G_006534_D1MODE_VBLANK_STAT(x)               (((x) >> 12) & 0x1)
+#define   C_006534_D1MODE_VBLANK_STAT                  0xFFFFEFFF
+#define   S_006534_D1MODE_VBLANK_INTERRUPT(x)          (((x) & 0x1) << 16)
+#define   G_006534_D1MODE_VBLANK_INTERRUPT(x)          (((x) >> 16) & 0x1)
+#define   C_006534_D1MODE_VBLANK_INTERRUPT             0xFFFEFFFF
+#define R_006540_DxMODE_INT_MASK                     0x006540
+#define   S_006540_D1MODE_VBLANK_INT_MASK(x)           (((x) & 0x1) << 0)
+#define   G_006540_D1MODE_VBLANK_INT_MASK(x)           (((x) >> 0) & 0x1)
+#define   C_006540_D1MODE_VBLANK_INT_MASK              0xFFFFFFFE
+#define   S_006540_D1MODE_VLINE_INT_MASK(x)            (((x) & 0x1) << 4)
+#define   G_006540_D1MODE_VLINE_INT_MASK(x)            (((x) >> 4) & 0x1)
+#define   C_006540_D1MODE_VLINE_INT_MASK               0xFFFFFFEF
+#define   S_006540_D2MODE_VBLANK_INT_MASK(x)           (((x) & 0x1) << 8)
+#define   G_006540_D2MODE_VBLANK_INT_MASK(x)           (((x) >> 8) & 0x1)
+#define   C_006540_D2MODE_VBLANK_INT_MASK              0xFFFFFEFF
+#define   S_006540_D2MODE_VLINE_INT_MASK(x)            (((x) & 0x1) << 12)
+#define   G_006540_D2MODE_VLINE_INT_MASK(x)            (((x) >> 12) & 0x1)
+#define   C_006540_D2MODE_VLINE_INT_MASK               0xFFFFEFFF
+#define   S_006540_D1MODE_VBLANK_CP_SEL(x)             (((x) & 0x1) << 30)
+#define   G_006540_D1MODE_VBLANK_CP_SEL(x)             (((x) >> 30) & 0x1)
+#define   C_006540_D1MODE_VBLANK_CP_SEL                0xBFFFFFFF
+#define   S_006540_D2MODE_VBLANK_CP_SEL(x)             (((x) & 0x1) << 31)
+#define   G_006540_D2MODE_VBLANK_CP_SEL(x)             (((x) >> 31) & 0x1)
+#define   C_006540_D2MODE_VBLANK_CP_SEL                0x7FFFFFFF
+#define R_0068A4_D2CRTC_STATUS_FRAME_COUNT           0x0068A4
+#define   S_0068A4_D2CRTC_FRAME_COUNT(x)               (((x) & 0xFFFFFF) << 0)
+#define   G_0068A4_D2CRTC_FRAME_COUNT(x)               (((x) >> 0) & 0xFFFFFF)
+#define   C_0068A4_D2CRTC_FRAME_COUNT                  0xFF000000
+#define R_006D34_D2MODE_VBLANK_STATUS                0x006D34
+#define   S_006D34_D2MODE_VBLANK_OCCURRED(x)           (((x) & 0x1) << 0)
+#define   G_006D34_D2MODE_VBLANK_OCCURRED(x)           (((x) >> 0) & 0x1)
+#define   C_006D34_D2MODE_VBLANK_OCCURRED              0xFFFFFFFE
+#define   S_006D34_D2MODE_VBLANK_ACK(x)                (((x) & 0x1) << 4)
+#define   G_006D34_D2MODE_VBLANK_ACK(x)                (((x) >> 4) & 0x1)
+#define   C_006D34_D2MODE_VBLANK_ACK                   0xFFFFFFEF
+#define   S_006D34_D2MODE_VBLANK_STAT(x)               (((x) & 0x1) << 12)
+#define   G_006D34_D2MODE_VBLANK_STAT(x)               (((x) >> 12) & 0x1)
+#define   C_006D34_D2MODE_VBLANK_STAT                  0xFFFFEFFF
+#define   S_006D34_D2MODE_VBLANK_INTERRUPT(x)          (((x) & 0x1) << 16)
+#define   G_006D34_D2MODE_VBLANK_INTERRUPT(x)          (((x) >> 16) & 0x1)
+#define   C_006D34_D2MODE_VBLANK_INTERRUPT             0xFFFEFFFF
+#define R_007EDC_DISP_INTERRUPT_STATUS               0x007EDC
+#define   S_007EDC_LB_D1_VBLANK_INTERRUPT(x)           (((x) & 0x1) << 4)
+#define   G_007EDC_LB_D1_VBLANK_INTERRUPT(x)           (((x) >> 4) & 0x1)
+#define   C_007EDC_LB_D1_VBLANK_INTERRUPT              0xFFFFFFEF
+#define   S_007EDC_LB_D2_VBLANK_INTERRUPT(x)           (((x) & 0x1) << 5)
+#define   G_007EDC_LB_D2_VBLANK_INTERRUPT(x)           (((x) >> 5) & 0x1)
+#define   C_007EDC_LB_D2_VBLANK_INTERRUPT              0xFFFFFFDF
+#define   S_007EDC_DACA_AUTODETECT_INTERRUPT(x)        (((x) & 0x1) << 16)
+#define   G_007EDC_DACA_AUTODETECT_INTERRUPT(x)        (((x) >> 16) & 0x1)
+#define   C_007EDC_DACA_AUTODETECT_INTERRUPT           0xFFFEFFFF
+#define   S_007EDC_DACB_AUTODETECT_INTERRUPT(x)        (((x) & 0x1) << 17)
+#define   G_007EDC_DACB_AUTODETECT_INTERRUPT(x)        (((x) >> 17) & 0x1)
+#define   C_007EDC_DACB_AUTODETECT_INTERRUPT           0xFFFDFFFF
+#define   S_007EDC_DC_HOT_PLUG_DETECT1_INTERRUPT(x)    (((x) & 0x1) << 18)
+#define   G_007EDC_DC_HOT_PLUG_DETECT1_INTERRUPT(x)    (((x) >> 18) & 0x1)
+#define   C_007EDC_DC_HOT_PLUG_DETECT1_INTERRUPT       0xFFFBFFFF
+#define   S_007EDC_DC_HOT_PLUG_DETECT2_INTERRUPT(x)    (((x) & 0x1) << 19)
+#define   G_007EDC_DC_HOT_PLUG_DETECT2_INTERRUPT(x)    (((x) >> 19) & 0x1)
+#define   C_007EDC_DC_HOT_PLUG_DETECT2_INTERRUPT       0xFFF7FFFF
+#define R_007828_DACA_AUTODETECT_CONTROL               0x007828
+#define   S_007828_DACA_AUTODETECT_MODE(x)             (((x) & 0x3) << 0)
+#define   G_007828_DACA_AUTODETECT_MODE(x)             (((x) >> 0) & 0x3)
+#define   C_007828_DACA_AUTODETECT_MODE                0xFFFFFFFC
+#define   S_007828_DACA_AUTODETECT_FRAME_TIME_COUNTER(x) (((x) & 0xff) << 8)
+#define   G_007828_DACA_AUTODETECT_FRAME_TIME_COUNTER(x) (((x) >> 8) & 0xff)
+#define   C_007828_DACA_AUTODETECT_FRAME_TIME_COUNTER  0xFFFF00FF
+#define   S_007828_DACA_AUTODETECT_CHECK_MASK(x)       (((x) & 0x3) << 16)
+#define   G_007828_DACA_AUTODETECT_CHECK_MASK(x)       (((x) >> 16) & 0x3)
+#define   C_007828_DACA_AUTODETECT_CHECK_MASK          0xFFFCFFFF
+#define R_007838_DACA_AUTODETECT_INT_CONTROL           0x007838
+#define   S_007838_DACA_AUTODETECT_ACK(x)              (((x) & 0x1) << 0)
+#define   C_007838_DACA_DACA_AUTODETECT_ACK            0xFFFFFFFE
+#define   S_007838_DACA_AUTODETECT_INT_ENABLE(x)       (((x) & 0x1) << 16)
+#define   G_007838_DACA_AUTODETECT_INT_ENABLE(x)       (((x) >> 16) & 0x1)
+#define   C_007838_DACA_AUTODETECT_INT_ENABLE          0xFFFCFFFF
+#define R_007A28_DACB_AUTODETECT_CONTROL               0x007A28
+#define   S_007A28_DACB_AUTODETECT_MODE(x)             (((x) & 0x3) << 0)
+#define   G_007A28_DACB_AUTODETECT_MODE(x)             (((x) >> 0) & 0x3)
+#define   C_007A28_DACB_AUTODETECT_MODE                0xFFFFFFFC
+#define   S_007A28_DACB_AUTODETECT_FRAME_TIME_COUNTER(x) (((x) & 0xff) << 8)
+#define   G_007A28_DACB_AUTODETECT_FRAME_TIME_COUNTER(x) (((x) >> 8) & 0xff)
+#define   C_007A28_DACB_AUTODETECT_FRAME_TIME_COUNTER  0xFFFF00FF
+#define   S_007A28_DACB_AUTODETECT_CHECK_MASK(x)       (((x) & 0x3) << 16)
+#define   G_007A28_DACB_AUTODETECT_CHECK_MASK(x)       (((x) >> 16) & 0x3)
+#define   C_007A28_DACB_AUTODETECT_CHECK_MASK          0xFFFCFFFF
+#define R_007A38_DACB_AUTODETECT_INT_CONTROL           0x007A38
+#define   S_007A38_DACB_AUTODETECT_ACK(x)              (((x) & 0x1) << 0)
+#define   C_007A38_DACB_DACA_AUTODETECT_ACK            0xFFFFFFFE
+#define   S_007A38_DACB_AUTODETECT_INT_ENABLE(x)       (((x) & 0x1) << 16)
+#define   G_007A38_DACB_AUTODETECT_INT_ENABLE(x)       (((x) >> 16) & 0x1)
+#define   C_007A38_DACB_AUTODETECT_INT_ENABLE          0xFFFCFFFF
+#define R_007D00_DC_HOT_PLUG_DETECT1_CONTROL           0x007D00
+#define   S_007D00_DC_HOT_PLUG_DETECT1_EN(x)           (((x) & 0x1) << 0)
+#define   G_007D00_DC_HOT_PLUG_DETECT1_EN(x)           (((x) >> 0) & 0x1)
+#define   C_007D00_DC_HOT_PLUG_DETECT1_EN              0xFFFFFFFE
+#define R_007D04_DC_HOT_PLUG_DETECT1_INT_STATUS        0x007D04
+#define   S_007D04_DC_HOT_PLUG_DETECT1_INT_STATUS(x)   (((x) & 0x1) << 0)
+#define   G_007D04_DC_HOT_PLUG_DETECT1_INT_STATUS(x)   (((x) >> 0) & 0x1)
+#define   C_007D04_DC_HOT_PLUG_DETECT1_INT_STATUS      0xFFFFFFFE
+#define   S_007D04_DC_HOT_PLUG_DETECT1_SENSE(x)        (((x) & 0x1) << 1)
+#define   G_007D04_DC_HOT_PLUG_DETECT1_SENSE(x)        (((x) >> 1) & 0x1)
+#define   C_007D04_DC_HOT_PLUG_DETECT1_SENSE           0xFFFFFFFD
+#define R_007D08_DC_HOT_PLUG_DETECT1_INT_CONTROL       0x007D08
+#define   S_007D08_DC_HOT_PLUG_DETECT1_INT_ACK(x)      (((x) & 0x1) << 0)
+#define   C_007D08_DC_HOT_PLUG_DETECT1_INT_ACK         0xFFFFFFFE
+#define   S_007D08_DC_HOT_PLUG_DETECT1_INT_POLARITY(x) (((x) & 0x1) << 8)
+#define   G_007D08_DC_HOT_PLUG_DETECT1_INT_POLARITY(x) (((x) >> 8) & 0x1)
+#define   C_007D08_DC_HOT_PLUG_DETECT1_INT_POLARITY    0xFFFFFEFF
+#define   S_007D08_DC_HOT_PLUG_DETECT1_INT_EN(x)       (((x) & 0x1) << 16)
+#define   G_007D08_DC_HOT_PLUG_DETECT1_INT_EN(x)       (((x) >> 16) & 0x1)
+#define   C_007D08_DC_HOT_PLUG_DETECT1_INT_EN          0xFFFEFFFF
+#define R_007D10_DC_HOT_PLUG_DETECT2_CONTROL           0x007D10
+#define   S_007D10_DC_HOT_PLUG_DETECT2_EN(x)           (((x) & 0x1) << 0)
+#define   G_007D10_DC_HOT_PLUG_DETECT2_EN(x)           (((x) >> 0) & 0x1)
+#define   C_007D10_DC_HOT_PLUG_DETECT2_EN              0xFFFFFFFE
+#define R_007D14_DC_HOT_PLUG_DETECT2_INT_STATUS        0x007D14
+#define   S_007D14_DC_HOT_PLUG_DETECT2_INT_STATUS(x)   (((x) & 0x1) << 0)
+#define   G_007D14_DC_HOT_PLUG_DETECT2_INT_STATUS(x)   (((x) >> 0) & 0x1)
+#define   C_007D14_DC_HOT_PLUG_DETECT2_INT_STATUS      0xFFFFFFFE
+#define   S_007D14_DC_HOT_PLUG_DETECT2_SENSE(x)        (((x) & 0x1) << 1)
+#define   G_007D14_DC_HOT_PLUG_DETECT2_SENSE(x)        (((x) >> 1) & 0x1)
+#define   C_007D14_DC_HOT_PLUG_DETECT2_SENSE           0xFFFFFFFD
+#define R_007D18_DC_HOT_PLUG_DETECT2_INT_CONTROL       0x007D18
+#define   S_007D18_DC_HOT_PLUG_DETECT2_INT_ACK(x)      (((x) & 0x1) << 0)
+#define   C_007D18_DC_HOT_PLUG_DETECT2_INT_ACK         0xFFFFFFFE
+#define   S_007D18_DC_HOT_PLUG_DETECT2_INT_POLARITY(x) (((x) & 0x1) << 8)
+#define   G_007D18_DC_HOT_PLUG_DETECT2_INT_POLARITY(x) (((x) >> 8) & 0x1)
+#define   C_007D18_DC_HOT_PLUG_DETECT2_INT_POLARITY    0xFFFFFEFF
+#define   S_007D18_DC_HOT_PLUG_DETECT2_INT_EN(x)       (((x) & 0x1) << 16)
+#define   G_007D18_DC_HOT_PLUG_DETECT2_INT_EN(x)       (((x) >> 16) & 0x1)
+#define   C_007D18_DC_HOT_PLUG_DETECT2_INT_EN          0xFFFEFFFF
+
+/* MC registers */
+#define R_000000_MC_STATUS                           0x000000
+#define   S_000000_MC_IDLE(x)                          (((x) & 0x1) << 0)
+#define   G_000000_MC_IDLE(x)                          (((x) >> 0) & 0x1)
+#define   C_000000_MC_IDLE                             0xFFFFFFFE
+#define R_000004_MC_FB_LOCATION                      0x000004
+#define   S_000004_MC_FB_START(x)                      (((x) & 0xFFFF) << 0)
+#define   G_000004_MC_FB_START(x)                      (((x) >> 0) & 0xFFFF)
+#define   C_000004_MC_FB_START                         0xFFFF0000
+#define   S_000004_MC_FB_TOP(x)                        (((x) & 0xFFFF) << 16)
+#define   G_000004_MC_FB_TOP(x)                        (((x) >> 16) & 0xFFFF)
+#define   C_000004_MC_FB_TOP                           0x0000FFFF
+#define R_000005_MC_AGP_LOCATION                     0x000005
+#define   S_000005_MC_AGP_START(x)                     (((x) & 0xFFFF) << 0)
+#define   G_000005_MC_AGP_START(x)                     (((x) >> 0) & 0xFFFF)
+#define   C_000005_MC_AGP_START                        0xFFFF0000
+#define   S_000005_MC_AGP_TOP(x)                       (((x) & 0xFFFF) << 16)
+#define   G_000005_MC_AGP_TOP(x)                       (((x) >> 16) & 0xFFFF)
+#define   C_000005_MC_AGP_TOP                          0x0000FFFF
+#define R_000006_AGP_BASE                            0x000006
+#define   S_000006_AGP_BASE_ADDR(x)                    (((x) & 0xFFFFFFFF) << 0)
+#define   G_000006_AGP_BASE_ADDR(x)                    (((x) >> 0) & 0xFFFFFFFF)
+#define   C_000006_AGP_BASE_ADDR                       0x00000000
+#define R_000007_AGP_BASE_2                          0x000007
+#define   S_000007_AGP_BASE_ADDR_2(x)                  (((x) & 0xF) << 0)
+#define   G_000007_AGP_BASE_ADDR_2(x)                  (((x) >> 0) & 0xF)
+#define   C_000007_AGP_BASE_ADDR_2                     0xFFFFFFF0
+#define R_000009_MC_CNTL1                            0x000009
+#define   S_000009_ENABLE_PAGE_TABLES(x)               (((x) & 0x1) << 26)
+#define   G_000009_ENABLE_PAGE_TABLES(x)               (((x) >> 26) & 0x1)
+#define   C_000009_ENABLE_PAGE_TABLES                  0xFBFFFFFF
+/* FIXME don't know the various field size need feedback from AMD */
+#define R_000100_MC_PT0_CNTL                         0x000100
+#define   S_000100_ENABLE_PT(x)                        (((x) & 0x1) << 0)
+#define   G_000100_ENABLE_PT(x)                        (((x) >> 0) & 0x1)
+#define   C_000100_ENABLE_PT                           0xFFFFFFFE
+#define   S_000100_EFFECTIVE_L2_CACHE_SIZE(x)          (((x) & 0x7) << 15)
+#define   G_000100_EFFECTIVE_L2_CACHE_SIZE(x)          (((x) >> 15) & 0x7)
+#define   C_000100_EFFECTIVE_L2_CACHE_SIZE             0xFFFC7FFF
+#define   S_000100_EFFECTIVE_L2_QUEUE_SIZE(x)          (((x) & 0x7) << 21)
+#define   G_000100_EFFECTIVE_L2_QUEUE_SIZE(x)          (((x) >> 21) & 0x7)
+#define   C_000100_EFFECTIVE_L2_QUEUE_SIZE             0xFF1FFFFF
+#define   S_000100_INVALIDATE_ALL_L1_TLBS(x)           (((x) & 0x1) << 28)
+#define   G_000100_INVALIDATE_ALL_L1_TLBS(x)           (((x) >> 28) & 0x1)
+#define   C_000100_INVALIDATE_ALL_L1_TLBS              0xEFFFFFFF
+#define   S_000100_INVALIDATE_L2_CACHE(x)              (((x) & 0x1) << 29)
+#define   G_000100_INVALIDATE_L2_CACHE(x)              (((x) >> 29) & 0x1)
+#define   C_000100_INVALIDATE_L2_CACHE                 0xDFFFFFFF
+#define R_000102_MC_PT0_CONTEXT0_CNTL                0x000102
+#define   S_000102_ENABLE_PAGE_TABLE(x)                (((x) & 0x1) << 0)
+#define   G_000102_ENABLE_PAGE_TABLE(x)                (((x) >> 0) & 0x1)
+#define   C_000102_ENABLE_PAGE_TABLE                   0xFFFFFFFE
+#define   S_000102_PAGE_TABLE_DEPTH(x)                 (((x) & 0x3) << 1)
+#define   G_000102_PAGE_TABLE_DEPTH(x)                 (((x) >> 1) & 0x3)
+#define   C_000102_PAGE_TABLE_DEPTH                    0xFFFFFFF9
+#define   V_000102_PAGE_TABLE_FLAT                     0
+/* R600 documentation suggest that this should be a number of pages */
+#define R_000112_MC_PT0_SYSTEM_APERTURE_LOW_ADDR     0x000112
+#define R_000114_MC_PT0_SYSTEM_APERTURE_HIGH_ADDR    0x000114
+#define R_00011C_MC_PT0_CONTEXT0_DEFAULT_READ_ADDR   0x00011C
+#define R_00012C_MC_PT0_CONTEXT0_FLAT_BASE_ADDR      0x00012C
+#define R_00013C_MC_PT0_CONTEXT0_FLAT_START_ADDR     0x00013C
+#define R_00014C_MC_PT0_CONTEXT0_FLAT_END_ADDR       0x00014C
+#define R_00016C_MC_PT0_CLIENT0_CNTL                 0x00016C
+#define   S_00016C_ENABLE_TRANSLATION_MODE_OVERRIDE(x) (((x) & 0x1) << 0)
+#define   G_00016C_ENABLE_TRANSLATION_MODE_OVERRIDE(x) (((x) >> 0) & 0x1)
+#define   C_00016C_ENABLE_TRANSLATION_MODE_OVERRIDE    0xFFFFFFFE
+#define   S_00016C_TRANSLATION_MODE_OVERRIDE(x)        (((x) & 0x1) << 1)
+#define   G_00016C_TRANSLATION_MODE_OVERRIDE(x)        (((x) >> 1) & 0x1)
+#define   C_00016C_TRANSLATION_MODE_OVERRIDE           0xFFFFFFFD
+#define   S_00016C_SYSTEM_ACCESS_MODE_MASK(x)          (((x) & 0x3) << 8)
+#define   G_00016C_SYSTEM_ACCESS_MODE_MASK(x)          (((x) >> 8) & 0x3)
+#define   C_00016C_SYSTEM_ACCESS_MODE_MASK             0xFFFFFCFF
+#define   V_00016C_SYSTEM_ACCESS_MODE_PA_ONLY          0
+#define   V_00016C_SYSTEM_ACCESS_MODE_USE_SYS_MAP      1
+#define   V_00016C_SYSTEM_ACCESS_MODE_IN_SYS           2
+#define   V_00016C_SYSTEM_ACCESS_MODE_NOT_IN_SYS       3
+#define   S_00016C_SYSTEM_APERTURE_UNMAPPED_ACCESS(x)  (((x) & 0x1) << 10)
+#define   G_00016C_SYSTEM_APERTURE_UNMAPPED_ACCESS(x)  (((x) >> 10) & 0x1)
+#define   C_00016C_SYSTEM_APERTURE_UNMAPPED_ACCESS     0xFFFFFBFF
+#define   V_00016C_SYSTEM_APERTURE_UNMAPPED_PASSTHROUGH  0
+#define   V_00016C_SYSTEM_APERTURE_UNMAPPED_DEFAULT_PAGE 1
+#define   S_00016C_EFFECTIVE_L1_CACHE_SIZE(x)          (((x) & 0x7) << 11)
+#define   G_00016C_EFFECTIVE_L1_CACHE_SIZE(x)          (((x) >> 11) & 0x7)
+#define   C_00016C_EFFECTIVE_L1_CACHE_SIZE             0xFFFFC7FF
+#define   S_00016C_ENABLE_FRAGMENT_PROCESSING(x)       (((x) & 0x1) << 14)
+#define   G_00016C_ENABLE_FRAGMENT_PROCESSING(x)       (((x) >> 14) & 0x1)
+#define   C_00016C_ENABLE_FRAGMENT_PROCESSING          0xFFFFBFFF
+#define   S_00016C_EFFECTIVE_L1_QUEUE_SIZE(x)          (((x) & 0x7) << 15)
+#define   G_00016C_EFFECTIVE_L1_QUEUE_SIZE(x)          (((x) >> 15) & 0x7)
+#define   C_00016C_EFFECTIVE_L1_QUEUE_SIZE             0xFFFC7FFF
+#define   S_00016C_INVALIDATE_L1_TLB(x)                (((x) & 0x1) << 20)
+#define   G_00016C_INVALIDATE_L1_TLB(x)                (((x) >> 20) & 0x1)
+#define   C_00016C_INVALIDATE_L1_TLB                   0xFFEFFFFF
+
+#endif
diff --git a/drivers/gpu/drm/radeon/rs690.c b/drivers/gpu/drm/radeon/rs690.c
index 0f585ca8276d..83b9174f76f2 100644
--- a/drivers/gpu/drm/radeon/rs690.c
+++ b/drivers/gpu/drm/radeon/rs690.c
@@ -26,106 +26,29 @@
  *          Jerome Glisse
  */
 #include "drmP.h"
-#include "radeon_reg.h"
 #include "radeon.h"
-#include "rs690r.h"
 #include "atom.h"
-#include "atom-bits.h"
-
-/* rs690,rs740 depends on : */
-void r100_hdp_reset(struct radeon_device *rdev);
-int r300_mc_wait_for_idle(struct radeon_device *rdev);
-void r420_pipes_init(struct radeon_device *rdev);
-void rs400_gart_disable(struct radeon_device *rdev);
-int rs400_gart_enable(struct radeon_device *rdev);
-void rs400_gart_adjust_size(struct radeon_device *rdev);
-void rs600_mc_disable_clients(struct radeon_device *rdev);
-void rs600_disable_vga(struct radeon_device *rdev);
-
-/* This files gather functions specifics to :
- * rs690,rs740
- *
- * Some of these functions might be used by newer ASICs.
- */
-void rs690_gpu_init(struct radeon_device *rdev);
-int rs690_mc_wait_for_idle(struct radeon_device *rdev);
-
-
-/*
- * MC functions.
- */
-int rs690_mc_init(struct radeon_device *rdev)
-{
-	uint32_t tmp;
-	int r;
-
-	if (r100_debugfs_rbbm_init(rdev)) {
-		DRM_ERROR("Failed to register debugfs file for RBBM !\n");
-	}
-
-	rs690_gpu_init(rdev);
-	rs400_gart_disable(rdev);
-
-	/* Setup GPU memory space */
-	rdev->mc.gtt_location = rdev->mc.mc_vram_size;
-	rdev->mc.gtt_location += (rdev->mc.gtt_size - 1);
-	rdev->mc.gtt_location &= ~(rdev->mc.gtt_size - 1);
-	rdev->mc.vram_location = 0xFFFFFFFFUL;
-	r = radeon_mc_setup(rdev);
-	if (r) {
-		return r;
-	}
-
-	/* Program GPU memory space */
-	rs600_mc_disable_clients(rdev);
-	if (rs690_mc_wait_for_idle(rdev)) {
-		printk(KERN_WARNING "Failed to wait MC idle while "
-		       "programming pipes. Bad things might happen.\n");
-	}
-	tmp = rdev->mc.vram_location + rdev->mc.mc_vram_size - 1;
-	tmp = REG_SET(RS690_MC_FB_TOP, tmp >> 16);
-	tmp |= REG_SET(RS690_MC_FB_START, rdev->mc.vram_location >> 16);
-	WREG32_MC(RS690_MCCFG_FB_LOCATION, tmp);
-	/* FIXME: Does this reg exist on RS480,RS740 ? */
-	WREG32(0x310, rdev->mc.vram_location);
-	WREG32(RS690_HDP_FB_LOCATION, rdev->mc.vram_location >> 16);
-	return 0;
-}
-
-void rs690_mc_fini(struct radeon_device *rdev)
-{
-}
-
+#include "rs690d.h"
 
-/*
- * Global GPU functions
- */
-int rs690_mc_wait_for_idle(struct radeon_device *rdev)
+static int rs690_mc_wait_for_idle(struct radeon_device *rdev)
 {
 	unsigned i;
 	uint32_t tmp;
 
 	for (i = 0; i < rdev->usec_timeout; i++) {
 		/* read MC_STATUS */
-		tmp = RREG32_MC(RS690_MC_STATUS);
-		if (tmp & RS690_MC_STATUS_IDLE) {
+		tmp = RREG32_MC(R_000090_MC_SYSTEM_STATUS);
+		if (G_000090_MC_SYSTEM_IDLE(tmp))
 			return 0;
-		}
-		DRM_UDELAY(1);
+		udelay(1);
 	}
 	return -1;
 }
 
-void rs690_errata(struct radeon_device *rdev)
-{
-	rdev->pll_errata = 0;
-}
-
-void rs690_gpu_init(struct radeon_device *rdev)
+static void rs690_gpu_init(struct radeon_device *rdev)
 {
 	/* FIXME: HDP same place on rs690 ? */
 	r100_hdp_reset(rdev);
-	rs600_disable_vga(rdev);
 	/* FIXME: is this correct ? */
 	r420_pipes_init(rdev);
 	if (rs690_mc_wait_for_idle(rdev)) {
@@ -134,10 +57,6 @@ void rs690_gpu_init(struct radeon_device *rdev)
 	}
 }
 
-
-/*
- * VRAM info.
- */
 void rs690_pm_info(struct radeon_device *rdev)
 {
 	int index = GetIndexIntoMasterTable(DATA, IntegratedSystemInfo);
@@ -210,26 +129,21 @@ void rs690_pm_info(struct radeon_device *rdev)
 	rdev->pm.sideport_bandwidth.full = rfixed_div(rdev->pm.sideport_bandwidth, tmp);
 }
 
-void rs690_vram_info(struct radeon_device *rdev)
+void rs690_mc_init(struct radeon_device *rdev)
 {
-	uint32_t tmp;
 	fixed20_12 a;
+	u64 base;
 
 	rs400_gart_adjust_size(rdev);
-	/* DDR for all card after R300 & IGP */
 	rdev->mc.vram_is_ddr = true;
-	/* FIXME: is this correct for RS690/RS740 ? */
-	tmp = RREG32(RADEON_MEM_CNTL);
-	if (tmp & R300_MEM_NUM_CHANNELS_MASK) {
-		rdev->mc.vram_width = 128;
-	} else {
-		rdev->mc.vram_width = 64;
-	}
+	rdev->mc.vram_width = 128;
 	rdev->mc.real_vram_size = RREG32(RADEON_CONFIG_MEMSIZE);
 	rdev->mc.mc_vram_size = rdev->mc.real_vram_size;
-
 	rdev->mc.aper_base = drm_get_resource_start(rdev->ddev, 0);
 	rdev->mc.aper_size = drm_get_resource_len(rdev->ddev, 0);
+	rdev->mc.visible_vram_size = rdev->mc.aper_size;
+	base = RREG32_MC(R_000100_MCCFG_FB_LOCATION);
+	base = G_000100_MC_FB_START(base) << 16;
 	rs690_pm_info(rdev);
 	/* FIXME: we should enforce default clock in case GPU is not in
 	 * default setup
@@ -240,6 +154,9 @@ void rs690_vram_info(struct radeon_device *rdev)
 	a.full = rfixed_const(16);
 	/* core_bandwidth = sclk(Mhz) * 16 */
 	rdev->pm.core_bandwidth.full = rfixed_div(rdev->pm.sclk, a);
+	rdev->mc.igp_sideport_enabled = radeon_atombios_sideport_present(rdev);
+	radeon_vram_location(rdev, &rdev->mc, base);
+	radeon_gtt_location(rdev, &rdev->mc);
 }
 
 void rs690_line_buffer_adjust(struct radeon_device *rdev,
@@ -251,39 +168,39 @@ void rs690_line_buffer_adjust(struct radeon_device *rdev,
 	/*
 	 * Line Buffer Setup
 	 * There is a single line buffer shared by both display controllers.
-	 * DC_LB_MEMORY_SPLIT controls how that line buffer is shared between
+	 * R_006520_DC_LB_MEMORY_SPLIT controls how that line buffer is shared between
 	 * the display controllers.  The paritioning can either be done
 	 * manually or via one of four preset allocations specified in bits 1:0:
 	 *  0 - line buffer is divided in half and shared between crtc
 	 *  1 - D1 gets 3/4 of the line buffer, D2 gets 1/4
 	 *  2 - D1 gets the whole buffer
 	 *  3 - D1 gets 1/4 of the line buffer, D2 gets 3/4
-	 * Setting bit 2 of DC_LB_MEMORY_SPLIT controls switches to manual
+	 * Setting bit 2 of R_006520_DC_LB_MEMORY_SPLIT controls switches to manual
 	 * allocation mode. In manual allocation mode, D1 always starts at 0,
 	 * D1 end/2 is specified in bits 14:4; D2 allocation follows D1.
 	 */
-	tmp = RREG32(DC_LB_MEMORY_SPLIT) & ~DC_LB_MEMORY_SPLIT_MASK;
-	tmp &= ~DC_LB_MEMORY_SPLIT_SHIFT_MODE;
+	tmp = RREG32(R_006520_DC_LB_MEMORY_SPLIT) & C_006520_DC_LB_MEMORY_SPLIT;
+	tmp &= ~C_006520_DC_LB_MEMORY_SPLIT_MODE;
 	/* auto */
 	if (mode1 && mode2) {
 		if (mode1->hdisplay > mode2->hdisplay) {
 			if (mode1->hdisplay > 2560)
-				tmp |= DC_LB_MEMORY_SPLIT_D1_3Q_D2_1Q;
+				tmp |= V_006520_DC_LB_MEMORY_SPLIT_D1_3Q_D2_1Q;
 			else
-				tmp |= DC_LB_MEMORY_SPLIT_D1HALF_D2HALF;
+				tmp |= V_006520_DC_LB_MEMORY_SPLIT_D1HALF_D2HALF;
 		} else if (mode2->hdisplay > mode1->hdisplay) {
 			if (mode2->hdisplay > 2560)
-				tmp |= DC_LB_MEMORY_SPLIT_D1_1Q_D2_3Q;
+				tmp |= V_006520_DC_LB_MEMORY_SPLIT_D1_1Q_D2_3Q;
 			else
-				tmp |= DC_LB_MEMORY_SPLIT_D1HALF_D2HALF;
+				tmp |= V_006520_DC_LB_MEMORY_SPLIT_D1HALF_D2HALF;
 		} else
-			tmp |= AVIVO_DC_LB_MEMORY_SPLIT_D1HALF_D2HALF;
+			tmp |= V_006520_DC_LB_MEMORY_SPLIT_D1HALF_D2HALF;
 	} else if (mode1) {
-		tmp |= DC_LB_MEMORY_SPLIT_D1_ONLY;
+		tmp |= V_006520_DC_LB_MEMORY_SPLIT_D1_ONLY;
 	} else if (mode2) {
-		tmp |= DC_LB_MEMORY_SPLIT_D1_1Q_D2_3Q;
+		tmp |= V_006520_DC_LB_MEMORY_SPLIT_D1_1Q_D2_3Q;
 	}
-	WREG32(DC_LB_MEMORY_SPLIT, tmp);
+	WREG32(R_006520_DC_LB_MEMORY_SPLIT, tmp);
 }
 
 struct rs690_watermark {
@@ -325,8 +242,9 @@ void rs690_crtc_bandwidth_compute(struct radeon_device *rdev,
 
 	b.full = rfixed_const(mode->crtc_hdisplay);
 	c.full = rfixed_const(256);
-	a.full = rfixed_mul(wm->num_line_pair, b);
-	request_fifo_depth.full = rfixed_div(a, c);
+	a.full = rfixed_div(b, c);
+	request_fifo_depth.full = rfixed_mul(a, wm->num_line_pair);
+	request_fifo_depth.full = rfixed_ceil(request_fifo_depth);
 	if (a.full < rfixed_const(4)) {
 		wm->lb_request_fifo_depth = 4;
 	} else {
@@ -455,6 +373,7 @@ void rs690_crtc_bandwidth_compute(struct radeon_device *rdev,
 	a.full = rfixed_const(16);
 	wm->priority_mark_max.full = rfixed_const(crtc->base.mode.crtc_hdisplay);
 	wm->priority_mark_max.full = rfixed_div(wm->priority_mark_max, a);
+	wm->priority_mark_max.full = rfixed_ceil(wm->priority_mark_max);
 
 	/* Determine estimated width */
 	estimated_width.full = tolerable_latency.full - wm->worst_case_latency.full;
@@ -464,6 +383,7 @@ void rs690_crtc_bandwidth_compute(struct radeon_device *rdev,
 	} else {
 		a.full = rfixed_const(16);
 		wm->priority_mark.full = rfixed_div(estimated_width, a);
+		wm->priority_mark.full = rfixed_ceil(wm->priority_mark);
 		wm->priority_mark.full = wm->priority_mark_max.full - wm->priority_mark.full;
 	}
 }
@@ -488,28 +408,28 @@ void rs690_bandwidth_update(struct radeon_device *rdev)
 	 * option.
 	 */
 	if (rdev->disp_priority == 2) {
-		tmp = RREG32_MC(MC_INIT_MISC_LAT_TIMER);
-		tmp &= ~MC_DISP1R_INIT_LAT_MASK;
-		tmp &= ~MC_DISP0R_INIT_LAT_MASK;
-		if (mode1)
-			tmp |= (1 << MC_DISP1R_INIT_LAT_SHIFT);
+		tmp = RREG32_MC(R_000104_MC_INIT_MISC_LAT_TIMER);
+		tmp &= C_000104_MC_DISP0R_INIT_LAT;
+		tmp &= C_000104_MC_DISP1R_INIT_LAT;
 		if (mode0)
-			tmp |= (1 << MC_DISP0R_INIT_LAT_SHIFT);
-		WREG32_MC(MC_INIT_MISC_LAT_TIMER, tmp);
+			tmp |= S_000104_MC_DISP0R_INIT_LAT(1);
+		if (mode1)
+			tmp |= S_000104_MC_DISP1R_INIT_LAT(1);
+		WREG32_MC(R_000104_MC_INIT_MISC_LAT_TIMER, tmp);
 	}
 	rs690_line_buffer_adjust(rdev, mode0, mode1);
 
 	if ((rdev->family == CHIP_RS690) || (rdev->family == CHIP_RS740))
-		WREG32(DCP_CONTROL, 0);
+		WREG32(R_006C9C_DCP_CONTROL, 0);
 	if ((rdev->family == CHIP_RS780) || (rdev->family == CHIP_RS880))
-		WREG32(DCP_CONTROL, 2);
+		WREG32(R_006C9C_DCP_CONTROL, 2);
 
 	rs690_crtc_bandwidth_compute(rdev, rdev->mode_info.crtcs[0], &wm0);
 	rs690_crtc_bandwidth_compute(rdev, rdev->mode_info.crtcs[1], &wm1);
 
 	tmp = (wm0.lb_request_fifo_depth - 1);
 	tmp |= (wm1.lb_request_fifo_depth - 1) << 16;
-	WREG32(LB_MAX_REQ_OUTSTANDING, tmp);
+	WREG32(R_006D58_LB_MAX_REQ_OUTSTANDING, tmp);
 
 	if (mode0 && mode1) {
 		if (rfixed_trunc(wm0.dbpp) > 64)
@@ -562,10 +482,10 @@ void rs690_bandwidth_update(struct radeon_device *rdev)
 			priority_mark12.full = 0;
 		if (wm1.priority_mark_max.full > priority_mark12.full)
 			priority_mark12.full = wm1.priority_mark_max.full;
-		WREG32(D1MODE_PRIORITY_A_CNT, rfixed_trunc(priority_mark02));
-		WREG32(D1MODE_PRIORITY_B_CNT, rfixed_trunc(priority_mark02));
-		WREG32(D2MODE_PRIORITY_A_CNT, rfixed_trunc(priority_mark12));
-		WREG32(D2MODE_PRIORITY_B_CNT, rfixed_trunc(priority_mark12));
+		WREG32(R_006548_D1MODE_PRIORITY_A_CNT, rfixed_trunc(priority_mark02));
+		WREG32(R_00654C_D1MODE_PRIORITY_B_CNT, rfixed_trunc(priority_mark02));
+		WREG32(R_006D48_D2MODE_PRIORITY_A_CNT, rfixed_trunc(priority_mark12));
+		WREG32(R_006D4C_D2MODE_PRIORITY_B_CNT, rfixed_trunc(priority_mark12));
 	} else if (mode0) {
 		if (rfixed_trunc(wm0.dbpp) > 64)
 			a.full = rfixed_mul(wm0.dbpp, wm0.num_line_pair);
@@ -592,10 +512,12 @@ void rs690_bandwidth_update(struct radeon_device *rdev)
 			priority_mark02.full = 0;
 		if (wm0.priority_mark_max.full > priority_mark02.full)
 			priority_mark02.full = wm0.priority_mark_max.full;
-		WREG32(D1MODE_PRIORITY_A_CNT, rfixed_trunc(priority_mark02));
-		WREG32(D1MODE_PRIORITY_B_CNT, rfixed_trunc(priority_mark02));
-		WREG32(D2MODE_PRIORITY_A_CNT, MODE_PRIORITY_OFF);
-		WREG32(D2MODE_PRIORITY_B_CNT, MODE_PRIORITY_OFF);
+		WREG32(R_006548_D1MODE_PRIORITY_A_CNT, rfixed_trunc(priority_mark02));
+		WREG32(R_00654C_D1MODE_PRIORITY_B_CNT, rfixed_trunc(priority_mark02));
+		WREG32(R_006D48_D2MODE_PRIORITY_A_CNT,
+			S_006D48_D2MODE_PRIORITY_A_OFF(1));
+		WREG32(R_006D4C_D2MODE_PRIORITY_B_CNT,
+			S_006D4C_D2MODE_PRIORITY_B_OFF(1));
 	} else {
 		if (rfixed_trunc(wm1.dbpp) > 64)
 			a.full = rfixed_mul(wm1.dbpp, wm1.num_line_pair);
@@ -622,30 +544,200 @@ void rs690_bandwidth_update(struct radeon_device *rdev)
 			priority_mark12.full = 0;
 		if (wm1.priority_mark_max.full > priority_mark12.full)
 			priority_mark12.full = wm1.priority_mark_max.full;
-		WREG32(D1MODE_PRIORITY_A_CNT, MODE_PRIORITY_OFF);
-		WREG32(D1MODE_PRIORITY_B_CNT, MODE_PRIORITY_OFF);
-		WREG32(D2MODE_PRIORITY_A_CNT, rfixed_trunc(priority_mark12));
-		WREG32(D2MODE_PRIORITY_B_CNT, rfixed_trunc(priority_mark12));
+		WREG32(R_006548_D1MODE_PRIORITY_A_CNT,
+			S_006548_D1MODE_PRIORITY_A_OFF(1));
+		WREG32(R_00654C_D1MODE_PRIORITY_B_CNT,
+			S_00654C_D1MODE_PRIORITY_B_OFF(1));
+		WREG32(R_006D48_D2MODE_PRIORITY_A_CNT, rfixed_trunc(priority_mark12));
+		WREG32(R_006D4C_D2MODE_PRIORITY_B_CNT, rfixed_trunc(priority_mark12));
 	}
 }
 
-/*
- * Indirect registers accessor
- */
 uint32_t rs690_mc_rreg(struct radeon_device *rdev, uint32_t reg)
 {
 	uint32_t r;
 
-	WREG32(RS690_MC_INDEX, (reg & RS690_MC_INDEX_MASK));
-	r = RREG32(RS690_MC_DATA);
-	WREG32(RS690_MC_INDEX, RS690_MC_INDEX_MASK);
+	WREG32(R_000078_MC_INDEX, S_000078_MC_IND_ADDR(reg));
+	r = RREG32(R_00007C_MC_DATA);
+	WREG32(R_000078_MC_INDEX, ~C_000078_MC_IND_ADDR);
 	return r;
 }
 
 void rs690_mc_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v)
 {
-	WREG32(RS690_MC_INDEX,
-	       RS690_MC_INDEX_WR_EN | ((reg) & RS690_MC_INDEX_MASK));
-	WREG32(RS690_MC_DATA, v);
-	WREG32(RS690_MC_INDEX, RS690_MC_INDEX_WR_ACK);
+	WREG32(R_000078_MC_INDEX, S_000078_MC_IND_ADDR(reg) |
+		S_000078_MC_IND_WR_EN(1));
+	WREG32(R_00007C_MC_DATA, v);
+	WREG32(R_000078_MC_INDEX, 0x7F);
+}
+
+void rs690_mc_program(struct radeon_device *rdev)
+{
+	struct rv515_mc_save save;
+
+	/* Stops all mc clients */
+	rv515_mc_stop(rdev, &save);
+
+	/* Wait for mc idle */
+	if (rs690_mc_wait_for_idle(rdev))
+		dev_warn(rdev->dev, "Wait MC idle timeout before updating MC.\n");
+	/* Program MC, should be a 32bits limited address space */
+	WREG32_MC(R_000100_MCCFG_FB_LOCATION,
+			S_000100_MC_FB_START(rdev->mc.vram_start >> 16) |
+			S_000100_MC_FB_TOP(rdev->mc.vram_end >> 16));
+	WREG32(R_000134_HDP_FB_LOCATION,
+		S_000134_HDP_FB_START(rdev->mc.vram_start >> 16));
+
+	rv515_mc_resume(rdev, &save);
+}
+
+static int rs690_startup(struct radeon_device *rdev)
+{
+	int r;
+
+	rs690_mc_program(rdev);
+	/* Resume clock */
+	rv515_clock_startup(rdev);
+	/* Initialize GPU configuration (# pipes, ...) */
+	rs690_gpu_init(rdev);
+	/* Initialize GART (initialize after TTM so we can allocate
+	 * memory through TTM but finalize after TTM) */
+	r = rs400_gart_enable(rdev);
+	if (r)
+		return r;
+	/* Enable IRQ */
+	rs600_irq_set(rdev);
+	rdev->config.r300.hdp_cntl = RREG32(RADEON_HOST_PATH_CNTL);
+	/* 1M ring buffer */
+	r = r100_cp_init(rdev, 1024 * 1024);
+	if (r) {
+		dev_err(rdev->dev, "failled initializing CP (%d).\n", r);
+		return r;
+	}
+	r = r100_wb_init(rdev);
+	if (r)
+		dev_err(rdev->dev, "failled initializing WB (%d).\n", r);
+	r = r100_ib_init(rdev);
+	if (r) {
+		dev_err(rdev->dev, "failled initializing IB (%d).\n", r);
+		return r;
+	}
+	return 0;
+}
+
+int rs690_resume(struct radeon_device *rdev)
+{
+	/* Make sur GART are not working */
+	rs400_gart_disable(rdev);
+	/* Resume clock before doing reset */
+	rv515_clock_startup(rdev);
+	/* Reset gpu before posting otherwise ATOM will enter infinite loop */
+	if (radeon_gpu_reset(rdev)) {
+		dev_warn(rdev->dev, "GPU reset failed ! (0xE40=0x%08X, 0x7C0=0x%08X)\n",
+			RREG32(R_000E40_RBBM_STATUS),
+			RREG32(R_0007C0_CP_STAT));
+	}
+	/* post */
+	atom_asic_init(rdev->mode_info.atom_context);
+	/* Resume clock after posting */
+	rv515_clock_startup(rdev);
+	/* Initialize surface registers */
+	radeon_surface_init(rdev);
+	return rs690_startup(rdev);
+}
+
+int rs690_suspend(struct radeon_device *rdev)
+{
+	r100_cp_disable(rdev);
+	r100_wb_disable(rdev);
+	rs600_irq_disable(rdev);
+	rs400_gart_disable(rdev);
+	return 0;
+}
+
+void rs690_fini(struct radeon_device *rdev)
+{
+	r100_cp_fini(rdev);
+	r100_wb_fini(rdev);
+	r100_ib_fini(rdev);
+	radeon_gem_fini(rdev);
+	rs400_gart_fini(rdev);
+	radeon_irq_kms_fini(rdev);
+	radeon_fence_driver_fini(rdev);
+	radeon_bo_fini(rdev);
+	radeon_atombios_fini(rdev);
+	kfree(rdev->bios);
+	rdev->bios = NULL;
+}
+
+int rs690_init(struct radeon_device *rdev)
+{
+	int r;
+
+	/* Disable VGA */
+	rv515_vga_render_disable(rdev);
+	/* Initialize scratch registers */
+	radeon_scratch_init(rdev);
+	/* Initialize surface registers */
+	radeon_surface_init(rdev);
+	/* TODO: disable VGA need to use VGA request */
+	/* BIOS*/
+	if (!radeon_get_bios(rdev)) {
+		if (ASIC_IS_AVIVO(rdev))
+			return -EINVAL;
+	}
+	if (rdev->is_atom_bios) {
+		r = radeon_atombios_init(rdev);
+		if (r)
+			return r;
+	} else {
+		dev_err(rdev->dev, "Expecting atombios for RV515 GPU\n");
+		return -EINVAL;
+	}
+	/* Reset gpu before posting otherwise ATOM will enter infinite loop */
+	if (radeon_gpu_reset(rdev)) {
+		dev_warn(rdev->dev,
+			"GPU reset failed ! (0xE40=0x%08X, 0x7C0=0x%08X)\n",
+			RREG32(R_000E40_RBBM_STATUS),
+			RREG32(R_0007C0_CP_STAT));
+	}
+	/* check if cards are posted or not */
+	if (radeon_boot_test_post_card(rdev) == false)
+		return -EINVAL;
+
+	/* Initialize clocks */
+	radeon_get_clock_info(rdev->ddev);
+	/* Initialize power management */
+	radeon_pm_init(rdev);
+	/* initialize memory controller */
+	rs690_mc_init(rdev);
+	rv515_debugfs(rdev);
+	/* Fence driver */
+	r = radeon_fence_driver_init(rdev);
+	if (r)
+		return r;
+	r = radeon_irq_kms_init(rdev);
+	if (r)
+		return r;
+	/* Memory manager */
+	r = radeon_bo_init(rdev);
+	if (r)
+		return r;
+	r = rs400_gart_init(rdev);
+	if (r)
+		return r;
+	rs600_set_safe_registers(rdev);
+	rdev->accel_working = true;
+	r = rs690_startup(rdev);
+	if (r) {
+		/* Somethings want wront with the accel init stop accel */
+		dev_err(rdev->dev, "Disabling GPU acceleration\n");
+		r100_cp_fini(rdev);
+		r100_wb_fini(rdev);
+		r100_ib_fini(rdev);
+		rs400_gart_fini(rdev);
+		radeon_irq_kms_fini(rdev);
+		rdev->accel_working = false;
+	}
+	return 0;
 }
diff --git a/drivers/gpu/drm/radeon/rs690d.h b/drivers/gpu/drm/radeon/rs690d.h
new file mode 100644
index 000000000000..62d31e7a897f
--- /dev/null
+++ b/drivers/gpu/drm/radeon/rs690d.h
@@ -0,0 +1,307 @@
+/*
+ * Copyright 2008 Advanced Micro Devices, Inc.
+ * Copyright 2008 Red Hat Inc.
+ * Copyright 2009 Jerome Glisse.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Dave Airlie
+ *          Alex Deucher
+ *          Jerome Glisse
+ */
+#ifndef __RS690D_H__
+#define __RS690D_H__
+
+/* Registers */
+#define R_000078_MC_INDEX                            0x000078
+#define   S_000078_MC_IND_ADDR(x)                      (((x) & 0x1FF) << 0)
+#define   G_000078_MC_IND_ADDR(x)                      (((x) >> 0) & 0x1FF)
+#define   C_000078_MC_IND_ADDR                         0xFFFFFE00
+#define   S_000078_MC_IND_WR_EN(x)                     (((x) & 0x1) << 9)
+#define   G_000078_MC_IND_WR_EN(x)                     (((x) >> 9) & 0x1)
+#define   C_000078_MC_IND_WR_EN                        0xFFFFFDFF
+#define R_00007C_MC_DATA                             0x00007C
+#define   S_00007C_MC_DATA(x)                          (((x) & 0xFFFFFFFF) << 0)
+#define   G_00007C_MC_DATA(x)                          (((x) >> 0) & 0xFFFFFFFF)
+#define   C_00007C_MC_DATA                             0x00000000
+#define R_0000F8_CONFIG_MEMSIZE                      0x0000F8
+#define   S_0000F8_CONFIG_MEMSIZE(x)                   (((x) & 0xFFFFFFFF) << 0)
+#define   G_0000F8_CONFIG_MEMSIZE(x)                   (((x) >> 0) & 0xFFFFFFFF)
+#define   C_0000F8_CONFIG_MEMSIZE                      0x00000000
+#define R_000134_HDP_FB_LOCATION                     0x000134
+#define   S_000134_HDP_FB_START(x)                     (((x) & 0xFFFF) << 0)
+#define   G_000134_HDP_FB_START(x)                     (((x) >> 0) & 0xFFFF)
+#define   C_000134_HDP_FB_START                        0xFFFF0000
+#define R_0007C0_CP_STAT                             0x0007C0
+#define   S_0007C0_MRU_BUSY(x)                         (((x) & 0x1) << 0)
+#define   G_0007C0_MRU_BUSY(x)                         (((x) >> 0) & 0x1)
+#define   C_0007C0_MRU_BUSY                            0xFFFFFFFE
+#define   S_0007C0_MWU_BUSY(x)                         (((x) & 0x1) << 1)
+#define   G_0007C0_MWU_BUSY(x)                         (((x) >> 1) & 0x1)
+#define   C_0007C0_MWU_BUSY                            0xFFFFFFFD
+#define   S_0007C0_RSIU_BUSY(x)                        (((x) & 0x1) << 2)
+#define   G_0007C0_RSIU_BUSY(x)                        (((x) >> 2) & 0x1)
+#define   C_0007C0_RSIU_BUSY                           0xFFFFFFFB
+#define   S_0007C0_RCIU_BUSY(x)                        (((x) & 0x1) << 3)
+#define   G_0007C0_RCIU_BUSY(x)                        (((x) >> 3) & 0x1)
+#define   C_0007C0_RCIU_BUSY                           0xFFFFFFF7
+#define   S_0007C0_CSF_PRIMARY_BUSY(x)                 (((x) & 0x1) << 9)
+#define   G_0007C0_CSF_PRIMARY_BUSY(x)                 (((x) >> 9) & 0x1)
+#define   C_0007C0_CSF_PRIMARY_BUSY                    0xFFFFFDFF
+#define   S_0007C0_CSF_INDIRECT_BUSY(x)                (((x) & 0x1) << 10)
+#define   G_0007C0_CSF_INDIRECT_BUSY(x)                (((x) >> 10) & 0x1)
+#define   C_0007C0_CSF_INDIRECT_BUSY                   0xFFFFFBFF
+#define   S_0007C0_CSQ_PRIMARY_BUSY(x)                 (((x) & 0x1) << 11)
+#define   G_0007C0_CSQ_PRIMARY_BUSY(x)                 (((x) >> 11) & 0x1)
+#define   C_0007C0_CSQ_PRIMARY_BUSY                    0xFFFFF7FF
+#define   S_0007C0_CSQ_INDIRECT_BUSY(x)                (((x) & 0x1) << 12)
+#define   G_0007C0_CSQ_INDIRECT_BUSY(x)                (((x) >> 12) & 0x1)
+#define   C_0007C0_CSQ_INDIRECT_BUSY                   0xFFFFEFFF
+#define   S_0007C0_CSI_BUSY(x)                         (((x) & 0x1) << 13)
+#define   G_0007C0_CSI_BUSY(x)                         (((x) >> 13) & 0x1)
+#define   C_0007C0_CSI_BUSY                            0xFFFFDFFF
+#define   S_0007C0_CSF_INDIRECT2_BUSY(x)               (((x) & 0x1) << 14)
+#define   G_0007C0_CSF_INDIRECT2_BUSY(x)               (((x) >> 14) & 0x1)
+#define   C_0007C0_CSF_INDIRECT2_BUSY                  0xFFFFBFFF
+#define   S_0007C0_CSQ_INDIRECT2_BUSY(x)               (((x) & 0x1) << 15)
+#define   G_0007C0_CSQ_INDIRECT2_BUSY(x)               (((x) >> 15) & 0x1)
+#define   C_0007C0_CSQ_INDIRECT2_BUSY                  0xFFFF7FFF
+#define   S_0007C0_GUIDMA_BUSY(x)                      (((x) & 0x1) << 28)
+#define   G_0007C0_GUIDMA_BUSY(x)                      (((x) >> 28) & 0x1)
+#define   C_0007C0_GUIDMA_BUSY                         0xEFFFFFFF
+#define   S_0007C0_VIDDMA_BUSY(x)                      (((x) & 0x1) << 29)
+#define   G_0007C0_VIDDMA_BUSY(x)                      (((x) >> 29) & 0x1)
+#define   C_0007C0_VIDDMA_BUSY                         0xDFFFFFFF
+#define   S_0007C0_CMDSTRM_BUSY(x)                     (((x) & 0x1) << 30)
+#define   G_0007C0_CMDSTRM_BUSY(x)                     (((x) >> 30) & 0x1)
+#define   C_0007C0_CMDSTRM_BUSY                        0xBFFFFFFF
+#define   S_0007C0_CP_BUSY(x)                          (((x) & 0x1) << 31)
+#define   G_0007C0_CP_BUSY(x)                          (((x) >> 31) & 0x1)
+#define   C_0007C0_CP_BUSY                             0x7FFFFFFF
+#define R_000E40_RBBM_STATUS                         0x000E40
+#define   S_000E40_CMDFIFO_AVAIL(x)                    (((x) & 0x7F) << 0)
+#define   G_000E40_CMDFIFO_AVAIL(x)                    (((x) >> 0) & 0x7F)
+#define   C_000E40_CMDFIFO_AVAIL                       0xFFFFFF80
+#define   S_000E40_HIRQ_ON_RBB(x)                      (((x) & 0x1) << 8)
+#define   G_000E40_HIRQ_ON_RBB(x)                      (((x) >> 8) & 0x1)
+#define   C_000E40_HIRQ_ON_RBB                         0xFFFFFEFF
+#define   S_000E40_CPRQ_ON_RBB(x)                      (((x) & 0x1) << 9)
+#define   G_000E40_CPRQ_ON_RBB(x)                      (((x) >> 9) & 0x1)
+#define   C_000E40_CPRQ_ON_RBB                         0xFFFFFDFF
+#define   S_000E40_CFRQ_ON_RBB(x)                      (((x) & 0x1) << 10)
+#define   G_000E40_CFRQ_ON_RBB(x)                      (((x) >> 10) & 0x1)
+#define   C_000E40_CFRQ_ON_RBB                         0xFFFFFBFF
+#define   S_000E40_HIRQ_IN_RTBUF(x)                    (((x) & 0x1) << 11)
+#define   G_000E40_HIRQ_IN_RTBUF(x)                    (((x) >> 11) & 0x1)
+#define   C_000E40_HIRQ_IN_RTBUF                       0xFFFFF7FF
+#define   S_000E40_CPRQ_IN_RTBUF(x)                    (((x) & 0x1) << 12)
+#define   G_000E40_CPRQ_IN_RTBUF(x)                    (((x) >> 12) & 0x1)
+#define   C_000E40_CPRQ_IN_RTBUF                       0xFFFFEFFF
+#define   S_000E40_CFRQ_IN_RTBUF(x)                    (((x) & 0x1) << 13)
+#define   G_000E40_CFRQ_IN_RTBUF(x)                    (((x) >> 13) & 0x1)
+#define   C_000E40_CFRQ_IN_RTBUF                       0xFFFFDFFF
+#define   S_000E40_CF_PIPE_BUSY(x)                     (((x) & 0x1) << 14)
+#define   G_000E40_CF_PIPE_BUSY(x)                     (((x) >> 14) & 0x1)
+#define   C_000E40_CF_PIPE_BUSY                        0xFFFFBFFF
+#define   S_000E40_ENG_EV_BUSY(x)                      (((x) & 0x1) << 15)
+#define   G_000E40_ENG_EV_BUSY(x)                      (((x) >> 15) & 0x1)
+#define   C_000E40_ENG_EV_BUSY                         0xFFFF7FFF
+#define   S_000E40_CP_CMDSTRM_BUSY(x)                  (((x) & 0x1) << 16)
+#define   G_000E40_CP_CMDSTRM_BUSY(x)                  (((x) >> 16) & 0x1)
+#define   C_000E40_CP_CMDSTRM_BUSY                     0xFFFEFFFF
+#define   S_000E40_E2_BUSY(x)                          (((x) & 0x1) << 17)
+#define   G_000E40_E2_BUSY(x)                          (((x) >> 17) & 0x1)
+#define   C_000E40_E2_BUSY                             0xFFFDFFFF
+#define   S_000E40_RB2D_BUSY(x)                        (((x) & 0x1) << 18)
+#define   G_000E40_RB2D_BUSY(x)                        (((x) >> 18) & 0x1)
+#define   C_000E40_RB2D_BUSY                           0xFFFBFFFF
+#define   S_000E40_RB3D_BUSY(x)                        (((x) & 0x1) << 19)
+#define   G_000E40_RB3D_BUSY(x)                        (((x) >> 19) & 0x1)
+#define   C_000E40_RB3D_BUSY                           0xFFF7FFFF
+#define   S_000E40_VAP_BUSY(x)                         (((x) & 0x1) << 20)
+#define   G_000E40_VAP_BUSY(x)                         (((x) >> 20) & 0x1)
+#define   C_000E40_VAP_BUSY                            0xFFEFFFFF
+#define   S_000E40_RE_BUSY(x)                          (((x) & 0x1) << 21)
+#define   G_000E40_RE_BUSY(x)                          (((x) >> 21) & 0x1)
+#define   C_000E40_RE_BUSY                             0xFFDFFFFF
+#define   S_000E40_TAM_BUSY(x)                         (((x) & 0x1) << 22)
+#define   G_000E40_TAM_BUSY(x)                         (((x) >> 22) & 0x1)
+#define   C_000E40_TAM_BUSY                            0xFFBFFFFF
+#define   S_000E40_TDM_BUSY(x)                         (((x) & 0x1) << 23)
+#define   G_000E40_TDM_BUSY(x)                         (((x) >> 23) & 0x1)
+#define   C_000E40_TDM_BUSY                            0xFF7FFFFF
+#define   S_000E40_PB_BUSY(x)                          (((x) & 0x1) << 24)
+#define   G_000E40_PB_BUSY(x)                          (((x) >> 24) & 0x1)
+#define   C_000E40_PB_BUSY                             0xFEFFFFFF
+#define   S_000E40_TIM_BUSY(x)                         (((x) & 0x1) << 25)
+#define   G_000E40_TIM_BUSY(x)                         (((x) >> 25) & 0x1)
+#define   C_000E40_TIM_BUSY                            0xFDFFFFFF
+#define   S_000E40_GA_BUSY(x)                          (((x) & 0x1) << 26)
+#define   G_000E40_GA_BUSY(x)                          (((x) >> 26) & 0x1)
+#define   C_000E40_GA_BUSY                             0xFBFFFFFF
+#define   S_000E40_CBA2D_BUSY(x)                       (((x) & 0x1) << 27)
+#define   G_000E40_CBA2D_BUSY(x)                       (((x) >> 27) & 0x1)
+#define   C_000E40_CBA2D_BUSY                          0xF7FFFFFF
+#define   S_000E40_GUI_ACTIVE(x)                       (((x) & 0x1) << 31)
+#define   G_000E40_GUI_ACTIVE(x)                       (((x) >> 31) & 0x1)
+#define   C_000E40_GUI_ACTIVE                          0x7FFFFFFF
+#define R_006520_DC_LB_MEMORY_SPLIT                  0x006520
+#define   S_006520_DC_LB_MEMORY_SPLIT(x)               (((x) & 0x3) << 0)
+#define   G_006520_DC_LB_MEMORY_SPLIT(x)               (((x) >> 0) & 0x3)
+#define   C_006520_DC_LB_MEMORY_SPLIT                  0xFFFFFFFC
+#define   S_006520_DC_LB_MEMORY_SPLIT_MODE(x)          (((x) & 0x1) << 2)
+#define   G_006520_DC_LB_MEMORY_SPLIT_MODE(x)          (((x) >> 2) & 0x1)
+#define   C_006520_DC_LB_MEMORY_SPLIT_MODE             0xFFFFFFFB
+#define   V_006520_DC_LB_MEMORY_SPLIT_D1HALF_D2HALF    0
+#define   V_006520_DC_LB_MEMORY_SPLIT_D1_3Q_D2_1Q      1
+#define   V_006520_DC_LB_MEMORY_SPLIT_D1_ONLY          2
+#define   V_006520_DC_LB_MEMORY_SPLIT_D1_1Q_D2_3Q      3
+#define   S_006520_DC_LB_DISP1_END_ADR(x)              (((x) & 0x7FF) << 4)
+#define   G_006520_DC_LB_DISP1_END_ADR(x)              (((x) >> 4) & 0x7FF)
+#define   C_006520_DC_LB_DISP1_END_ADR                 0xFFFF800F
+#define R_006548_D1MODE_PRIORITY_A_CNT               0x006548
+#define   S_006548_D1MODE_PRIORITY_MARK_A(x)           (((x) & 0x7FFF) << 0)
+#define   G_006548_D1MODE_PRIORITY_MARK_A(x)           (((x) >> 0) & 0x7FFF)
+#define   C_006548_D1MODE_PRIORITY_MARK_A              0xFFFF8000
+#define   S_006548_D1MODE_PRIORITY_A_OFF(x)            (((x) & 0x1) << 16)
+#define   G_006548_D1MODE_PRIORITY_A_OFF(x)            (((x) >> 16) & 0x1)
+#define   C_006548_D1MODE_PRIORITY_A_OFF               0xFFFEFFFF
+#define   S_006548_D1MODE_PRIORITY_A_FORCE_MASK(x)     (((x) & 0x1) << 24)
+#define   G_006548_D1MODE_PRIORITY_A_FORCE_MASK(x)     (((x) >> 24) & 0x1)
+#define   C_006548_D1MODE_PRIORITY_A_FORCE_MASK        0xFEFFFFFF
+#define R_00654C_D1MODE_PRIORITY_B_CNT               0x00654C
+#define   S_00654C_D1MODE_PRIORITY_MARK_B(x)           (((x) & 0x7FFF) << 0)
+#define   G_00654C_D1MODE_PRIORITY_MARK_B(x)           (((x) >> 0) & 0x7FFF)
+#define   C_00654C_D1MODE_PRIORITY_MARK_B              0xFFFF8000
+#define   S_00654C_D1MODE_PRIORITY_B_OFF(x)            (((x) & 0x1) << 16)
+#define   G_00654C_D1MODE_PRIORITY_B_OFF(x)            (((x) >> 16) & 0x1)
+#define   C_00654C_D1MODE_PRIORITY_B_OFF               0xFFFEFFFF
+#define   S_00654C_D1MODE_PRIORITY_B_ALWAYS_ON(x)      (((x) & 0x1) << 20)
+#define   G_00654C_D1MODE_PRIORITY_B_ALWAYS_ON(x)      (((x) >> 20) & 0x1)
+#define   C_00654C_D1MODE_PRIORITY_B_ALWAYS_ON         0xFFEFFFFF
+#define   S_00654C_D1MODE_PRIORITY_B_FORCE_MASK(x)     (((x) & 0x1) << 24)
+#define   G_00654C_D1MODE_PRIORITY_B_FORCE_MASK(x)     (((x) >> 24) & 0x1)
+#define   C_00654C_D1MODE_PRIORITY_B_FORCE_MASK        0xFEFFFFFF
+#define R_006C9C_DCP_CONTROL                         0x006C9C
+#define R_006D48_D2MODE_PRIORITY_A_CNT               0x006D48
+#define   S_006D48_D2MODE_PRIORITY_MARK_A(x)           (((x) & 0x7FFF) << 0)
+#define   G_006D48_D2MODE_PRIORITY_MARK_A(x)           (((x) >> 0) & 0x7FFF)
+#define   C_006D48_D2MODE_PRIORITY_MARK_A              0xFFFF8000
+#define   S_006D48_D2MODE_PRIORITY_A_OFF(x)            (((x) & 0x1) << 16)
+#define   G_006D48_D2MODE_PRIORITY_A_OFF(x)            (((x) >> 16) & 0x1)
+#define   C_006D48_D2MODE_PRIORITY_A_OFF               0xFFFEFFFF
+#define   S_006D48_D2MODE_PRIORITY_A_ALWAYS_ON(x)      (((x) & 0x1) << 20)
+#define   G_006D48_D2MODE_PRIORITY_A_ALWAYS_ON(x)      (((x) >> 20) & 0x1)
+#define   C_006D48_D2MODE_PRIORITY_A_ALWAYS_ON         0xFFEFFFFF
+#define   S_006D48_D2MODE_PRIORITY_A_FORCE_MASK(x)     (((x) & 0x1) << 24)
+#define   G_006D48_D2MODE_PRIORITY_A_FORCE_MASK(x)     (((x) >> 24) & 0x1)
+#define   C_006D48_D2MODE_PRIORITY_A_FORCE_MASK        0xFEFFFFFF
+#define R_006D4C_D2MODE_PRIORITY_B_CNT               0x006D4C
+#define   S_006D4C_D2MODE_PRIORITY_MARK_B(x)           (((x) & 0x7FFF) << 0)
+#define   G_006D4C_D2MODE_PRIORITY_MARK_B(x)           (((x) >> 0) & 0x7FFF)
+#define   C_006D4C_D2MODE_PRIORITY_MARK_B              0xFFFF8000
+#define   S_006D4C_D2MODE_PRIORITY_B_OFF(x)            (((x) & 0x1) << 16)
+#define   G_006D4C_D2MODE_PRIORITY_B_OFF(x)            (((x) >> 16) & 0x1)
+#define   C_006D4C_D2MODE_PRIORITY_B_OFF               0xFFFEFFFF
+#define   S_006D4C_D2MODE_PRIORITY_B_ALWAYS_ON(x)      (((x) & 0x1) << 20)
+#define   G_006D4C_D2MODE_PRIORITY_B_ALWAYS_ON(x)      (((x) >> 20) & 0x1)
+#define   C_006D4C_D2MODE_PRIORITY_B_ALWAYS_ON         0xFFEFFFFF
+#define   S_006D4C_D2MODE_PRIORITY_B_FORCE_MASK(x)     (((x) & 0x1) << 24)
+#define   G_006D4C_D2MODE_PRIORITY_B_FORCE_MASK(x)     (((x) >> 24) & 0x1)
+#define   C_006D4C_D2MODE_PRIORITY_B_FORCE_MASK        0xFEFFFFFF
+#define R_006D58_LB_MAX_REQ_OUTSTANDING              0x006D58
+#define   S_006D58_LB_D1_MAX_REQ_OUTSTANDING(x)        (((x) & 0xF) << 0)
+#define   G_006D58_LB_D1_MAX_REQ_OUTSTANDING(x)        (((x) >> 0) & 0xF)
+#define   C_006D58_LB_D1_MAX_REQ_OUTSTANDING           0xFFFFFFF0
+#define   S_006D58_LB_D2_MAX_REQ_OUTSTANDING(x)        (((x) & 0xF) << 16)
+#define   G_006D58_LB_D2_MAX_REQ_OUTSTANDING(x)        (((x) >> 16) & 0xF)
+#define   C_006D58_LB_D2_MAX_REQ_OUTSTANDING           0xFFF0FFFF
+
+
+#define R_000090_MC_SYSTEM_STATUS                    0x000090
+#define   S_000090_MC_SYSTEM_IDLE(x)                   (((x) & 0x1) << 0)
+#define   G_000090_MC_SYSTEM_IDLE(x)                   (((x) >> 0) & 0x1)
+#define   C_000090_MC_SYSTEM_IDLE                      0xFFFFFFFE
+#define   S_000090_MC_SEQUENCER_IDLE(x)                (((x) & 0x1) << 1)
+#define   G_000090_MC_SEQUENCER_IDLE(x)                (((x) >> 1) & 0x1)
+#define   C_000090_MC_SEQUENCER_IDLE                   0xFFFFFFFD
+#define   S_000090_MC_ARBITER_IDLE(x)                  (((x) & 0x1) << 2)
+#define   G_000090_MC_ARBITER_IDLE(x)                  (((x) >> 2) & 0x1)
+#define   C_000090_MC_ARBITER_IDLE                     0xFFFFFFFB
+#define   S_000090_MC_SELECT_PM(x)                     (((x) & 0x1) << 3)
+#define   G_000090_MC_SELECT_PM(x)                     (((x) >> 3) & 0x1)
+#define   C_000090_MC_SELECT_PM                        0xFFFFFFF7
+#define   S_000090_RESERVED4(x)                        (((x) & 0xF) << 4)
+#define   G_000090_RESERVED4(x)                        (((x) >> 4) & 0xF)
+#define   C_000090_RESERVED4                           0xFFFFFF0F
+#define   S_000090_RESERVED8(x)                        (((x) & 0xF) << 8)
+#define   G_000090_RESERVED8(x)                        (((x) >> 8) & 0xF)
+#define   C_000090_RESERVED8                           0xFFFFF0FF
+#define   S_000090_RESERVED12(x)                       (((x) & 0xF) << 12)
+#define   G_000090_RESERVED12(x)                       (((x) >> 12) & 0xF)
+#define   C_000090_RESERVED12                          0xFFFF0FFF
+#define   S_000090_MCA_INIT_EXECUTED(x)                (((x) & 0x1) << 16)
+#define   G_000090_MCA_INIT_EXECUTED(x)                (((x) >> 16) & 0x1)
+#define   C_000090_MCA_INIT_EXECUTED                   0xFFFEFFFF
+#define   S_000090_MCA_IDLE(x)                         (((x) & 0x1) << 17)
+#define   G_000090_MCA_IDLE(x)                         (((x) >> 17) & 0x1)
+#define   C_000090_MCA_IDLE                            0xFFFDFFFF
+#define   S_000090_MCA_SEQ_IDLE(x)                     (((x) & 0x1) << 18)
+#define   G_000090_MCA_SEQ_IDLE(x)                     (((x) >> 18) & 0x1)
+#define   C_000090_MCA_SEQ_IDLE                        0xFFFBFFFF
+#define   S_000090_MCA_ARB_IDLE(x)                     (((x) & 0x1) << 19)
+#define   G_000090_MCA_ARB_IDLE(x)                     (((x) >> 19) & 0x1)
+#define   C_000090_MCA_ARB_IDLE                        0xFFF7FFFF
+#define   S_000090_RESERVED20(x)                       (((x) & 0xFFF) << 20)
+#define   G_000090_RESERVED20(x)                       (((x) >> 20) & 0xFFF)
+#define   C_000090_RESERVED20                          0x000FFFFF
+#define R_000100_MCCFG_FB_LOCATION                   0x000100
+#define   S_000100_MC_FB_START(x)                      (((x) & 0xFFFF) << 0)
+#define   G_000100_MC_FB_START(x)                      (((x) >> 0) & 0xFFFF)
+#define   C_000100_MC_FB_START                         0xFFFF0000
+#define   S_000100_MC_FB_TOP(x)                        (((x) & 0xFFFF) << 16)
+#define   G_000100_MC_FB_TOP(x)                        (((x) >> 16) & 0xFFFF)
+#define   C_000100_MC_FB_TOP                           0x0000FFFF
+#define R_000104_MC_INIT_MISC_LAT_TIMER              0x000104
+#define   S_000104_MC_CPR_INIT_LAT(x)                  (((x) & 0xF) << 0)
+#define   G_000104_MC_CPR_INIT_LAT(x)                  (((x) >> 0) & 0xF)
+#define   C_000104_MC_CPR_INIT_LAT                     0xFFFFFFF0
+#define   S_000104_MC_VF_INIT_LAT(x)                   (((x) & 0xF) << 4)
+#define   G_000104_MC_VF_INIT_LAT(x)                   (((x) >> 4) & 0xF)
+#define   C_000104_MC_VF_INIT_LAT                      0xFFFFFF0F
+#define   S_000104_MC_DISP0R_INIT_LAT(x)               (((x) & 0xF) << 8)
+#define   G_000104_MC_DISP0R_INIT_LAT(x)               (((x) >> 8) & 0xF)
+#define   C_000104_MC_DISP0R_INIT_LAT                  0xFFFFF0FF
+#define   S_000104_MC_DISP1R_INIT_LAT(x)               (((x) & 0xF) << 12)
+#define   G_000104_MC_DISP1R_INIT_LAT(x)               (((x) >> 12) & 0xF)
+#define   C_000104_MC_DISP1R_INIT_LAT                  0xFFFF0FFF
+#define   S_000104_MC_FIXED_INIT_LAT(x)                (((x) & 0xF) << 16)
+#define   G_000104_MC_FIXED_INIT_LAT(x)                (((x) >> 16) & 0xF)
+#define   C_000104_MC_FIXED_INIT_LAT                   0xFFF0FFFF
+#define   S_000104_MC_E2R_INIT_LAT(x)                  (((x) & 0xF) << 20)
+#define   G_000104_MC_E2R_INIT_LAT(x)                  (((x) >> 20) & 0xF)
+#define   C_000104_MC_E2R_INIT_LAT                     0xFF0FFFFF
+#define   S_000104_SAME_PAGE_PRIO(x)                   (((x) & 0xF) << 24)
+#define   G_000104_SAME_PAGE_PRIO(x)                   (((x) >> 24) & 0xF)
+#define   C_000104_SAME_PAGE_PRIO                      0xF0FFFFFF
+#define   S_000104_MC_GLOBW_INIT_LAT(x)                (((x) & 0xF) << 28)
+#define   G_000104_MC_GLOBW_INIT_LAT(x)                (((x) >> 28) & 0xF)
+#define   C_000104_MC_GLOBW_INIT_LAT                   0x0FFFFFFF
+
+#endif
diff --git a/drivers/gpu/drm/radeon/rs690r.h b/drivers/gpu/drm/radeon/rs690r.h
deleted file mode 100644
index c0d9faa2175b..000000000000
--- a/drivers/gpu/drm/radeon/rs690r.h
+++ /dev/null
@@ -1,99 +0,0 @@
-/*
- * Copyright 2008 Advanced Micro Devices, Inc.
- * Copyright 2008 Red Hat Inc.
- * Copyright 2009 Jerome Glisse.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in
- * all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
- * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
- * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
- * OTHER DEALINGS IN THE SOFTWARE.
- *
- * Authors: Dave Airlie
- *          Alex Deucher
- *          Jerome Glisse
- */
-#ifndef RS690R_H
-#define RS690R_H
-
-/* RS690/RS740 registers */
-#define MC_INDEX			0x0078
-#	define MC_INDEX_MASK			0x1FF
-#	define MC_INDEX_WR_EN			(1 << 9)
-#	define MC_INDEX_WR_ACK			0x7F
-#define MC_DATA				0x007C
-#define HDP_FB_LOCATION			0x0134
-#define DC_LB_MEMORY_SPLIT		0x6520
-#define		DC_LB_MEMORY_SPLIT_MASK			0x00000003
-#define		DC_LB_MEMORY_SPLIT_SHIFT		0
-#define		DC_LB_MEMORY_SPLIT_D1HALF_D2HALF	0
-#define		DC_LB_MEMORY_SPLIT_D1_3Q_D2_1Q		1
-#define		DC_LB_MEMORY_SPLIT_D1_ONLY		2
-#define		DC_LB_MEMORY_SPLIT_D1_1Q_D2_3Q		3
-#define		DC_LB_MEMORY_SPLIT_SHIFT_MODE		(1 << 2)
-#define		DC_LB_DISP1_END_ADR_SHIFT		4
-#define		DC_LB_DISP1_END_ADR_MASK		0x00007FF0
-#define D1MODE_PRIORITY_A_CNT		0x6548
-#define		MODE_PRIORITY_MARK_MASK			0x00007FFF
-#define		MODE_PRIORITY_OFF			(1 << 16)
-#define		MODE_PRIORITY_ALWAYS_ON			(1 << 20)
-#define		MODE_PRIORITY_FORCE_MASK		(1 << 24)
-#define D1MODE_PRIORITY_B_CNT		0x654C
-#define LB_MAX_REQ_OUTSTANDING		0x6D58
-#define		LB_D1_MAX_REQ_OUTSTANDING_MASK		0x0000000F
-#define		LB_D1_MAX_REQ_OUTSTANDING_SHIFT		0
-#define		LB_D2_MAX_REQ_OUTSTANDING_MASK		0x000F0000
-#define		LB_D2_MAX_REQ_OUTSTANDING_SHIFT		16
-#define DCP_CONTROL			0x6C9C
-#define D2MODE_PRIORITY_A_CNT		0x6D48
-#define D2MODE_PRIORITY_B_CNT		0x6D4C
-
-/* MC indirect registers */
-#define MC_STATUS_IDLE				(1 << 0)
-#define MC_MISC_CNTL			0x18
-#define		DISABLE_GTW			(1 << 1)
-#define		GART_INDEX_REG_EN		(1 << 12)
-#define		BLOCK_GFX_D3_EN			(1 << 14)
-#define GART_FEATURE_ID			0x2B
-#define		HANG_EN				(1 << 11)
-#define		TLB_ENABLE			(1 << 18)
-#define		P2P_ENABLE			(1 << 19)
-#define		GTW_LAC_EN			(1 << 25)
-#define		LEVEL2_GART			(0 << 30)
-#define		LEVEL1_GART			(1 << 30)
-#define		PDC_EN				(1 << 31)
-#define GART_BASE			0x2C
-#define GART_CACHE_CNTRL		0x2E
-#	define GART_CACHE_INVALIDATE		(1 << 0)
-#define MC_STATUS			0x90
-#define MCCFG_FB_LOCATION		0x100
-#define		MC_FB_START_MASK		0x0000FFFF
-#define		MC_FB_START_SHIFT		0
-#define		MC_FB_TOP_MASK			0xFFFF0000
-#define		MC_FB_TOP_SHIFT			16
-#define MCCFG_AGP_LOCATION		0x101
-#define		MC_AGP_START_MASK		0x0000FFFF
-#define		MC_AGP_START_SHIFT		0
-#define		MC_AGP_TOP_MASK			0xFFFF0000
-#define		MC_AGP_TOP_SHIFT		16
-#define MCCFG_AGP_BASE			0x102
-#define MCCFG_AGP_BASE_2		0x103
-#define MC_INIT_MISC_LAT_TIMER		0x104
-#define		MC_DISP0R_INIT_LAT_SHIFT	8
-#define		MC_DISP0R_INIT_LAT_MASK		0x00000F00
-#define		MC_DISP1R_INIT_LAT_SHIFT	12
-#define		MC_DISP1R_INIT_LAT_MASK		0x0000F000
-
-#endif
diff --git a/drivers/gpu/drm/radeon/rv200d.h b/drivers/gpu/drm/radeon/rv200d.h
new file mode 100644
index 000000000000..c5b398330c26
--- /dev/null
+++ b/drivers/gpu/drm/radeon/rv200d.h
@@ -0,0 +1,36 @@
+/*
+ * Copyright 2008 Advanced Micro Devices, Inc.
+ * Copyright 2008 Red Hat Inc.
+ * Copyright 2009 Jerome Glisse.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Dave Airlie
+ *          Alex Deucher
+ *          Jerome Glisse
+ */
+#ifndef __RV200D_H__
+#define __RV200D_H__
+
+#define R_00015C_AGP_BASE_2                          0x00015C
+#define   S_00015C_AGP_BASE_ADDR_2(x)                  (((x) & 0xF) << 0)
+#define   G_00015C_AGP_BASE_ADDR_2(x)                  (((x) >> 0) & 0xF)
+#define   C_00015C_AGP_BASE_ADDR_2                     0xFFFFFFF0
+
+#endif
diff --git a/drivers/gpu/drm/radeon/rv250d.h b/drivers/gpu/drm/radeon/rv250d.h
new file mode 100644
index 000000000000..e5a70b06fe1f
--- /dev/null
+++ b/drivers/gpu/drm/radeon/rv250d.h
@@ -0,0 +1,123 @@
+/*
+ * Copyright 2008 Advanced Micro Devices, Inc.
+ * Copyright 2008 Red Hat Inc.
+ * Copyright 2009 Jerome Glisse.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Dave Airlie
+ *          Alex Deucher
+ *          Jerome Glisse
+ */
+#ifndef __RV250D_H__
+#define __RV250D_H__
+
+#define R_00000D_SCLK_CNTL_M6                        0x00000D
+#define   S_00000D_SCLK_SRC_SEL(x)                     (((x) & 0x7) << 0)
+#define   G_00000D_SCLK_SRC_SEL(x)                     (((x) >> 0) & 0x7)
+#define   C_00000D_SCLK_SRC_SEL                        0xFFFFFFF8
+#define   S_00000D_CP_MAX_DYN_STOP_LAT(x)              (((x) & 0x1) << 3)
+#define   G_00000D_CP_MAX_DYN_STOP_LAT(x)              (((x) >> 3) & 0x1)
+#define   C_00000D_CP_MAX_DYN_STOP_LAT                 0xFFFFFFF7
+#define   S_00000D_HDP_MAX_DYN_STOP_LAT(x)             (((x) & 0x1) << 4)
+#define   G_00000D_HDP_MAX_DYN_STOP_LAT(x)             (((x) >> 4) & 0x1)
+#define   C_00000D_HDP_MAX_DYN_STOP_LAT                0xFFFFFFEF
+#define   S_00000D_TV_MAX_DYN_STOP_LAT(x)              (((x) & 0x1) << 5)
+#define   G_00000D_TV_MAX_DYN_STOP_LAT(x)              (((x) >> 5) & 0x1)
+#define   C_00000D_TV_MAX_DYN_STOP_LAT                 0xFFFFFFDF
+#define   S_00000D_E2_MAX_DYN_STOP_LAT(x)              (((x) & 0x1) << 6)
+#define   G_00000D_E2_MAX_DYN_STOP_LAT(x)              (((x) >> 6) & 0x1)
+#define   C_00000D_E2_MAX_DYN_STOP_LAT                 0xFFFFFFBF
+#define   S_00000D_SE_MAX_DYN_STOP_LAT(x)              (((x) & 0x1) << 7)
+#define   G_00000D_SE_MAX_DYN_STOP_LAT(x)              (((x) >> 7) & 0x1)
+#define   C_00000D_SE_MAX_DYN_STOP_LAT                 0xFFFFFF7F
+#define   S_00000D_IDCT_MAX_DYN_STOP_LAT(x)            (((x) & 0x1) << 8)
+#define   G_00000D_IDCT_MAX_DYN_STOP_LAT(x)            (((x) >> 8) & 0x1)
+#define   C_00000D_IDCT_MAX_DYN_STOP_LAT               0xFFFFFEFF
+#define   S_00000D_VIP_MAX_DYN_STOP_LAT(x)             (((x) & 0x1) << 9)
+#define   G_00000D_VIP_MAX_DYN_STOP_LAT(x)             (((x) >> 9) & 0x1)
+#define   C_00000D_VIP_MAX_DYN_STOP_LAT                0xFFFFFDFF
+#define   S_00000D_RE_MAX_DYN_STOP_LAT(x)              (((x) & 0x1) << 10)
+#define   G_00000D_RE_MAX_DYN_STOP_LAT(x)              (((x) >> 10) & 0x1)
+#define   C_00000D_RE_MAX_DYN_STOP_LAT                 0xFFFFFBFF
+#define   S_00000D_PB_MAX_DYN_STOP_LAT(x)              (((x) & 0x1) << 11)
+#define   G_00000D_PB_MAX_DYN_STOP_LAT(x)              (((x) >> 11) & 0x1)
+#define   C_00000D_PB_MAX_DYN_STOP_LAT                 0xFFFFF7FF
+#define   S_00000D_TAM_MAX_DYN_STOP_LAT(x)             (((x) & 0x1) << 12)
+#define   G_00000D_TAM_MAX_DYN_STOP_LAT(x)             (((x) >> 12) & 0x1)
+#define   C_00000D_TAM_MAX_DYN_STOP_LAT                0xFFFFEFFF
+#define   S_00000D_TDM_MAX_DYN_STOP_LAT(x)             (((x) & 0x1) << 13)
+#define   G_00000D_TDM_MAX_DYN_STOP_LAT(x)             (((x) >> 13) & 0x1)
+#define   C_00000D_TDM_MAX_DYN_STOP_LAT                0xFFFFDFFF
+#define   S_00000D_RB_MAX_DYN_STOP_LAT(x)              (((x) & 0x1) << 14)
+#define   G_00000D_RB_MAX_DYN_STOP_LAT(x)              (((x) >> 14) & 0x1)
+#define   C_00000D_RB_MAX_DYN_STOP_LAT                 0xFFFFBFFF
+#define   S_00000D_FORCE_DISP2(x)                      (((x) & 0x1) << 15)
+#define   G_00000D_FORCE_DISP2(x)                      (((x) >> 15) & 0x1)
+#define   C_00000D_FORCE_DISP2                         0xFFFF7FFF
+#define   S_00000D_FORCE_CP(x)                         (((x) & 0x1) << 16)
+#define   G_00000D_FORCE_CP(x)                         (((x) >> 16) & 0x1)
+#define   C_00000D_FORCE_CP                            0xFFFEFFFF
+#define   S_00000D_FORCE_HDP(x)                        (((x) & 0x1) << 17)
+#define   G_00000D_FORCE_HDP(x)                        (((x) >> 17) & 0x1)
+#define   C_00000D_FORCE_HDP                           0xFFFDFFFF
+#define   S_00000D_FORCE_DISP1(x)                      (((x) & 0x1) << 18)
+#define   G_00000D_FORCE_DISP1(x)                      (((x) >> 18) & 0x1)
+#define   C_00000D_FORCE_DISP1                         0xFFFBFFFF
+#define   S_00000D_FORCE_TOP(x)                        (((x) & 0x1) << 19)
+#define   G_00000D_FORCE_TOP(x)                        (((x) >> 19) & 0x1)
+#define   C_00000D_FORCE_TOP                           0xFFF7FFFF
+#define   S_00000D_FORCE_E2(x)                         (((x) & 0x1) << 20)
+#define   G_00000D_FORCE_E2(x)                         (((x) >> 20) & 0x1)
+#define   C_00000D_FORCE_E2                            0xFFEFFFFF
+#define   S_00000D_FORCE_SE(x)                         (((x) & 0x1) << 21)
+#define   G_00000D_FORCE_SE(x)                         (((x) >> 21) & 0x1)
+#define   C_00000D_FORCE_SE                            0xFFDFFFFF
+#define   S_00000D_FORCE_IDCT(x)                       (((x) & 0x1) << 22)
+#define   G_00000D_FORCE_IDCT(x)                       (((x) >> 22) & 0x1)
+#define   C_00000D_FORCE_IDCT                          0xFFBFFFFF
+#define   S_00000D_FORCE_VIP(x)                        (((x) & 0x1) << 23)
+#define   G_00000D_FORCE_VIP(x)                        (((x) >> 23) & 0x1)
+#define   C_00000D_FORCE_VIP                           0xFF7FFFFF
+#define   S_00000D_FORCE_RE(x)                         (((x) & 0x1) << 24)
+#define   G_00000D_FORCE_RE(x)                         (((x) >> 24) & 0x1)
+#define   C_00000D_FORCE_RE                            0xFEFFFFFF
+#define   S_00000D_FORCE_PB(x)                         (((x) & 0x1) << 25)
+#define   G_00000D_FORCE_PB(x)                         (((x) >> 25) & 0x1)
+#define   C_00000D_FORCE_PB                            0xFDFFFFFF
+#define   S_00000D_FORCE_TAM(x)                        (((x) & 0x1) << 26)
+#define   G_00000D_FORCE_TAM(x)                        (((x) >> 26) & 0x1)
+#define   C_00000D_FORCE_TAM                           0xFBFFFFFF
+#define   S_00000D_FORCE_TDM(x)                        (((x) & 0x1) << 27)
+#define   G_00000D_FORCE_TDM(x)                        (((x) >> 27) & 0x1)
+#define   C_00000D_FORCE_TDM                           0xF7FFFFFF
+#define   S_00000D_FORCE_RB(x)                         (((x) & 0x1) << 28)
+#define   G_00000D_FORCE_RB(x)                         (((x) >> 28) & 0x1)
+#define   C_00000D_FORCE_RB                            0xEFFFFFFF
+#define   S_00000D_FORCE_TV_SCLK(x)                    (((x) & 0x1) << 29)
+#define   G_00000D_FORCE_TV_SCLK(x)                    (((x) >> 29) & 0x1)
+#define   C_00000D_FORCE_TV_SCLK                       0xDFFFFFFF
+#define   S_00000D_FORCE_SUBPIC(x)                     (((x) & 0x1) << 30)
+#define   G_00000D_FORCE_SUBPIC(x)                     (((x) >> 30) & 0x1)
+#define   C_00000D_FORCE_SUBPIC                        0xBFFFFFFF
+#define   S_00000D_FORCE_OV0(x)                        (((x) & 0x1) << 31)
+#define   G_00000D_FORCE_OV0(x)                        (((x) >> 31) & 0x1)
+#define   C_00000D_FORCE_OV0                           0x7FFFFFFF
+
+#endif
diff --git a/drivers/gpu/drm/radeon/rv350d.h b/drivers/gpu/drm/radeon/rv350d.h
new file mode 100644
index 000000000000..c75c5ed9e654
--- /dev/null
+++ b/drivers/gpu/drm/radeon/rv350d.h
@@ -0,0 +1,52 @@
+/*
+ * Copyright 2008 Advanced Micro Devices, Inc.
+ * Copyright 2008 Red Hat Inc.
+ * Copyright 2009 Jerome Glisse.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Dave Airlie
+ *          Alex Deucher
+ *          Jerome Glisse
+ */
+#ifndef __RV350D_H__
+#define __RV350D_H__
+
+/* RV350, RV380 registers */
+/* #define R_00000D_SCLK_CNTL                           0x00000D */
+#define   S_00000D_FORCE_VAP(x)                        (((x) & 0x1) << 21)
+#define   G_00000D_FORCE_VAP(x)                        (((x) >> 21) & 0x1)
+#define   C_00000D_FORCE_VAP                           0xFFDFFFFF
+#define   S_00000D_FORCE_SR(x)                         (((x) & 0x1) << 25)
+#define   G_00000D_FORCE_SR(x)                         (((x) >> 25) & 0x1)
+#define   C_00000D_FORCE_SR                            0xFDFFFFFF
+#define   S_00000D_FORCE_PX(x)                         (((x) & 0x1) << 26)
+#define   G_00000D_FORCE_PX(x)                         (((x) >> 26) & 0x1)
+#define   C_00000D_FORCE_PX                            0xFBFFFFFF
+#define   S_00000D_FORCE_TX(x)                         (((x) & 0x1) << 27)
+#define   G_00000D_FORCE_TX(x)                         (((x) >> 27) & 0x1)
+#define   C_00000D_FORCE_TX                            0xF7FFFFFF
+#define   S_00000D_FORCE_US(x)                         (((x) & 0x1) << 28)
+#define   G_00000D_FORCE_US(x)                         (((x) >> 28) & 0x1)
+#define   C_00000D_FORCE_US                            0xEFFFFFFF
+#define   S_00000D_FORCE_SU(x)                         (((x) & 0x1) << 30)
+#define   G_00000D_FORCE_SU(x)                         (((x) >> 30) & 0x1)
+#define   C_00000D_FORCE_SU                            0xBFFFFFFF
+
+#endif
diff --git a/drivers/gpu/drm/radeon/rv515.c b/drivers/gpu/drm/radeon/rv515.c
index fd799748e7d8..bea747da123f 100644
--- a/drivers/gpu/drm/radeon/rv515.c
+++ b/drivers/gpu/drm/radeon/rv515.c
@@ -29,37 +29,17 @@
 #include "drmP.h"
 #include "rv515d.h"
 #include "radeon.h"
-
+#include "atom.h"
 #include "rv515_reg_safe.h"
-/* rv515 depends on : */
-void r100_hdp_reset(struct radeon_device *rdev);
-int r100_cp_reset(struct radeon_device *rdev);
-int r100_rb2d_reset(struct radeon_device *rdev);
-int r100_gui_wait_for_idle(struct radeon_device *rdev);
-int r100_cp_init(struct radeon_device *rdev, unsigned ring_size);
-void r420_pipes_init(struct radeon_device *rdev);
-void rs600_mc_disable_clients(struct radeon_device *rdev);
-void rs600_disable_vga(struct radeon_device *rdev);
-
-/* This files gather functions specifics to:
- * rv515
- *
- * Some of these functions might be used by newer ASICs.
- */
+
+/* This files gather functions specifics to: rv515 */
 int rv515_debugfs_pipes_info_init(struct radeon_device *rdev);
 int rv515_debugfs_ga_info_init(struct radeon_device *rdev);
 void rv515_gpu_init(struct radeon_device *rdev);
 int rv515_mc_wait_for_idle(struct radeon_device *rdev);
 
-
-/*
- * MC
- */
-int rv515_mc_init(struct radeon_device *rdev)
+void rv515_debugfs(struct radeon_device *rdev)
 {
-	uint32_t tmp;
-	int r;
-
 	if (r100_debugfs_rbbm_init(rdev)) {
 		DRM_ERROR("Failed to register debugfs file for RBBM !\n");
 	}
@@ -69,67 +49,8 @@ int rv515_mc_init(struct radeon_device *rdev)
 	if (rv515_debugfs_ga_info_init(rdev)) {
 		DRM_ERROR("Failed to register debugfs file for pipes !\n");
 	}
-
-	rv515_gpu_init(rdev);
-	rv370_pcie_gart_disable(rdev);
-
-	/* Setup GPU memory space */
-	rdev->mc.vram_location = 0xFFFFFFFFUL;
-	rdev->mc.gtt_location = 0xFFFFFFFFUL;
-	if (rdev->flags & RADEON_IS_AGP) {
-		r = radeon_agp_init(rdev);
-		if (r) {
-			printk(KERN_WARNING "[drm] Disabling AGP\n");
-			rdev->flags &= ~RADEON_IS_AGP;
-			rdev->mc.gtt_size = radeon_gart_size * 1024 * 1024;
-		} else {
-			rdev->mc.gtt_location = rdev->mc.agp_base;
-		}
-	}
-	r = radeon_mc_setup(rdev);
-	if (r) {
-		return r;
-	}
-
-	/* Program GPU memory space */
-	rs600_mc_disable_clients(rdev);
-	if (rv515_mc_wait_for_idle(rdev)) {
-		printk(KERN_WARNING "Failed to wait MC idle while "
-		       "programming pipes. Bad things might happen.\n");
-	}
-	/* Write VRAM size in case we are limiting it */
-	WREG32(RADEON_CONFIG_MEMSIZE, rdev->mc.real_vram_size);
-	tmp = REG_SET(MC_FB_START, rdev->mc.vram_location >> 16);
-	WREG32(0x134, tmp);
-	tmp = rdev->mc.vram_location + rdev->mc.mc_vram_size - 1;
-	tmp = REG_SET(MC_FB_TOP, tmp >> 16);
-	tmp |= REG_SET(MC_FB_START, rdev->mc.vram_location >> 16);
-	WREG32_MC(MC_FB_LOCATION, tmp);
-	WREG32(HDP_FB_LOCATION, rdev->mc.vram_location >> 16);
-	WREG32(0x310, rdev->mc.vram_location);
-	if (rdev->flags & RADEON_IS_AGP) {
-		tmp = rdev->mc.gtt_location + rdev->mc.gtt_size - 1;
-		tmp = REG_SET(MC_AGP_TOP, tmp >> 16);
-		tmp |= REG_SET(MC_AGP_START, rdev->mc.gtt_location >> 16);
-		WREG32_MC(MC_AGP_LOCATION, tmp);
-		WREG32_MC(MC_AGP_BASE, rdev->mc.agp_base);
-		WREG32_MC(MC_AGP_BASE_2, 0);
-	} else {
-		WREG32_MC(MC_AGP_LOCATION, 0x0FFFFFFF);
-		WREG32_MC(MC_AGP_BASE, 0);
-		WREG32_MC(MC_AGP_BASE_2, 0);
-	}
-	return 0;
-}
-
-void rv515_mc_fini(struct radeon_device *rdev)
-{
 }
 
-
-/*
- * Global GPU functions
- */
 void rv515_ring_start(struct radeon_device *rdev)
 {
 	int r;
@@ -198,11 +119,6 @@ void rv515_ring_start(struct radeon_device *rdev)
 	radeon_ring_unlock_commit(rdev);
 }
 
-void rv515_errata(struct radeon_device *rdev)
-{
-	rdev->pll_errata = 0;
-}
-
 int rv515_mc_wait_for_idle(struct radeon_device *rdev)
 {
 	unsigned i;
@@ -219,6 +135,12 @@ int rv515_mc_wait_for_idle(struct radeon_device *rdev)
 	return -1;
 }
 
+void rv515_vga_render_disable(struct radeon_device *rdev)
+{
+	WREG32(R_000300_VGA_RENDER_CONTROL,
+		RREG32(R_000300_VGA_RENDER_CONTROL) & C_000300_VGA_VSTATUS_CNTL);
+}
+
 void rv515_gpu_init(struct radeon_device *rdev)
 {
 	unsigned pipe_select_current, gb_pipe_select, tmp;
@@ -231,7 +153,7 @@ void rv515_gpu_init(struct radeon_device *rdev)
 		       "reseting GPU. Bad things might happen.\n");
 	}
 
-	rs600_disable_vga(rdev);
+	rv515_vga_render_disable(rdev);
 
 	r420_pipes_init(rdev);
 	gb_pipe_select = RREG32(0x402C);
@@ -335,10 +257,6 @@ int rv515_gpu_reset(struct radeon_device *rdev)
 	return 0;
 }
 
-
-/*
- * VRAM info
- */
 static void rv515_vram_get_type(struct radeon_device *rdev)
 {
 	uint32_t tmp;
@@ -359,13 +277,15 @@ static void rv515_vram_get_type(struct radeon_device *rdev)
 	}
 }
 
-void rv515_vram_info(struct radeon_device *rdev)
+void rv515_mc_init(struct radeon_device *rdev)
 {
 	fixed20_12 a;
 
 	rv515_vram_get_type(rdev);
-
 	r100_vram_init_sizes(rdev);
+	radeon_vram_location(rdev, &rdev->mc, 0);
+	if (!(rdev->flags & RADEON_IS_AGP))
+		radeon_gtt_location(rdev, &rdev->mc);
 	/* FIXME: we should enforce default clock in case GPU is not in
 	 * default setup
 	 */
@@ -374,10 +294,6 @@ void rv515_vram_info(struct radeon_device *rdev)
 	rdev->pm.sclk.full = rfixed_div(rdev->pm.sclk, a);
 }
 
-
-/*
- * Indirect registers accessor
- */
 uint32_t rv515_mc_rreg(struct radeon_device *rdev, uint32_t reg)
 {
 	uint32_t r;
@@ -395,9 +311,6 @@ void rv515_mc_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v)
 	WREG32(MC_IND_INDEX, 0);
 }
 
-/*
- * Debugfs info
- */
 #if defined(CONFIG_DEBUG_FS)
 static int rv515_debugfs_pipes_info(struct seq_file *m, void *data)
 {
@@ -459,13 +372,261 @@ int rv515_debugfs_ga_info_init(struct radeon_device *rdev)
 #endif
 }
 
-/*
- * Asic initialization
- */
-int rv515_init(struct radeon_device *rdev)
+void rv515_mc_stop(struct radeon_device *rdev, struct rv515_mc_save *save)
+{
+	save->d1vga_control = RREG32(R_000330_D1VGA_CONTROL);
+	save->d2vga_control = RREG32(R_000338_D2VGA_CONTROL);
+	save->vga_render_control = RREG32(R_000300_VGA_RENDER_CONTROL);
+	save->vga_hdp_control = RREG32(R_000328_VGA_HDP_CONTROL);
+	save->d1crtc_control = RREG32(R_006080_D1CRTC_CONTROL);
+	save->d2crtc_control = RREG32(R_006880_D2CRTC_CONTROL);
+
+	/* Stop all video */
+	WREG32(R_0068E8_D2CRTC_UPDATE_LOCK, 0);
+	WREG32(R_000300_VGA_RENDER_CONTROL, 0);
+	WREG32(R_0060E8_D1CRTC_UPDATE_LOCK, 1);
+	WREG32(R_0068E8_D2CRTC_UPDATE_LOCK, 1);
+	WREG32(R_006080_D1CRTC_CONTROL, 0);
+	WREG32(R_006880_D2CRTC_CONTROL, 0);
+	WREG32(R_0060E8_D1CRTC_UPDATE_LOCK, 0);
+	WREG32(R_0068E8_D2CRTC_UPDATE_LOCK, 0);
+	WREG32(R_000330_D1VGA_CONTROL, 0);
+	WREG32(R_000338_D2VGA_CONTROL, 0);
+}
+
+void rv515_mc_resume(struct radeon_device *rdev, struct rv515_mc_save *save)
+{
+	WREG32(R_006110_D1GRPH_PRIMARY_SURFACE_ADDRESS, rdev->mc.vram_start);
+	WREG32(R_006118_D1GRPH_SECONDARY_SURFACE_ADDRESS, rdev->mc.vram_start);
+	WREG32(R_006910_D2GRPH_PRIMARY_SURFACE_ADDRESS, rdev->mc.vram_start);
+	WREG32(R_006918_D2GRPH_SECONDARY_SURFACE_ADDRESS, rdev->mc.vram_start);
+	WREG32(R_000310_VGA_MEMORY_BASE_ADDRESS, rdev->mc.vram_start);
+	/* Unlock host access */
+	WREG32(R_000328_VGA_HDP_CONTROL, save->vga_hdp_control);
+	mdelay(1);
+	/* Restore video state */
+	WREG32(R_000330_D1VGA_CONTROL, save->d1vga_control);
+	WREG32(R_000338_D2VGA_CONTROL, save->d2vga_control);
+	WREG32(R_0060E8_D1CRTC_UPDATE_LOCK, 1);
+	WREG32(R_0068E8_D2CRTC_UPDATE_LOCK, 1);
+	WREG32(R_006080_D1CRTC_CONTROL, save->d1crtc_control);
+	WREG32(R_006880_D2CRTC_CONTROL, save->d2crtc_control);
+	WREG32(R_0060E8_D1CRTC_UPDATE_LOCK, 0);
+	WREG32(R_0068E8_D2CRTC_UPDATE_LOCK, 0);
+	WREG32(R_000300_VGA_RENDER_CONTROL, save->vga_render_control);
+}
+
+void rv515_mc_program(struct radeon_device *rdev)
+{
+	struct rv515_mc_save save;
+
+	/* Stops all mc clients */
+	rv515_mc_stop(rdev, &save);
+
+	/* Wait for mc idle */
+	if (rv515_mc_wait_for_idle(rdev))
+		dev_warn(rdev->dev, "Wait MC idle timeout before updating MC.\n");
+	/* Write VRAM size in case we are limiting it */
+	WREG32(R_0000F8_CONFIG_MEMSIZE, rdev->mc.real_vram_size);
+	/* Program MC, should be a 32bits limited address space */
+	WREG32_MC(R_000001_MC_FB_LOCATION,
+			S_000001_MC_FB_START(rdev->mc.vram_start >> 16) |
+			S_000001_MC_FB_TOP(rdev->mc.vram_end >> 16));
+	WREG32(R_000134_HDP_FB_LOCATION,
+		S_000134_HDP_FB_START(rdev->mc.vram_start >> 16));
+	if (rdev->flags & RADEON_IS_AGP) {
+		WREG32_MC(R_000002_MC_AGP_LOCATION,
+			S_000002_MC_AGP_START(rdev->mc.gtt_start >> 16) |
+			S_000002_MC_AGP_TOP(rdev->mc.gtt_end >> 16));
+		WREG32_MC(R_000003_MC_AGP_BASE, lower_32_bits(rdev->mc.agp_base));
+		WREG32_MC(R_000004_MC_AGP_BASE_2,
+			S_000004_AGP_BASE_ADDR_2(upper_32_bits(rdev->mc.agp_base)));
+	} else {
+		WREG32_MC(R_000002_MC_AGP_LOCATION, 0xFFFFFFFF);
+		WREG32_MC(R_000003_MC_AGP_BASE, 0);
+		WREG32_MC(R_000004_MC_AGP_BASE_2, 0);
+	}
+
+	rv515_mc_resume(rdev, &save);
+}
+
+void rv515_clock_startup(struct radeon_device *rdev)
+{
+	if (radeon_dynclks != -1 && radeon_dynclks)
+		radeon_atom_set_clock_gating(rdev, 1);
+	/* We need to force on some of the block */
+	WREG32_PLL(R_00000F_CP_DYN_CNTL,
+		RREG32_PLL(R_00000F_CP_DYN_CNTL) | S_00000F_CP_FORCEON(1));
+	WREG32_PLL(R_000011_E2_DYN_CNTL,
+		RREG32_PLL(R_000011_E2_DYN_CNTL) | S_000011_E2_FORCEON(1));
+	WREG32_PLL(R_000013_IDCT_DYN_CNTL,
+		RREG32_PLL(R_000013_IDCT_DYN_CNTL) | S_000013_IDCT_FORCEON(1));
+}
+
+static int rv515_startup(struct radeon_device *rdev)
+{
+	int r;
+
+	rv515_mc_program(rdev);
+	/* Resume clock */
+	rv515_clock_startup(rdev);
+	/* Initialize GPU configuration (# pipes, ...) */
+	rv515_gpu_init(rdev);
+	/* Initialize GART (initialize after TTM so we can allocate
+	 * memory through TTM but finalize after TTM) */
+	if (rdev->flags & RADEON_IS_PCIE) {
+		r = rv370_pcie_gart_enable(rdev);
+		if (r)
+			return r;
+	}
+	/* Enable IRQ */
+	rs600_irq_set(rdev);
+	rdev->config.r300.hdp_cntl = RREG32(RADEON_HOST_PATH_CNTL);
+	/* 1M ring buffer */
+	r = r100_cp_init(rdev, 1024 * 1024);
+	if (r) {
+		dev_err(rdev->dev, "failled initializing CP (%d).\n", r);
+		return r;
+	}
+	r = r100_wb_init(rdev);
+	if (r)
+		dev_err(rdev->dev, "failled initializing WB (%d).\n", r);
+	r = r100_ib_init(rdev);
+	if (r) {
+		dev_err(rdev->dev, "failled initializing IB (%d).\n", r);
+		return r;
+	}
+	return 0;
+}
+
+int rv515_resume(struct radeon_device *rdev)
+{
+	/* Make sur GART are not working */
+	if (rdev->flags & RADEON_IS_PCIE)
+		rv370_pcie_gart_disable(rdev);
+	/* Resume clock before doing reset */
+	rv515_clock_startup(rdev);
+	/* Reset gpu before posting otherwise ATOM will enter infinite loop */
+	if (radeon_gpu_reset(rdev)) {
+		dev_warn(rdev->dev, "GPU reset failed ! (0xE40=0x%08X, 0x7C0=0x%08X)\n",
+			RREG32(R_000E40_RBBM_STATUS),
+			RREG32(R_0007C0_CP_STAT));
+	}
+	/* post */
+	atom_asic_init(rdev->mode_info.atom_context);
+	/* Resume clock after posting */
+	rv515_clock_startup(rdev);
+	/* Initialize surface registers */
+	radeon_surface_init(rdev);
+	return rv515_startup(rdev);
+}
+
+int rv515_suspend(struct radeon_device *rdev)
+{
+	r100_cp_disable(rdev);
+	r100_wb_disable(rdev);
+	rs600_irq_disable(rdev);
+	if (rdev->flags & RADEON_IS_PCIE)
+		rv370_pcie_gart_disable(rdev);
+	return 0;
+}
+
+void rv515_set_safe_registers(struct radeon_device *rdev)
 {
 	rdev->config.r300.reg_safe_bm = rv515_reg_safe_bm;
 	rdev->config.r300.reg_safe_bm_size = ARRAY_SIZE(rv515_reg_safe_bm);
+}
+
+void rv515_fini(struct radeon_device *rdev)
+{
+	r100_cp_fini(rdev);
+	r100_wb_fini(rdev);
+	r100_ib_fini(rdev);
+	radeon_gem_fini(rdev);
+	rv370_pcie_gart_fini(rdev);
+	radeon_agp_fini(rdev);
+	radeon_irq_kms_fini(rdev);
+	radeon_fence_driver_fini(rdev);
+	radeon_bo_fini(rdev);
+	radeon_atombios_fini(rdev);
+	kfree(rdev->bios);
+	rdev->bios = NULL;
+}
+
+int rv515_init(struct radeon_device *rdev)
+{
+	int r;
+
+	/* Initialize scratch registers */
+	radeon_scratch_init(rdev);
+	/* Initialize surface registers */
+	radeon_surface_init(rdev);
+	/* TODO: disable VGA need to use VGA request */
+	/* BIOS*/
+	if (!radeon_get_bios(rdev)) {
+		if (ASIC_IS_AVIVO(rdev))
+			return -EINVAL;
+	}
+	if (rdev->is_atom_bios) {
+		r = radeon_atombios_init(rdev);
+		if (r)
+			return r;
+	} else {
+		dev_err(rdev->dev, "Expecting atombios for RV515 GPU\n");
+		return -EINVAL;
+	}
+	/* Reset gpu before posting otherwise ATOM will enter infinite loop */
+	if (radeon_gpu_reset(rdev)) {
+		dev_warn(rdev->dev,
+			"GPU reset failed ! (0xE40=0x%08X, 0x7C0=0x%08X)\n",
+			RREG32(R_000E40_RBBM_STATUS),
+			RREG32(R_0007C0_CP_STAT));
+	}
+	/* check if cards are posted or not */
+	if (radeon_boot_test_post_card(rdev) == false)
+		return -EINVAL;
+	/* Initialize clocks */
+	radeon_get_clock_info(rdev->ddev);
+	/* Initialize power management */
+	radeon_pm_init(rdev);
+	/* initialize AGP */
+	if (rdev->flags & RADEON_IS_AGP) {
+		r = radeon_agp_init(rdev);
+		if (r) {
+			radeon_agp_disable(rdev);
+		}
+	}
+	/* initialize memory controller */
+	rv515_mc_init(rdev);
+	rv515_debugfs(rdev);
+	/* Fence driver */
+	r = radeon_fence_driver_init(rdev);
+	if (r)
+		return r;
+	r = radeon_irq_kms_init(rdev);
+	if (r)
+		return r;
+	/* Memory manager */
+	r = radeon_bo_init(rdev);
+	if (r)
+		return r;
+	r = rv370_pcie_gart_init(rdev);
+	if (r)
+		return r;
+	rv515_set_safe_registers(rdev);
+	rdev->accel_working = true;
+	r = rv515_startup(rdev);
+	if (r) {
+		/* Somethings want wront with the accel init stop accel */
+		dev_err(rdev->dev, "Disabling GPU acceleration\n");
+		r100_cp_fini(rdev);
+		r100_wb_fini(rdev);
+		r100_ib_fini(rdev);
+		radeon_irq_kms_fini(rdev);
+		rv370_pcie_gart_fini(rdev);
+		radeon_agp_fini(rdev);
+		rdev->accel_working = false;
+	}
 	return 0;
 }
 
@@ -734,8 +895,9 @@ void rv515_crtc_bandwidth_compute(struct radeon_device *rdev,
 
 	b.full = rfixed_const(mode->crtc_hdisplay);
 	c.full = rfixed_const(256);
-	a.full = rfixed_mul(wm->num_line_pair, b);
-	request_fifo_depth.full = rfixed_div(a, c);
+	a.full = rfixed_div(b, c);
+	request_fifo_depth.full = rfixed_mul(a, wm->num_line_pair);
+	request_fifo_depth.full = rfixed_ceil(request_fifo_depth);
 	if (a.full < rfixed_const(4)) {
 		wm->lb_request_fifo_depth = 4;
 	} else {
@@ -837,15 +999,17 @@ void rv515_crtc_bandwidth_compute(struct radeon_device *rdev,
 	a.full = rfixed_const(16);
 	wm->priority_mark_max.full = rfixed_const(crtc->base.mode.crtc_hdisplay);
 	wm->priority_mark_max.full = rfixed_div(wm->priority_mark_max, a);
+	wm->priority_mark_max.full = rfixed_ceil(wm->priority_mark_max);
 
 	/* Determine estimated width */
 	estimated_width.full = tolerable_latency.full - wm->worst_case_latency.full;
 	estimated_width.full = rfixed_div(estimated_width, consumption_time);
 	if (rfixed_trunc(estimated_width) > crtc->base.mode.crtc_hdisplay) {
-		wm->priority_mark.full = rfixed_const(10);
+		wm->priority_mark.full = wm->priority_mark_max.full;
 	} else {
 		a.full = rfixed_const(16);
 		wm->priority_mark.full = rfixed_div(estimated_width, a);
+		wm->priority_mark.full = rfixed_ceil(wm->priority_mark);
 		wm->priority_mark.full = wm->priority_mark_max.full - wm->priority_mark.full;
 	}
 }
diff --git a/drivers/gpu/drm/radeon/rv515d.h b/drivers/gpu/drm/radeon/rv515d.h
index a65e17ec1c08..fc216e49384d 100644
--- a/drivers/gpu/drm/radeon/rv515d.h
+++ b/drivers/gpu/drm/radeon/rv515d.h
@@ -216,5 +216,388 @@
 #define CP_PACKET0_GET_ONE_REG_WR(h) (((h) >> 15) & 1)
 #define CP_PACKET3_GET_OPCODE(h) (((h) >> 8) & 0xFF)
 
-#endif
+/* Registers */
+#define R_0000F8_CONFIG_MEMSIZE                      0x0000F8
+#define   S_0000F8_CONFIG_MEMSIZE(x)                   (((x) & 0xFFFFFFFF) << 0)
+#define   G_0000F8_CONFIG_MEMSIZE(x)                   (((x) >> 0) & 0xFFFFFFFF)
+#define   C_0000F8_CONFIG_MEMSIZE                      0x00000000
+#define R_000134_HDP_FB_LOCATION                     0x000134
+#define   S_000134_HDP_FB_START(x)                     (((x) & 0xFFFF) << 0)
+#define   G_000134_HDP_FB_START(x)                     (((x) >> 0) & 0xFFFF)
+#define   C_000134_HDP_FB_START                        0xFFFF0000
+#define R_000300_VGA_RENDER_CONTROL                  0x000300
+#define   S_000300_VGA_BLINK_RATE(x)                   (((x) & 0x1F) << 0)
+#define   G_000300_VGA_BLINK_RATE(x)                   (((x) >> 0) & 0x1F)
+#define   C_000300_VGA_BLINK_RATE                      0xFFFFFFE0
+#define   S_000300_VGA_BLINK_MODE(x)                   (((x) & 0x3) << 5)
+#define   G_000300_VGA_BLINK_MODE(x)                   (((x) >> 5) & 0x3)
+#define   C_000300_VGA_BLINK_MODE                      0xFFFFFF9F
+#define   S_000300_VGA_CURSOR_BLINK_INVERT(x)          (((x) & 0x1) << 7)
+#define   G_000300_VGA_CURSOR_BLINK_INVERT(x)          (((x) >> 7) & 0x1)
+#define   C_000300_VGA_CURSOR_BLINK_INVERT             0xFFFFFF7F
+#define   S_000300_VGA_EXTD_ADDR_COUNT_ENABLE(x)       (((x) & 0x1) << 8)
+#define   G_000300_VGA_EXTD_ADDR_COUNT_ENABLE(x)       (((x) >> 8) & 0x1)
+#define   C_000300_VGA_EXTD_ADDR_COUNT_ENABLE          0xFFFFFEFF
+#define   S_000300_VGA_VSTATUS_CNTL(x)                 (((x) & 0x3) << 16)
+#define   G_000300_VGA_VSTATUS_CNTL(x)                 (((x) >> 16) & 0x3)
+#define   C_000300_VGA_VSTATUS_CNTL                    0xFFFCFFFF
+#define   S_000300_VGA_LOCK_8DOT(x)                    (((x) & 0x1) << 24)
+#define   G_000300_VGA_LOCK_8DOT(x)                    (((x) >> 24) & 0x1)
+#define   C_000300_VGA_LOCK_8DOT                       0xFEFFFFFF
+#define   S_000300_VGAREG_LINECMP_COMPATIBILITY_SEL(x) (((x) & 0x1) << 25)
+#define   G_000300_VGAREG_LINECMP_COMPATIBILITY_SEL(x) (((x) >> 25) & 0x1)
+#define   C_000300_VGAREG_LINECMP_COMPATIBILITY_SEL    0xFDFFFFFF
+#define R_000310_VGA_MEMORY_BASE_ADDRESS             0x000310
+#define   S_000310_VGA_MEMORY_BASE_ADDRESS(x)          (((x) & 0xFFFFFFFF) << 0)
+#define   G_000310_VGA_MEMORY_BASE_ADDRESS(x)          (((x) >> 0) & 0xFFFFFFFF)
+#define   C_000310_VGA_MEMORY_BASE_ADDRESS             0x00000000
+#define R_000328_VGA_HDP_CONTROL                     0x000328
+#define   S_000328_VGA_MEM_PAGE_SELECT_EN(x)           (((x) & 0x1) << 0)
+#define   G_000328_VGA_MEM_PAGE_SELECT_EN(x)           (((x) >> 0) & 0x1)
+#define   C_000328_VGA_MEM_PAGE_SELECT_EN              0xFFFFFFFE
+#define   S_000328_VGA_RBBM_LOCK_DISABLE(x)            (((x) & 0x1) << 8)
+#define   G_000328_VGA_RBBM_LOCK_DISABLE(x)            (((x) >> 8) & 0x1)
+#define   C_000328_VGA_RBBM_LOCK_DISABLE               0xFFFFFEFF
+#define   S_000328_VGA_SOFT_RESET(x)                   (((x) & 0x1) << 16)
+#define   G_000328_VGA_SOFT_RESET(x)                   (((x) >> 16) & 0x1)
+#define   C_000328_VGA_SOFT_RESET                      0xFFFEFFFF
+#define   S_000328_VGA_TEST_RESET_CONTROL(x)           (((x) & 0x1) << 24)
+#define   G_000328_VGA_TEST_RESET_CONTROL(x)           (((x) >> 24) & 0x1)
+#define   C_000328_VGA_TEST_RESET_CONTROL              0xFEFFFFFF
+#define R_000330_D1VGA_CONTROL                       0x000330
+#define   S_000330_D1VGA_MODE_ENABLE(x)                (((x) & 0x1) << 0)
+#define   G_000330_D1VGA_MODE_ENABLE(x)                (((x) >> 0) & 0x1)
+#define   C_000330_D1VGA_MODE_ENABLE                   0xFFFFFFFE
+#define   S_000330_D1VGA_TIMING_SELECT(x)              (((x) & 0x1) << 8)
+#define   G_000330_D1VGA_TIMING_SELECT(x)              (((x) >> 8) & 0x1)
+#define   C_000330_D1VGA_TIMING_SELECT                 0xFFFFFEFF
+#define   S_000330_D1VGA_SYNC_POLARITY_SELECT(x)       (((x) & 0x1) << 9)
+#define   G_000330_D1VGA_SYNC_POLARITY_SELECT(x)       (((x) >> 9) & 0x1)
+#define   C_000330_D1VGA_SYNC_POLARITY_SELECT          0xFFFFFDFF
+#define   S_000330_D1VGA_OVERSCAN_TIMING_SELECT(x)     (((x) & 0x1) << 10)
+#define   G_000330_D1VGA_OVERSCAN_TIMING_SELECT(x)     (((x) >> 10) & 0x1)
+#define   C_000330_D1VGA_OVERSCAN_TIMING_SELECT        0xFFFFFBFF
+#define   S_000330_D1VGA_OVERSCAN_COLOR_EN(x)          (((x) & 0x1) << 16)
+#define   G_000330_D1VGA_OVERSCAN_COLOR_EN(x)          (((x) >> 16) & 0x1)
+#define   C_000330_D1VGA_OVERSCAN_COLOR_EN             0xFFFEFFFF
+#define   S_000330_D1VGA_ROTATE(x)                     (((x) & 0x3) << 24)
+#define   G_000330_D1VGA_ROTATE(x)                     (((x) >> 24) & 0x3)
+#define   C_000330_D1VGA_ROTATE                        0xFCFFFFFF
+#define R_000338_D2VGA_CONTROL                       0x000338
+#define   S_000338_D2VGA_MODE_ENABLE(x)                (((x) & 0x1) << 0)
+#define   G_000338_D2VGA_MODE_ENABLE(x)                (((x) >> 0) & 0x1)
+#define   C_000338_D2VGA_MODE_ENABLE                   0xFFFFFFFE
+#define   S_000338_D2VGA_TIMING_SELECT(x)              (((x) & 0x1) << 8)
+#define   G_000338_D2VGA_TIMING_SELECT(x)              (((x) >> 8) & 0x1)
+#define   C_000338_D2VGA_TIMING_SELECT                 0xFFFFFEFF
+#define   S_000338_D2VGA_SYNC_POLARITY_SELECT(x)       (((x) & 0x1) << 9)
+#define   G_000338_D2VGA_SYNC_POLARITY_SELECT(x)       (((x) >> 9) & 0x1)
+#define   C_000338_D2VGA_SYNC_POLARITY_SELECT          0xFFFFFDFF
+#define   S_000338_D2VGA_OVERSCAN_TIMING_SELECT(x)     (((x) & 0x1) << 10)
+#define   G_000338_D2VGA_OVERSCAN_TIMING_SELECT(x)     (((x) >> 10) & 0x1)
+#define   C_000338_D2VGA_OVERSCAN_TIMING_SELECT        0xFFFFFBFF
+#define   S_000338_D2VGA_OVERSCAN_COLOR_EN(x)          (((x) & 0x1) << 16)
+#define   G_000338_D2VGA_OVERSCAN_COLOR_EN(x)          (((x) >> 16) & 0x1)
+#define   C_000338_D2VGA_OVERSCAN_COLOR_EN             0xFFFEFFFF
+#define   S_000338_D2VGA_ROTATE(x)                     (((x) & 0x3) << 24)
+#define   G_000338_D2VGA_ROTATE(x)                     (((x) >> 24) & 0x3)
+#define   C_000338_D2VGA_ROTATE                        0xFCFFFFFF
+#define R_0007C0_CP_STAT                             0x0007C0
+#define   S_0007C0_MRU_BUSY(x)                         (((x) & 0x1) << 0)
+#define   G_0007C0_MRU_BUSY(x)                         (((x) >> 0) & 0x1)
+#define   C_0007C0_MRU_BUSY                            0xFFFFFFFE
+#define   S_0007C0_MWU_BUSY(x)                         (((x) & 0x1) << 1)
+#define   G_0007C0_MWU_BUSY(x)                         (((x) >> 1) & 0x1)
+#define   C_0007C0_MWU_BUSY                            0xFFFFFFFD
+#define   S_0007C0_RSIU_BUSY(x)                        (((x) & 0x1) << 2)
+#define   G_0007C0_RSIU_BUSY(x)                        (((x) >> 2) & 0x1)
+#define   C_0007C0_RSIU_BUSY                           0xFFFFFFFB
+#define   S_0007C0_RCIU_BUSY(x)                        (((x) & 0x1) << 3)
+#define   G_0007C0_RCIU_BUSY(x)                        (((x) >> 3) & 0x1)
+#define   C_0007C0_RCIU_BUSY                           0xFFFFFFF7
+#define   S_0007C0_CSF_PRIMARY_BUSY(x)                 (((x) & 0x1) << 9)
+#define   G_0007C0_CSF_PRIMARY_BUSY(x)                 (((x) >> 9) & 0x1)
+#define   C_0007C0_CSF_PRIMARY_BUSY                    0xFFFFFDFF
+#define   S_0007C0_CSF_INDIRECT_BUSY(x)                (((x) & 0x1) << 10)
+#define   G_0007C0_CSF_INDIRECT_BUSY(x)                (((x) >> 10) & 0x1)
+#define   C_0007C0_CSF_INDIRECT_BUSY                   0xFFFFFBFF
+#define   S_0007C0_CSQ_PRIMARY_BUSY(x)                 (((x) & 0x1) << 11)
+#define   G_0007C0_CSQ_PRIMARY_BUSY(x)                 (((x) >> 11) & 0x1)
+#define   C_0007C0_CSQ_PRIMARY_BUSY                    0xFFFFF7FF
+#define   S_0007C0_CSQ_INDIRECT_BUSY(x)                (((x) & 0x1) << 12)
+#define   G_0007C0_CSQ_INDIRECT_BUSY(x)                (((x) >> 12) & 0x1)
+#define   C_0007C0_CSQ_INDIRECT_BUSY                   0xFFFFEFFF
+#define   S_0007C0_CSI_BUSY(x)                         (((x) & 0x1) << 13)
+#define   G_0007C0_CSI_BUSY(x)                         (((x) >> 13) & 0x1)
+#define   C_0007C0_CSI_BUSY                            0xFFFFDFFF
+#define   S_0007C0_CSF_INDIRECT2_BUSY(x)               (((x) & 0x1) << 14)
+#define   G_0007C0_CSF_INDIRECT2_BUSY(x)               (((x) >> 14) & 0x1)
+#define   C_0007C0_CSF_INDIRECT2_BUSY                  0xFFFFBFFF
+#define   S_0007C0_CSQ_INDIRECT2_BUSY(x)               (((x) & 0x1) << 15)
+#define   G_0007C0_CSQ_INDIRECT2_BUSY(x)               (((x) >> 15) & 0x1)
+#define   C_0007C0_CSQ_INDIRECT2_BUSY                  0xFFFF7FFF
+#define   S_0007C0_GUIDMA_BUSY(x)                      (((x) & 0x1) << 28)
+#define   G_0007C0_GUIDMA_BUSY(x)                      (((x) >> 28) & 0x1)
+#define   C_0007C0_GUIDMA_BUSY                         0xEFFFFFFF
+#define   S_0007C0_VIDDMA_BUSY(x)                      (((x) & 0x1) << 29)
+#define   G_0007C0_VIDDMA_BUSY(x)                      (((x) >> 29) & 0x1)
+#define   C_0007C0_VIDDMA_BUSY                         0xDFFFFFFF
+#define   S_0007C0_CMDSTRM_BUSY(x)                     (((x) & 0x1) << 30)
+#define   G_0007C0_CMDSTRM_BUSY(x)                     (((x) >> 30) & 0x1)
+#define   C_0007C0_CMDSTRM_BUSY                        0xBFFFFFFF
+#define   S_0007C0_CP_BUSY(x)                          (((x) & 0x1) << 31)
+#define   G_0007C0_CP_BUSY(x)                          (((x) >> 31) & 0x1)
+#define   C_0007C0_CP_BUSY                             0x7FFFFFFF
+#define R_000E40_RBBM_STATUS                         0x000E40
+#define   S_000E40_CMDFIFO_AVAIL(x)                    (((x) & 0x7F) << 0)
+#define   G_000E40_CMDFIFO_AVAIL(x)                    (((x) >> 0) & 0x7F)
+#define   C_000E40_CMDFIFO_AVAIL                       0xFFFFFF80
+#define   S_000E40_HIRQ_ON_RBB(x)                      (((x) & 0x1) << 8)
+#define   G_000E40_HIRQ_ON_RBB(x)                      (((x) >> 8) & 0x1)
+#define   C_000E40_HIRQ_ON_RBB                         0xFFFFFEFF
+#define   S_000E40_CPRQ_ON_RBB(x)                      (((x) & 0x1) << 9)
+#define   G_000E40_CPRQ_ON_RBB(x)                      (((x) >> 9) & 0x1)
+#define   C_000E40_CPRQ_ON_RBB                         0xFFFFFDFF
+#define   S_000E40_CFRQ_ON_RBB(x)                      (((x) & 0x1) << 10)
+#define   G_000E40_CFRQ_ON_RBB(x)                      (((x) >> 10) & 0x1)
+#define   C_000E40_CFRQ_ON_RBB                         0xFFFFFBFF
+#define   S_000E40_HIRQ_IN_RTBUF(x)                    (((x) & 0x1) << 11)
+#define   G_000E40_HIRQ_IN_RTBUF(x)                    (((x) >> 11) & 0x1)
+#define   C_000E40_HIRQ_IN_RTBUF                       0xFFFFF7FF
+#define   S_000E40_CPRQ_IN_RTBUF(x)                    (((x) & 0x1) << 12)
+#define   G_000E40_CPRQ_IN_RTBUF(x)                    (((x) >> 12) & 0x1)
+#define   C_000E40_CPRQ_IN_RTBUF                       0xFFFFEFFF
+#define   S_000E40_CFRQ_IN_RTBUF(x)                    (((x) & 0x1) << 13)
+#define   G_000E40_CFRQ_IN_RTBUF(x)                    (((x) >> 13) & 0x1)
+#define   C_000E40_CFRQ_IN_RTBUF                       0xFFFFDFFF
+#define   S_000E40_CF_PIPE_BUSY(x)                     (((x) & 0x1) << 14)
+#define   G_000E40_CF_PIPE_BUSY(x)                     (((x) >> 14) & 0x1)
+#define   C_000E40_CF_PIPE_BUSY                        0xFFFFBFFF
+#define   S_000E40_ENG_EV_BUSY(x)                      (((x) & 0x1) << 15)
+#define   G_000E40_ENG_EV_BUSY(x)                      (((x) >> 15) & 0x1)
+#define   C_000E40_ENG_EV_BUSY                         0xFFFF7FFF
+#define   S_000E40_CP_CMDSTRM_BUSY(x)                  (((x) & 0x1) << 16)
+#define   G_000E40_CP_CMDSTRM_BUSY(x)                  (((x) >> 16) & 0x1)
+#define   C_000E40_CP_CMDSTRM_BUSY                     0xFFFEFFFF
+#define   S_000E40_E2_BUSY(x)                          (((x) & 0x1) << 17)
+#define   G_000E40_E2_BUSY(x)                          (((x) >> 17) & 0x1)
+#define   C_000E40_E2_BUSY                             0xFFFDFFFF
+#define   S_000E40_RB2D_BUSY(x)                        (((x) & 0x1) << 18)
+#define   G_000E40_RB2D_BUSY(x)                        (((x) >> 18) & 0x1)
+#define   C_000E40_RB2D_BUSY                           0xFFFBFFFF
+#define   S_000E40_RB3D_BUSY(x)                        (((x) & 0x1) << 19)
+#define   G_000E40_RB3D_BUSY(x)                        (((x) >> 19) & 0x1)
+#define   C_000E40_RB3D_BUSY                           0xFFF7FFFF
+#define   S_000E40_VAP_BUSY(x)                         (((x) & 0x1) << 20)
+#define   G_000E40_VAP_BUSY(x)                         (((x) >> 20) & 0x1)
+#define   C_000E40_VAP_BUSY                            0xFFEFFFFF
+#define   S_000E40_RE_BUSY(x)                          (((x) & 0x1) << 21)
+#define   G_000E40_RE_BUSY(x)                          (((x) >> 21) & 0x1)
+#define   C_000E40_RE_BUSY                             0xFFDFFFFF
+#define   S_000E40_TAM_BUSY(x)                         (((x) & 0x1) << 22)
+#define   G_000E40_TAM_BUSY(x)                         (((x) >> 22) & 0x1)
+#define   C_000E40_TAM_BUSY                            0xFFBFFFFF
+#define   S_000E40_TDM_BUSY(x)                         (((x) & 0x1) << 23)
+#define   G_000E40_TDM_BUSY(x)                         (((x) >> 23) & 0x1)
+#define   C_000E40_TDM_BUSY                            0xFF7FFFFF
+#define   S_000E40_PB_BUSY(x)                          (((x) & 0x1) << 24)
+#define   G_000E40_PB_BUSY(x)                          (((x) >> 24) & 0x1)
+#define   C_000E40_PB_BUSY                             0xFEFFFFFF
+#define   S_000E40_TIM_BUSY(x)                         (((x) & 0x1) << 25)
+#define   G_000E40_TIM_BUSY(x)                         (((x) >> 25) & 0x1)
+#define   C_000E40_TIM_BUSY                            0xFDFFFFFF
+#define   S_000E40_GA_BUSY(x)                          (((x) & 0x1) << 26)
+#define   G_000E40_GA_BUSY(x)                          (((x) >> 26) & 0x1)
+#define   C_000E40_GA_BUSY                             0xFBFFFFFF
+#define   S_000E40_CBA2D_BUSY(x)                       (((x) & 0x1) << 27)
+#define   G_000E40_CBA2D_BUSY(x)                       (((x) >> 27) & 0x1)
+#define   C_000E40_CBA2D_BUSY                          0xF7FFFFFF
+#define   S_000E40_RBBM_HIBUSY(x)                      (((x) & 0x1) << 28)
+#define   G_000E40_RBBM_HIBUSY(x)                      (((x) >> 28) & 0x1)
+#define   C_000E40_RBBM_HIBUSY                         0xEFFFFFFF
+#define   S_000E40_SKID_CFBUSY(x)                      (((x) & 0x1) << 29)
+#define   G_000E40_SKID_CFBUSY(x)                      (((x) >> 29) & 0x1)
+#define   C_000E40_SKID_CFBUSY                         0xDFFFFFFF
+#define   S_000E40_VAP_VF_BUSY(x)                      (((x) & 0x1) << 30)
+#define   G_000E40_VAP_VF_BUSY(x)                      (((x) >> 30) & 0x1)
+#define   C_000E40_VAP_VF_BUSY                         0xBFFFFFFF
+#define   S_000E40_GUI_ACTIVE(x)                       (((x) & 0x1) << 31)
+#define   G_000E40_GUI_ACTIVE(x)                       (((x) >> 31) & 0x1)
+#define   C_000E40_GUI_ACTIVE                          0x7FFFFFFF
+#define R_006080_D1CRTC_CONTROL                      0x006080
+#define   S_006080_D1CRTC_MASTER_EN(x)                 (((x) & 0x1) << 0)
+#define   G_006080_D1CRTC_MASTER_EN(x)                 (((x) >> 0) & 0x1)
+#define   C_006080_D1CRTC_MASTER_EN                    0xFFFFFFFE
+#define   S_006080_D1CRTC_SYNC_RESET_SEL(x)            (((x) & 0x1) << 4)
+#define   G_006080_D1CRTC_SYNC_RESET_SEL(x)            (((x) >> 4) & 0x1)
+#define   C_006080_D1CRTC_SYNC_RESET_SEL               0xFFFFFFEF
+#define   S_006080_D1CRTC_DISABLE_POINT_CNTL(x)        (((x) & 0x3) << 8)
+#define   G_006080_D1CRTC_DISABLE_POINT_CNTL(x)        (((x) >> 8) & 0x3)
+#define   C_006080_D1CRTC_DISABLE_POINT_CNTL           0xFFFFFCFF
+#define   S_006080_D1CRTC_CURRENT_MASTER_EN_STATE(x)   (((x) & 0x1) << 16)
+#define   G_006080_D1CRTC_CURRENT_MASTER_EN_STATE(x)   (((x) >> 16) & 0x1)
+#define   C_006080_D1CRTC_CURRENT_MASTER_EN_STATE      0xFFFEFFFF
+#define   S_006080_D1CRTC_DISP_READ_REQUEST_DISABLE(x) (((x) & 0x1) << 24)
+#define   G_006080_D1CRTC_DISP_READ_REQUEST_DISABLE(x) (((x) >> 24) & 0x1)
+#define   C_006080_D1CRTC_DISP_READ_REQUEST_DISABLE    0xFEFFFFFF
+#define R_0060E8_D1CRTC_UPDATE_LOCK                  0x0060E8
+#define   S_0060E8_D1CRTC_UPDATE_LOCK(x)               (((x) & 0x1) << 0)
+#define   G_0060E8_D1CRTC_UPDATE_LOCK(x)               (((x) >> 0) & 0x1)
+#define   C_0060E8_D1CRTC_UPDATE_LOCK                  0xFFFFFFFE
+#define R_006110_D1GRPH_PRIMARY_SURFACE_ADDRESS      0x006110
+#define   S_006110_D1GRPH_PRIMARY_SURFACE_ADDRESS(x)   (((x) & 0xFFFFFFFF) << 0)
+#define   G_006110_D1GRPH_PRIMARY_SURFACE_ADDRESS(x)   (((x) >> 0) & 0xFFFFFFFF)
+#define   C_006110_D1GRPH_PRIMARY_SURFACE_ADDRESS      0x00000000
+#define R_006118_D1GRPH_SECONDARY_SURFACE_ADDRESS    0x006118
+#define   S_006118_D1GRPH_SECONDARY_SURFACE_ADDRESS(x) (((x) & 0xFFFFFFFF) << 0)
+#define   G_006118_D1GRPH_SECONDARY_SURFACE_ADDRESS(x) (((x) >> 0) & 0xFFFFFFFF)
+#define   C_006118_D1GRPH_SECONDARY_SURFACE_ADDRESS    0x00000000
+#define R_006880_D2CRTC_CONTROL                      0x006880
+#define   S_006880_D2CRTC_MASTER_EN(x)                 (((x) & 0x1) << 0)
+#define   G_006880_D2CRTC_MASTER_EN(x)                 (((x) >> 0) & 0x1)
+#define   C_006880_D2CRTC_MASTER_EN                    0xFFFFFFFE
+#define   S_006880_D2CRTC_SYNC_RESET_SEL(x)            (((x) & 0x1) << 4)
+#define   G_006880_D2CRTC_SYNC_RESET_SEL(x)            (((x) >> 4) & 0x1)
+#define   C_006880_D2CRTC_SYNC_RESET_SEL               0xFFFFFFEF
+#define   S_006880_D2CRTC_DISABLE_POINT_CNTL(x)        (((x) & 0x3) << 8)
+#define   G_006880_D2CRTC_DISABLE_POINT_CNTL(x)        (((x) >> 8) & 0x3)
+#define   C_006880_D2CRTC_DISABLE_POINT_CNTL           0xFFFFFCFF
+#define   S_006880_D2CRTC_CURRENT_MASTER_EN_STATE(x)   (((x) & 0x1) << 16)
+#define   G_006880_D2CRTC_CURRENT_MASTER_EN_STATE(x)   (((x) >> 16) & 0x1)
+#define   C_006880_D2CRTC_CURRENT_MASTER_EN_STATE      0xFFFEFFFF
+#define   S_006880_D2CRTC_DISP_READ_REQUEST_DISABLE(x) (((x) & 0x1) << 24)
+#define   G_006880_D2CRTC_DISP_READ_REQUEST_DISABLE(x) (((x) >> 24) & 0x1)
+#define   C_006880_D2CRTC_DISP_READ_REQUEST_DISABLE    0xFEFFFFFF
+#define R_0068E8_D2CRTC_UPDATE_LOCK                  0x0068E8
+#define   S_0068E8_D2CRTC_UPDATE_LOCK(x)               (((x) & 0x1) << 0)
+#define   G_0068E8_D2CRTC_UPDATE_LOCK(x)               (((x) >> 0) & 0x1)
+#define   C_0068E8_D2CRTC_UPDATE_LOCK                  0xFFFFFFFE
+#define R_006910_D2GRPH_PRIMARY_SURFACE_ADDRESS      0x006910
+#define   S_006910_D2GRPH_PRIMARY_SURFACE_ADDRESS(x)   (((x) & 0xFFFFFFFF) << 0)
+#define   G_006910_D2GRPH_PRIMARY_SURFACE_ADDRESS(x)   (((x) >> 0) & 0xFFFFFFFF)
+#define   C_006910_D2GRPH_PRIMARY_SURFACE_ADDRESS      0x00000000
+#define R_006918_D2GRPH_SECONDARY_SURFACE_ADDRESS    0x006918
+#define   S_006918_D2GRPH_SECONDARY_SURFACE_ADDRESS(x) (((x) & 0xFFFFFFFF) << 0)
+#define   G_006918_D2GRPH_SECONDARY_SURFACE_ADDRESS(x) (((x) >> 0) & 0xFFFFFFFF)
+#define   C_006918_D2GRPH_SECONDARY_SURFACE_ADDRESS    0x00000000
+
+
+#define R_000001_MC_FB_LOCATION                      0x000001
+#define   S_000001_MC_FB_START(x)                      (((x) & 0xFFFF) << 0)
+#define   G_000001_MC_FB_START(x)                      (((x) >> 0) & 0xFFFF)
+#define   C_000001_MC_FB_START                         0xFFFF0000
+#define   S_000001_MC_FB_TOP(x)                        (((x) & 0xFFFF) << 16)
+#define   G_000001_MC_FB_TOP(x)                        (((x) >> 16) & 0xFFFF)
+#define   C_000001_MC_FB_TOP                           0x0000FFFF
+#define R_000002_MC_AGP_LOCATION                     0x000002
+#define   S_000002_MC_AGP_START(x)                     (((x) & 0xFFFF) << 0)
+#define   G_000002_MC_AGP_START(x)                     (((x) >> 0) & 0xFFFF)
+#define   C_000002_MC_AGP_START                        0xFFFF0000
+#define   S_000002_MC_AGP_TOP(x)                       (((x) & 0xFFFF) << 16)
+#define   G_000002_MC_AGP_TOP(x)                       (((x) >> 16) & 0xFFFF)
+#define   C_000002_MC_AGP_TOP                          0x0000FFFF
+#define R_000003_MC_AGP_BASE                         0x000003
+#define   S_000003_AGP_BASE_ADDR(x)                    (((x) & 0xFFFFFFFF) << 0)
+#define   G_000003_AGP_BASE_ADDR(x)                    (((x) >> 0) & 0xFFFFFFFF)
+#define   C_000003_AGP_BASE_ADDR                       0x00000000
+#define R_000004_MC_AGP_BASE_2                       0x000004
+#define   S_000004_AGP_BASE_ADDR_2(x)                  (((x) & 0xF) << 0)
+#define   G_000004_AGP_BASE_ADDR_2(x)                  (((x) >> 0) & 0xF)
+#define   C_000004_AGP_BASE_ADDR_2                     0xFFFFFFF0
 
+
+#define R_00000F_CP_DYN_CNTL                         0x00000F
+#define   S_00000F_CP_FORCEON(x)                       (((x) & 0x1) << 0)
+#define   G_00000F_CP_FORCEON(x)                       (((x) >> 0) & 0x1)
+#define   C_00000F_CP_FORCEON                          0xFFFFFFFE
+#define   S_00000F_CP_MAX_DYN_STOP_LAT(x)              (((x) & 0x1) << 1)
+#define   G_00000F_CP_MAX_DYN_STOP_LAT(x)              (((x) >> 1) & 0x1)
+#define   C_00000F_CP_MAX_DYN_STOP_LAT                 0xFFFFFFFD
+#define   S_00000F_CP_CLOCK_STATUS(x)                  (((x) & 0x1) << 2)
+#define   G_00000F_CP_CLOCK_STATUS(x)                  (((x) >> 2) & 0x1)
+#define   C_00000F_CP_CLOCK_STATUS                     0xFFFFFFFB
+#define   S_00000F_CP_PROG_SHUTOFF(x)                  (((x) & 0x1) << 3)
+#define   G_00000F_CP_PROG_SHUTOFF(x)                  (((x) >> 3) & 0x1)
+#define   C_00000F_CP_PROG_SHUTOFF                     0xFFFFFFF7
+#define   S_00000F_CP_PROG_DELAY_VALUE(x)              (((x) & 0xFF) << 4)
+#define   G_00000F_CP_PROG_DELAY_VALUE(x)              (((x) >> 4) & 0xFF)
+#define   C_00000F_CP_PROG_DELAY_VALUE                 0xFFFFF00F
+#define   S_00000F_CP_LOWER_POWER_IDLE(x)              (((x) & 0xFF) << 12)
+#define   G_00000F_CP_LOWER_POWER_IDLE(x)              (((x) >> 12) & 0xFF)
+#define   C_00000F_CP_LOWER_POWER_IDLE                 0xFFF00FFF
+#define   S_00000F_CP_LOWER_POWER_IGNORE(x)            (((x) & 0x1) << 20)
+#define   G_00000F_CP_LOWER_POWER_IGNORE(x)            (((x) >> 20) & 0x1)
+#define   C_00000F_CP_LOWER_POWER_IGNORE               0xFFEFFFFF
+#define   S_00000F_CP_NORMAL_POWER_IGNORE(x)           (((x) & 0x1) << 21)
+#define   G_00000F_CP_NORMAL_POWER_IGNORE(x)           (((x) >> 21) & 0x1)
+#define   C_00000F_CP_NORMAL_POWER_IGNORE              0xFFDFFFFF
+#define   S_00000F_SPARE(x)                            (((x) & 0x3) << 22)
+#define   G_00000F_SPARE(x)                            (((x) >> 22) & 0x3)
+#define   C_00000F_SPARE                               0xFF3FFFFF
+#define   S_00000F_CP_NORMAL_POWER_BUSY(x)             (((x) & 0xFF) << 24)
+#define   G_00000F_CP_NORMAL_POWER_BUSY(x)             (((x) >> 24) & 0xFF)
+#define   C_00000F_CP_NORMAL_POWER_BUSY                0x00FFFFFF
+#define R_000011_E2_DYN_CNTL                         0x000011
+#define   S_000011_E2_FORCEON(x)                       (((x) & 0x1) << 0)
+#define   G_000011_E2_FORCEON(x)                       (((x) >> 0) & 0x1)
+#define   C_000011_E2_FORCEON                          0xFFFFFFFE
+#define   S_000011_E2_MAX_DYN_STOP_LAT(x)              (((x) & 0x1) << 1)
+#define   G_000011_E2_MAX_DYN_STOP_LAT(x)              (((x) >> 1) & 0x1)
+#define   C_000011_E2_MAX_DYN_STOP_LAT                 0xFFFFFFFD
+#define   S_000011_E2_CLOCK_STATUS(x)                  (((x) & 0x1) << 2)
+#define   G_000011_E2_CLOCK_STATUS(x)                  (((x) >> 2) & 0x1)
+#define   C_000011_E2_CLOCK_STATUS                     0xFFFFFFFB
+#define   S_000011_E2_PROG_SHUTOFF(x)                  (((x) & 0x1) << 3)
+#define   G_000011_E2_PROG_SHUTOFF(x)                  (((x) >> 3) & 0x1)
+#define   C_000011_E2_PROG_SHUTOFF                     0xFFFFFFF7
+#define   S_000011_E2_PROG_DELAY_VALUE(x)              (((x) & 0xFF) << 4)
+#define   G_000011_E2_PROG_DELAY_VALUE(x)              (((x) >> 4) & 0xFF)
+#define   C_000011_E2_PROG_DELAY_VALUE                 0xFFFFF00F
+#define   S_000011_E2_LOWER_POWER_IDLE(x)              (((x) & 0xFF) << 12)
+#define   G_000011_E2_LOWER_POWER_IDLE(x)              (((x) >> 12) & 0xFF)
+#define   C_000011_E2_LOWER_POWER_IDLE                 0xFFF00FFF
+#define   S_000011_E2_LOWER_POWER_IGNORE(x)            (((x) & 0x1) << 20)
+#define   G_000011_E2_LOWER_POWER_IGNORE(x)            (((x) >> 20) & 0x1)
+#define   C_000011_E2_LOWER_POWER_IGNORE               0xFFEFFFFF
+#define   S_000011_E2_NORMAL_POWER_IGNORE(x)           (((x) & 0x1) << 21)
+#define   G_000011_E2_NORMAL_POWER_IGNORE(x)           (((x) >> 21) & 0x1)
+#define   C_000011_E2_NORMAL_POWER_IGNORE              0xFFDFFFFF
+#define   S_000011_SPARE(x)                            (((x) & 0x3) << 22)
+#define   G_000011_SPARE(x)                            (((x) >> 22) & 0x3)
+#define   C_000011_SPARE                               0xFF3FFFFF
+#define   S_000011_E2_NORMAL_POWER_BUSY(x)             (((x) & 0xFF) << 24)
+#define   G_000011_E2_NORMAL_POWER_BUSY(x)             (((x) >> 24) & 0xFF)
+#define   C_000011_E2_NORMAL_POWER_BUSY                0x00FFFFFF
+#define R_000013_IDCT_DYN_CNTL                       0x000013
+#define   S_000013_IDCT_FORCEON(x)                     (((x) & 0x1) << 0)
+#define   G_000013_IDCT_FORCEON(x)                     (((x) >> 0) & 0x1)
+#define   C_000013_IDCT_FORCEON                        0xFFFFFFFE
+#define   S_000013_IDCT_MAX_DYN_STOP_LAT(x)            (((x) & 0x1) << 1)
+#define   G_000013_IDCT_MAX_DYN_STOP_LAT(x)            (((x) >> 1) & 0x1)
+#define   C_000013_IDCT_MAX_DYN_STOP_LAT               0xFFFFFFFD
+#define   S_000013_IDCT_CLOCK_STATUS(x)                (((x) & 0x1) << 2)
+#define   G_000013_IDCT_CLOCK_STATUS(x)                (((x) >> 2) & 0x1)
+#define   C_000013_IDCT_CLOCK_STATUS                   0xFFFFFFFB
+#define   S_000013_IDCT_PROG_SHUTOFF(x)                (((x) & 0x1) << 3)
+#define   G_000013_IDCT_PROG_SHUTOFF(x)                (((x) >> 3) & 0x1)
+#define   C_000013_IDCT_PROG_SHUTOFF                   0xFFFFFFF7
+#define   S_000013_IDCT_PROG_DELAY_VALUE(x)            (((x) & 0xFF) << 4)
+#define   G_000013_IDCT_PROG_DELAY_VALUE(x)            (((x) >> 4) & 0xFF)
+#define   C_000013_IDCT_PROG_DELAY_VALUE               0xFFFFF00F
+#define   S_000013_IDCT_LOWER_POWER_IDLE(x)            (((x) & 0xFF) << 12)
+#define   G_000013_IDCT_LOWER_POWER_IDLE(x)            (((x) >> 12) & 0xFF)
+#define   C_000013_IDCT_LOWER_POWER_IDLE               0xFFF00FFF
+#define   S_000013_IDCT_LOWER_POWER_IGNORE(x)          (((x) & 0x1) << 20)
+#define   G_000013_IDCT_LOWER_POWER_IGNORE(x)          (((x) >> 20) & 0x1)
+#define   C_000013_IDCT_LOWER_POWER_IGNORE             0xFFEFFFFF
+#define   S_000013_IDCT_NORMAL_POWER_IGNORE(x)         (((x) & 0x1) << 21)
+#define   G_000013_IDCT_NORMAL_POWER_IGNORE(x)         (((x) >> 21) & 0x1)
+#define   C_000013_IDCT_NORMAL_POWER_IGNORE            0xFFDFFFFF
+#define   S_000013_SPARE(x)                            (((x) & 0x3) << 22)
+#define   G_000013_SPARE(x)                            (((x) >> 22) & 0x3)
+#define   C_000013_SPARE                               0xFF3FFFFF
+#define   S_000013_IDCT_NORMAL_POWER_BUSY(x)           (((x) & 0xFF) << 24)
+#define   G_000013_IDCT_NORMAL_POWER_BUSY(x)           (((x) >> 24) & 0xFF)
+#define   C_000013_IDCT_NORMAL_POWER_BUSY              0x00FFFFFF
+
+#endif
diff --git a/drivers/gpu/drm/radeon/rv770.c b/drivers/gpu/drm/radeon/rv770.c
index b574c73a5109..37887dee12af 100644
--- a/drivers/gpu/drm/radeon/rv770.c
+++ b/drivers/gpu/drm/radeon/rv770.c
@@ -31,8 +31,8 @@
 #include "radeon.h"
 #include "radeon_drm.h"
 #include "rv770d.h"
-#include "avivod.h"
 #include "atom.h"
+#include "avivod.h"
 
 #define R700_PFP_UCODE_SIZE 848
 #define R700_PM4_UCODE_SIZE 1360
@@ -56,6 +56,7 @@ int rv770_pcie_gart_enable(struct radeon_device *rdev)
 	r = radeon_gart_table_vram_pin(rdev);
 	if (r)
 		return r;
+	radeon_gart_restore(rdev);
 	/* Setup L2 cache */
 	WREG32(VM_L2_CNTL, ENABLE_L2_CACHE | ENABLE_L2_FRAGMENT_PROCESSING |
 				ENABLE_L2_PTE_CACHE_LRU_UPDATE_BY_WRITE |
@@ -75,7 +76,7 @@ int rv770_pcie_gart_enable(struct radeon_device *rdev)
 	WREG32(MC_VM_MB_L1_TLB2_CNTL, tmp);
 	WREG32(MC_VM_MB_L1_TLB3_CNTL, tmp);
 	WREG32(VM_CONTEXT0_PAGE_TABLE_START_ADDR, rdev->mc.gtt_start >> 12);
-	WREG32(VM_CONTEXT0_PAGE_TABLE_END_ADDR, (rdev->mc.gtt_end - 1) >> 12);
+	WREG32(VM_CONTEXT0_PAGE_TABLE_END_ADDR, rdev->mc.gtt_end >> 12);
 	WREG32(VM_CONTEXT0_PAGE_TABLE_BASE_ADDR, rdev->gart.table_addr >> 12);
 	WREG32(VM_CONTEXT0_CNTL, ENABLE_CONTEXT | PAGE_TABLE_DEPTH(0) |
 				RANGE_PROTECTION_FAULT_ENABLE_DEFAULT);
@@ -92,7 +93,7 @@ int rv770_pcie_gart_enable(struct radeon_device *rdev)
 void rv770_pcie_gart_disable(struct radeon_device *rdev)
 {
 	u32 tmp;
-	int i;
+	int i, r;
 
 	/* Disable all tables */
 	for (i = 0; i < 7; i++)
@@ -113,8 +114,12 @@ void rv770_pcie_gart_disable(struct radeon_device *rdev)
 	WREG32(MC_VM_MB_L1_TLB2_CNTL, tmp);
 	WREG32(MC_VM_MB_L1_TLB3_CNTL, tmp);
 	if (rdev->gart.table.vram.robj) {
-		radeon_object_kunmap(rdev->gart.table.vram.robj);
-		radeon_object_unpin(rdev->gart.table.vram.robj);
+		r = radeon_bo_reserve(rdev->gart.table.vram.robj, false);
+		if (likely(r == 0)) {
+			radeon_bo_kunmap(rdev->gart.table.vram.robj);
+			radeon_bo_unpin(rdev->gart.table.vram.robj);
+			radeon_bo_unreserve(rdev->gart.table.vram.robj);
+		}
 	}
 }
 
@@ -126,17 +131,36 @@ void rv770_pcie_gart_fini(struct radeon_device *rdev)
 }
 
 
-/*
- * MC
- */
-static void rv770_mc_resume(struct radeon_device *rdev)
+void rv770_agp_enable(struct radeon_device *rdev)
 {
-	u32 d1vga_control, d2vga_control;
-	u32 vga_render_control, vga_hdp_control;
-	u32 d1crtc_control, d2crtc_control;
-	u32 new_d1grph_primary, new_d1grph_secondary;
-	u32 new_d2grph_primary, new_d2grph_secondary;
-	u64 old_vram_start;
+	u32 tmp;
+	int i;
+
+	/* Setup L2 cache */
+	WREG32(VM_L2_CNTL, ENABLE_L2_CACHE | ENABLE_L2_FRAGMENT_PROCESSING |
+				ENABLE_L2_PTE_CACHE_LRU_UPDATE_BY_WRITE |
+				EFFECTIVE_L2_QUEUE_SIZE(7));
+	WREG32(VM_L2_CNTL2, 0);
+	WREG32(VM_L2_CNTL3, BANK_SELECT(0) | CACHE_UPDATE_MODE(2));
+	/* Setup TLB control */
+	tmp = ENABLE_L1_TLB | ENABLE_L1_FRAGMENT_PROCESSING |
+		SYSTEM_ACCESS_MODE_NOT_IN_SYS |
+		SYSTEM_APERTURE_UNMAPPED_ACCESS_PASS_THRU |
+		EFFECTIVE_L1_TLB_SIZE(5) | EFFECTIVE_L1_QUEUE_SIZE(5);
+	WREG32(MC_VM_MD_L1_TLB0_CNTL, tmp);
+	WREG32(MC_VM_MD_L1_TLB1_CNTL, tmp);
+	WREG32(MC_VM_MD_L1_TLB2_CNTL, tmp);
+	WREG32(MC_VM_MB_L1_TLB0_CNTL, tmp);
+	WREG32(MC_VM_MB_L1_TLB1_CNTL, tmp);
+	WREG32(MC_VM_MB_L1_TLB2_CNTL, tmp);
+	WREG32(MC_VM_MB_L1_TLB3_CNTL, tmp);
+	for (i = 0; i < 7; i++)
+		WREG32(VM_CONTEXT0_CNTL + (i * 4), 0);
+}
+
+static void rv770_mc_program(struct radeon_device *rdev)
+{
+	struct rv515_mc_save save;
 	u32 tmp;
 	int i, j;
 
@@ -150,53 +174,42 @@ static void rv770_mc_resume(struct radeon_device *rdev)
 	}
 	WREG32(HDP_REG_COHERENCY_FLUSH_CNTL, 0);
 
-	d1vga_control = RREG32(D1VGA_CONTROL);
-	d2vga_control = RREG32(D2VGA_CONTROL);
-	vga_render_control = RREG32(VGA_RENDER_CONTROL);
-	vga_hdp_control = RREG32(VGA_HDP_CONTROL);
-	d1crtc_control = RREG32(D1CRTC_CONTROL);
-	d2crtc_control = RREG32(D2CRTC_CONTROL);
-	old_vram_start = (u64)(RREG32(MC_VM_FB_LOCATION) & 0xFFFF) << 24;
-	new_d1grph_primary = RREG32(D1GRPH_PRIMARY_SURFACE_ADDRESS);
-	new_d1grph_secondary = RREG32(D1GRPH_SECONDARY_SURFACE_ADDRESS);
-	new_d1grph_primary += rdev->mc.vram_start - old_vram_start;
-	new_d1grph_secondary += rdev->mc.vram_start - old_vram_start;
-	new_d2grph_primary = RREG32(D2GRPH_PRIMARY_SURFACE_ADDRESS);
-	new_d2grph_secondary = RREG32(D2GRPH_SECONDARY_SURFACE_ADDRESS);
-	new_d2grph_primary += rdev->mc.vram_start - old_vram_start;
-	new_d2grph_secondary += rdev->mc.vram_start - old_vram_start;
-
-	/* Stop all video */
-	WREG32(D1VGA_CONTROL, 0);
-	WREG32(D2VGA_CONTROL, 0);
-	WREG32(VGA_RENDER_CONTROL, 0);
-	WREG32(D1CRTC_UPDATE_LOCK, 1);
-	WREG32(D2CRTC_UPDATE_LOCK, 1);
-	WREG32(D1CRTC_CONTROL, 0);
-	WREG32(D2CRTC_CONTROL, 0);
-	WREG32(D1CRTC_UPDATE_LOCK, 0);
-	WREG32(D2CRTC_UPDATE_LOCK, 0);
-
-	mdelay(1);
+	rv515_mc_stop(rdev, &save);
 	if (r600_mc_wait_for_idle(rdev)) {
-		printk(KERN_WARNING "[drm] MC not idle !\n");
+		dev_warn(rdev->dev, "Wait for MC idle timedout !\n");
 	}
-
 	/* Lockout access through VGA aperture*/
 	WREG32(VGA_HDP_CONTROL, VGA_MEMORY_DISABLE);
-
 	/* Update configuration */
-	WREG32(MC_VM_SYSTEM_APERTURE_LOW_ADDR, rdev->mc.vram_start >> 12);
-	WREG32(MC_VM_SYSTEM_APERTURE_HIGH_ADDR, (rdev->mc.vram_end - 1) >> 12);
+	if (rdev->flags & RADEON_IS_AGP) {
+		if (rdev->mc.vram_start < rdev->mc.gtt_start) {
+			/* VRAM before AGP */
+			WREG32(MC_VM_SYSTEM_APERTURE_LOW_ADDR,
+				rdev->mc.vram_start >> 12);
+			WREG32(MC_VM_SYSTEM_APERTURE_HIGH_ADDR,
+				rdev->mc.gtt_end >> 12);
+		} else {
+			/* VRAM after AGP */
+			WREG32(MC_VM_SYSTEM_APERTURE_LOW_ADDR,
+				rdev->mc.gtt_start >> 12);
+			WREG32(MC_VM_SYSTEM_APERTURE_HIGH_ADDR,
+				rdev->mc.vram_end >> 12);
+		}
+	} else {
+		WREG32(MC_VM_SYSTEM_APERTURE_LOW_ADDR,
+			rdev->mc.vram_start >> 12);
+		WREG32(MC_VM_SYSTEM_APERTURE_HIGH_ADDR,
+			rdev->mc.vram_end >> 12);
+	}
 	WREG32(MC_VM_SYSTEM_APERTURE_DEFAULT_ADDR, 0);
-	tmp = (((rdev->mc.vram_end - 1) >> 24) & 0xFFFF) << 16;
+	tmp = ((rdev->mc.vram_end >> 24) & 0xFFFF) << 16;
 	tmp |= ((rdev->mc.vram_start >> 24) & 0xFFFF);
 	WREG32(MC_VM_FB_LOCATION, tmp);
 	WREG32(HDP_NONSURFACE_BASE, (rdev->mc.vram_start >> 8));
 	WREG32(HDP_NONSURFACE_INFO, (2 << 7));
 	WREG32(HDP_NONSURFACE_SIZE, (rdev->mc.mc_vram_size - 1) | 0x3FF);
 	if (rdev->flags & RADEON_IS_AGP) {
-		WREG32(MC_VM_AGP_TOP, (rdev->mc.gtt_end - 1) >> 16);
+		WREG32(MC_VM_AGP_TOP, rdev->mc.gtt_end >> 16);
 		WREG32(MC_VM_AGP_BOT, rdev->mc.gtt_start >> 16);
 		WREG32(MC_VM_AGP_BASE, rdev->mc.agp_base >> 22);
 	} else {
@@ -204,34 +217,13 @@ static void rv770_mc_resume(struct radeon_device *rdev)
 		WREG32(MC_VM_AGP_TOP, 0x0FFFFFFF);
 		WREG32(MC_VM_AGP_BOT, 0x0FFFFFFF);
 	}
-	WREG32(D1GRPH_PRIMARY_SURFACE_ADDRESS, new_d1grph_primary);
-	WREG32(D1GRPH_SECONDARY_SURFACE_ADDRESS, new_d1grph_secondary);
-	WREG32(D2GRPH_PRIMARY_SURFACE_ADDRESS, new_d2grph_primary);
-	WREG32(D2GRPH_SECONDARY_SURFACE_ADDRESS, new_d2grph_secondary);
-	WREG32(VGA_MEMORY_BASE_ADDRESS, rdev->mc.vram_start);
-
-	/* Unlock host access */
-	WREG32(VGA_HDP_CONTROL, vga_hdp_control);
-
-	mdelay(1);
 	if (r600_mc_wait_for_idle(rdev)) {
-		printk(KERN_WARNING "[drm] MC not idle !\n");
+		dev_warn(rdev->dev, "Wait for MC idle timedout !\n");
 	}
-
-	/* Restore video state */
-	WREG32(D1CRTC_UPDATE_LOCK, 1);
-	WREG32(D2CRTC_UPDATE_LOCK, 1);
-	WREG32(D1CRTC_CONTROL, d1crtc_control);
-	WREG32(D2CRTC_CONTROL, d2crtc_control);
-	WREG32(D1CRTC_UPDATE_LOCK, 0);
-	WREG32(D2CRTC_UPDATE_LOCK, 0);
-	WREG32(D1VGA_CONTROL, d1vga_control);
-	WREG32(D2VGA_CONTROL, d2vga_control);
-	WREG32(VGA_RENDER_CONTROL, vga_render_control);
-
+	rv515_mc_resume(rdev, &save);
 	/* we need to own VRAM, so turn off the VGA renderer here
 	 * to stop it overwriting our objects */
-	radeon_avivo_vga_render_disable(rdev);
+	rv515_vga_render_disable(rdev);
 }
 
 
@@ -282,9 +274,10 @@ static int rv770_cp_load_microcode(struct radeon_device *rdev)
 /*
  * Core functions
  */
-static u32 r700_get_tile_pipe_to_backend_map(u32 num_tile_pipes,
-						u32 num_backends,
-						u32 backend_disable_mask)
+static u32 r700_get_tile_pipe_to_backend_map(struct radeon_device *rdev,
+					     u32 num_tile_pipes,
+					     u32 num_backends,
+					     u32 backend_disable_mask)
 {
 	u32 backend_map = 0;
 	u32 enabled_backends_mask;
@@ -293,6 +286,7 @@ static u32 r700_get_tile_pipe_to_backend_map(u32 num_tile_pipes,
 	u32 swizzle_pipe[R7XX_MAX_PIPES];
 	u32 cur_backend;
 	u32 i;
+	bool force_no_swizzle;
 
 	if (num_tile_pipes > R7XX_MAX_PIPES)
 		num_tile_pipes = R7XX_MAX_PIPES;
@@ -322,6 +316,18 @@ static u32 r700_get_tile_pipe_to_backend_map(u32 num_tile_pipes,
 	if (enabled_backends_count != num_backends)
 		num_backends = enabled_backends_count;
 
+	switch (rdev->family) {
+	case CHIP_RV770:
+	case CHIP_RV730:
+		force_no_swizzle = false;
+		break;
+	case CHIP_RV710:
+	case CHIP_RV740:
+	default:
+		force_no_swizzle = true;
+		break;
+	}
+
 	memset((uint8_t *)&swizzle_pipe[0], 0, sizeof(u32) * R7XX_MAX_PIPES);
 	switch (num_tile_pipes) {
 	case 1:
@@ -332,49 +338,100 @@ static u32 r700_get_tile_pipe_to_backend_map(u32 num_tile_pipes,
 		swizzle_pipe[1] = 1;
 		break;
 	case 3:
-		swizzle_pipe[0] = 0;
-		swizzle_pipe[1] = 2;
-		swizzle_pipe[2] = 1;
+		if (force_no_swizzle) {
+			swizzle_pipe[0] = 0;
+			swizzle_pipe[1] = 1;
+			swizzle_pipe[2] = 2;
+		} else {
+			swizzle_pipe[0] = 0;
+			swizzle_pipe[1] = 2;
+			swizzle_pipe[2] = 1;
+		}
 		break;
 	case 4:
-		swizzle_pipe[0] = 0;
-		swizzle_pipe[1] = 2;
-		swizzle_pipe[2] = 3;
-		swizzle_pipe[3] = 1;
+		if (force_no_swizzle) {
+			swizzle_pipe[0] = 0;
+			swizzle_pipe[1] = 1;
+			swizzle_pipe[2] = 2;
+			swizzle_pipe[3] = 3;
+		} else {
+			swizzle_pipe[0] = 0;
+			swizzle_pipe[1] = 2;
+			swizzle_pipe[2] = 3;
+			swizzle_pipe[3] = 1;
+		}
 		break;
 	case 5:
-		swizzle_pipe[0] = 0;
-		swizzle_pipe[1] = 2;
-		swizzle_pipe[2] = 4;
-		swizzle_pipe[3] = 1;
-		swizzle_pipe[4] = 3;
+		if (force_no_swizzle) {
+			swizzle_pipe[0] = 0;
+			swizzle_pipe[1] = 1;
+			swizzle_pipe[2] = 2;
+			swizzle_pipe[3] = 3;
+			swizzle_pipe[4] = 4;
+		} else {
+			swizzle_pipe[0] = 0;
+			swizzle_pipe[1] = 2;
+			swizzle_pipe[2] = 4;
+			swizzle_pipe[3] = 1;
+			swizzle_pipe[4] = 3;
+		}
 		break;
 	case 6:
-		swizzle_pipe[0] = 0;
-		swizzle_pipe[1] = 2;
-		swizzle_pipe[2] = 4;
-		swizzle_pipe[3] = 5;
-		swizzle_pipe[4] = 3;
-		swizzle_pipe[5] = 1;
+		if (force_no_swizzle) {
+			swizzle_pipe[0] = 0;
+			swizzle_pipe[1] = 1;
+			swizzle_pipe[2] = 2;
+			swizzle_pipe[3] = 3;
+			swizzle_pipe[4] = 4;
+			swizzle_pipe[5] = 5;
+		} else {
+			swizzle_pipe[0] = 0;
+			swizzle_pipe[1] = 2;
+			swizzle_pipe[2] = 4;
+			swizzle_pipe[3] = 5;
+			swizzle_pipe[4] = 3;
+			swizzle_pipe[5] = 1;
+		}
 		break;
 	case 7:
-		swizzle_pipe[0] = 0;
-		swizzle_pipe[1] = 2;
-		swizzle_pipe[2] = 4;
-		swizzle_pipe[3] = 6;
-		swizzle_pipe[4] = 3;
-		swizzle_pipe[5] = 1;
-		swizzle_pipe[6] = 5;
+		if (force_no_swizzle) {
+			swizzle_pipe[0] = 0;
+			swizzle_pipe[1] = 1;
+			swizzle_pipe[2] = 2;
+			swizzle_pipe[3] = 3;
+			swizzle_pipe[4] = 4;
+			swizzle_pipe[5] = 5;
+			swizzle_pipe[6] = 6;
+		} else {
+			swizzle_pipe[0] = 0;
+			swizzle_pipe[1] = 2;
+			swizzle_pipe[2] = 4;
+			swizzle_pipe[3] = 6;
+			swizzle_pipe[4] = 3;
+			swizzle_pipe[5] = 1;
+			swizzle_pipe[6] = 5;
+		}
 		break;
 	case 8:
-		swizzle_pipe[0] = 0;
-		swizzle_pipe[1] = 2;
-		swizzle_pipe[2] = 4;
-		swizzle_pipe[3] = 6;
-		swizzle_pipe[4] = 3;
-		swizzle_pipe[5] = 1;
-		swizzle_pipe[6] = 7;
-		swizzle_pipe[7] = 5;
+		if (force_no_swizzle) {
+			swizzle_pipe[0] = 0;
+			swizzle_pipe[1] = 1;
+			swizzle_pipe[2] = 2;
+			swizzle_pipe[3] = 3;
+			swizzle_pipe[4] = 4;
+			swizzle_pipe[5] = 5;
+			swizzle_pipe[6] = 6;
+			swizzle_pipe[7] = 7;
+		} else {
+			swizzle_pipe[0] = 0;
+			swizzle_pipe[1] = 2;
+			swizzle_pipe[2] = 4;
+			swizzle_pipe[3] = 6;
+			swizzle_pipe[4] = 3;
+			swizzle_pipe[5] = 1;
+			swizzle_pipe[6] = 7;
+			swizzle_pipe[7] = 5;
+		}
 		break;
 	}
 
@@ -394,8 +451,10 @@ static u32 r700_get_tile_pipe_to_backend_map(u32 num_tile_pipes,
 static void rv770_gpu_init(struct radeon_device *rdev)
 {
 	int i, j, num_qd_pipes;
+	u32 ta_aux_cntl;
 	u32 sx_debug_1;
 	u32 smx_dc_ctl0;
+	u32 db_debug3;
 	u32 num_gs_verts_per_thread;
 	u32 vgt_gs_per_es;
 	u32 gs_prim_buffer_depth = 0;
@@ -524,6 +583,7 @@ static void rv770_gpu_init(struct radeon_device *rdev)
 
 	switch (rdev->config.rv770.max_tile_pipes) {
 	case 1:
+	default:
 		gb_tiling_config |= PIPE_TILING(0);
 		break;
 	case 2:
@@ -535,18 +595,19 @@ static void rv770_gpu_init(struct radeon_device *rdev)
 	case 8:
 		gb_tiling_config |= PIPE_TILING(3);
 		break;
-	default:
-		break;
 	}
+	rdev->config.rv770.tiling_npipes = rdev->config.rv770.max_tile_pipes;
 
 	if (rdev->family == CHIP_RV770)
 		gb_tiling_config |= BANK_TILING(1);
 	else
-		gb_tiling_config |= BANK_TILING((mc_arb_ramcfg & NOOFBANK_SHIFT) >> NOOFBANK_MASK);
+		gb_tiling_config |= BANK_TILING((mc_arb_ramcfg & NOOFBANK_MASK) >> NOOFBANK_SHIFT);
+	rdev->config.rv770.tiling_nbanks = 4 << ((gb_tiling_config >> 4) & 0x3);
 
 	gb_tiling_config |= GROUP_SIZE(0);
+	rdev->config.rv770.tiling_group_size = 256;
 
-	if (((mc_arb_ramcfg & NOOFROWS_MASK) & NOOFROWS_SHIFT) > 3) {
+	if (((mc_arb_ramcfg & NOOFROWS_MASK) >> NOOFROWS_SHIFT) > 3) {
 		gb_tiling_config |= ROW_TILING(3);
 		gb_tiling_config |= SAMPLE_SPLIT(3);
 	} else {
@@ -558,18 +619,27 @@ static void rv770_gpu_init(struct radeon_device *rdev)
 
 	gb_tiling_config |= BANK_SWAPS(1);
 
-	backend_map = r700_get_tile_pipe_to_backend_map(rdev->config.rv770.max_tile_pipes,
-							rdev->config.rv770.max_backends,
-							(0xff << rdev->config.rv770.max_backends) & 0xff);
-	gb_tiling_config |= BACKEND_MAP(backend_map);
+	cc_rb_backend_disable = RREG32(CC_RB_BACKEND_DISABLE) & 0x00ff0000;
+	cc_rb_backend_disable |=
+		BACKEND_DISABLE((R7XX_MAX_BACKENDS_MASK << rdev->config.rv770.max_backends) & R7XX_MAX_BACKENDS_MASK);
 
-	cc_gc_shader_pipe_config =
+	cc_gc_shader_pipe_config = RREG32(CC_GC_SHADER_PIPE_CONFIG) & 0xffffff00;
+	cc_gc_shader_pipe_config |=
 		INACTIVE_QD_PIPES((R7XX_MAX_PIPES_MASK << rdev->config.rv770.max_pipes) & R7XX_MAX_PIPES_MASK);
 	cc_gc_shader_pipe_config |=
 		INACTIVE_SIMDS((R7XX_MAX_SIMDS_MASK << rdev->config.rv770.max_simds) & R7XX_MAX_SIMDS_MASK);
 
-	cc_rb_backend_disable =
-		BACKEND_DISABLE((R7XX_MAX_BACKENDS_MASK << rdev->config.rv770.max_backends) & R7XX_MAX_BACKENDS_MASK);
+	if (rdev->family == CHIP_RV740)
+		backend_map = 0x28;
+	else
+		backend_map = r700_get_tile_pipe_to_backend_map(rdev,
+								rdev->config.rv770.max_tile_pipes,
+								(R7XX_MAX_BACKENDS -
+								 r600_count_pipe_bits((cc_rb_backend_disable &
+										       R7XX_MAX_BACKENDS_MASK) >> 16)),
+								(cc_rb_backend_disable >> 16));
+	gb_tiling_config |= BACKEND_MAP(backend_map);
+
 
 	WREG32(GB_TILING_CONFIG, gb_tiling_config);
 	WREG32(DCP_TILING_CONFIG, (gb_tiling_config & 0xffff));
@@ -577,29 +647,24 @@ static void rv770_gpu_init(struct radeon_device *rdev)
 
 	WREG32(CC_RB_BACKEND_DISABLE,      cc_rb_backend_disable);
 	WREG32(CC_GC_SHADER_PIPE_CONFIG,   cc_gc_shader_pipe_config);
-	WREG32(GC_USER_SHADER_PIPE_CONFIG, cc_gc_shader_pipe_config);
+	WREG32(CC_SYS_RB_BACKEND_DISABLE,  cc_rb_backend_disable);
 
-	WREG32(CC_SYS_RB_BACKEND_DISABLE, cc_rb_backend_disable);
 	WREG32(CGTS_SYS_TCC_DISABLE, 0);
 	WREG32(CGTS_TCC_DISABLE, 0);
-	WREG32(CGTS_USER_SYS_TCC_DISABLE, 0);
-	WREG32(CGTS_USER_TCC_DISABLE, 0);
 
 	num_qd_pipes =
-		R7XX_MAX_BACKENDS - r600_count_pipe_bits(cc_gc_shader_pipe_config & INACTIVE_QD_PIPES_MASK);
+		R7XX_MAX_PIPES - r600_count_pipe_bits((cc_gc_shader_pipe_config & INACTIVE_QD_PIPES_MASK) >> 8);
 	WREG32(VGT_OUT_DEALLOC_CNTL, (num_qd_pipes * 4) & DEALLOC_DIST_MASK);
 	WREG32(VGT_VERTEX_REUSE_BLOCK_CNTL, ((num_qd_pipes * 4) - 2) & VTX_REUSE_DEPTH_MASK);
 
 	/* set HW defaults for 3D engine */
 	WREG32(CP_QUEUE_THRESHOLDS, (ROQ_IB1_START(0x16) |
-						ROQ_IB2_START(0x2b)));
+				     ROQ_IB2_START(0x2b)));
 
 	WREG32(CP_MEQ_THRESHOLDS, STQ_SPLIT(0x30));
 
-	WREG32(TA_CNTL_AUX, (DISABLE_CUBE_ANISO |
-					SYNC_GRADIENT |
-					SYNC_WALKER |
-					SYNC_ALIGNER));
+	ta_aux_cntl = RREG32(TA_CNTL_AUX);
+	WREG32(TA_CNTL_AUX, ta_aux_cntl | DISABLE_CUBE_ANISO);
 
 	sx_debug_1 = RREG32(SX_DEBUG_1);
 	sx_debug_1 |= ENABLE_NEW_SMX_ADDRESS;
@@ -610,26 +675,40 @@ static void rv770_gpu_init(struct radeon_device *rdev)
 	smx_dc_ctl0 |= CACHE_DEPTH((rdev->config.rv770.sx_num_of_sets * 64) - 1);
 	WREG32(SMX_DC_CTL0, smx_dc_ctl0);
 
-	WREG32(SMX_EVENT_CTL, (ES_FLUSH_CTL(4) |
-					  GS_FLUSH_CTL(4) |
-					  ACK_FLUSH_CTL(3) |
-					  SYNC_FLUSH_CTL));
+	if (rdev->family != CHIP_RV740)
+		WREG32(SMX_EVENT_CTL, (ES_FLUSH_CTL(4) |
+				       GS_FLUSH_CTL(4) |
+				       ACK_FLUSH_CTL(3) |
+				       SYNC_FLUSH_CTL));
 
-	if (rdev->family == CHIP_RV770)
-		WREG32(DB_DEBUG3, DB_CLK_OFF_DELAY(0x1f));
-	else {
+	db_debug3 = RREG32(DB_DEBUG3);
+	db_debug3 &= ~DB_CLK_OFF_DELAY(0x1f);
+	switch (rdev->family) {
+	case CHIP_RV770:
+	case CHIP_RV740:
+		db_debug3 |= DB_CLK_OFF_DELAY(0x1f);
+		break;
+	case CHIP_RV710:
+	case CHIP_RV730:
+	default:
+		db_debug3 |= DB_CLK_OFF_DELAY(2);
+		break;
+	}
+	WREG32(DB_DEBUG3, db_debug3);
+
+	if (rdev->family != CHIP_RV770) {
 		db_debug4 = RREG32(DB_DEBUG4);
 		db_debug4 |= DISABLE_TILE_COVERED_FOR_PS_ITER;
 		WREG32(DB_DEBUG4, db_debug4);
 	}
 
 	WREG32(SX_EXPORT_BUFFER_SIZES, (COLOR_BUFFER_SIZE((rdev->config.rv770.sx_max_export_size / 4) - 1) |
-						   POSITION_BUFFER_SIZE((rdev->config.rv770.sx_max_export_pos_size / 4) - 1) |
-						   SMX_BUFFER_SIZE((rdev->config.rv770.sx_max_export_smx_size / 4) - 1)));
+					POSITION_BUFFER_SIZE((rdev->config.rv770.sx_max_export_pos_size / 4) - 1) |
+					SMX_BUFFER_SIZE((rdev->config.rv770.sx_max_export_smx_size / 4) - 1)));
 
 	WREG32(PA_SC_FIFO_SIZE, (SC_PRIM_FIFO_SIZE(rdev->config.rv770.sc_prim_fifo_size) |
-						 SC_HIZ_TILE_FIFO_SIZE(rdev->config.rv770.sc_hiz_tile_fifo_size) |
-						 SC_EARLYZ_TILE_FIFO_SIZE(rdev->config.rv770.sc_earlyz_tile_fifo_fize)));
+				 SC_HIZ_TILE_FIFO_SIZE(rdev->config.rv770.sc_hiz_tile_fifo_size) |
+				 SC_EARLYZ_TILE_FIFO_SIZE(rdev->config.rv770.sc_earlyz_tile_fifo_fize)));
 
 	WREG32(PA_SC_MULTI_CHIP_CNTL, 0);
 
@@ -646,10 +725,10 @@ static void rv770_gpu_init(struct radeon_device *rdev)
 			    ALU_UPDATE_FIFO_HIWATER(0x8));
 	switch (rdev->family) {
 	case CHIP_RV770:
-		sq_ms_fifo_sizes |= FETCH_FIFO_HIWATER(0x1);
-		break;
 	case CHIP_RV730:
 	case CHIP_RV710:
+		sq_ms_fifo_sizes |= FETCH_FIFO_HIWATER(0x1);
+		break;
 	case CHIP_RV740:
 	default:
 		sq_ms_fifo_sizes |= FETCH_FIFO_HIWATER(0x4);
@@ -787,55 +866,48 @@ int rv770_mc_init(struct radeon_device *rdev)
 {
 	fixed20_12 a;
 	u32 tmp;
-	int r;
+	int chansize, numchan;
 
 	/* Get VRAM informations */
-	/* FIXME: Don't know how to determine vram width, need to check
-	 * vram_width usage
-	 */
-	rdev->mc.vram_width = 128;
 	rdev->mc.vram_is_ddr = true;
+	tmp = RREG32(MC_ARB_RAMCFG);
+	if (tmp & CHANSIZE_OVERRIDE) {
+		chansize = 16;
+	} else if (tmp & CHANSIZE_MASK) {
+		chansize = 64;
+	} else {
+		chansize = 32;
+	}
+	tmp = RREG32(MC_SHARED_CHMAP);
+	switch ((tmp & NOOFCHAN_MASK) >> NOOFCHAN_SHIFT) {
+	case 0:
+	default:
+		numchan = 1;
+		break;
+	case 1:
+		numchan = 2;
+		break;
+	case 2:
+		numchan = 4;
+		break;
+	case 3:
+		numchan = 8;
+		break;
+	}
+	rdev->mc.vram_width = numchan * chansize;
 	/* Could aper size report 0 ? */
 	rdev->mc.aper_base = drm_get_resource_start(rdev->ddev, 0);
 	rdev->mc.aper_size = drm_get_resource_len(rdev->ddev, 0);
 	/* Setup GPU memory space */
 	rdev->mc.mc_vram_size = RREG32(CONFIG_MEMSIZE);
 	rdev->mc.real_vram_size = RREG32(CONFIG_MEMSIZE);
-	if (rdev->flags & RADEON_IS_AGP) {
-		r = radeon_agp_init(rdev);
-		if (r)
-			return r;
-		/* gtt_size is setup by radeon_agp_init */
-		rdev->mc.gtt_location = rdev->mc.agp_base;
-		tmp = 0xFFFFFFFFUL - rdev->mc.agp_base - rdev->mc.gtt_size;
-		/* Try to put vram before or after AGP because we
-		 * we want SYSTEM_APERTURE to cover both VRAM and
-		 * AGP so that GPU can catch out of VRAM/AGP access
-		 */
-		if (rdev->mc.gtt_location > rdev->mc.mc_vram_size) {
-			/* Enought place before */
-			rdev->mc.vram_location = rdev->mc.gtt_location -
-							rdev->mc.mc_vram_size;
-		} else if (tmp > rdev->mc.mc_vram_size) {
-			/* Enought place after */
-			rdev->mc.vram_location = rdev->mc.gtt_location +
-							rdev->mc.gtt_size;
-		} else {
-			/* Try to setup VRAM then AGP might not
-			 * not work on some card
-			 */
-			rdev->mc.vram_location = 0x00000000UL;
-			rdev->mc.gtt_location = rdev->mc.mc_vram_size;
-		}
-	} else {
-		rdev->mc.vram_location = 0x00000000UL;
-		rdev->mc.gtt_location = rdev->mc.mc_vram_size;
-		rdev->mc.gtt_size = radeon_gart_size * 1024 * 1024;
+	rdev->mc.visible_vram_size = rdev->mc.aper_size;
+	/* FIXME remove this once we support unmappable VRAM */
+	if (rdev->mc.mc_vram_size > rdev->mc.aper_size) {
+		rdev->mc.mc_vram_size = rdev->mc.aper_size;
+		rdev->mc.real_vram_size = rdev->mc.aper_size;
 	}
-	rdev->mc.vram_start = rdev->mc.vram_location;
-	rdev->mc.vram_end = rdev->mc.vram_location + rdev->mc.mc_vram_size;
-	rdev->mc.gtt_start = rdev->mc.gtt_location;
-	rdev->mc.gtt_end = rdev->mc.gtt_location + rdev->mc.gtt_size;
+	r600_vram_gtt_location(rdev, &rdev->mc);
 	/* FIXME: we should enforce default clock in case GPU is not in
 	 * default setup
 	 */
@@ -844,6 +916,7 @@ int rv770_mc_init(struct radeon_device *rdev)
 	rdev->pm.sclk.full = rfixed_div(rdev->pm.sclk, a);
 	return 0;
 }
+
 int rv770_gpu_reset(struct radeon_device *rdev)
 {
 	/* FIXME: implement any rv770 specific bits */
@@ -854,19 +927,50 @@ static int rv770_startup(struct radeon_device *rdev)
 {
 	int r;
 
-	radeon_gpu_reset(rdev);
-	rv770_mc_resume(rdev);
-	r = rv770_pcie_gart_enable(rdev);
-	if (r)
-		return r;
-	rv770_gpu_init(rdev);
+	if (!rdev->me_fw || !rdev->pfp_fw || !rdev->rlc_fw) {
+		r = r600_init_microcode(rdev);
+		if (r) {
+			DRM_ERROR("Failed to load firmware!\n");
+			return r;
+		}
+	}
 
-	r = radeon_object_pin(rdev->r600_blit.shader_obj, RADEON_GEM_DOMAIN_VRAM,
-			      &rdev->r600_blit.shader_gpu_addr);
+	rv770_mc_program(rdev);
+	if (rdev->flags & RADEON_IS_AGP) {
+		rv770_agp_enable(rdev);
+	} else {
+		r = rv770_pcie_gart_enable(rdev);
+		if (r)
+			return r;
+	}
+	rv770_gpu_init(rdev);
+	r = r600_blit_init(rdev);
+	if (r) {
+		r600_blit_fini(rdev);
+		rdev->asic->copy = NULL;
+		dev_warn(rdev->dev, "failed blitter (%d) falling back to memcpy\n", r);
+	}
+	/* pin copy shader into vram */
+	if (rdev->r600_blit.shader_obj) {
+		r = radeon_bo_reserve(rdev->r600_blit.shader_obj, false);
+		if (unlikely(r != 0))
+			return r;
+		r = radeon_bo_pin(rdev->r600_blit.shader_obj, RADEON_GEM_DOMAIN_VRAM,
+				&rdev->r600_blit.shader_gpu_addr);
+		radeon_bo_unreserve(rdev->r600_blit.shader_obj);
+		if (r) {
+			DRM_ERROR("failed to pin blit object %d\n", r);
+			return r;
+		}
+	}
+	/* Enable IRQ */
+	r = r600_irq_init(rdev);
 	if (r) {
-		DRM_ERROR("failed to pin blit object %d\n", r);
+		DRM_ERROR("radeon: IH init failed (%d).\n", r);
+		radeon_irq_kms_fini(rdev);
 		return r;
 	}
+	r600_irq_set(rdev);
 
 	r = radeon_ring_init(rdev, rdev->cp.ring_size);
 	if (r)
@@ -877,9 +981,8 @@ static int rv770_startup(struct radeon_device *rdev)
 	r = r600_cp_resume(rdev);
 	if (r)
 		return r;
-	r = r600_wb_init(rdev);
-	if (r)
-		return r;
+	/* write back buffer are not vital so don't worry about failure */
+	r600_wb_enable(rdev);
 	return 0;
 }
 
@@ -887,15 +990,12 @@ int rv770_resume(struct radeon_device *rdev)
 {
 	int r;
 
-	if (radeon_gpu_reset(rdev)) {
-		/* FIXME: what do we want to do here ? */
-	}
+	/* Do not reset GPU before posting, on rv770 hw unlike on r500 hw,
+	 * posting will perform necessary task to bring back GPU into good
+	 * shape.
+	 */
 	/* post card */
-	if (rdev->is_atom_bios) {
-		atom_asic_init(rdev->mode_info.atom_context);
-	} else {
-		radeon_combios_asic_init(rdev->ddev);
-	}
+	atom_asic_init(rdev->mode_info.atom_context);
 	/* Initialize clocks */
 	r = radeon_clocks_init(rdev);
 	if (r) {
@@ -908,7 +1008,7 @@ int rv770_resume(struct radeon_device *rdev)
 		return r;
 	}
 
-	r = radeon_ib_test(rdev);
+	r = r600_ib_test(rdev);
 	if (r) {
 		DRM_ERROR("radeon: failled testing IB (%d).\n", r);
 		return r;
@@ -919,13 +1019,22 @@ int rv770_resume(struct radeon_device *rdev)
 
 int rv770_suspend(struct radeon_device *rdev)
 {
+	int r;
+
 	/* FIXME: we should wait for ring to be empty */
 	r700_cp_stop(rdev);
 	rdev->cp.ready = false;
+	r600_irq_suspend(rdev);
+	r600_wb_disable(rdev);
 	rv770_pcie_gart_disable(rdev);
-
 	/* unpin shaders bo */
-        radeon_object_unpin(rdev->r600_blit.shader_obj);
+	if (rdev->r600_blit.shader_obj) {
+		r = radeon_bo_reserve(rdev->r600_blit.shader_obj, false);
+		if (likely(r == 0)) {
+			radeon_bo_unpin(rdev->r600_blit.shader_obj);
+			radeon_bo_unreserve(rdev->r600_blit.shader_obj);
+		}
+	}
 	return 0;
 }
 
@@ -939,7 +1048,6 @@ int rv770_init(struct radeon_device *rdev)
 {
 	int r;
 
-	rdev->new_init_path = true;
 	r = radeon_dummy_page_init(rdev);
 	if (r)
 		return r;
@@ -953,13 +1061,19 @@ int rv770_init(struct radeon_device *rdev)
 			return -EINVAL;
 	}
 	/* Must be an ATOMBIOS */
-	if (!rdev->is_atom_bios)
+	if (!rdev->is_atom_bios) {
+		dev_err(rdev->dev, "Expecting atombios for R600 GPU\n");
 		return -EINVAL;
+	}
 	r = radeon_atombios_init(rdev);
 	if (r)
 		return r;
 	/* Post card if necessary */
-	if (!r600_card_posted(rdev) && rdev->bios) {
+	if (!r600_card_posted(rdev)) {
+		if (!rdev->bios) {
+			dev_err(rdev->dev, "Card not posted and no BIOS - ignoring\n");
+			return -EINVAL;
+		}
 		DRM_INFO("GPU not posted. posting now...\n");
 		atom_asic_init(rdev->mode_info.atom_context);
 	}
@@ -967,70 +1081,67 @@ int rv770_init(struct radeon_device *rdev)
 	r600_scratch_init(rdev);
 	/* Initialize surface registers */
 	radeon_surface_init(rdev);
+	/* Initialize clocks */
 	radeon_get_clock_info(rdev->ddev);
 	r = radeon_clocks_init(rdev);
 	if (r)
 		return r;
+	/* Initialize power management */
+	radeon_pm_init(rdev);
 	/* Fence driver */
 	r = radeon_fence_driver_init(rdev);
 	if (r)
 		return r;
+	/* initialize AGP */
+	if (rdev->flags & RADEON_IS_AGP) {
+		r = radeon_agp_init(rdev);
+		if (r)
+			radeon_agp_disable(rdev);
+	}
 	r = rv770_mc_init(rdev);
-	if (r) {
-		if (rdev->flags & RADEON_IS_AGP) {
-			/* Retry with disabling AGP */
-			rv770_fini(rdev);
-			rdev->flags &= ~RADEON_IS_AGP;
-			return rv770_init(rdev);
-		}
+	if (r)
 		return r;
-	}
 	/* Memory manager */
-	r = radeon_object_init(rdev);
+	r = radeon_bo_init(rdev);
+	if (r)
+		return r;
+
+	r = radeon_irq_kms_init(rdev);
 	if (r)
 		return r;
+
 	rdev->cp.ring_obj = NULL;
 	r600_ring_init(rdev, 1024 * 1024);
 
-	if (!rdev->me_fw || !rdev->pfp_fw) {
-		r = r600_cp_init_microcode(rdev);
-		if (r) {
-			DRM_ERROR("Failed to load firmware!\n");
-			return r;
-		}
-	}
+	rdev->ih.ring_obj = NULL;
+	r600_ih_ring_init(rdev, 64 * 1024);
 
 	r = r600_pcie_gart_init(rdev);
 	if (r)
 		return r;
 
 	rdev->accel_working = true;
-	r = r600_blit_init(rdev);
-	if (r) {
-		DRM_ERROR("radeon: failled blitter (%d).\n", r);
-		rdev->accel_working = false;
-	}
-
 	r = rv770_startup(rdev);
 	if (r) {
-		if (rdev->flags & RADEON_IS_AGP) {
-			/* Retry with disabling AGP */
-			rv770_fini(rdev);
-			rdev->flags &= ~RADEON_IS_AGP;
-			return rv770_init(rdev);
-		}
+		dev_err(rdev->dev, "disabling GPU acceleration\n");
+		r600_cp_fini(rdev);
+		r600_wb_fini(rdev);
+		r600_irq_fini(rdev);
+		radeon_irq_kms_fini(rdev);
+		rv770_pcie_gart_fini(rdev);
 		rdev->accel_working = false;
 	}
 	if (rdev->accel_working) {
 		r = radeon_ib_pool_init(rdev);
 		if (r) {
-			DRM_ERROR("radeon: failled initializing IB pool (%d).\n", r);
-			rdev->accel_working = false;
-		}
-		r = radeon_ib_test(rdev);
-		if (r) {
-			DRM_ERROR("radeon: failled testing IB (%d).\n", r);
+			dev_err(rdev->dev, "IB initialization failed (%d).\n", r);
 			rdev->accel_working = false;
+		} else {
+			r = r600_ib_test(rdev);
+			if (r) {
+				dev_err(rdev->dev, "IB test failed (%d).\n", r);
+				rdev->accel_working = false;
+			}
 		}
 	}
 	return 0;
@@ -1038,24 +1149,18 @@ int rv770_init(struct radeon_device *rdev)
 
 void rv770_fini(struct radeon_device *rdev)
 {
-	rv770_suspend(rdev);
-
 	r600_blit_fini(rdev);
-	radeon_ring_fini(rdev);
+	r600_cp_fini(rdev);
+	r600_wb_fini(rdev);
+	r600_irq_fini(rdev);
+	radeon_irq_kms_fini(rdev);
 	rv770_pcie_gart_fini(rdev);
 	radeon_gem_fini(rdev);
 	radeon_fence_driver_fini(rdev);
 	radeon_clocks_fini(rdev);
-#if __OS_HAS_AGP
-	if (rdev->flags & RADEON_IS_AGP)
-		radeon_agp_fini(rdev);
-#endif
-	radeon_object_fini(rdev);
-	if (rdev->is_atom_bios) {
-		radeon_atombios_fini(rdev);
-	} else {
-		radeon_combios_fini(rdev);
-	}
+	radeon_agp_fini(rdev);
+	radeon_bo_fini(rdev);
+	radeon_atombios_fini(rdev);
 	kfree(rdev->bios);
 	rdev->bios = NULL;
 	radeon_dummy_page_fini(rdev);
diff --git a/drivers/gpu/drm/radeon/rv770d.h b/drivers/gpu/drm/radeon/rv770d.h
index 4b9c3d6396ff..9506f8cb99e0 100644
--- a/drivers/gpu/drm/radeon/rv770d.h
+++ b/drivers/gpu/drm/radeon/rv770d.h
@@ -129,6 +129,10 @@
 #define HDP_REG_COHERENCY_FLUSH_CNTL			0x54A0
 #define	HDP_TILING_CONFIG				0x2F3C
 
+#define MC_SHARED_CHMAP						0x2004
+#define		NOOFCHAN_SHIFT					12
+#define		NOOFCHAN_MASK					0x00003000
+
 #define	MC_ARB_RAMCFG					0x2760
 #define		NOOFBANK_SHIFT					0
 #define		NOOFBANK_MASK					0x00000003
@@ -142,6 +146,7 @@
 #define		CHANSIZE_MASK					0x00000100
 #define		BURSTLENGTH_SHIFT				9
 #define		BURSTLENGTH_MASK				0x00000200
+#define		CHANSIZE_OVERRIDE				(1 << 11)
 #define	MC_VM_AGP_TOP					0x2028
 #define	MC_VM_AGP_BOT					0x202C
 #define	MC_VM_AGP_BASE					0x2030
@@ -338,4 +343,6 @@
 
 #define	WAIT_UNTIL					0x8040
 
+#define	SRBM_STATUS				        0x0E50
+
 #endif
diff --git a/drivers/gpu/drm/savage/savage_drv.c b/drivers/gpu/drm/savage/savage_drv.c
index eee52aa92a7c..021de44c15ab 100644
--- a/drivers/gpu/drm/savage/savage_drv.c
+++ b/drivers/gpu/drm/savage/savage_drv.c
@@ -50,7 +50,7 @@ static struct drm_driver driver = {
 		 .owner = THIS_MODULE,
 		 .open = drm_open,
 		 .release = drm_release,
-		 .ioctl = drm_ioctl,
+		 .unlocked_ioctl = drm_ioctl,
 		 .mmap = drm_mmap,
 		 .poll = drm_poll,
 		 .fasync = drm_fasync,
diff --git a/drivers/gpu/drm/sis/sis_drv.c b/drivers/gpu/drm/sis/sis_drv.c
index e725cc0b1155..4fd1f067d380 100644
--- a/drivers/gpu/drm/sis/sis_drv.c
+++ b/drivers/gpu/drm/sis/sis_drv.c
@@ -80,7 +80,7 @@ static struct drm_driver driver = {
 		 .owner = THIS_MODULE,
 		 .open = drm_open,
 		 .release = drm_release,
-		 .ioctl = drm_ioctl,
+		 .unlocked_ioctl = drm_ioctl,
 		 .mmap = drm_mmap,
 		 .poll = drm_poll,
 		 .fasync = drm_fasync,
diff --git a/drivers/gpu/drm/tdfx/tdfx_drv.c b/drivers/gpu/drm/tdfx/tdfx_drv.c
index 012ff2e356b2..ec5a43e65722 100644
--- a/drivers/gpu/drm/tdfx/tdfx_drv.c
+++ b/drivers/gpu/drm/tdfx/tdfx_drv.c
@@ -48,7 +48,7 @@ static struct drm_driver driver = {
 		 .owner = THIS_MODULE,
 		 .open = drm_open,
 		 .release = drm_release,
-		 .ioctl = drm_ioctl,
+		 .unlocked_ioctl = drm_ioctl,
 		 .mmap = drm_mmap,
 		 .poll = drm_poll,
 		 .fasync = drm_fasync,
diff --git a/drivers/gpu/drm/ttm/Makefile b/drivers/gpu/drm/ttm/Makefile
index b0a9de7a57c2..1e138f5bae09 100644
--- a/drivers/gpu/drm/ttm/Makefile
+++ b/drivers/gpu/drm/ttm/Makefile
@@ -3,6 +3,7 @@
 
 ccflags-y := -Iinclude/drm
 ttm-y := ttm_agp_backend.o ttm_memory.o ttm_tt.o ttm_bo.o \
-	ttm_bo_util.o ttm_bo_vm.o ttm_module.o ttm_global.o
+	ttm_bo_util.o ttm_bo_vm.o ttm_module.o ttm_global.o \
+	ttm_object.o ttm_lock.o ttm_execbuf_util.o
 
 obj-$(CONFIG_DRM_TTM) += ttm.o
diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
index 87c06252d464..89c38c49066f 100644
--- a/drivers/gpu/drm/ttm/ttm_bo.c
+++ b/drivers/gpu/drm/ttm/ttm_bo.c
@@ -27,6 +27,14 @@
 /*
  * Authors: Thomas Hellstrom <thellstrom-at-vmware-dot-com>
  */
+/* Notes:
+ *
+ * We store bo pointer in drm_mm_node struct so we know which bo own a
+ * specific node. There is no protection on the pointer, thus to make
+ * sure things don't go berserk you have to access this pointer while
+ * holding the global lru lock and make sure anytime you free a node you
+ * reset the pointer to NULL.
+ */
 
 #include "ttm/ttm_module.h"
 #include "ttm/ttm_bo_driver.h"
@@ -51,6 +59,59 @@ static struct attribute ttm_bo_count = {
 	.mode = S_IRUGO
 };
 
+static inline int ttm_mem_type_from_flags(uint32_t flags, uint32_t *mem_type)
+{
+	int i;
+
+	for (i = 0; i <= TTM_PL_PRIV5; i++)
+		if (flags & (1 << i)) {
+			*mem_type = i;
+			return 0;
+		}
+	return -EINVAL;
+}
+
+static void ttm_mem_type_debug(struct ttm_bo_device *bdev, int mem_type)
+{
+	struct ttm_mem_type_manager *man = &bdev->man[mem_type];
+
+	printk(KERN_ERR TTM_PFX "    has_type: %d\n", man->has_type);
+	printk(KERN_ERR TTM_PFX "    use_type: %d\n", man->use_type);
+	printk(KERN_ERR TTM_PFX "    flags: 0x%08X\n", man->flags);
+	printk(KERN_ERR TTM_PFX "    gpu_offset: 0x%08lX\n", man->gpu_offset);
+	printk(KERN_ERR TTM_PFX "    io_offset: 0x%08lX\n", man->io_offset);
+	printk(KERN_ERR TTM_PFX "    io_size: %ld\n", man->io_size);
+	printk(KERN_ERR TTM_PFX "    size: %llu\n", man->size);
+	printk(KERN_ERR TTM_PFX "    available_caching: 0x%08X\n",
+		man->available_caching);
+	printk(KERN_ERR TTM_PFX "    default_caching: 0x%08X\n",
+		man->default_caching);
+	if (mem_type != TTM_PL_SYSTEM) {
+		spin_lock(&bdev->glob->lru_lock);
+		drm_mm_debug_table(&man->manager, TTM_PFX);
+		spin_unlock(&bdev->glob->lru_lock);
+	}
+}
+
+static void ttm_bo_mem_space_debug(struct ttm_buffer_object *bo,
+					struct ttm_placement *placement)
+{
+	int i, ret, mem_type;
+
+	printk(KERN_ERR TTM_PFX "No space for %p (%lu pages, %luK, %luM)\n",
+		bo, bo->mem.num_pages, bo->mem.size >> 10,
+		bo->mem.size >> 20);
+	for (i = 0; i < placement->num_placement; i++) {
+		ret = ttm_mem_type_from_flags(placement->placement[i],
+						&mem_type);
+		if (ret)
+			return;
+		printk(KERN_ERR TTM_PFX "  placement[%d]=0x%08X (%d)\n",
+			i, placement->placement[i], mem_type);
+		ttm_mem_type_debug(bo->bdev, mem_type);
+	}
+}
+
 static ssize_t ttm_bo_global_show(struct kobject *kobj,
 				  struct attribute *attr,
 				  char *buffer)
@@ -67,7 +128,7 @@ static struct attribute *ttm_bo_global_attrs[] = {
 	NULL
 };
 
-static struct sysfs_ops ttm_bo_global_ops = {
+static const struct sysfs_ops ttm_bo_global_ops = {
 	.show = &ttm_bo_global_show
 };
 
@@ -117,12 +178,13 @@ int ttm_bo_wait_unreserved(struct ttm_buffer_object *bo, bool interruptible)
 		ret = wait_event_interruptible(bo->event_queue,
 					       atomic_read(&bo->reserved) == 0);
 		if (unlikely(ret != 0))
-			return -ERESTART;
+			return ret;
 	} else {
 		wait_event(bo->event_queue, atomic_read(&bo->reserved) == 0);
 	}
 	return 0;
 }
+EXPORT_SYMBOL(ttm_bo_wait_unreserved);
 
 static void ttm_bo_add_to_lru(struct ttm_buffer_object *bo)
 {
@@ -247,7 +309,6 @@ EXPORT_SYMBOL(ttm_bo_unreserve);
 /*
  * Call bo->mutex locked.
  */
-
 static int ttm_bo_add_ttm(struct ttm_buffer_object *bo, bool zero_alloc)
 {
 	struct ttm_bo_device *bdev = bo->bdev;
@@ -275,9 +336,10 @@ static int ttm_bo_add_ttm(struct ttm_buffer_object *bo, bool zero_alloc)
 		bo->ttm = ttm_tt_create(bdev, bo->num_pages << PAGE_SHIFT,
 					page_flags | TTM_PAGE_FLAG_USER,
 					glob->dummy_read_page);
-		if (unlikely(bo->ttm == NULL))
+		if (unlikely(bo->ttm == NULL)) {
 			ret = -ENOMEM;
-		break;
+			break;
+		}
 
 		ret = ttm_tt_set_user(bo->ttm, current,
 				      bo->buffer_start, bo->num_pages);
@@ -328,14 +390,8 @@ static int ttm_bo_handle_move_mem(struct ttm_buffer_object *bo,
 		}
 
 		if (bo->mem.mem_type == TTM_PL_SYSTEM) {
-
-			struct ttm_mem_reg *old_mem = &bo->mem;
-			uint32_t save_flags = old_mem->placement;
-
-			*old_mem = *mem;
+			bo->mem = *mem;
 			mem->mm_node = NULL;
-			ttm_flag_masked(&save_flags, mem->placement,
-					TTM_PL_MASK_MEMTYPE);
 			goto moved;
 		}
 
@@ -370,7 +426,8 @@ moved:
 		    bdev->man[bo->mem.mem_type].gpu_offset;
 		bo->cur_placement = bo->mem.placement;
 		spin_unlock(&bo->lock);
-	}
+	} else
+		bo->offset = 0;
 
 	return 0;
 
@@ -408,6 +465,8 @@ static int ttm_bo_cleanup_refs(struct ttm_buffer_object *bo, bool remove_all)
 		spin_unlock(&bo->lock);
 
 		spin_lock(&glob->lru_lock);
+		put_count = ttm_bo_del_from_lru(bo);
+
 		ret = ttm_bo_reserve_locked(bo, false, false, false, 0);
 		BUG_ON(ret);
 		if (bo->ttm)
@@ -415,19 +474,19 @@ static int ttm_bo_cleanup_refs(struct ttm_buffer_object *bo, bool remove_all)
 
 		if (!list_empty(&bo->ddestroy)) {
 			list_del_init(&bo->ddestroy);
-			kref_put(&bo->list_kref, ttm_bo_ref_bug);
+			++put_count;
 		}
 		if (bo->mem.mm_node) {
+			bo->mem.mm_node->private = NULL;
 			drm_mm_put_block(bo->mem.mm_node);
 			bo->mem.mm_node = NULL;
 		}
-		put_count = ttm_bo_del_from_lru(bo);
 		spin_unlock(&glob->lru_lock);
 
 		atomic_set(&bo->reserved, 0);
 
 		while (put_count--)
-			kref_put(&bo->list_kref, ttm_bo_release_list);
+			kref_put(&bo->list_kref, ttm_bo_ref_bug);
 
 		return 0;
 	}
@@ -465,52 +524,44 @@ static int ttm_bo_cleanup_refs(struct ttm_buffer_object *bo, bool remove_all)
 static int ttm_bo_delayed_delete(struct ttm_bo_device *bdev, bool remove_all)
 {
 	struct ttm_bo_global *glob = bdev->glob;
-	struct ttm_buffer_object *entry, *nentry;
-	struct list_head *list, *next;
-	int ret;
+	struct ttm_buffer_object *entry = NULL;
+	int ret = 0;
 
 	spin_lock(&glob->lru_lock);
-	list_for_each_safe(list, next, &bdev->ddestroy) {
-		entry = list_entry(list, struct ttm_buffer_object, ddestroy);
-		nentry = NULL;
+	if (list_empty(&bdev->ddestroy))
+		goto out_unlock;
 
-		/*
-		 * Protect the next list entry from destruction while we
-		 * unlock the lru_lock.
-		 */
+	entry = list_first_entry(&bdev->ddestroy,
+		struct ttm_buffer_object, ddestroy);
+	kref_get(&entry->list_kref);
 
-		if (next != &bdev->ddestroy) {
-			nentry = list_entry(next, struct ttm_buffer_object,
-					    ddestroy);
+	for (;;) {
+		struct ttm_buffer_object *nentry = NULL;
+
+		if (entry->ddestroy.next != &bdev->ddestroy) {
+			nentry = list_first_entry(&entry->ddestroy,
+				struct ttm_buffer_object, ddestroy);
 			kref_get(&nentry->list_kref);
 		}
-		kref_get(&entry->list_kref);
 
 		spin_unlock(&glob->lru_lock);
 		ret = ttm_bo_cleanup_refs(entry, remove_all);
 		kref_put(&entry->list_kref, ttm_bo_release_list);
+		entry = nentry;
+
+		if (ret || !entry)
+			goto out;
 
 		spin_lock(&glob->lru_lock);
-		if (nentry) {
-			bool next_onlist = !list_empty(next);
-			spin_unlock(&glob->lru_lock);
-			kref_put(&nentry->list_kref, ttm_bo_release_list);
-			spin_lock(&glob->lru_lock);
-			/*
-			 * Someone might have raced us and removed the
-			 * next entry from the list. We don't bother restarting
-			 * list traversal.
-			 */
-
-			if (!next_onlist)
-				break;
-		}
-		if (ret)
+		if (list_empty(&entry->ddestroy))
 			break;
 	}
-	ret = !list_empty(&bdev->ddestroy);
-	spin_unlock(&glob->lru_lock);
 
+out_unlock:
+	spin_unlock(&glob->lru_lock);
+out:
+	if (entry)
+		kref_put(&entry->list_kref, ttm_bo_release_list);
 	return ret;
 }
 
@@ -554,24 +605,21 @@ void ttm_bo_unref(struct ttm_buffer_object **p_bo)
 }
 EXPORT_SYMBOL(ttm_bo_unref);
 
-static int ttm_bo_evict(struct ttm_buffer_object *bo, unsigned mem_type,
-			bool interruptible, bool no_wait)
+static int ttm_bo_evict(struct ttm_buffer_object *bo, bool interruptible,
+			bool no_wait)
 {
-	int ret = 0;
 	struct ttm_bo_device *bdev = bo->bdev;
 	struct ttm_bo_global *glob = bo->glob;
 	struct ttm_mem_reg evict_mem;
-	uint32_t proposed_placement;
-
-	if (bo->mem.mem_type != mem_type)
-		goto out;
+	struct ttm_placement placement;
+	int ret = 0;
 
 	spin_lock(&bo->lock);
 	ret = ttm_bo_wait(bo, false, interruptible, no_wait);
 	spin_unlock(&bo->lock);
 
 	if (unlikely(ret != 0)) {
-		if (ret != -ERESTART) {
+		if (ret != -ERESTARTSYS) {
 			printk(KERN_ERR TTM_PFX
 			       "Failed to expire sync object before "
 			       "buffer eviction.\n");
@@ -584,116 +632,165 @@ static int ttm_bo_evict(struct ttm_buffer_object *bo, unsigned mem_type,
 	evict_mem = bo->mem;
 	evict_mem.mm_node = NULL;
 
-	proposed_placement = bdev->driver->evict_flags(bo);
-
-	ret = ttm_bo_mem_space(bo, proposed_placement,
-			       &evict_mem, interruptible, no_wait);
-	if (unlikely(ret != 0 && ret != -ERESTART))
-		ret = ttm_bo_mem_space(bo, TTM_PL_FLAG_SYSTEM,
-				       &evict_mem, interruptible, no_wait);
-
+	placement.fpfn = 0;
+	placement.lpfn = 0;
+	placement.num_placement = 0;
+	placement.num_busy_placement = 0;
+	bdev->driver->evict_flags(bo, &placement);
+	ret = ttm_bo_mem_space(bo, &placement, &evict_mem, interruptible,
+				no_wait);
 	if (ret) {
-		if (ret != -ERESTART)
+		if (ret != -ERESTARTSYS) {
 			printk(KERN_ERR TTM_PFX
 			       "Failed to find memory space for "
 			       "buffer 0x%p eviction.\n", bo);
+			ttm_bo_mem_space_debug(bo, &placement);
+		}
 		goto out;
 	}
 
 	ret = ttm_bo_handle_move_mem(bo, &evict_mem, true, interruptible,
 				     no_wait);
 	if (ret) {
-		if (ret != -ERESTART)
+		if (ret != -ERESTARTSYS)
 			printk(KERN_ERR TTM_PFX "Buffer eviction failed\n");
+		spin_lock(&glob->lru_lock);
+		if (evict_mem.mm_node) {
+			evict_mem.mm_node->private = NULL;
+			drm_mm_put_block(evict_mem.mm_node);
+			evict_mem.mm_node = NULL;
+		}
+		spin_unlock(&glob->lru_lock);
 		goto out;
 	}
-
-	spin_lock(&glob->lru_lock);
-	if (evict_mem.mm_node) {
-		drm_mm_put_block(evict_mem.mm_node);
-		evict_mem.mm_node = NULL;
-	}
-	spin_unlock(&glob->lru_lock);
 	bo->evicted = true;
 out:
 	return ret;
 }
 
-/**
- * Repeatedly evict memory from the LRU for @mem_type until we create enough
- * space, or we've evicted everything and there isn't enough space.
- */
-static int ttm_bo_mem_force_space(struct ttm_bo_device *bdev,
-				  struct ttm_mem_reg *mem,
-				  uint32_t mem_type,
-				  bool interruptible, bool no_wait)
+static int ttm_mem_evict_first(struct ttm_bo_device *bdev,
+				uint32_t mem_type,
+				bool interruptible, bool no_wait)
 {
 	struct ttm_bo_global *glob = bdev->glob;
-	struct drm_mm_node *node;
-	struct ttm_buffer_object *entry;
 	struct ttm_mem_type_manager *man = &bdev->man[mem_type];
-	struct list_head *lru;
-	unsigned long num_pages = mem->num_pages;
-	int put_count = 0;
-	int ret;
-
-retry_pre_get:
-	ret = drm_mm_pre_get(&man->manager);
-	if (unlikely(ret != 0))
-		return ret;
+	struct ttm_buffer_object *bo;
+	int ret, put_count = 0;
 
+retry:
 	spin_lock(&glob->lru_lock);
-	do {
-		node = drm_mm_search_free(&man->manager, num_pages,
-					  mem->page_alignment, 1);
-		if (node)
-			break;
+	if (list_empty(&man->lru)) {
+		spin_unlock(&glob->lru_lock);
+		return -EBUSY;
+	}
 
-		lru = &man->lru;
-		if (list_empty(lru))
-			break;
+	bo = list_first_entry(&man->lru, struct ttm_buffer_object, lru);
+	kref_get(&bo->list_kref);
 
-		entry = list_first_entry(lru, struct ttm_buffer_object, lru);
-		kref_get(&entry->list_kref);
+	ret = ttm_bo_reserve_locked(bo, false, true, false, 0);
 
-		ret =
-		    ttm_bo_reserve_locked(entry, interruptible, no_wait,
-					  false, 0);
+	if (unlikely(ret == -EBUSY)) {
+		spin_unlock(&glob->lru_lock);
+		if (likely(!no_wait))
+			ret = ttm_bo_wait_unreserved(bo, interruptible);
 
-		if (likely(ret == 0))
-			put_count = ttm_bo_del_from_lru(entry);
+		kref_put(&bo->list_kref, ttm_bo_release_list);
 
-		spin_unlock(&glob->lru_lock);
+		/**
+		 * We *need* to retry after releasing the lru lock.
+		 */
 
 		if (unlikely(ret != 0))
 			return ret;
+		goto retry;
+	}
 
-		while (put_count--)
-			kref_put(&entry->list_kref, ttm_bo_ref_bug);
+	put_count = ttm_bo_del_from_lru(bo);
+	spin_unlock(&glob->lru_lock);
+
+	BUG_ON(ret != 0);
+
+	while (put_count--)
+		kref_put(&bo->list_kref, ttm_bo_ref_bug);
 
-		ret = ttm_bo_evict(entry, mem_type, interruptible, no_wait);
+	ret = ttm_bo_evict(bo, interruptible, no_wait);
+	ttm_bo_unreserve(bo);
 
-		ttm_bo_unreserve(entry);
+	kref_put(&bo->list_kref, ttm_bo_release_list);
+	return ret;
+}
 
-		kref_put(&entry->list_kref, ttm_bo_release_list);
-		if (ret)
+static int ttm_bo_man_get_node(struct ttm_buffer_object *bo,
+				struct ttm_mem_type_manager *man,
+				struct ttm_placement *placement,
+				struct ttm_mem_reg *mem,
+				struct drm_mm_node **node)
+{
+	struct ttm_bo_global *glob = bo->glob;
+	unsigned long lpfn;
+	int ret;
+
+	lpfn = placement->lpfn;
+	if (!lpfn)
+		lpfn = man->size;
+	*node = NULL;
+	do {
+		ret = drm_mm_pre_get(&man->manager);
+		if (unlikely(ret))
 			return ret;
 
 		spin_lock(&glob->lru_lock);
-	} while (1);
-
-	if (!node) {
+		*node = drm_mm_search_free_in_range(&man->manager,
+					mem->num_pages, mem->page_alignment,
+					placement->fpfn, lpfn, 1);
+		if (unlikely(*node == NULL)) {
+			spin_unlock(&glob->lru_lock);
+			return 0;
+		}
+		*node = drm_mm_get_block_atomic_range(*node, mem->num_pages,
+							mem->page_alignment,
+							placement->fpfn,
+							lpfn);
 		spin_unlock(&glob->lru_lock);
-		return -ENOMEM;
-	}
+	} while (*node == NULL);
+	return 0;
+}
 
-	node = drm_mm_get_block_atomic(node, num_pages, mem->page_alignment);
-	if (unlikely(!node)) {
-		spin_unlock(&glob->lru_lock);
-		goto retry_pre_get;
-	}
+/**
+ * Repeatedly evict memory from the LRU for @mem_type until we create enough
+ * space, or we've evicted everything and there isn't enough space.
+ */
+static int ttm_bo_mem_force_space(struct ttm_buffer_object *bo,
+					uint32_t mem_type,
+					struct ttm_placement *placement,
+					struct ttm_mem_reg *mem,
+					bool interruptible, bool no_wait)
+{
+	struct ttm_bo_device *bdev = bo->bdev;
+	struct ttm_bo_global *glob = bdev->glob;
+	struct ttm_mem_type_manager *man = &bdev->man[mem_type];
+	struct drm_mm_node *node;
+	int ret;
 
-	spin_unlock(&glob->lru_lock);
+	do {
+		ret = ttm_bo_man_get_node(bo, man, placement, mem, &node);
+		if (unlikely(ret != 0))
+			return ret;
+		if (node)
+			break;
+		spin_lock(&glob->lru_lock);
+		if (list_empty(&man->lru)) {
+			spin_unlock(&glob->lru_lock);
+			break;
+		}
+		spin_unlock(&glob->lru_lock);
+		ret = ttm_mem_evict_first(bdev, mem_type, interruptible,
+						no_wait);
+		if (unlikely(ret != 0))
+			return ret;
+	} while (1);
+	if (node == NULL)
+		return -ENOMEM;
 	mem->mm_node = node;
 	mem->mem_type = mem_type;
 	return 0;
@@ -724,7 +821,6 @@ static uint32_t ttm_bo_select_caching(struct ttm_mem_type_manager *man,
 	return result;
 }
 
-
 static bool ttm_bo_mt_compatible(struct ttm_mem_type_manager *man,
 				 bool disallow_fixed,
 				 uint32_t mem_type,
@@ -757,66 +853,55 @@ static bool ttm_bo_mt_compatible(struct ttm_mem_type_manager *man,
  * space.
  */
 int ttm_bo_mem_space(struct ttm_buffer_object *bo,
-		     uint32_t proposed_placement,
-		     struct ttm_mem_reg *mem,
-		     bool interruptible, bool no_wait)
+			struct ttm_placement *placement,
+			struct ttm_mem_reg *mem,
+			bool interruptible, bool no_wait)
 {
 	struct ttm_bo_device *bdev = bo->bdev;
-	struct ttm_bo_global *glob = bo->glob;
 	struct ttm_mem_type_manager *man;
-
-	uint32_t num_prios = bdev->driver->num_mem_type_prio;
-	const uint32_t *prios = bdev->driver->mem_type_prio;
-	uint32_t i;
 	uint32_t mem_type = TTM_PL_SYSTEM;
 	uint32_t cur_flags = 0;
 	bool type_found = false;
 	bool type_ok = false;
-	bool has_eagain = false;
+	bool has_erestartsys = false;
 	struct drm_mm_node *node = NULL;
-	int ret;
+	int i, ret;
 
 	mem->mm_node = NULL;
-	for (i = 0; i < num_prios; ++i) {
-		mem_type = prios[i];
+	for (i = 0; i < placement->num_placement; ++i) {
+		ret = ttm_mem_type_from_flags(placement->placement[i],
+						&mem_type);
+		if (ret)
+			return ret;
 		man = &bdev->man[mem_type];
 
 		type_ok = ttm_bo_mt_compatible(man,
-					       bo->type == ttm_bo_type_user,
-					       mem_type, proposed_placement,
-					       &cur_flags);
+						bo->type == ttm_bo_type_user,
+						mem_type,
+						placement->placement[i],
+						&cur_flags);
 
 		if (!type_ok)
 			continue;
 
 		cur_flags = ttm_bo_select_caching(man, bo->mem.placement,
 						  cur_flags);
+		/*
+		 * Use the access and other non-mapping-related flag bits from
+		 * the memory placement flags to the current flags
+		 */
+		ttm_flag_masked(&cur_flags, placement->placement[i],
+				~TTM_PL_MASK_MEMTYPE);
 
 		if (mem_type == TTM_PL_SYSTEM)
 			break;
 
 		if (man->has_type && man->use_type) {
 			type_found = true;
-			do {
-				ret = drm_mm_pre_get(&man->manager);
-				if (unlikely(ret))
-					return ret;
-
-				spin_lock(&glob->lru_lock);
-				node = drm_mm_search_free(&man->manager,
-							  mem->num_pages,
-							  mem->page_alignment,
-							  1);
-				if (unlikely(!node)) {
-					spin_unlock(&glob->lru_lock);
-					break;
-				}
-				node = drm_mm_get_block_atomic(node,
-							       mem->num_pages,
-							       mem->
-							       page_alignment);
-				spin_unlock(&glob->lru_lock);
-			} while (!node);
+			ret = ttm_bo_man_get_node(bo, man, placement, mem,
+							&node);
+			if (unlikely(ret))
+				return ret;
 		}
 		if (node)
 			break;
@@ -826,67 +911,74 @@ int ttm_bo_mem_space(struct ttm_buffer_object *bo,
 		mem->mm_node = node;
 		mem->mem_type = mem_type;
 		mem->placement = cur_flags;
+		if (node)
+			node->private = bo;
 		return 0;
 	}
 
 	if (!type_found)
 		return -EINVAL;
 
-	num_prios = bdev->driver->num_mem_busy_prio;
-	prios = bdev->driver->mem_busy_prio;
-
-	for (i = 0; i < num_prios; ++i) {
-		mem_type = prios[i];
+	for (i = 0; i < placement->num_busy_placement; ++i) {
+		ret = ttm_mem_type_from_flags(placement->busy_placement[i],
+						&mem_type);
+		if (ret)
+			return ret;
 		man = &bdev->man[mem_type];
-
 		if (!man->has_type)
 			continue;
-
 		if (!ttm_bo_mt_compatible(man,
-					  bo->type == ttm_bo_type_user,
-					  mem_type,
-					  proposed_placement, &cur_flags))
+						bo->type == ttm_bo_type_user,
+						mem_type,
+						placement->busy_placement[i],
+						&cur_flags))
 			continue;
 
 		cur_flags = ttm_bo_select_caching(man, bo->mem.placement,
 						  cur_flags);
+		/*
+		 * Use the access and other non-mapping-related flag bits from
+		 * the memory placement flags to the current flags
+		 */
+		ttm_flag_masked(&cur_flags, placement->busy_placement[i],
+				~TTM_PL_MASK_MEMTYPE);
 
-		ret = ttm_bo_mem_force_space(bdev, mem, mem_type,
-					     interruptible, no_wait);
 
-		if (ret == 0 && mem->mm_node) {
+		if (mem_type == TTM_PL_SYSTEM) {
+			mem->mem_type = mem_type;
 			mem->placement = cur_flags;
+			mem->mm_node = NULL;
 			return 0;
 		}
 
-		if (ret == -ERESTART)
-			has_eagain = true;
+		ret = ttm_bo_mem_force_space(bo, mem_type, placement, mem,
+						interruptible, no_wait);
+		if (ret == 0 && mem->mm_node) {
+			mem->placement = cur_flags;
+			mem->mm_node->private = bo;
+			return 0;
+		}
+		if (ret == -ERESTARTSYS)
+			has_erestartsys = true;
 	}
-
-	ret = (has_eagain) ? -ERESTART : -ENOMEM;
+	ret = (has_erestartsys) ? -ERESTARTSYS : -ENOMEM;
 	return ret;
 }
 EXPORT_SYMBOL(ttm_bo_mem_space);
 
 int ttm_bo_wait_cpu(struct ttm_buffer_object *bo, bool no_wait)
 {
-	int ret = 0;
-
 	if ((atomic_read(&bo->cpu_writers) > 0) && no_wait)
 		return -EBUSY;
 
-	ret = wait_event_interruptible(bo->event_queue,
-				       atomic_read(&bo->cpu_writers) == 0);
-
-	if (ret == -ERESTARTSYS)
-		ret = -ERESTART;
-
-	return ret;
+	return wait_event_interruptible(bo->event_queue,
+					atomic_read(&bo->cpu_writers) == 0);
 }
+EXPORT_SYMBOL(ttm_bo_wait_cpu);
 
 int ttm_bo_move_buffer(struct ttm_buffer_object *bo,
-		       uint32_t proposed_placement,
-		       bool interruptible, bool no_wait)
+			struct ttm_placement *placement,
+			bool interruptible, bool no_wait)
 {
 	struct ttm_bo_global *glob = bo->glob;
 	int ret = 0;
@@ -899,147 +991,138 @@ int ttm_bo_move_buffer(struct ttm_buffer_object *bo,
 	 * Have the driver move function wait for idle when necessary,
 	 * instead of doing it here.
 	 */
-
 	spin_lock(&bo->lock);
 	ret = ttm_bo_wait(bo, false, interruptible, no_wait);
 	spin_unlock(&bo->lock);
-
 	if (ret)
 		return ret;
-
 	mem.num_pages = bo->num_pages;
 	mem.size = mem.num_pages << PAGE_SHIFT;
 	mem.page_alignment = bo->mem.page_alignment;
-
 	/*
 	 * Determine where to move the buffer.
 	 */
-
-	ret = ttm_bo_mem_space(bo, proposed_placement, &mem,
-			       interruptible, no_wait);
+	ret = ttm_bo_mem_space(bo, placement, &mem, interruptible, no_wait);
 	if (ret)
 		goto out_unlock;
-
 	ret = ttm_bo_handle_move_mem(bo, &mem, false, interruptible, no_wait);
-
 out_unlock:
 	if (ret && mem.mm_node) {
 		spin_lock(&glob->lru_lock);
+		mem.mm_node->private = NULL;
 		drm_mm_put_block(mem.mm_node);
 		spin_unlock(&glob->lru_lock);
 	}
 	return ret;
 }
 
-static int ttm_bo_mem_compat(uint32_t proposed_placement,
+static int ttm_bo_mem_compat(struct ttm_placement *placement,
 			     struct ttm_mem_reg *mem)
 {
-	if ((proposed_placement & mem->placement & TTM_PL_MASK_MEM) == 0)
-		return 0;
-	if ((proposed_placement & mem->placement & TTM_PL_MASK_CACHING) == 0)
-		return 0;
-
-	return 1;
+	int i;
+	struct drm_mm_node *node = mem->mm_node;
+
+	if (node && placement->lpfn != 0 &&
+	    (node->start < placement->fpfn ||
+	     node->start + node->size > placement->lpfn))
+		return -1;
+
+	for (i = 0; i < placement->num_placement; i++) {
+		if ((placement->placement[i] & mem->placement &
+			TTM_PL_MASK_CACHING) &&
+			(placement->placement[i] & mem->placement &
+			TTM_PL_MASK_MEM))
+			return i;
+	}
+	return -1;
 }
 
-int ttm_buffer_object_validate(struct ttm_buffer_object *bo,
-			       uint32_t proposed_placement,
-			       bool interruptible, bool no_wait)
+int ttm_bo_validate(struct ttm_buffer_object *bo,
+			struct ttm_placement *placement,
+			bool interruptible, bool no_wait)
 {
 	int ret;
 
 	BUG_ON(!atomic_read(&bo->reserved));
-	bo->proposed_placement = proposed_placement;
-
-	TTM_DEBUG("Proposed placement 0x%08lx, Old flags 0x%08lx\n",
-		  (unsigned long)proposed_placement,
-		  (unsigned long)bo->mem.placement);
-
+	/* Check that range is valid */
+	if (placement->lpfn || placement->fpfn)
+		if (placement->fpfn > placement->lpfn ||
+			(placement->lpfn - placement->fpfn) < bo->num_pages)
+			return -EINVAL;
 	/*
 	 * Check whether we need to move buffer.
 	 */
-
-	if (!ttm_bo_mem_compat(bo->proposed_placement, &bo->mem)) {
-		ret = ttm_bo_move_buffer(bo, bo->proposed_placement,
-					 interruptible, no_wait);
-		if (ret) {
-			if (ret != -ERESTART)
-				printk(KERN_ERR TTM_PFX
-				       "Failed moving buffer. "
-				       "Proposed placement 0x%08x\n",
-				       bo->proposed_placement);
-			if (ret == -ENOMEM)
-				printk(KERN_ERR TTM_PFX
-				       "Out of aperture space or "
-				       "DRM memory quota.\n");
+	ret = ttm_bo_mem_compat(placement, &bo->mem);
+	if (ret < 0) {
+		ret = ttm_bo_move_buffer(bo, placement, interruptible, no_wait);
+		if (ret)
 			return ret;
-		}
+	} else {
+		/*
+		 * Use the access and other non-mapping-related flag bits from
+		 * the compatible memory placement flags to the active flags
+		 */
+		ttm_flag_masked(&bo->mem.placement, placement->placement[ret],
+				~TTM_PL_MASK_MEMTYPE);
 	}
-
 	/*
 	 * We might need to add a TTM.
 	 */
-
 	if (bo->mem.mem_type == TTM_PL_SYSTEM && bo->ttm == NULL) {
 		ret = ttm_bo_add_ttm(bo, true);
 		if (ret)
 			return ret;
 	}
-	/*
-	 * Validation has succeeded, move the access and other
-	 * non-mapping-related flag bits from the proposed flags to
-	 * the active flags
-	 */
-
-	ttm_flag_masked(&bo->mem.placement, bo->proposed_placement,
-			~TTM_PL_MASK_MEMTYPE);
-
 	return 0;
 }
-EXPORT_SYMBOL(ttm_buffer_object_validate);
+EXPORT_SYMBOL(ttm_bo_validate);
 
-int
-ttm_bo_check_placement(struct ttm_buffer_object *bo,
-		       uint32_t set_flags, uint32_t clr_flags)
+int ttm_bo_check_placement(struct ttm_buffer_object *bo,
+				struct ttm_placement *placement)
 {
-	uint32_t new_mask = set_flags | clr_flags;
-
-	if ((bo->type == ttm_bo_type_user) &&
-	    (clr_flags & TTM_PL_FLAG_CACHED)) {
-		printk(KERN_ERR TTM_PFX
-		       "User buffers require cache-coherent memory.\n");
-		return -EINVAL;
-	}
-
-	if (!capable(CAP_SYS_ADMIN)) {
-		if (new_mask & TTM_PL_FLAG_NO_EVICT) {
-			printk(KERN_ERR TTM_PFX "Need to be root to modify"
-			       " NO_EVICT status.\n");
+	int i;
+
+	if (placement->fpfn || placement->lpfn) {
+		if (bo->mem.num_pages > (placement->lpfn - placement->fpfn)) {
+			printk(KERN_ERR TTM_PFX "Page number range to small "
+				"Need %lu pages, range is [%u, %u]\n",
+				bo->mem.num_pages, placement->fpfn,
+				placement->lpfn);
 			return -EINVAL;
 		}
-
-		if ((clr_flags & bo->mem.placement & TTM_PL_MASK_MEMTYPE) &&
-		    (bo->mem.placement & TTM_PL_FLAG_NO_EVICT)) {
-			printk(KERN_ERR TTM_PFX
-			       "Incompatible memory specification"
-			       " for NO_EVICT buffer.\n");
-			return -EINVAL;
+	}
+	for (i = 0; i < placement->num_placement; i++) {
+		if (!capable(CAP_SYS_ADMIN)) {
+			if (placement->placement[i] & TTM_PL_FLAG_NO_EVICT) {
+				printk(KERN_ERR TTM_PFX "Need to be root to "
+					"modify NO_EVICT status.\n");
+				return -EINVAL;
+			}
+		}
+	}
+	for (i = 0; i < placement->num_busy_placement; i++) {
+		if (!capable(CAP_SYS_ADMIN)) {
+			if (placement->busy_placement[i] & TTM_PL_FLAG_NO_EVICT) {
+				printk(KERN_ERR TTM_PFX "Need to be root to "
+					"modify NO_EVICT status.\n");
+				return -EINVAL;
+			}
 		}
 	}
 	return 0;
 }
 
-int ttm_buffer_object_init(struct ttm_bo_device *bdev,
-			   struct ttm_buffer_object *bo,
-			   unsigned long size,
-			   enum ttm_bo_type type,
-			   uint32_t flags,
-			   uint32_t page_alignment,
-			   unsigned long buffer_start,
-			   bool interruptible,
-			   struct file *persistant_swap_storage,
-			   size_t acc_size,
-			   void (*destroy) (struct ttm_buffer_object *))
+int ttm_bo_init(struct ttm_bo_device *bdev,
+		struct ttm_buffer_object *bo,
+		unsigned long size,
+		enum ttm_bo_type type,
+		struct ttm_placement *placement,
+		uint32_t page_alignment,
+		unsigned long buffer_start,
+		bool interruptible,
+		struct file *persistant_swap_storage,
+		size_t acc_size,
+		void (*destroy) (struct ttm_buffer_object *))
 {
 	int ret = 0;
 	unsigned long num_pages;
@@ -1065,6 +1148,7 @@ int ttm_buffer_object_init(struct ttm_bo_device *bdev,
 	bo->glob = bdev->glob;
 	bo->type = type;
 	bo->num_pages = num_pages;
+	bo->mem.size = num_pages << PAGE_SHIFT;
 	bo->mem.mem_type = TTM_PL_SYSTEM;
 	bo->mem.num_pages = bo->num_pages;
 	bo->mem.mm_node = NULL;
@@ -1077,29 +1161,21 @@ int ttm_buffer_object_init(struct ttm_bo_device *bdev,
 	bo->acc_size = acc_size;
 	atomic_inc(&bo->glob->bo_count);
 
-	ret = ttm_bo_check_placement(bo, flags, 0ULL);
+	ret = ttm_bo_check_placement(bo, placement);
 	if (unlikely(ret != 0))
 		goto out_err;
 
 	/*
-	 * If no caching attributes are set, accept any form of caching.
-	 */
-
-	if ((flags & TTM_PL_MASK_CACHING) == 0)
-		flags |= TTM_PL_MASK_CACHING;
-
-	/*
 	 * For ttm_bo_type_device buffers, allocate
 	 * address space from the device.
 	 */
-
 	if (bo->type == ttm_bo_type_device) {
 		ret = ttm_bo_setup_vm(bo);
 		if (ret)
 			goto out_err;
 	}
 
-	ret = ttm_buffer_object_validate(bo, flags, interruptible, false);
+	ret = ttm_bo_validate(bo, placement, interruptible, false);
 	if (ret)
 		goto out_err;
 
@@ -1112,7 +1188,7 @@ out_err:
 
 	return ret;
 }
-EXPORT_SYMBOL(ttm_buffer_object_init);
+EXPORT_SYMBOL(ttm_bo_init);
 
 static inline size_t ttm_bo_size(struct ttm_bo_global *glob,
 				 unsigned long num_pages)
@@ -1123,19 +1199,19 @@ static inline size_t ttm_bo_size(struct ttm_bo_global *glob,
 	return glob->ttm_bo_size + 2 * page_array_size;
 }
 
-int ttm_buffer_object_create(struct ttm_bo_device *bdev,
-			     unsigned long size,
-			     enum ttm_bo_type type,
-			     uint32_t flags,
-			     uint32_t page_alignment,
-			     unsigned long buffer_start,
-			     bool interruptible,
-			     struct file *persistant_swap_storage,
-			     struct ttm_buffer_object **p_bo)
+int ttm_bo_create(struct ttm_bo_device *bdev,
+			unsigned long size,
+			enum ttm_bo_type type,
+			struct ttm_placement *placement,
+			uint32_t page_alignment,
+			unsigned long buffer_start,
+			bool interruptible,
+			struct file *persistant_swap_storage,
+			struct ttm_buffer_object **p_bo)
 {
 	struct ttm_buffer_object *bo;
-	int ret;
 	struct ttm_mem_global *mem_glob = bdev->glob->mem_glob;
+	int ret;
 
 	size_t acc_size =
 	    ttm_bo_size(bdev->glob, (size + PAGE_SIZE - 1) >> PAGE_SHIFT);
@@ -1150,76 +1226,41 @@ int ttm_buffer_object_create(struct ttm_bo_device *bdev,
 		return -ENOMEM;
 	}
 
-	ret = ttm_buffer_object_init(bdev, bo, size, type, flags,
-				     page_alignment, buffer_start,
-				     interruptible,
-				     persistant_swap_storage, acc_size, NULL);
+	ret = ttm_bo_init(bdev, bo, size, type, placement, page_alignment,
+				buffer_start, interruptible,
+				persistant_swap_storage, acc_size, NULL);
 	if (likely(ret == 0))
 		*p_bo = bo;
 
 	return ret;
 }
 
-static int ttm_bo_leave_list(struct ttm_buffer_object *bo,
-			     uint32_t mem_type, bool allow_errors)
-{
-	int ret;
-
-	spin_lock(&bo->lock);
-	ret = ttm_bo_wait(bo, false, false, false);
-	spin_unlock(&bo->lock);
-
-	if (ret && allow_errors)
-		goto out;
-
-	if (bo->mem.mem_type == mem_type)
-		ret = ttm_bo_evict(bo, mem_type, false, false);
-
-	if (ret) {
-		if (allow_errors) {
-			goto out;
-		} else {
-			ret = 0;
-			printk(KERN_ERR TTM_PFX "Cleanup eviction failed\n");
-		}
-	}
-
-out:
-	return ret;
-}
-
 static int ttm_bo_force_list_clean(struct ttm_bo_device *bdev,
-				   struct list_head *head,
-				   unsigned mem_type, bool allow_errors)
+					unsigned mem_type, bool allow_errors)
 {
+	struct ttm_mem_type_manager *man = &bdev->man[mem_type];
 	struct ttm_bo_global *glob = bdev->glob;
-	struct ttm_buffer_object *entry;
 	int ret;
-	int put_count;
 
 	/*
 	 * Can't use standard list traversal since we're unlocking.
 	 */
 
 	spin_lock(&glob->lru_lock);
-
-	while (!list_empty(head)) {
-		entry = list_first_entry(head, struct ttm_buffer_object, lru);
-		kref_get(&entry->list_kref);
-		ret = ttm_bo_reserve_locked(entry, false, false, false, 0);
-		put_count = ttm_bo_del_from_lru(entry);
+	while (!list_empty(&man->lru)) {
 		spin_unlock(&glob->lru_lock);
-		while (put_count--)
-			kref_put(&entry->list_kref, ttm_bo_ref_bug);
-		BUG_ON(ret);
-		ret = ttm_bo_leave_list(entry, mem_type, allow_errors);
-		ttm_bo_unreserve(entry);
-		kref_put(&entry->list_kref, ttm_bo_release_list);
+		ret = ttm_mem_evict_first(bdev, mem_type, false, false);
+		if (ret) {
+			if (allow_errors) {
+				return ret;
+			} else {
+				printk(KERN_ERR TTM_PFX
+					"Cleanup eviction failed\n");
+			}
+		}
 		spin_lock(&glob->lru_lock);
 	}
-
 	spin_unlock(&glob->lru_lock);
-
 	return 0;
 }
 
@@ -1246,7 +1287,7 @@ int ttm_bo_clean_mm(struct ttm_bo_device *bdev, unsigned mem_type)
 
 	ret = 0;
 	if (mem_type > 0) {
-		ttm_bo_force_list_clean(bdev, &man->lru, mem_type, false);
+		ttm_bo_force_list_clean(bdev, mem_type, false);
 
 		spin_lock(&glob->lru_lock);
 		if (drm_mm_clean(&man->manager))
@@ -1279,12 +1320,12 @@ int ttm_bo_evict_mm(struct ttm_bo_device *bdev, unsigned mem_type)
 		return 0;
 	}
 
-	return ttm_bo_force_list_clean(bdev, &man->lru, mem_type, true);
+	return ttm_bo_force_list_clean(bdev, mem_type, true);
 }
 EXPORT_SYMBOL(ttm_bo_evict_mm);
 
 int ttm_bo_init_mm(struct ttm_bo_device *bdev, unsigned type,
-		   unsigned long p_offset, unsigned long p_size)
+			unsigned long p_size)
 {
 	int ret = -EINVAL;
 	struct ttm_mem_type_manager *man;
@@ -1314,7 +1355,7 @@ int ttm_bo_init_mm(struct ttm_bo_device *bdev, unsigned type,
 			       type);
 			return ret;
 		}
-		ret = drm_mm_init(&man->manager, p_offset, p_size);
+		ret = drm_mm_init(&man->manager, 0, p_size);
 		if (ret)
 			return ret;
 	}
@@ -1463,7 +1504,7 @@ int ttm_bo_device_init(struct ttm_bo_device *bdev,
 	 * Initialize the system memory buffer type.
 	 * Other types need to be driver / IOCTL initialized.
 	 */
-	ret = ttm_bo_init_mm(bdev, TTM_PL_SYSTEM, 0, 0);
+	ret = ttm_bo_init_mm(bdev, TTM_PL_SYSTEM, 0);
 	if (unlikely(ret != 0))
 		goto out_no_sys;
 
@@ -1693,7 +1734,7 @@ int ttm_bo_block_reservation(struct ttm_buffer_object *bo, bool interruptible,
 			ret = wait_event_interruptible
 			    (bo->event_queue, atomic_read(&bo->reserved) == 0);
 			if (unlikely(ret != 0))
-				return -ERESTART;
+				return ret;
 		} else {
 			wait_event(bo->event_queue,
 				   atomic_read(&bo->reserved) == 0);
@@ -1722,12 +1763,14 @@ int ttm_bo_synccpu_write_grab(struct ttm_buffer_object *bo, bool no_wait)
 	ttm_bo_unreserve(bo);
 	return ret;
 }
+EXPORT_SYMBOL(ttm_bo_synccpu_write_grab);
 
 void ttm_bo_synccpu_write_release(struct ttm_buffer_object *bo)
 {
 	if (atomic_dec_and_test(&bo->cpu_writers))
 		wake_up_all(&bo->event_queue);
 }
+EXPORT_SYMBOL(ttm_bo_synccpu_write_release);
 
 /**
  * A buffer object shrink method that tries to swap out the first
@@ -1808,6 +1851,9 @@ static int ttm_bo_swapout(struct ttm_mem_shrink *shrink)
 	 * anyone tries to access a ttm page.
 	 */
 
+	if (bo->bdev->driver->swap_notify)
+		bo->bdev->driver->swap_notify(bo);
+
 	ret = ttm_tt_swapout(bo->ttm, bo->persistant_swap_storage);
 out:
 
@@ -1828,3 +1874,4 @@ void ttm_bo_swapout_all(struct ttm_bo_device *bdev)
 	while (ttm_bo_swapout(&bdev->glob->shrink) == 0)
 		;
 }
+EXPORT_SYMBOL(ttm_bo_swapout_all);
diff --git a/drivers/gpu/drm/ttm/ttm_bo_util.c b/drivers/gpu/drm/ttm/ttm_bo_util.c
index c70927ecda21..5ca37a58a98c 100644
--- a/drivers/gpu/drm/ttm/ttm_bo_util.c
+++ b/drivers/gpu/drm/ttm/ttm_bo_util.c
@@ -53,7 +53,6 @@ int ttm_bo_move_ttm(struct ttm_buffer_object *bo,
 {
 	struct ttm_tt *ttm = bo->ttm;
 	struct ttm_mem_reg *old_mem = &bo->mem;
-	uint32_t save_flags = old_mem->placement;
 	int ret;
 
 	if (old_mem->mem_type != TTM_PL_SYSTEM) {
@@ -62,7 +61,6 @@ int ttm_bo_move_ttm(struct ttm_buffer_object *bo,
 		ttm_flag_masked(&old_mem->placement, TTM_PL_FLAG_SYSTEM,
 				TTM_PL_MASK_MEM);
 		old_mem->mem_type = TTM_PL_SYSTEM;
-		save_flags = old_mem->placement;
 	}
 
 	ret = ttm_tt_set_placement_caching(ttm, new_mem->placement);
@@ -77,7 +75,7 @@ int ttm_bo_move_ttm(struct ttm_buffer_object *bo,
 
 	*old_mem = *new_mem;
 	new_mem->mm_node = NULL;
-	ttm_flag_masked(&save_flags, new_mem->placement, TTM_PL_MASK_MEMTYPE);
+
 	return 0;
 }
 EXPORT_SYMBOL(ttm_bo_move_ttm);
@@ -219,7 +217,6 @@ int ttm_bo_move_memcpy(struct ttm_buffer_object *bo,
 	void *old_iomap;
 	void *new_iomap;
 	int ret;
-	uint32_t save_flags = old_mem->placement;
 	unsigned long i;
 	unsigned long page;
 	unsigned long add = 0;
@@ -270,7 +267,6 @@ out2:
 
 	*old_mem = *new_mem;
 	new_mem->mm_node = NULL;
-	ttm_flag_masked(&save_flags, new_mem->placement, TTM_PL_MASK_MEMTYPE);
 
 	if ((man->flags & TTM_MEMTYPE_FLAG_FIXED) && (ttm != NULL)) {
 		ttm_tt_unbind(ttm);
@@ -369,6 +365,7 @@ pgprot_t ttm_io_prot(uint32_t caching_flags, pgprot_t tmp)
 #endif
 	return tmp;
 }
+EXPORT_SYMBOL(ttm_io_prot);
 
 static int ttm_bo_ioremap(struct ttm_buffer_object *bo,
 			  unsigned long bus_base,
@@ -427,7 +424,7 @@ static int ttm_bo_kmap_ttm(struct ttm_buffer_object *bo,
 
 		/*
 		 * We need to use vmap to get the desired page protection
-		 * or to make the buffer object look contigous.
+		 * or to make the buffer object look contiguous.
 		 */
 		prot = (mem->placement & TTM_PL_FLAG_CACHED) ?
 			PAGE_KERNEL :
@@ -536,7 +533,6 @@ int ttm_bo_move_accel_cleanup(struct ttm_buffer_object *bo,
 	struct ttm_mem_type_manager *man = &bdev->man[new_mem->mem_type];
 	struct ttm_mem_reg *old_mem = &bo->mem;
 	int ret;
-	uint32_t save_flags = old_mem->placement;
 	struct ttm_buffer_object *ghost_obj;
 	void *tmp_obj = NULL;
 
@@ -597,7 +593,7 @@ int ttm_bo_move_accel_cleanup(struct ttm_buffer_object *bo,
 
 	*old_mem = *new_mem;
 	new_mem->mm_node = NULL;
-	ttm_flag_masked(&save_flags, new_mem->placement, TTM_PL_MASK_MEMTYPE);
+
 	return 0;
 }
 EXPORT_SYMBOL(ttm_bo_move_accel_cleanup);
diff --git a/drivers/gpu/drm/ttm/ttm_bo_vm.c b/drivers/gpu/drm/ttm/ttm_bo_vm.c
index 1c040d040338..668dbe8b8dd3 100644
--- a/drivers/gpu/drm/ttm/ttm_bo_vm.c
+++ b/drivers/gpu/drm/ttm/ttm_bo_vm.c
@@ -114,7 +114,7 @@ static int ttm_bo_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 		ret = ttm_bo_wait(bo, false, true, false);
 		spin_unlock(&bo->lock);
 		if (unlikely(ret != 0)) {
-			retval = (ret != -ERESTART) ?
+			retval = (ret != -ERESTARTSYS) ?
 			    VM_FAULT_SIGBUS : VM_FAULT_NOPAGE;
 			goto out_unlock;
 		}
@@ -320,7 +320,7 @@ ssize_t ttm_bo_io(struct ttm_bo_device *bdev, struct file *filp,
 		return -EFAULT;
 
 	driver = bo->bdev->driver;
-	if (unlikely(driver->verify_access)) {
+	if (unlikely(!driver->verify_access)) {
 		ret = -EPERM;
 		goto out_unref;
 	}
@@ -349,9 +349,6 @@ ssize_t ttm_bo_io(struct ttm_bo_device *bdev, struct file *filp,
 	switch (ret) {
 	case 0:
 		break;
-	case -ERESTART:
-		ret = -EINTR;
-		goto out_unref;
 	case -EBUSY:
 		ret = -EAGAIN;
 		goto out_unref;
@@ -421,8 +418,6 @@ ssize_t ttm_bo_fbdev_io(struct ttm_buffer_object *bo, const char __user *wbuf,
 	switch (ret) {
 	case 0:
 		break;
-	case -ERESTART:
-		return -EINTR;
 	case -EBUSY:
 		return -EAGAIN;
 	default:
diff --git a/drivers/gpu/drm/ttm/ttm_execbuf_util.c b/drivers/gpu/drm/ttm/ttm_execbuf_util.c
new file mode 100644
index 000000000000..c285c2902d15
--- /dev/null
+++ b/drivers/gpu/drm/ttm/ttm_execbuf_util.c
@@ -0,0 +1,117 @@
+/**************************************************************************
+ *
+ * Copyright (c) 2006-2009 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "ttm/ttm_execbuf_util.h"
+#include "ttm/ttm_bo_driver.h"
+#include "ttm/ttm_placement.h"
+#include <linux/wait.h>
+#include <linux/sched.h>
+#include <linux/module.h>
+
+void ttm_eu_backoff_reservation(struct list_head *list)
+{
+	struct ttm_validate_buffer *entry;
+
+	list_for_each_entry(entry, list, head) {
+		struct ttm_buffer_object *bo = entry->bo;
+		if (!entry->reserved)
+			continue;
+
+		entry->reserved = false;
+		ttm_bo_unreserve(bo);
+	}
+}
+EXPORT_SYMBOL(ttm_eu_backoff_reservation);
+
+/*
+ * Reserve buffers for validation.
+ *
+ * If a buffer in the list is marked for CPU access, we back off and
+ * wait for that buffer to become free for GPU access.
+ *
+ * If a buffer is reserved for another validation, the validator with
+ * the highest validation sequence backs off and waits for that buffer
+ * to become unreserved. This prevents deadlocks when validating multiple
+ * buffers in different orders.
+ */
+
+int ttm_eu_reserve_buffers(struct list_head *list, uint32_t val_seq)
+{
+	struct ttm_validate_buffer *entry;
+	int ret;
+
+retry:
+	list_for_each_entry(entry, list, head) {
+		struct ttm_buffer_object *bo = entry->bo;
+
+		entry->reserved = false;
+		ret = ttm_bo_reserve(bo, true, false, true, val_seq);
+		if (ret != 0) {
+			ttm_eu_backoff_reservation(list);
+			if (ret == -EAGAIN) {
+				ret = ttm_bo_wait_unreserved(bo, true);
+				if (unlikely(ret != 0))
+					return ret;
+				goto retry;
+			} else
+				return ret;
+		}
+
+		entry->reserved = true;
+		if (unlikely(atomic_read(&bo->cpu_writers) > 0)) {
+			ttm_eu_backoff_reservation(list);
+			ret = ttm_bo_wait_cpu(bo, false);
+			if (ret)
+				return ret;
+			goto retry;
+		}
+	}
+	return 0;
+}
+EXPORT_SYMBOL(ttm_eu_reserve_buffers);
+
+void ttm_eu_fence_buffer_objects(struct list_head *list, void *sync_obj)
+{
+	struct ttm_validate_buffer *entry;
+
+	list_for_each_entry(entry, list, head) {
+		struct ttm_buffer_object *bo = entry->bo;
+		struct ttm_bo_driver *driver = bo->bdev->driver;
+		void *old_sync_obj;
+
+		spin_lock(&bo->lock);
+		old_sync_obj = bo->sync_obj;
+		bo->sync_obj = driver->sync_obj_ref(sync_obj);
+		bo->sync_obj_arg = entry->new_sync_obj_arg;
+		spin_unlock(&bo->lock);
+		ttm_bo_unreserve(bo);
+		entry->reserved = false;
+		if (old_sync_obj)
+			driver->sync_obj_unref(&old_sync_obj);
+	}
+}
+EXPORT_SYMBOL(ttm_eu_fence_buffer_objects);
diff --git a/drivers/gpu/drm/ttm/ttm_global.c b/drivers/gpu/drm/ttm/ttm_global.c
index 541744d00d3e..b17007178a36 100644
--- a/drivers/gpu/drm/ttm/ttm_global.c
+++ b/drivers/gpu/drm/ttm/ttm_global.c
@@ -82,8 +82,8 @@ int ttm_global_item_ref(struct ttm_global_reference *ref)
 		if (unlikely(ret != 0))
 			goto out_err;
 
-		++item->refcount;
 	}
+	++item->refcount;
 	ref->object = item->object;
 	object = item->object;
 	mutex_unlock(&item->mutex);
diff --git a/drivers/gpu/drm/ttm/ttm_lock.c b/drivers/gpu/drm/ttm/ttm_lock.c
new file mode 100644
index 000000000000..3d172ef04ee1
--- /dev/null
+++ b/drivers/gpu/drm/ttm/ttm_lock.c
@@ -0,0 +1,313 @@
+/**************************************************************************
+ *
+ * Copyright (c) 2007-2009 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+/*
+ * Authors: Thomas Hellstrom <thellstrom-at-vmware-dot-com>
+ */
+
+#include "ttm/ttm_lock.h"
+#include "ttm/ttm_module.h"
+#include <asm/atomic.h>
+#include <linux/errno.h>
+#include <linux/wait.h>
+#include <linux/sched.h>
+#include <linux/module.h>
+
+#define TTM_WRITE_LOCK_PENDING    (1 << 0)
+#define TTM_VT_LOCK_PENDING       (1 << 1)
+#define TTM_SUSPEND_LOCK_PENDING  (1 << 2)
+#define TTM_VT_LOCK               (1 << 3)
+#define TTM_SUSPEND_LOCK          (1 << 4)
+
+void ttm_lock_init(struct ttm_lock *lock)
+{
+	spin_lock_init(&lock->lock);
+	init_waitqueue_head(&lock->queue);
+	lock->rw = 0;
+	lock->flags = 0;
+	lock->kill_takers = false;
+	lock->signal = SIGKILL;
+}
+EXPORT_SYMBOL(ttm_lock_init);
+
+void ttm_read_unlock(struct ttm_lock *lock)
+{
+	spin_lock(&lock->lock);
+	if (--lock->rw == 0)
+		wake_up_all(&lock->queue);
+	spin_unlock(&lock->lock);
+}
+EXPORT_SYMBOL(ttm_read_unlock);
+
+static bool __ttm_read_lock(struct ttm_lock *lock)
+{
+	bool locked = false;
+
+	spin_lock(&lock->lock);
+	if (unlikely(lock->kill_takers)) {
+		send_sig(lock->signal, current, 0);
+		spin_unlock(&lock->lock);
+		return false;
+	}
+	if (lock->rw >= 0 && lock->flags == 0) {
+		++lock->rw;
+		locked = true;
+	}
+	spin_unlock(&lock->lock);
+	return locked;
+}
+
+int ttm_read_lock(struct ttm_lock *lock, bool interruptible)
+{
+	int ret = 0;
+
+	if (interruptible)
+		ret = wait_event_interruptible(lock->queue,
+					       __ttm_read_lock(lock));
+	else
+		wait_event(lock->queue, __ttm_read_lock(lock));
+	return ret;
+}
+EXPORT_SYMBOL(ttm_read_lock);
+
+static bool __ttm_read_trylock(struct ttm_lock *lock, bool *locked)
+{
+	bool block = true;
+
+	*locked = false;
+
+	spin_lock(&lock->lock);
+	if (unlikely(lock->kill_takers)) {
+		send_sig(lock->signal, current, 0);
+		spin_unlock(&lock->lock);
+		return false;
+	}
+	if (lock->rw >= 0 && lock->flags == 0) {
+		++lock->rw;
+		block = false;
+		*locked = true;
+	} else if (lock->flags == 0) {
+		block = false;
+	}
+	spin_unlock(&lock->lock);
+
+	return !block;
+}
+
+int ttm_read_trylock(struct ttm_lock *lock, bool interruptible)
+{
+	int ret = 0;
+	bool locked;
+
+	if (interruptible)
+		ret = wait_event_interruptible
+			(lock->queue, __ttm_read_trylock(lock, &locked));
+	else
+		wait_event(lock->queue, __ttm_read_trylock(lock, &locked));
+
+	if (unlikely(ret != 0)) {
+		BUG_ON(locked);
+		return ret;
+	}
+
+	return (locked) ? 0 : -EBUSY;
+}
+
+void ttm_write_unlock(struct ttm_lock *lock)
+{
+	spin_lock(&lock->lock);
+	lock->rw = 0;
+	wake_up_all(&lock->queue);
+	spin_unlock(&lock->lock);
+}
+EXPORT_SYMBOL(ttm_write_unlock);
+
+static bool __ttm_write_lock(struct ttm_lock *lock)
+{
+	bool locked = false;
+
+	spin_lock(&lock->lock);
+	if (unlikely(lock->kill_takers)) {
+		send_sig(lock->signal, current, 0);
+		spin_unlock(&lock->lock);
+		return false;
+	}
+	if (lock->rw == 0 && ((lock->flags & ~TTM_WRITE_LOCK_PENDING) == 0)) {
+		lock->rw = -1;
+		lock->flags &= ~TTM_WRITE_LOCK_PENDING;
+		locked = true;
+	} else {
+		lock->flags |= TTM_WRITE_LOCK_PENDING;
+	}
+	spin_unlock(&lock->lock);
+	return locked;
+}
+
+int ttm_write_lock(struct ttm_lock *lock, bool interruptible)
+{
+	int ret = 0;
+
+	if (interruptible) {
+		ret = wait_event_interruptible(lock->queue,
+					       __ttm_write_lock(lock));
+		if (unlikely(ret != 0)) {
+			spin_lock(&lock->lock);
+			lock->flags &= ~TTM_WRITE_LOCK_PENDING;
+			wake_up_all(&lock->queue);
+			spin_unlock(&lock->lock);
+		}
+	} else
+		wait_event(lock->queue, __ttm_read_lock(lock));
+
+	return ret;
+}
+EXPORT_SYMBOL(ttm_write_lock);
+
+void ttm_write_lock_downgrade(struct ttm_lock *lock)
+{
+	spin_lock(&lock->lock);
+	lock->rw = 1;
+	wake_up_all(&lock->queue);
+	spin_unlock(&lock->lock);
+}
+
+static int __ttm_vt_unlock(struct ttm_lock *lock)
+{
+	int ret = 0;
+
+	spin_lock(&lock->lock);
+	if (unlikely(!(lock->flags & TTM_VT_LOCK)))
+		ret = -EINVAL;
+	lock->flags &= ~TTM_VT_LOCK;
+	wake_up_all(&lock->queue);
+	spin_unlock(&lock->lock);
+	printk(KERN_INFO TTM_PFX "vt unlock.\n");
+
+	return ret;
+}
+
+static void ttm_vt_lock_remove(struct ttm_base_object **p_base)
+{
+	struct ttm_base_object *base = *p_base;
+	struct ttm_lock *lock = container_of(base, struct ttm_lock, base);
+	int ret;
+
+	*p_base = NULL;
+	ret = __ttm_vt_unlock(lock);
+	BUG_ON(ret != 0);
+}
+
+static bool __ttm_vt_lock(struct ttm_lock *lock)
+{
+	bool locked = false;
+
+	spin_lock(&lock->lock);
+	if (lock->rw == 0) {
+		lock->flags &= ~TTM_VT_LOCK_PENDING;
+		lock->flags |= TTM_VT_LOCK;
+		locked = true;
+	} else {
+		lock->flags |= TTM_VT_LOCK_PENDING;
+	}
+	spin_unlock(&lock->lock);
+	return locked;
+}
+
+int ttm_vt_lock(struct ttm_lock *lock,
+		bool interruptible,
+		struct ttm_object_file *tfile)
+{
+	int ret = 0;
+
+	if (interruptible) {
+		ret = wait_event_interruptible(lock->queue,
+					       __ttm_vt_lock(lock));
+		if (unlikely(ret != 0)) {
+			spin_lock(&lock->lock);
+			lock->flags &= ~TTM_VT_LOCK_PENDING;
+			wake_up_all(&lock->queue);
+			spin_unlock(&lock->lock);
+			return ret;
+		}
+	} else
+		wait_event(lock->queue, __ttm_vt_lock(lock));
+
+	/*
+	 * Add a base-object, the destructor of which will
+	 * make sure the lock is released if the client dies
+	 * while holding it.
+	 */
+
+	ret = ttm_base_object_init(tfile, &lock->base, false,
+				   ttm_lock_type, &ttm_vt_lock_remove, NULL);
+	if (ret)
+		(void)__ttm_vt_unlock(lock);
+	else {
+		lock->vt_holder = tfile;
+		printk(KERN_INFO TTM_PFX "vt lock.\n");
+	}
+
+	return ret;
+}
+EXPORT_SYMBOL(ttm_vt_lock);
+
+int ttm_vt_unlock(struct ttm_lock *lock)
+{
+	return ttm_ref_object_base_unref(lock->vt_holder,
+					 lock->base.hash.key, TTM_REF_USAGE);
+}
+EXPORT_SYMBOL(ttm_vt_unlock);
+
+void ttm_suspend_unlock(struct ttm_lock *lock)
+{
+	spin_lock(&lock->lock);
+	lock->flags &= ~TTM_SUSPEND_LOCK;
+	wake_up_all(&lock->queue);
+	spin_unlock(&lock->lock);
+}
+EXPORT_SYMBOL(ttm_suspend_unlock);
+
+static bool __ttm_suspend_lock(struct ttm_lock *lock)
+{
+	bool locked = false;
+
+	spin_lock(&lock->lock);
+	if (lock->rw == 0) {
+		lock->flags &= ~TTM_SUSPEND_LOCK_PENDING;
+		lock->flags |= TTM_SUSPEND_LOCK;
+		locked = true;
+	} else {
+		lock->flags |= TTM_SUSPEND_LOCK_PENDING;
+	}
+	spin_unlock(&lock->lock);
+	return locked;
+}
+
+void ttm_suspend_lock(struct ttm_lock *lock)
+{
+	wait_event(lock->queue, __ttm_suspend_lock(lock));
+}
+EXPORT_SYMBOL(ttm_suspend_lock);
diff --git a/drivers/gpu/drm/ttm/ttm_memory.c b/drivers/gpu/drm/ttm/ttm_memory.c
index 072c281a6bb5..eb143e04d402 100644
--- a/drivers/gpu/drm/ttm/ttm_memory.c
+++ b/drivers/gpu/drm/ttm/ttm_memory.c
@@ -152,7 +152,7 @@ static struct attribute *ttm_mem_zone_attrs[] = {
 	NULL
 };
 
-static struct sysfs_ops ttm_mem_zone_ops = {
+static const struct sysfs_ops ttm_mem_zone_ops = {
 	.show = &ttm_mem_zone_show,
 	.store = &ttm_mem_zone_store
 };
@@ -274,16 +274,17 @@ static int ttm_mem_init_kernel_zone(struct ttm_mem_global *glob,
 static int ttm_mem_init_highmem_zone(struct ttm_mem_global *glob,
 				     const struct sysinfo *si)
 {
-	struct ttm_mem_zone *zone = kzalloc(sizeof(*zone), GFP_KERNEL);
+	struct ttm_mem_zone *zone;
 	uint64_t mem;
 	int ret;
 
-	if (unlikely(!zone))
-		return -ENOMEM;
-
 	if (si->totalhigh == 0)
 		return 0;
 
+	zone = kzalloc(sizeof(*zone), GFP_KERNEL);
+	if (unlikely(!zone))
+		return -ENOMEM;
+
 	mem = si->totalram;
 	mem *= si->mem_unit;
 
@@ -322,8 +323,10 @@ static int ttm_mem_init_dma32_zone(struct ttm_mem_global *glob,
 	 * No special dma32 zone needed.
 	 */
 
-	if (mem <= ((uint64_t) 1ULL << 32))
+	if (mem <= ((uint64_t) 1ULL << 32)) {
+		kfree(zone);
 		return 0;
+	}
 
 	/*
 	 * Limit max dma32 memory to 4GB for now
@@ -460,6 +463,7 @@ void ttm_mem_global_free(struct ttm_mem_global *glob,
 {
 	return ttm_mem_global_free_zone(glob, NULL, amount);
 }
+EXPORT_SYMBOL(ttm_mem_global_free);
 
 static int ttm_mem_global_reserve(struct ttm_mem_global *glob,
 				  struct ttm_mem_zone *single_zone,
@@ -533,6 +537,7 @@ int ttm_mem_global_alloc(struct ttm_mem_global *glob, uint64_t memory,
 	return ttm_mem_global_alloc_zone(glob, NULL, memory, no_wait,
 					 interruptible);
 }
+EXPORT_SYMBOL(ttm_mem_global_alloc);
 
 int ttm_mem_global_alloc_page(struct ttm_mem_global *glob,
 			      struct page *page,
@@ -588,3 +593,4 @@ size_t ttm_round_pot(size_t size)
 	}
 	return 0;
 }
+EXPORT_SYMBOL(ttm_round_pot);
diff --git a/drivers/gpu/drm/ttm/ttm_object.c b/drivers/gpu/drm/ttm/ttm_object.c
new file mode 100644
index 000000000000..75e9d6f86ba4
--- /dev/null
+++ b/drivers/gpu/drm/ttm/ttm_object.c
@@ -0,0 +1,452 @@
+/**************************************************************************
+ *
+ * Copyright (c) 2009 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+/*
+ * Authors: Thomas Hellstrom <thellstrom-at-vmware-dot-com>
+ */
+/** @file ttm_ref_object.c
+ *
+ * Base- and reference object implementation for the various
+ * ttm objects. Implements reference counting, minimal security checks
+ * and release on file close.
+ */
+
+/**
+ * struct ttm_object_file
+ *
+ * @tdev: Pointer to the ttm_object_device.
+ *
+ * @lock: Lock that protects the ref_list list and the
+ * ref_hash hash tables.
+ *
+ * @ref_list: List of ttm_ref_objects to be destroyed at
+ * file release.
+ *
+ * @ref_hash: Hash tables of ref objects, one per ttm_ref_type,
+ * for fast lookup of ref objects given a base object.
+ */
+
+#include "ttm/ttm_object.h"
+#include "ttm/ttm_module.h"
+#include <linux/list.h>
+#include <linux/spinlock.h>
+#include <linux/slab.h>
+#include <linux/module.h>
+#include <asm/atomic.h>
+
+struct ttm_object_file {
+	struct ttm_object_device *tdev;
+	rwlock_t lock;
+	struct list_head ref_list;
+	struct drm_open_hash ref_hash[TTM_REF_NUM];
+	struct kref refcount;
+};
+
+/**
+ * struct ttm_object_device
+ *
+ * @object_lock: lock that protects the object_hash hash table.
+ *
+ * @object_hash: hash table for fast lookup of object global names.
+ *
+ * @object_count: Per device object count.
+ *
+ * This is the per-device data structure needed for ttm object management.
+ */
+
+struct ttm_object_device {
+	rwlock_t object_lock;
+	struct drm_open_hash object_hash;
+	atomic_t object_count;
+	struct ttm_mem_global *mem_glob;
+};
+
+/**
+ * struct ttm_ref_object
+ *
+ * @hash: Hash entry for the per-file object reference hash.
+ *
+ * @head: List entry for the per-file list of ref-objects.
+ *
+ * @kref: Ref count.
+ *
+ * @obj: Base object this ref object is referencing.
+ *
+ * @ref_type: Type of ref object.
+ *
+ * This is similar to an idr object, but it also has a hash table entry
+ * that allows lookup with a pointer to the referenced object as a key. In
+ * that way, one can easily detect whether a base object is referenced by
+ * a particular ttm_object_file. It also carries a ref count to avoid creating
+ * multiple ref objects if a ttm_object_file references the same base
+ * object more than once.
+ */
+
+struct ttm_ref_object {
+	struct drm_hash_item hash;
+	struct list_head head;
+	struct kref kref;
+	enum ttm_ref_type ref_type;
+	struct ttm_base_object *obj;
+	struct ttm_object_file *tfile;
+};
+
+static inline struct ttm_object_file *
+ttm_object_file_ref(struct ttm_object_file *tfile)
+{
+	kref_get(&tfile->refcount);
+	return tfile;
+}
+
+static void ttm_object_file_destroy(struct kref *kref)
+{
+	struct ttm_object_file *tfile =
+		container_of(kref, struct ttm_object_file, refcount);
+
+	kfree(tfile);
+}
+
+
+static inline void ttm_object_file_unref(struct ttm_object_file **p_tfile)
+{
+	struct ttm_object_file *tfile = *p_tfile;
+
+	*p_tfile = NULL;
+	kref_put(&tfile->refcount, ttm_object_file_destroy);
+}
+
+
+int ttm_base_object_init(struct ttm_object_file *tfile,
+			 struct ttm_base_object *base,
+			 bool shareable,
+			 enum ttm_object_type object_type,
+			 void (*refcount_release) (struct ttm_base_object **),
+			 void (*ref_obj_release) (struct ttm_base_object *,
+						  enum ttm_ref_type ref_type))
+{
+	struct ttm_object_device *tdev = tfile->tdev;
+	int ret;
+
+	base->shareable = shareable;
+	base->tfile = ttm_object_file_ref(tfile);
+	base->refcount_release = refcount_release;
+	base->ref_obj_release = ref_obj_release;
+	base->object_type = object_type;
+	write_lock(&tdev->object_lock);
+	kref_init(&base->refcount);
+	ret = drm_ht_just_insert_please(&tdev->object_hash,
+					&base->hash,
+					(unsigned long)base, 31, 0, 0);
+	write_unlock(&tdev->object_lock);
+	if (unlikely(ret != 0))
+		goto out_err0;
+
+	ret = ttm_ref_object_add(tfile, base, TTM_REF_USAGE, NULL);
+	if (unlikely(ret != 0))
+		goto out_err1;
+
+	ttm_base_object_unref(&base);
+
+	return 0;
+out_err1:
+	(void)drm_ht_remove_item(&tdev->object_hash, &base->hash);
+out_err0:
+	return ret;
+}
+EXPORT_SYMBOL(ttm_base_object_init);
+
+static void ttm_release_base(struct kref *kref)
+{
+	struct ttm_base_object *base =
+	    container_of(kref, struct ttm_base_object, refcount);
+	struct ttm_object_device *tdev = base->tfile->tdev;
+
+	(void)drm_ht_remove_item(&tdev->object_hash, &base->hash);
+	write_unlock(&tdev->object_lock);
+	if (base->refcount_release) {
+		ttm_object_file_unref(&base->tfile);
+		base->refcount_release(&base);
+	}
+	write_lock(&tdev->object_lock);
+}
+
+void ttm_base_object_unref(struct ttm_base_object **p_base)
+{
+	struct ttm_base_object *base = *p_base;
+	struct ttm_object_device *tdev = base->tfile->tdev;
+
+	*p_base = NULL;
+
+	/*
+	 * Need to take the lock here to avoid racing with
+	 * users trying to look up the object.
+	 */
+
+	write_lock(&tdev->object_lock);
+	(void)kref_put(&base->refcount, &ttm_release_base);
+	write_unlock(&tdev->object_lock);
+}
+EXPORT_SYMBOL(ttm_base_object_unref);
+
+struct ttm_base_object *ttm_base_object_lookup(struct ttm_object_file *tfile,
+					       uint32_t key)
+{
+	struct ttm_object_device *tdev = tfile->tdev;
+	struct ttm_base_object *base;
+	struct drm_hash_item *hash;
+	int ret;
+
+	read_lock(&tdev->object_lock);
+	ret = drm_ht_find_item(&tdev->object_hash, key, &hash);
+
+	if (likely(ret == 0)) {
+		base = drm_hash_entry(hash, struct ttm_base_object, hash);
+		kref_get(&base->refcount);
+	}
+	read_unlock(&tdev->object_lock);
+
+	if (unlikely(ret != 0))
+		return NULL;
+
+	if (tfile != base->tfile && !base->shareable) {
+		printk(KERN_ERR TTM_PFX
+		       "Attempted access of non-shareable object.\n");
+		ttm_base_object_unref(&base);
+		return NULL;
+	}
+
+	return base;
+}
+EXPORT_SYMBOL(ttm_base_object_lookup);
+
+int ttm_ref_object_add(struct ttm_object_file *tfile,
+		       struct ttm_base_object *base,
+		       enum ttm_ref_type ref_type, bool *existed)
+{
+	struct drm_open_hash *ht = &tfile->ref_hash[ref_type];
+	struct ttm_ref_object *ref;
+	struct drm_hash_item *hash;
+	struct ttm_mem_global *mem_glob = tfile->tdev->mem_glob;
+	int ret = -EINVAL;
+
+	if (existed != NULL)
+		*existed = true;
+
+	while (ret == -EINVAL) {
+		read_lock(&tfile->lock);
+		ret = drm_ht_find_item(ht, base->hash.key, &hash);
+
+		if (ret == 0) {
+			ref = drm_hash_entry(hash, struct ttm_ref_object, hash);
+			kref_get(&ref->kref);
+			read_unlock(&tfile->lock);
+			break;
+		}
+
+		read_unlock(&tfile->lock);
+		ret = ttm_mem_global_alloc(mem_glob, sizeof(*ref),
+					   false, false);
+		if (unlikely(ret != 0))
+			return ret;
+		ref = kmalloc(sizeof(*ref), GFP_KERNEL);
+		if (unlikely(ref == NULL)) {
+			ttm_mem_global_free(mem_glob, sizeof(*ref));
+			return -ENOMEM;
+		}
+
+		ref->hash.key = base->hash.key;
+		ref->obj = base;
+		ref->tfile = tfile;
+		ref->ref_type = ref_type;
+		kref_init(&ref->kref);
+
+		write_lock(&tfile->lock);
+		ret = drm_ht_insert_item(ht, &ref->hash);
+
+		if (likely(ret == 0)) {
+			list_add_tail(&ref->head, &tfile->ref_list);
+			kref_get(&base->refcount);
+			write_unlock(&tfile->lock);
+			if (existed != NULL)
+				*existed = false;
+			break;
+		}
+
+		write_unlock(&tfile->lock);
+		BUG_ON(ret != -EINVAL);
+
+		ttm_mem_global_free(mem_glob, sizeof(*ref));
+		kfree(ref);
+	}
+
+	return ret;
+}
+EXPORT_SYMBOL(ttm_ref_object_add);
+
+static void ttm_ref_object_release(struct kref *kref)
+{
+	struct ttm_ref_object *ref =
+	    container_of(kref, struct ttm_ref_object, kref);
+	struct ttm_base_object *base = ref->obj;
+	struct ttm_object_file *tfile = ref->tfile;
+	struct drm_open_hash *ht;
+	struct ttm_mem_global *mem_glob = tfile->tdev->mem_glob;
+
+	ht = &tfile->ref_hash[ref->ref_type];
+	(void)drm_ht_remove_item(ht, &ref->hash);
+	list_del(&ref->head);
+	write_unlock(&tfile->lock);
+
+	if (ref->ref_type != TTM_REF_USAGE && base->ref_obj_release)
+		base->ref_obj_release(base, ref->ref_type);
+
+	ttm_base_object_unref(&ref->obj);
+	ttm_mem_global_free(mem_glob, sizeof(*ref));
+	kfree(ref);
+	write_lock(&tfile->lock);
+}
+
+int ttm_ref_object_base_unref(struct ttm_object_file *tfile,
+			      unsigned long key, enum ttm_ref_type ref_type)
+{
+	struct drm_open_hash *ht = &tfile->ref_hash[ref_type];
+	struct ttm_ref_object *ref;
+	struct drm_hash_item *hash;
+	int ret;
+
+	write_lock(&tfile->lock);
+	ret = drm_ht_find_item(ht, key, &hash);
+	if (unlikely(ret != 0)) {
+		write_unlock(&tfile->lock);
+		return -EINVAL;
+	}
+	ref = drm_hash_entry(hash, struct ttm_ref_object, hash);
+	kref_put(&ref->kref, ttm_ref_object_release);
+	write_unlock(&tfile->lock);
+	return 0;
+}
+EXPORT_SYMBOL(ttm_ref_object_base_unref);
+
+void ttm_object_file_release(struct ttm_object_file **p_tfile)
+{
+	struct ttm_ref_object *ref;
+	struct list_head *list;
+	unsigned int i;
+	struct ttm_object_file *tfile = *p_tfile;
+
+	*p_tfile = NULL;
+	write_lock(&tfile->lock);
+
+	/*
+	 * Since we release the lock within the loop, we have to
+	 * restart it from the beginning each time.
+	 */
+
+	while (!list_empty(&tfile->ref_list)) {
+		list = tfile->ref_list.next;
+		ref = list_entry(list, struct ttm_ref_object, head);
+		ttm_ref_object_release(&ref->kref);
+	}
+
+	for (i = 0; i < TTM_REF_NUM; ++i)
+		drm_ht_remove(&tfile->ref_hash[i]);
+
+	write_unlock(&tfile->lock);
+	ttm_object_file_unref(&tfile);
+}
+EXPORT_SYMBOL(ttm_object_file_release);
+
+struct ttm_object_file *ttm_object_file_init(struct ttm_object_device *tdev,
+					     unsigned int hash_order)
+{
+	struct ttm_object_file *tfile = kmalloc(sizeof(*tfile), GFP_KERNEL);
+	unsigned int i;
+	unsigned int j = 0;
+	int ret;
+
+	if (unlikely(tfile == NULL))
+		return NULL;
+
+	rwlock_init(&tfile->lock);
+	tfile->tdev = tdev;
+	kref_init(&tfile->refcount);
+	INIT_LIST_HEAD(&tfile->ref_list);
+
+	for (i = 0; i < TTM_REF_NUM; ++i) {
+		ret = drm_ht_create(&tfile->ref_hash[i], hash_order);
+		if (ret) {
+			j = i;
+			goto out_err;
+		}
+	}
+
+	return tfile;
+out_err:
+	for (i = 0; i < j; ++i)
+		drm_ht_remove(&tfile->ref_hash[i]);
+
+	kfree(tfile);
+
+	return NULL;
+}
+EXPORT_SYMBOL(ttm_object_file_init);
+
+struct ttm_object_device *ttm_object_device_init(struct ttm_mem_global
+						 *mem_glob,
+						 unsigned int hash_order)
+{
+	struct ttm_object_device *tdev = kmalloc(sizeof(*tdev), GFP_KERNEL);
+	int ret;
+
+	if (unlikely(tdev == NULL))
+		return NULL;
+
+	tdev->mem_glob = mem_glob;
+	rwlock_init(&tdev->object_lock);
+	atomic_set(&tdev->object_count, 0);
+	ret = drm_ht_create(&tdev->object_hash, hash_order);
+
+	if (likely(ret == 0))
+		return tdev;
+
+	kfree(tdev);
+	return NULL;
+}
+EXPORT_SYMBOL(ttm_object_device_init);
+
+void ttm_object_device_release(struct ttm_object_device **p_tdev)
+{
+	struct ttm_object_device *tdev = *p_tdev;
+
+	*p_tdev = NULL;
+
+	write_lock(&tdev->object_lock);
+	drm_ht_remove(&tdev->object_hash);
+	write_unlock(&tdev->object_lock);
+
+	kfree(tdev);
+}
+EXPORT_SYMBOL(ttm_object_device_release);
diff --git a/drivers/gpu/drm/ttm/ttm_tt.c b/drivers/gpu/drm/ttm/ttm_tt.c
index a55ee1a56c16..a759170763bb 100644
--- a/drivers/gpu/drm/ttm/ttm_tt.c
+++ b/drivers/gpu/drm/ttm/ttm_tt.c
@@ -192,26 +192,38 @@ int ttm_tt_populate(struct ttm_tt *ttm)
 	ttm->state = tt_unbound;
 	return 0;
 }
+EXPORT_SYMBOL(ttm_tt_populate);
 
 #ifdef CONFIG_X86
 static inline int ttm_tt_set_page_caching(struct page *p,
-					  enum ttm_caching_state c_state)
+					  enum ttm_caching_state c_old,
+					  enum ttm_caching_state c_new)
 {
+	int ret = 0;
+
 	if (PageHighMem(p))
 		return 0;
 
-	switch (c_state) {
-	case tt_cached:
-		return set_pages_wb(p, 1);
-	case tt_wc:
-	    return set_memory_wc((unsigned long) page_address(p), 1);
-	default:
-		return set_pages_uc(p, 1);
+	if (c_old != tt_cached) {
+		/* p isn't in the default caching state, set it to
+		 * writeback first to free its current memtype. */
+
+		ret = set_pages_wb(p, 1);
+		if (ret)
+			return ret;
 	}
+
+	if (c_new == tt_wc)
+		ret = set_memory_wc((unsigned long) page_address(p), 1);
+	else if (c_new == tt_uncached)
+		ret = set_pages_uc(p, 1);
+
+	return ret;
 }
 #else /* CONFIG_X86 */
 static inline int ttm_tt_set_page_caching(struct page *p,
-					  enum ttm_caching_state c_state)
+					  enum ttm_caching_state c_old,
+					  enum ttm_caching_state c_new)
 {
 	return 0;
 }
@@ -244,7 +256,9 @@ static int ttm_tt_set_caching(struct ttm_tt *ttm,
 	for (i = 0; i < ttm->num_pages; ++i) {
 		cur_page = ttm->pages[i];
 		if (likely(cur_page != NULL)) {
-			ret = ttm_tt_set_page_caching(cur_page, c_state);
+			ret = ttm_tt_set_page_caching(cur_page,
+						      ttm->caching_state,
+						      c_state);
 			if (unlikely(ret != 0))
 				goto out_err;
 		}
@@ -258,7 +272,7 @@ out_err:
 	for (j = 0; j < i; ++j) {
 		cur_page = ttm->pages[j];
 		if (likely(cur_page != NULL)) {
-			(void)ttm_tt_set_page_caching(cur_page,
+			(void)ttm_tt_set_page_caching(cur_page, c_state,
 						      ttm->caching_state);
 		}
 	}
@@ -279,6 +293,7 @@ int ttm_tt_set_placement_caching(struct ttm_tt *ttm, uint32_t placement)
 
 	return ttm_tt_set_caching(ttm, state);
 }
+EXPORT_SYMBOL(ttm_tt_set_placement_caching);
 
 static void ttm_tt_free_alloced_pages(struct ttm_tt *ttm)
 {
@@ -465,7 +480,7 @@ static int ttm_tt_swapin(struct ttm_tt *ttm)
 	void *from_virtual;
 	void *to_virtual;
 	int i;
-	int ret;
+	int ret = -ENOMEM;
 
 	if (ttm->page_flags & TTM_PAGE_FLAG_USER) {
 		ret = ttm_tt_set_user(ttm, ttm->tsk, ttm->start,
@@ -484,8 +499,10 @@ static int ttm_tt_swapin(struct ttm_tt *ttm)
 
 	for (i = 0; i < ttm->num_pages; ++i) {
 		from_page = read_mapping_page(swap_space, i, NULL);
-		if (IS_ERR(from_page))
+		if (IS_ERR(from_page)) {
+			ret = PTR_ERR(from_page);
 			goto out_err;
+		}
 		to_page = __ttm_tt_get_page(ttm, i);
 		if (unlikely(to_page == NULL))
 			goto out_err;
@@ -508,7 +525,7 @@ static int ttm_tt_swapin(struct ttm_tt *ttm)
 	return 0;
 out_err:
 	ttm_tt_free_alloced_pages(ttm);
-	return -ENOMEM;
+	return ret;
 }
 
 int ttm_tt_swapout(struct ttm_tt *ttm, struct file *persistant_swap_storage)
@@ -520,6 +537,7 @@ int ttm_tt_swapout(struct ttm_tt *ttm, struct file *persistant_swap_storage)
 	void *from_virtual;
 	void *to_virtual;
 	int i;
+	int ret = -ENOMEM;
 
 	BUG_ON(ttm->state != tt_unbound && ttm->state != tt_unpopulated);
 	BUG_ON(ttm->caching_state != tt_cached);
@@ -542,7 +560,7 @@ int ttm_tt_swapout(struct ttm_tt *ttm, struct file *persistant_swap_storage)
 						0);
 		if (unlikely(IS_ERR(swap_storage))) {
 			printk(KERN_ERR "Failed allocating swap storage.\n");
-			return -ENOMEM;
+			return PTR_ERR(swap_storage);
 		}
 	} else
 		swap_storage = persistant_swap_storage;
@@ -554,9 +572,10 @@ int ttm_tt_swapout(struct ttm_tt *ttm, struct file *persistant_swap_storage)
 		if (unlikely(from_page == NULL))
 			continue;
 		to_page = read_mapping_page(swap_space, i, NULL);
-		if (unlikely(to_page == NULL))
+		if (unlikely(IS_ERR(to_page))) {
+			ret = PTR_ERR(to_page);
 			goto out_err;
-
+		}
 		preempt_disable();
 		from_virtual = kmap_atomic(from_page, KM_USER0);
 		to_virtual = kmap_atomic(to_page, KM_USER1);
@@ -580,5 +599,5 @@ out_err:
 	if (!persistant_swap_storage)
 		fput(swap_storage);
 
-	return -ENOMEM;
+	return ret;
 }
diff --git a/drivers/gpu/drm/via/via_drv.c b/drivers/gpu/drm/via/via_drv.c
index bc2f51843005..7a1b210401e0 100644
--- a/drivers/gpu/drm/via/via_drv.c
+++ b/drivers/gpu/drm/via/via_drv.c
@@ -58,7 +58,7 @@ static struct drm_driver driver = {
 		.owner = THIS_MODULE,
 		.open = drm_open,
 		.release = drm_release,
-		.ioctl = drm_ioctl,
+		.unlocked_ioctl = drm_ioctl,
 		.mmap = drm_mmap,
 		.poll = drm_poll,
 		.fasync = drm_fasync,
diff --git a/drivers/gpu/drm/via/via_irq.c b/drivers/gpu/drm/via/via_irq.c
index 5935b8842e86..34079f251cd4 100644
--- a/drivers/gpu/drm/via/via_irq.c
+++ b/drivers/gpu/drm/via/via_irq.c
@@ -150,7 +150,7 @@ irqreturn_t via_driver_irq_handler(DRM_IRQ_ARGS)
 		cur_irq++;
 	}
 
-	/* Acknowlege interrupts */
+	/* Acknowledge interrupts */
 	VIA_WRITE(VIA_REG_INTERRUPT, status);
 
 
@@ -165,7 +165,7 @@ static __inline__ void viadrv_acknowledge_irqs(drm_via_private_t * dev_priv)
 	u32 status;
 
 	if (dev_priv) {
-		/* Acknowlege interrupts */
+		/* Acknowledge interrupts */
 		status = VIA_READ(VIA_REG_INTERRUPT);
 		VIA_WRITE(VIA_REG_INTERRUPT, status |
 			  dev_priv->irq_pending_mask);
diff --git a/drivers/gpu/drm/vmwgfx/Kconfig b/drivers/gpu/drm/vmwgfx/Kconfig
new file mode 100644
index 000000000000..f20b8bcbef39
--- /dev/null
+++ b/drivers/gpu/drm/vmwgfx/Kconfig
@@ -0,0 +1,13 @@
+config DRM_VMWGFX
+	tristate "DRM driver for VMware Virtual GPU"
+	depends on DRM && PCI
+	select FB_DEFERRED_IO
+	select FB_CFB_FILLRECT
+	select FB_CFB_COPYAREA
+	select FB_CFB_IMAGEBLIT
+	select DRM_TTM
+	help
+	  KMS enabled DRM driver for SVGA2 virtual hardware.
+
+	  If unsure say n. The compiled module will be
+	  called vmwgfx.ko
diff --git a/drivers/gpu/drm/vmwgfx/Makefile b/drivers/gpu/drm/vmwgfx/Makefile
new file mode 100644
index 000000000000..1a3cb6816d1c
--- /dev/null
+++ b/drivers/gpu/drm/vmwgfx/Makefile
@@ -0,0 +1,9 @@
+
+ccflags-y := -Iinclude/drm
+
+vmwgfx-y := vmwgfx_execbuf.o vmwgfx_gmr.o vmwgfx_kms.o vmwgfx_drv.o \
+	    vmwgfx_fb.o vmwgfx_ioctl.o vmwgfx_resource.o vmwgfx_buffer.o \
+	    vmwgfx_fifo.o vmwgfx_irq.o vmwgfx_ldu.o vmwgfx_ttm_glue.o \
+	    vmwgfx_overlay.o
+
+obj-$(CONFIG_DRM_VMWGFX) := vmwgfx.o
diff --git a/drivers/gpu/drm/vmwgfx/svga3d_reg.h b/drivers/gpu/drm/vmwgfx/svga3d_reg.h
new file mode 100644
index 000000000000..77cb45331000
--- /dev/null
+++ b/drivers/gpu/drm/vmwgfx/svga3d_reg.h
@@ -0,0 +1,1793 @@
+/**********************************************************
+ * Copyright 1998-2009 VMware, Inc.  All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use, copy,
+ * modify, merge, publish, distribute, sublicense, and/or sell copies
+ * of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ **********************************************************/
+
+/*
+ * svga3d_reg.h --
+ *
+ *       SVGA 3D hardware definitions
+ */
+
+#ifndef _SVGA3D_REG_H_
+#define _SVGA3D_REG_H_
+
+#include "svga_reg.h"
+
+
+/*
+ * 3D Hardware Version
+ *
+ *   The hardware version is stored in the SVGA_FIFO_3D_HWVERSION fifo
+ *   register.   Is set by the host and read by the guest.  This lets
+ *   us make new guest drivers which are backwards-compatible with old
+ *   SVGA hardware revisions.  It does not let us support old guest
+ *   drivers.  Good enough for now.
+ *
+ */
+
+#define SVGA3D_MAKE_HWVERSION(major, minor)      (((major) << 16) | ((minor) & 0xFF))
+#define SVGA3D_MAJOR_HWVERSION(version)          ((version) >> 16)
+#define SVGA3D_MINOR_HWVERSION(version)          ((version) & 0xFF)
+
+typedef enum {
+   SVGA3D_HWVERSION_WS5_RC1   = SVGA3D_MAKE_HWVERSION(0, 1),
+   SVGA3D_HWVERSION_WS5_RC2   = SVGA3D_MAKE_HWVERSION(0, 2),
+   SVGA3D_HWVERSION_WS51_RC1  = SVGA3D_MAKE_HWVERSION(0, 3),
+   SVGA3D_HWVERSION_WS6_B1    = SVGA3D_MAKE_HWVERSION(1, 1),
+   SVGA3D_HWVERSION_FUSION_11 = SVGA3D_MAKE_HWVERSION(1, 4),
+   SVGA3D_HWVERSION_WS65_B1   = SVGA3D_MAKE_HWVERSION(2, 0),
+   SVGA3D_HWVERSION_CURRENT   = SVGA3D_HWVERSION_WS65_B1,
+} SVGA3dHardwareVersion;
+
+/*
+ * Generic Types
+ */
+
+typedef uint32 SVGA3dBool; /* 32-bit Bool definition */
+#define SVGA3D_NUM_CLIPPLANES                   6
+#define SVGA3D_MAX_SIMULTANEOUS_RENDER_TARGETS  8
+
+
+/*
+ * Surface formats.
+ *
+ * If you modify this list, be sure to keep GLUtil.c in sync. It
+ * includes the internal format definition of each surface in
+ * GLUtil_ConvertSurfaceFormat, and it contains a table of
+ * human-readable names in GLUtil_GetFormatName.
+ */
+
+typedef enum SVGA3dSurfaceFormat {
+   SVGA3D_FORMAT_INVALID = 0,
+
+   SVGA3D_X8R8G8B8       = 1,
+   SVGA3D_A8R8G8B8       = 2,
+
+   SVGA3D_R5G6B5         = 3,
+   SVGA3D_X1R5G5B5       = 4,
+   SVGA3D_A1R5G5B5       = 5,
+   SVGA3D_A4R4G4B4       = 6,
+
+   SVGA3D_Z_D32          = 7,
+   SVGA3D_Z_D16          = 8,
+   SVGA3D_Z_D24S8        = 9,
+   SVGA3D_Z_D15S1        = 10,
+
+   SVGA3D_LUMINANCE8            = 11,
+   SVGA3D_LUMINANCE4_ALPHA4     = 12,
+   SVGA3D_LUMINANCE16           = 13,
+   SVGA3D_LUMINANCE8_ALPHA8     = 14,
+
+   SVGA3D_DXT1           = 15,
+   SVGA3D_DXT2           = 16,
+   SVGA3D_DXT3           = 17,
+   SVGA3D_DXT4           = 18,
+   SVGA3D_DXT5           = 19,
+
+   SVGA3D_BUMPU8V8       = 20,
+   SVGA3D_BUMPL6V5U5     = 21,
+   SVGA3D_BUMPX8L8V8U8   = 22,
+   SVGA3D_BUMPL8V8U8     = 23,
+
+   SVGA3D_ARGB_S10E5     = 24,   /* 16-bit floating-point ARGB */
+   SVGA3D_ARGB_S23E8     = 25,   /* 32-bit floating-point ARGB */
+
+   SVGA3D_A2R10G10B10    = 26,
+
+   /* signed formats */
+   SVGA3D_V8U8           = 27,
+   SVGA3D_Q8W8V8U8       = 28,
+   SVGA3D_CxV8U8         = 29,
+
+   /* mixed formats */
+   SVGA3D_X8L8V8U8       = 30,
+   SVGA3D_A2W10V10U10    = 31,
+
+   SVGA3D_ALPHA8         = 32,
+
+   /* Single- and dual-component floating point formats */
+   SVGA3D_R_S10E5        = 33,
+   SVGA3D_R_S23E8        = 34,
+   SVGA3D_RG_S10E5       = 35,
+   SVGA3D_RG_S23E8       = 36,
+
+   /*
+    * Any surface can be used as a buffer object, but SVGA3D_BUFFER is
+    * the most efficient format to use when creating new surfaces
+    * expressly for index or vertex data.
+    */
+   SVGA3D_BUFFER         = 37,
+
+   SVGA3D_Z_D24X8        = 38,
+
+   SVGA3D_V16U16         = 39,
+
+   SVGA3D_G16R16         = 40,
+   SVGA3D_A16B16G16R16   = 41,
+
+   /* Packed Video formats */
+   SVGA3D_UYVY           = 42,
+   SVGA3D_YUY2           = 43,
+
+   SVGA3D_FORMAT_MAX
+} SVGA3dSurfaceFormat;
+
+typedef uint32 SVGA3dColor; /* a, r, g, b */
+
+/*
+ * These match the D3DFORMAT_OP definitions used by Direct3D. We need
+ * them so that we can query the host for what the supported surface
+ * operations are (when we're using the D3D backend, in particular),
+ * and so we can send those operations to the guest.
+ */
+typedef enum {
+   SVGA3DFORMAT_OP_TEXTURE                               = 0x00000001,
+   SVGA3DFORMAT_OP_VOLUMETEXTURE                         = 0x00000002,
+   SVGA3DFORMAT_OP_CUBETEXTURE                           = 0x00000004,
+   SVGA3DFORMAT_OP_OFFSCREEN_RENDERTARGET                = 0x00000008,
+   SVGA3DFORMAT_OP_SAME_FORMAT_RENDERTARGET              = 0x00000010,
+   SVGA3DFORMAT_OP_ZSTENCIL                              = 0x00000040,
+   SVGA3DFORMAT_OP_ZSTENCIL_WITH_ARBITRARY_COLOR_DEPTH   = 0x00000080,
+
+/*
+ * This format can be used as a render target if the current display mode
+ * is the same depth if the alpha channel is ignored. e.g. if the device
+ * can render to A8R8G8B8 when the display mode is X8R8G8B8, then the
+ * format op list entry for A8R8G8B8 should have this cap.
+ */
+   SVGA3DFORMAT_OP_SAME_FORMAT_UP_TO_ALPHA_RENDERTARGET  = 0x00000100,
+
+/*
+ * This format contains DirectDraw support (including Flip).  This flag
+ * should not to be set on alpha formats.
+ */
+   SVGA3DFORMAT_OP_DISPLAYMODE                           = 0x00000400,
+
+/*
+ * The rasterizer can support some level of Direct3D support in this format
+ * and implies that the driver can create a Context in this mode (for some
+ * render target format).  When this flag is set, the SVGA3DFORMAT_OP_DISPLAYMODE
+ * flag must also be set.
+ */
+   SVGA3DFORMAT_OP_3DACCELERATION                        = 0x00000800,
+
+/*
+ * This is set for a private format when the driver has put the bpp in
+ * the structure.
+ */
+   SVGA3DFORMAT_OP_PIXELSIZE                             = 0x00001000,
+
+/*
+ * Indicates that this format can be converted to any RGB format for which
+ * SVGA3DFORMAT_OP_MEMBEROFGROUP_ARGB is specified
+ */
+   SVGA3DFORMAT_OP_CONVERT_TO_ARGB                       = 0x00002000,
+
+/*
+ * Indicates that this format can be used to create offscreen plain surfaces.
+ */
+   SVGA3DFORMAT_OP_OFFSCREENPLAIN                        = 0x00004000,
+
+/*
+ * Indicated that this format can be read as an SRGB texture (meaning that the
+ * sampler will linearize the looked up data)
+ */
+   SVGA3DFORMAT_OP_SRGBREAD                              = 0x00008000,
+
+/*
+ * Indicates that this format can be used in the bumpmap instructions
+ */
+   SVGA3DFORMAT_OP_BUMPMAP                               = 0x00010000,
+
+/*
+ * Indicates that this format can be sampled by the displacement map sampler
+ */
+   SVGA3DFORMAT_OP_DMAP                                  = 0x00020000,
+
+/*
+ * Indicates that this format cannot be used with texture filtering
+ */
+   SVGA3DFORMAT_OP_NOFILTER                              = 0x00040000,
+
+/*
+ * Indicates that format conversions are supported to this RGB format if
+ * SVGA3DFORMAT_OP_CONVERT_TO_ARGB is specified in the source format.
+ */
+   SVGA3DFORMAT_OP_MEMBEROFGROUP_ARGB                    = 0x00080000,
+
+/*
+ * Indicated that this format can be written as an SRGB target (meaning that the
+ * pixel pipe will DE-linearize data on output to format)
+ */
+   SVGA3DFORMAT_OP_SRGBWRITE                             = 0x00100000,
+
+/*
+ * Indicates that this format cannot be used with alpha blending
+ */
+   SVGA3DFORMAT_OP_NOALPHABLEND                          = 0x00200000,
+
+/*
+ * Indicates that the device can auto-generated sublevels for resources
+ * of this format
+ */
+   SVGA3DFORMAT_OP_AUTOGENMIPMAP                         = 0x00400000,
+
+/*
+ * Indicates that this format can be used by vertex texture sampler
+ */
+   SVGA3DFORMAT_OP_VERTEXTEXTURE                         = 0x00800000,
+
+/*
+ * Indicates that this format supports neither texture coordinate wrap
+ * modes, nor mipmapping
+ */
+   SVGA3DFORMAT_OP_NOTEXCOORDWRAPNORMIP                  = 0x01000000
+} SVGA3dFormatOp;
+
+/*
+ * This structure is a conversion of SVGA3DFORMAT_OP_*.
+ * Entries must be located at the same position.
+ */
+typedef union {
+   uint32 value;
+   struct {
+      uint32 texture : 1;
+      uint32 volumeTexture : 1;
+      uint32 cubeTexture : 1;
+      uint32 offscreenRenderTarget : 1;
+      uint32 sameFormatRenderTarget : 1;
+      uint32 unknown1 : 1;
+      uint32 zStencil : 1;
+      uint32 zStencilArbitraryDepth : 1;
+      uint32 sameFormatUpToAlpha : 1;
+      uint32 unknown2 : 1;
+      uint32 displayMode : 1;
+      uint32 acceleration3d : 1;
+      uint32 pixelSize : 1;
+      uint32 convertToARGB : 1;
+      uint32 offscreenPlain : 1;
+      uint32 sRGBRead : 1;
+      uint32 bumpMap : 1;
+      uint32 dmap : 1;
+      uint32 noFilter : 1;
+      uint32 memberOfGroupARGB : 1;
+      uint32 sRGBWrite : 1;
+      uint32 noAlphaBlend : 1;
+      uint32 autoGenMipMap : 1;
+      uint32 vertexTexture : 1;
+      uint32 noTexCoordWrapNorMip : 1;
+   };
+} SVGA3dSurfaceFormatCaps;
+
+/*
+ * SVGA_3D_CMD_SETRENDERSTATE Types.  All value types
+ * must fit in a uint32.
+ */
+
+typedef enum {
+   SVGA3D_RS_INVALID                   = 0,
+   SVGA3D_RS_ZENABLE                   = 1,     /* SVGA3dBool */
+   SVGA3D_RS_ZWRITEENABLE              = 2,     /* SVGA3dBool */
+   SVGA3D_RS_ALPHATESTENABLE           = 3,     /* SVGA3dBool */
+   SVGA3D_RS_DITHERENABLE              = 4,     /* SVGA3dBool */
+   SVGA3D_RS_BLENDENABLE               = 5,     /* SVGA3dBool */
+   SVGA3D_RS_FOGENABLE                 = 6,     /* SVGA3dBool */
+   SVGA3D_RS_SPECULARENABLE            = 7,     /* SVGA3dBool */
+   SVGA3D_RS_STENCILENABLE             = 8,     /* SVGA3dBool */
+   SVGA3D_RS_LIGHTINGENABLE            = 9,     /* SVGA3dBool */
+   SVGA3D_RS_NORMALIZENORMALS          = 10,    /* SVGA3dBool */
+   SVGA3D_RS_POINTSPRITEENABLE         = 11,    /* SVGA3dBool */
+   SVGA3D_RS_POINTSCALEENABLE          = 12,    /* SVGA3dBool */
+   SVGA3D_RS_STENCILREF                = 13,    /* uint32 */
+   SVGA3D_RS_STENCILMASK               = 14,    /* uint32 */
+   SVGA3D_RS_STENCILWRITEMASK          = 15,    /* uint32 */
+   SVGA3D_RS_FOGSTART                  = 16,    /* float */
+   SVGA3D_RS_FOGEND                    = 17,    /* float */
+   SVGA3D_RS_FOGDENSITY                = 18,    /* float */
+   SVGA3D_RS_POINTSIZE                 = 19,    /* float */
+   SVGA3D_RS_POINTSIZEMIN              = 20,    /* float */
+   SVGA3D_RS_POINTSIZEMAX              = 21,    /* float */
+   SVGA3D_RS_POINTSCALE_A              = 22,    /* float */
+   SVGA3D_RS_POINTSCALE_B              = 23,    /* float */
+   SVGA3D_RS_POINTSCALE_C              = 24,    /* float */
+   SVGA3D_RS_FOGCOLOR                  = 25,    /* SVGA3dColor */
+   SVGA3D_RS_AMBIENT                   = 26,    /* SVGA3dColor */
+   SVGA3D_RS_CLIPPLANEENABLE           = 27,    /* SVGA3dClipPlanes */
+   SVGA3D_RS_FOGMODE                   = 28,    /* SVGA3dFogMode */
+   SVGA3D_RS_FILLMODE                  = 29,    /* SVGA3dFillMode */
+   SVGA3D_RS_SHADEMODE                 = 30,    /* SVGA3dShadeMode */
+   SVGA3D_RS_LINEPATTERN               = 31,    /* SVGA3dLinePattern */
+   SVGA3D_RS_SRCBLEND                  = 32,    /* SVGA3dBlendOp */
+   SVGA3D_RS_DSTBLEND                  = 33,    /* SVGA3dBlendOp */
+   SVGA3D_RS_BLENDEQUATION             = 34,    /* SVGA3dBlendEquation */
+   SVGA3D_RS_CULLMODE                  = 35,    /* SVGA3dFace */
+   SVGA3D_RS_ZFUNC                     = 36,    /* SVGA3dCmpFunc */
+   SVGA3D_RS_ALPHAFUNC                 = 37,    /* SVGA3dCmpFunc */
+   SVGA3D_RS_STENCILFUNC               = 38,    /* SVGA3dCmpFunc */
+   SVGA3D_RS_STENCILFAIL               = 39,    /* SVGA3dStencilOp */
+   SVGA3D_RS_STENCILZFAIL              = 40,    /* SVGA3dStencilOp */
+   SVGA3D_RS_STENCILPASS               = 41,    /* SVGA3dStencilOp */
+   SVGA3D_RS_ALPHAREF                  = 42,    /* float (0.0 .. 1.0) */
+   SVGA3D_RS_FRONTWINDING              = 43,    /* SVGA3dFrontWinding */
+   SVGA3D_RS_COORDINATETYPE            = 44,    /* SVGA3dCoordinateType */
+   SVGA3D_RS_ZBIAS                     = 45,    /* float */
+   SVGA3D_RS_RANGEFOGENABLE            = 46,    /* SVGA3dBool */
+   SVGA3D_RS_COLORWRITEENABLE          = 47,    /* SVGA3dColorMask */
+   SVGA3D_RS_VERTEXMATERIALENABLE      = 48,    /* SVGA3dBool */
+   SVGA3D_RS_DIFFUSEMATERIALSOURCE     = 49,    /* SVGA3dVertexMaterial */
+   SVGA3D_RS_SPECULARMATERIALSOURCE    = 50,    /* SVGA3dVertexMaterial */
+   SVGA3D_RS_AMBIENTMATERIALSOURCE     = 51,    /* SVGA3dVertexMaterial */
+   SVGA3D_RS_EMISSIVEMATERIALSOURCE    = 52,    /* SVGA3dVertexMaterial */
+   SVGA3D_RS_TEXTUREFACTOR             = 53,    /* SVGA3dColor */
+   SVGA3D_RS_LOCALVIEWER               = 54,    /* SVGA3dBool */
+   SVGA3D_RS_SCISSORTESTENABLE         = 55,    /* SVGA3dBool */
+   SVGA3D_RS_BLENDCOLOR                = 56,    /* SVGA3dColor */
+   SVGA3D_RS_STENCILENABLE2SIDED       = 57,    /* SVGA3dBool */
+   SVGA3D_RS_CCWSTENCILFUNC            = 58,    /* SVGA3dCmpFunc */
+   SVGA3D_RS_CCWSTENCILFAIL            = 59,    /* SVGA3dStencilOp */
+   SVGA3D_RS_CCWSTENCILZFAIL           = 60,    /* SVGA3dStencilOp */
+   SVGA3D_RS_CCWSTENCILPASS            = 61,    /* SVGA3dStencilOp */
+   SVGA3D_RS_VERTEXBLEND               = 62,    /* SVGA3dVertexBlendFlags */
+   SVGA3D_RS_SLOPESCALEDEPTHBIAS       = 63,    /* float */
+   SVGA3D_RS_DEPTHBIAS                 = 64,    /* float */
+
+
+   /*
+    * Output Gamma Level
+    *
+    * Output gamma effects the gamma curve of colors that are output from the
+    * rendering pipeline.  A value of 1.0 specifies a linear color space. If the
+    * value is <= 0.0, gamma correction is ignored and linear color space is
+    * used.
+    */
+
+   SVGA3D_RS_OUTPUTGAMMA               = 65,    /* float */
+   SVGA3D_RS_ZVISIBLE                  = 66,    /* SVGA3dBool */
+   SVGA3D_RS_LASTPIXEL                 = 67,    /* SVGA3dBool */
+   SVGA3D_RS_CLIPPING                  = 68,    /* SVGA3dBool */
+   SVGA3D_RS_WRAP0                     = 69,    /* SVGA3dWrapFlags */
+   SVGA3D_RS_WRAP1                     = 70,    /* SVGA3dWrapFlags */
+   SVGA3D_RS_WRAP2                     = 71,    /* SVGA3dWrapFlags */
+   SVGA3D_RS_WRAP3                     = 72,    /* SVGA3dWrapFlags */
+   SVGA3D_RS_WRAP4                     = 73,    /* SVGA3dWrapFlags */
+   SVGA3D_RS_WRAP5                     = 74,    /* SVGA3dWrapFlags */
+   SVGA3D_RS_WRAP6                     = 75,    /* SVGA3dWrapFlags */
+   SVGA3D_RS_WRAP7                     = 76,    /* SVGA3dWrapFlags */
+   SVGA3D_RS_WRAP8                     = 77,    /* SVGA3dWrapFlags */
+   SVGA3D_RS_WRAP9                     = 78,    /* SVGA3dWrapFlags */
+   SVGA3D_RS_WRAP10                    = 79,    /* SVGA3dWrapFlags */
+   SVGA3D_RS_WRAP11                    = 80,    /* SVGA3dWrapFlags */
+   SVGA3D_RS_WRAP12                    = 81,    /* SVGA3dWrapFlags */
+   SVGA3D_RS_WRAP13                    = 82,    /* SVGA3dWrapFlags */
+   SVGA3D_RS_WRAP14                    = 83,    /* SVGA3dWrapFlags */
+   SVGA3D_RS_WRAP15                    = 84,    /* SVGA3dWrapFlags */
+   SVGA3D_RS_MULTISAMPLEANTIALIAS      = 85,    /* SVGA3dBool */
+   SVGA3D_RS_MULTISAMPLEMASK           = 86,    /* uint32 */
+   SVGA3D_RS_INDEXEDVERTEXBLENDENABLE  = 87,    /* SVGA3dBool */
+   SVGA3D_RS_TWEENFACTOR               = 88,    /* float */
+   SVGA3D_RS_ANTIALIASEDLINEENABLE     = 89,    /* SVGA3dBool */
+   SVGA3D_RS_COLORWRITEENABLE1         = 90,    /* SVGA3dColorMask */
+   SVGA3D_RS_COLORWRITEENABLE2         = 91,    /* SVGA3dColorMask */
+   SVGA3D_RS_COLORWRITEENABLE3         = 92,    /* SVGA3dColorMask */
+   SVGA3D_RS_SEPARATEALPHABLENDENABLE  = 93,    /* SVGA3dBool */
+   SVGA3D_RS_SRCBLENDALPHA             = 94,    /* SVGA3dBlendOp */
+   SVGA3D_RS_DSTBLENDALPHA             = 95,    /* SVGA3dBlendOp */
+   SVGA3D_RS_BLENDEQUATIONALPHA        = 96,    /* SVGA3dBlendEquation */
+   SVGA3D_RS_MAX
+} SVGA3dRenderStateName;
+
+typedef enum {
+   SVGA3D_VERTEXMATERIAL_NONE     = 0,    /* Use the value in the current material */
+   SVGA3D_VERTEXMATERIAL_DIFFUSE  = 1,    /* Use the value in the diffuse component */
+   SVGA3D_VERTEXMATERIAL_SPECULAR = 2,    /* Use the value in the specular component */
+} SVGA3dVertexMaterial;
+
+typedef enum {
+   SVGA3D_FILLMODE_INVALID = 0,
+   SVGA3D_FILLMODE_POINT   = 1,
+   SVGA3D_FILLMODE_LINE    = 2,
+   SVGA3D_FILLMODE_FILL    = 3,
+   SVGA3D_FILLMODE_MAX
+} SVGA3dFillModeType;
+
+
+typedef
+union {
+   struct {
+      uint16   mode;       /* SVGA3dFillModeType */
+      uint16   face;       /* SVGA3dFace */
+   };
+   uint32 uintValue;
+} SVGA3dFillMode;
+
+typedef enum {
+   SVGA3D_SHADEMODE_INVALID = 0,
+   SVGA3D_SHADEMODE_FLAT    = 1,
+   SVGA3D_SHADEMODE_SMOOTH  = 2,
+   SVGA3D_SHADEMODE_PHONG   = 3,     /* Not supported */
+   SVGA3D_SHADEMODE_MAX
+} SVGA3dShadeMode;
+
+typedef
+union {
+   struct {
+      uint16 repeat;
+      uint16 pattern;
+   };
+   uint32 uintValue;
+} SVGA3dLinePattern;
+
+typedef enum {
+   SVGA3D_BLENDOP_INVALID            = 0,
+   SVGA3D_BLENDOP_ZERO               = 1,
+   SVGA3D_BLENDOP_ONE                = 2,
+   SVGA3D_BLENDOP_SRCCOLOR           = 3,
+   SVGA3D_BLENDOP_INVSRCCOLOR        = 4,
+   SVGA3D_BLENDOP_SRCALPHA           = 5,
+   SVGA3D_BLENDOP_INVSRCALPHA        = 6,
+   SVGA3D_BLENDOP_DESTALPHA          = 7,
+   SVGA3D_BLENDOP_INVDESTALPHA       = 8,
+   SVGA3D_BLENDOP_DESTCOLOR          = 9,
+   SVGA3D_BLENDOP_INVDESTCOLOR       = 10,
+   SVGA3D_BLENDOP_SRCALPHASAT        = 11,
+   SVGA3D_BLENDOP_BLENDFACTOR        = 12,
+   SVGA3D_BLENDOP_INVBLENDFACTOR     = 13,
+   SVGA3D_BLENDOP_MAX
+} SVGA3dBlendOp;
+
+typedef enum {
+   SVGA3D_BLENDEQ_INVALID            = 0,
+   SVGA3D_BLENDEQ_ADD                = 1,
+   SVGA3D_BLENDEQ_SUBTRACT           = 2,
+   SVGA3D_BLENDEQ_REVSUBTRACT        = 3,
+   SVGA3D_BLENDEQ_MINIMUM            = 4,
+   SVGA3D_BLENDEQ_MAXIMUM            = 5,
+   SVGA3D_BLENDEQ_MAX
+} SVGA3dBlendEquation;
+
+typedef enum {
+   SVGA3D_FRONTWINDING_INVALID = 0,
+   SVGA3D_FRONTWINDING_CW      = 1,
+   SVGA3D_FRONTWINDING_CCW     = 2,
+   SVGA3D_FRONTWINDING_MAX
+} SVGA3dFrontWinding;
+
+typedef enum {
+   SVGA3D_FACE_INVALID  = 0,
+   SVGA3D_FACE_NONE     = 1,
+   SVGA3D_FACE_FRONT    = 2,
+   SVGA3D_FACE_BACK     = 3,
+   SVGA3D_FACE_FRONT_BACK = 4,
+   SVGA3D_FACE_MAX
+} SVGA3dFace;
+
+/*
+ * The order and the values should not be changed
+ */
+
+typedef enum {
+   SVGA3D_CMP_INVALID              = 0,
+   SVGA3D_CMP_NEVER                = 1,
+   SVGA3D_CMP_LESS                 = 2,
+   SVGA3D_CMP_EQUAL                = 3,
+   SVGA3D_CMP_LESSEQUAL            = 4,
+   SVGA3D_CMP_GREATER              = 5,
+   SVGA3D_CMP_NOTEQUAL             = 6,
+   SVGA3D_CMP_GREATEREQUAL         = 7,
+   SVGA3D_CMP_ALWAYS               = 8,
+   SVGA3D_CMP_MAX
+} SVGA3dCmpFunc;
+
+/*
+ * SVGA3D_FOGFUNC_* specifies the fog equation, or PER_VERTEX which allows
+ * the fog factor to be specified in the alpha component of the specular
+ * (a.k.a. secondary) vertex color.
+ */
+typedef enum {
+   SVGA3D_FOGFUNC_INVALID          = 0,
+   SVGA3D_FOGFUNC_EXP              = 1,
+   SVGA3D_FOGFUNC_EXP2             = 2,
+   SVGA3D_FOGFUNC_LINEAR           = 3,
+   SVGA3D_FOGFUNC_PER_VERTEX       = 4
+} SVGA3dFogFunction;
+
+/*
+ * SVGA3D_FOGTYPE_* specifies if fog factors are computed on a per-vertex
+ * or per-pixel basis.
+ */
+typedef enum {
+   SVGA3D_FOGTYPE_INVALID          = 0,
+   SVGA3D_FOGTYPE_VERTEX           = 1,
+   SVGA3D_FOGTYPE_PIXEL            = 2,
+   SVGA3D_FOGTYPE_MAX              = 3
+} SVGA3dFogType;
+
+/*
+ * SVGA3D_FOGBASE_* selects depth or range-based fog. Depth-based fog is
+ * computed using the eye Z value of each pixel (or vertex), whereas range-
+ * based fog is computed using the actual distance (range) to the eye.
+ */
+typedef enum {
+   SVGA3D_FOGBASE_INVALID          = 0,
+   SVGA3D_FOGBASE_DEPTHBASED       = 1,
+   SVGA3D_FOGBASE_RANGEBASED       = 2,
+   SVGA3D_FOGBASE_MAX              = 3
+} SVGA3dFogBase;
+
+typedef enum {
+   SVGA3D_STENCILOP_INVALID        = 0,
+   SVGA3D_STENCILOP_KEEP           = 1,
+   SVGA3D_STENCILOP_ZERO           = 2,
+   SVGA3D_STENCILOP_REPLACE        = 3,
+   SVGA3D_STENCILOP_INCRSAT        = 4,
+   SVGA3D_STENCILOP_DECRSAT        = 5,
+   SVGA3D_STENCILOP_INVERT         = 6,
+   SVGA3D_STENCILOP_INCR           = 7,
+   SVGA3D_STENCILOP_DECR           = 8,
+   SVGA3D_STENCILOP_MAX
+} SVGA3dStencilOp;
+
+typedef enum {
+   SVGA3D_CLIPPLANE_0              = (1 << 0),
+   SVGA3D_CLIPPLANE_1              = (1 << 1),
+   SVGA3D_CLIPPLANE_2              = (1 << 2),
+   SVGA3D_CLIPPLANE_3              = (1 << 3),
+   SVGA3D_CLIPPLANE_4              = (1 << 4),
+   SVGA3D_CLIPPLANE_5              = (1 << 5),
+} SVGA3dClipPlanes;
+
+typedef enum {
+   SVGA3D_CLEAR_COLOR              = 0x1,
+   SVGA3D_CLEAR_DEPTH              = 0x2,
+   SVGA3D_CLEAR_STENCIL            = 0x4
+} SVGA3dClearFlag;
+
+typedef enum {
+   SVGA3D_RT_DEPTH                 = 0,
+   SVGA3D_RT_STENCIL               = 1,
+   SVGA3D_RT_COLOR0                = 2,
+   SVGA3D_RT_COLOR1                = 3,
+   SVGA3D_RT_COLOR2                = 4,
+   SVGA3D_RT_COLOR3                = 5,
+   SVGA3D_RT_COLOR4                = 6,
+   SVGA3D_RT_COLOR5                = 7,
+   SVGA3D_RT_COLOR6                = 8,
+   SVGA3D_RT_COLOR7                = 9,
+   SVGA3D_RT_MAX,
+   SVGA3D_RT_INVALID               = ((uint32)-1),
+} SVGA3dRenderTargetType;
+
+#define SVGA3D_MAX_RT_COLOR (SVGA3D_RT_COLOR7 - SVGA3D_RT_COLOR0 + 1)
+
+typedef
+union {
+   struct {
+      uint32  red   : 1;
+      uint32  green : 1;
+      uint32  blue  : 1;
+      uint32  alpha : 1;
+   };
+   uint32 uintValue;
+} SVGA3dColorMask;
+
+typedef enum {
+   SVGA3D_VBLEND_DISABLE            = 0,
+   SVGA3D_VBLEND_1WEIGHT            = 1,
+   SVGA3D_VBLEND_2WEIGHT            = 2,
+   SVGA3D_VBLEND_3WEIGHT            = 3,
+} SVGA3dVertexBlendFlags;
+
+typedef enum {
+   SVGA3D_WRAPCOORD_0   = 1 << 0,
+   SVGA3D_WRAPCOORD_1   = 1 << 1,
+   SVGA3D_WRAPCOORD_2   = 1 << 2,
+   SVGA3D_WRAPCOORD_3   = 1 << 3,
+   SVGA3D_WRAPCOORD_ALL = 0xF,
+} SVGA3dWrapFlags;
+
+/*
+ * SVGA_3D_CMD_TEXTURESTATE Types.  All value types
+ * must fit in a uint32.
+ */
+
+typedef enum {
+   SVGA3D_TS_INVALID                    = 0,
+   SVGA3D_TS_BIND_TEXTURE               = 1,    /* SVGA3dSurfaceId */
+   SVGA3D_TS_COLOROP                    = 2,    /* SVGA3dTextureCombiner */
+   SVGA3D_TS_COLORARG1                  = 3,    /* SVGA3dTextureArgData */
+   SVGA3D_TS_COLORARG2                  = 4,    /* SVGA3dTextureArgData */
+   SVGA3D_TS_ALPHAOP                    = 5,    /* SVGA3dTextureCombiner */
+   SVGA3D_TS_ALPHAARG1                  = 6,    /* SVGA3dTextureArgData */
+   SVGA3D_TS_ALPHAARG2                  = 7,    /* SVGA3dTextureArgData */
+   SVGA3D_TS_ADDRESSU                   = 8,    /* SVGA3dTextureAddress */
+   SVGA3D_TS_ADDRESSV                   = 9,    /* SVGA3dTextureAddress */
+   SVGA3D_TS_MIPFILTER                  = 10,   /* SVGA3dTextureFilter */
+   SVGA3D_TS_MAGFILTER                  = 11,   /* SVGA3dTextureFilter */
+   SVGA3D_TS_MINFILTER                  = 12,   /* SVGA3dTextureFilter */
+   SVGA3D_TS_BORDERCOLOR                = 13,   /* SVGA3dColor */
+   SVGA3D_TS_TEXCOORDINDEX              = 14,   /* uint32 */
+   SVGA3D_TS_TEXTURETRANSFORMFLAGS      = 15,   /* SVGA3dTexTransformFlags */
+   SVGA3D_TS_TEXCOORDGEN                = 16,   /* SVGA3dTextureCoordGen */
+   SVGA3D_TS_BUMPENVMAT00               = 17,   /* float */
+   SVGA3D_TS_BUMPENVMAT01               = 18,   /* float */
+   SVGA3D_TS_BUMPENVMAT10               = 19,   /* float */
+   SVGA3D_TS_BUMPENVMAT11               = 20,   /* float */
+   SVGA3D_TS_TEXTURE_MIPMAP_LEVEL       = 21,   /* uint32 */
+   SVGA3D_TS_TEXTURE_LOD_BIAS           = 22,   /* float */
+   SVGA3D_TS_TEXTURE_ANISOTROPIC_LEVEL  = 23,   /* uint32 */
+   SVGA3D_TS_ADDRESSW                   = 24,   /* SVGA3dTextureAddress */
+
+
+   /*
+    * Sampler Gamma Level
+    *
+    * Sampler gamma effects the color of samples taken from the sampler.  A
+    * value of 1.0 will produce linear samples.  If the value is <= 0.0 the
+    * gamma value is ignored and a linear space is used.
+    */
+
+   SVGA3D_TS_GAMMA                      = 25,   /* float */
+   SVGA3D_TS_BUMPENVLSCALE              = 26,   /* float */
+   SVGA3D_TS_BUMPENVLOFFSET             = 27,   /* float */
+   SVGA3D_TS_COLORARG0                  = 28,   /* SVGA3dTextureArgData */
+   SVGA3D_TS_ALPHAARG0                  = 29,   /* SVGA3dTextureArgData */
+   SVGA3D_TS_MAX
+} SVGA3dTextureStateName;
+
+typedef enum {
+   SVGA3D_TC_INVALID                   = 0,
+   SVGA3D_TC_DISABLE                   = 1,
+   SVGA3D_TC_SELECTARG1                = 2,
+   SVGA3D_TC_SELECTARG2                = 3,
+   SVGA3D_TC_MODULATE                  = 4,
+   SVGA3D_TC_ADD                       = 5,
+   SVGA3D_TC_ADDSIGNED                 = 6,
+   SVGA3D_TC_SUBTRACT                  = 7,
+   SVGA3D_TC_BLENDTEXTUREALPHA         = 8,
+   SVGA3D_TC_BLENDDIFFUSEALPHA         = 9,
+   SVGA3D_TC_BLENDCURRENTALPHA         = 10,
+   SVGA3D_TC_BLENDFACTORALPHA          = 11,
+   SVGA3D_TC_MODULATE2X                = 12,
+   SVGA3D_TC_MODULATE4X                = 13,
+   SVGA3D_TC_DSDT                      = 14,
+   SVGA3D_TC_DOTPRODUCT3               = 15,
+   SVGA3D_TC_BLENDTEXTUREALPHAPM       = 16,
+   SVGA3D_TC_ADDSIGNED2X               = 17,
+   SVGA3D_TC_ADDSMOOTH                 = 18,
+   SVGA3D_TC_PREMODULATE               = 19,
+   SVGA3D_TC_MODULATEALPHA_ADDCOLOR    = 20,
+   SVGA3D_TC_MODULATECOLOR_ADDALPHA    = 21,
+   SVGA3D_TC_MODULATEINVALPHA_ADDCOLOR = 22,
+   SVGA3D_TC_MODULATEINVCOLOR_ADDALPHA = 23,
+   SVGA3D_TC_BUMPENVMAPLUMINANCE       = 24,
+   SVGA3D_TC_MULTIPLYADD               = 25,
+   SVGA3D_TC_LERP                      = 26,
+   SVGA3D_TC_MAX
+} SVGA3dTextureCombiner;
+
+#define SVGA3D_TC_CAP_BIT(svga3d_tc_op) (svga3d_tc_op ? (1 << (svga3d_tc_op - 1)) : 0)
+
+typedef enum {
+   SVGA3D_TEX_ADDRESS_INVALID    = 0,
+   SVGA3D_TEX_ADDRESS_WRAP       = 1,
+   SVGA3D_TEX_ADDRESS_MIRROR     = 2,
+   SVGA3D_TEX_ADDRESS_CLAMP      = 3,
+   SVGA3D_TEX_ADDRESS_BORDER     = 4,
+   SVGA3D_TEX_ADDRESS_MIRRORONCE = 5,
+   SVGA3D_TEX_ADDRESS_EDGE       = 6,
+   SVGA3D_TEX_ADDRESS_MAX
+} SVGA3dTextureAddress;
+
+/*
+ * SVGA3D_TEX_FILTER_NONE as the minification filter means mipmapping is
+ * disabled, and the rasterizer should use the magnification filter instead.
+ */
+typedef enum {
+   SVGA3D_TEX_FILTER_NONE           = 0,
+   SVGA3D_TEX_FILTER_NEAREST        = 1,
+   SVGA3D_TEX_FILTER_LINEAR         = 2,
+   SVGA3D_TEX_FILTER_ANISOTROPIC    = 3,
+   SVGA3D_TEX_FILTER_FLATCUBIC      = 4, // Deprecated, not implemented
+   SVGA3D_TEX_FILTER_GAUSSIANCUBIC  = 5, // Deprecated, not implemented
+   SVGA3D_TEX_FILTER_PYRAMIDALQUAD  = 6, // Not currently implemented
+   SVGA3D_TEX_FILTER_GAUSSIANQUAD   = 7, // Not currently implemented
+   SVGA3D_TEX_FILTER_MAX
+} SVGA3dTextureFilter;
+
+typedef enum {
+   SVGA3D_TEX_TRANSFORM_OFF    = 0,
+   SVGA3D_TEX_TRANSFORM_S      = (1 << 0),
+   SVGA3D_TEX_TRANSFORM_T      = (1 << 1),
+   SVGA3D_TEX_TRANSFORM_R      = (1 << 2),
+   SVGA3D_TEX_TRANSFORM_Q      = (1 << 3),
+   SVGA3D_TEX_PROJECTED        = (1 << 15),
+} SVGA3dTexTransformFlags;
+
+typedef enum {
+   SVGA3D_TEXCOORD_GEN_OFF              = 0,
+   SVGA3D_TEXCOORD_GEN_EYE_POSITION     = 1,
+   SVGA3D_TEXCOORD_GEN_EYE_NORMAL       = 2,
+   SVGA3D_TEXCOORD_GEN_REFLECTIONVECTOR = 3,
+   SVGA3D_TEXCOORD_GEN_SPHERE           = 4,
+   SVGA3D_TEXCOORD_GEN_MAX
+} SVGA3dTextureCoordGen;
+
+/*
+ * Texture argument constants for texture combiner
+ */
+typedef enum {
+   SVGA3D_TA_INVALID    = 0,
+   SVGA3D_TA_CONSTANT   = 1,
+   SVGA3D_TA_PREVIOUS   = 2,
+   SVGA3D_TA_DIFFUSE    = 3,
+   SVGA3D_TA_TEXTURE    = 4,
+   SVGA3D_TA_SPECULAR   = 5,
+   SVGA3D_TA_MAX
+} SVGA3dTextureArgData;
+
+#define SVGA3D_TM_MASK_LEN 4
+
+/* Modifiers for texture argument constants defined above. */
+typedef enum {
+   SVGA3D_TM_NONE       = 0,
+   SVGA3D_TM_ALPHA      = (1 << SVGA3D_TM_MASK_LEN),
+   SVGA3D_TM_ONE_MINUS  = (2 << SVGA3D_TM_MASK_LEN),
+} SVGA3dTextureArgModifier;
+
+#define SVGA3D_INVALID_ID         ((uint32)-1)
+#define SVGA3D_MAX_CLIP_PLANES    6
+
+/*
+ * This is the limit to the number of fixed-function texture
+ * transforms and texture coordinates we can support. It does *not*
+ * correspond to the number of texture image units (samplers) we
+ * support!
+ */
+#define SVGA3D_MAX_TEXTURE_COORDS 8
+
+/*
+ * Vertex declarations
+ *
+ * Notes:
+ *
+ * SVGA3D_DECLUSAGE_POSITIONT is for pre-transformed vertices. If you
+ * draw with any POSITIONT vertex arrays, the programmable vertex
+ * pipeline will be implicitly disabled. Drawing will take place as if
+ * no vertex shader was bound.
+ */
+
+typedef enum {
+   SVGA3D_DECLUSAGE_POSITION     = 0,
+   SVGA3D_DECLUSAGE_BLENDWEIGHT,       //  1
+   SVGA3D_DECLUSAGE_BLENDINDICES,      //  2
+   SVGA3D_DECLUSAGE_NORMAL,            //  3
+   SVGA3D_DECLUSAGE_PSIZE,             //  4
+   SVGA3D_DECLUSAGE_TEXCOORD,          //  5
+   SVGA3D_DECLUSAGE_TANGENT,           //  6
+   SVGA3D_DECLUSAGE_BINORMAL,          //  7
+   SVGA3D_DECLUSAGE_TESSFACTOR,        //  8
+   SVGA3D_DECLUSAGE_POSITIONT,         //  9
+   SVGA3D_DECLUSAGE_COLOR,             // 10
+   SVGA3D_DECLUSAGE_FOG,               // 11
+   SVGA3D_DECLUSAGE_DEPTH,             // 12
+   SVGA3D_DECLUSAGE_SAMPLE,            // 13
+   SVGA3D_DECLUSAGE_MAX
+} SVGA3dDeclUsage;
+
+typedef enum {
+   SVGA3D_DECLMETHOD_DEFAULT     = 0,
+   SVGA3D_DECLMETHOD_PARTIALU,
+   SVGA3D_DECLMETHOD_PARTIALV,
+   SVGA3D_DECLMETHOD_CROSSUV,          // Normal
+   SVGA3D_DECLMETHOD_UV,
+   SVGA3D_DECLMETHOD_LOOKUP,           // Lookup a displacement map
+   SVGA3D_DECLMETHOD_LOOKUPPRESAMPLED, // Lookup a pre-sampled displacement map
+} SVGA3dDeclMethod;
+
+typedef enum {
+   SVGA3D_DECLTYPE_FLOAT1        =  0,
+   SVGA3D_DECLTYPE_FLOAT2        =  1,
+   SVGA3D_DECLTYPE_FLOAT3        =  2,
+   SVGA3D_DECLTYPE_FLOAT4        =  3,
+   SVGA3D_DECLTYPE_D3DCOLOR      =  4,
+   SVGA3D_DECLTYPE_UBYTE4        =  5,
+   SVGA3D_DECLTYPE_SHORT2        =  6,
+   SVGA3D_DECLTYPE_SHORT4        =  7,
+   SVGA3D_DECLTYPE_UBYTE4N       =  8,
+   SVGA3D_DECLTYPE_SHORT2N       =  9,
+   SVGA3D_DECLTYPE_SHORT4N       = 10,
+   SVGA3D_DECLTYPE_USHORT2N      = 11,
+   SVGA3D_DECLTYPE_USHORT4N      = 12,
+   SVGA3D_DECLTYPE_UDEC3         = 13,
+   SVGA3D_DECLTYPE_DEC3N         = 14,
+   SVGA3D_DECLTYPE_FLOAT16_2     = 15,
+   SVGA3D_DECLTYPE_FLOAT16_4     = 16,
+   SVGA3D_DECLTYPE_MAX,
+} SVGA3dDeclType;
+
+/*
+ * This structure is used for the divisor for geometry instancing;
+ * it's a direct translation of the Direct3D equivalent.
+ */
+typedef union {
+   struct {
+      /*
+       * For index data, this number represents the number of instances to draw.
+       * For instance data, this number represents the number of
+       * instances/vertex in this stream
+       */
+      uint32 count : 30;
+
+      /*
+       * This is 1 if this is supposed to be the data that is repeated for
+       * every instance.
+       */
+      uint32 indexedData : 1;
+
+      /*
+       * This is 1 if this is supposed to be the per-instance data.
+       */
+      uint32 instanceData : 1;
+   };
+
+   uint32 value;
+} SVGA3dVertexDivisor;
+
+typedef enum {
+   SVGA3D_PRIMITIVE_INVALID                     = 0,
+   SVGA3D_PRIMITIVE_TRIANGLELIST                = 1,
+   SVGA3D_PRIMITIVE_POINTLIST                   = 2,
+   SVGA3D_PRIMITIVE_LINELIST                    = 3,
+   SVGA3D_PRIMITIVE_LINESTRIP                   = 4,
+   SVGA3D_PRIMITIVE_TRIANGLESTRIP               = 5,
+   SVGA3D_PRIMITIVE_TRIANGLEFAN                 = 6,
+   SVGA3D_PRIMITIVE_MAX
+} SVGA3dPrimitiveType;
+
+typedef enum {
+   SVGA3D_COORDINATE_INVALID                   = 0,
+   SVGA3D_COORDINATE_LEFTHANDED                = 1,
+   SVGA3D_COORDINATE_RIGHTHANDED               = 2,
+   SVGA3D_COORDINATE_MAX
+} SVGA3dCoordinateType;
+
+typedef enum {
+   SVGA3D_TRANSFORM_INVALID                     = 0,
+   SVGA3D_TRANSFORM_WORLD                       = 1,
+   SVGA3D_TRANSFORM_VIEW                        = 2,
+   SVGA3D_TRANSFORM_PROJECTION                  = 3,
+   SVGA3D_TRANSFORM_TEXTURE0                    = 4,
+   SVGA3D_TRANSFORM_TEXTURE1                    = 5,
+   SVGA3D_TRANSFORM_TEXTURE2                    = 6,
+   SVGA3D_TRANSFORM_TEXTURE3                    = 7,
+   SVGA3D_TRANSFORM_TEXTURE4                    = 8,
+   SVGA3D_TRANSFORM_TEXTURE5                    = 9,
+   SVGA3D_TRANSFORM_TEXTURE6                    = 10,
+   SVGA3D_TRANSFORM_TEXTURE7                    = 11,
+   SVGA3D_TRANSFORM_WORLD1                      = 12,
+   SVGA3D_TRANSFORM_WORLD2                      = 13,
+   SVGA3D_TRANSFORM_WORLD3                      = 14,
+   SVGA3D_TRANSFORM_MAX
+} SVGA3dTransformType;
+
+typedef enum {
+   SVGA3D_LIGHTTYPE_INVALID                     = 0,
+   SVGA3D_LIGHTTYPE_POINT                       = 1,
+   SVGA3D_LIGHTTYPE_SPOT1                       = 2, /* 1-cone, in degrees */
+   SVGA3D_LIGHTTYPE_SPOT2                       = 3, /* 2-cone, in radians */
+   SVGA3D_LIGHTTYPE_DIRECTIONAL                 = 4,
+   SVGA3D_LIGHTTYPE_MAX
+} SVGA3dLightType;
+
+typedef enum {
+   SVGA3D_CUBEFACE_POSX                         = 0,
+   SVGA3D_CUBEFACE_NEGX                         = 1,
+   SVGA3D_CUBEFACE_POSY                         = 2,
+   SVGA3D_CUBEFACE_NEGY                         = 3,
+   SVGA3D_CUBEFACE_POSZ                         = 4,
+   SVGA3D_CUBEFACE_NEGZ                         = 5,
+} SVGA3dCubeFace;
+
+typedef enum {
+   SVGA3D_SHADERTYPE_COMPILED_DX8               = 0,
+   SVGA3D_SHADERTYPE_VS                         = 1,
+   SVGA3D_SHADERTYPE_PS                         = 2,
+   SVGA3D_SHADERTYPE_MAX
+} SVGA3dShaderType;
+
+typedef enum {
+   SVGA3D_CONST_TYPE_FLOAT                      = 0,
+   SVGA3D_CONST_TYPE_INT                        = 1,
+   SVGA3D_CONST_TYPE_BOOL                       = 2,
+} SVGA3dShaderConstType;
+
+#define SVGA3D_MAX_SURFACE_FACES                6
+
+typedef enum {
+   SVGA3D_STRETCH_BLT_POINT                     = 0,
+   SVGA3D_STRETCH_BLT_LINEAR                    = 1,
+   SVGA3D_STRETCH_BLT_MAX
+} SVGA3dStretchBltMode;
+
+typedef enum {
+   SVGA3D_QUERYTYPE_OCCLUSION                   = 0,
+   SVGA3D_QUERYTYPE_MAX
+} SVGA3dQueryType;
+
+typedef enum {
+   SVGA3D_QUERYSTATE_PENDING     = 0,      /* Waiting on the host (set by guest) */
+   SVGA3D_QUERYSTATE_SUCCEEDED   = 1,      /* Completed successfully (set by host) */
+   SVGA3D_QUERYSTATE_FAILED      = 2,      /* Completed unsuccessfully (set by host) */
+   SVGA3D_QUERYSTATE_NEW         = 3,      /* Never submitted (For guest use only) */
+} SVGA3dQueryState;
+
+typedef enum {
+   SVGA3D_WRITE_HOST_VRAM        = 1,
+   SVGA3D_READ_HOST_VRAM         = 2,
+} SVGA3dTransferType;
+
+/*
+ * The maximum number vertex arrays we're guaranteed to support in
+ * SVGA_3D_CMD_DRAWPRIMITIVES.
+ */
+#define SVGA3D_MAX_VERTEX_ARRAYS   32
+
+/*
+ * Identifiers for commands in the command FIFO.
+ *
+ * IDs between 1000 and 1039 (inclusive) were used by obsolete versions of
+ * the SVGA3D protocol and remain reserved; they should not be used in the
+ * future.
+ *
+ * IDs between 1040 and 1999 (inclusive) are available for use by the
+ * current SVGA3D protocol.
+ *
+ * FIFO clients other than SVGA3D should stay below 1000, or at 2000
+ * and up.
+ */
+
+#define SVGA_3D_CMD_LEGACY_BASE            1000
+#define SVGA_3D_CMD_BASE                   1040
+
+#define SVGA_3D_CMD_SURFACE_DEFINE         SVGA_3D_CMD_BASE + 0
+#define SVGA_3D_CMD_SURFACE_DESTROY        SVGA_3D_CMD_BASE + 1
+#define SVGA_3D_CMD_SURFACE_COPY           SVGA_3D_CMD_BASE + 2
+#define SVGA_3D_CMD_SURFACE_STRETCHBLT     SVGA_3D_CMD_BASE + 3
+#define SVGA_3D_CMD_SURFACE_DMA            SVGA_3D_CMD_BASE + 4
+#define SVGA_3D_CMD_CONTEXT_DEFINE         SVGA_3D_CMD_BASE + 5
+#define SVGA_3D_CMD_CONTEXT_DESTROY        SVGA_3D_CMD_BASE + 6
+#define SVGA_3D_CMD_SETTRANSFORM           SVGA_3D_CMD_BASE + 7
+#define SVGA_3D_CMD_SETZRANGE              SVGA_3D_CMD_BASE + 8
+#define SVGA_3D_CMD_SETRENDERSTATE         SVGA_3D_CMD_BASE + 9
+#define SVGA_3D_CMD_SETRENDERTARGET        SVGA_3D_CMD_BASE + 10
+#define SVGA_3D_CMD_SETTEXTURESTATE        SVGA_3D_CMD_BASE + 11
+#define SVGA_3D_CMD_SETMATERIAL            SVGA_3D_CMD_BASE + 12
+#define SVGA_3D_CMD_SETLIGHTDATA           SVGA_3D_CMD_BASE + 13
+#define SVGA_3D_CMD_SETLIGHTENABLED        SVGA_3D_CMD_BASE + 14
+#define SVGA_3D_CMD_SETVIEWPORT            SVGA_3D_CMD_BASE + 15
+#define SVGA_3D_CMD_SETCLIPPLANE           SVGA_3D_CMD_BASE + 16
+#define SVGA_3D_CMD_CLEAR                  SVGA_3D_CMD_BASE + 17
+#define SVGA_3D_CMD_PRESENT                SVGA_3D_CMD_BASE + 18    // Deprecated
+#define SVGA_3D_CMD_SHADER_DEFINE          SVGA_3D_CMD_BASE + 19
+#define SVGA_3D_CMD_SHADER_DESTROY         SVGA_3D_CMD_BASE + 20
+#define SVGA_3D_CMD_SET_SHADER             SVGA_3D_CMD_BASE + 21
+#define SVGA_3D_CMD_SET_SHADER_CONST       SVGA_3D_CMD_BASE + 22
+#define SVGA_3D_CMD_DRAW_PRIMITIVES        SVGA_3D_CMD_BASE + 23
+#define SVGA_3D_CMD_SETSCISSORRECT         SVGA_3D_CMD_BASE + 24
+#define SVGA_3D_CMD_BEGIN_QUERY            SVGA_3D_CMD_BASE + 25
+#define SVGA_3D_CMD_END_QUERY              SVGA_3D_CMD_BASE + 26
+#define SVGA_3D_CMD_WAIT_FOR_QUERY         SVGA_3D_CMD_BASE + 27
+#define SVGA_3D_CMD_PRESENT_READBACK       SVGA_3D_CMD_BASE + 28    // Deprecated
+#define SVGA_3D_CMD_BLIT_SURFACE_TO_SCREEN SVGA_3D_CMD_BASE + 29
+#define SVGA_3D_CMD_MAX                    SVGA_3D_CMD_BASE + 30
+
+#define SVGA_3D_CMD_FUTURE_MAX             2000
+
+/*
+ * Common substructures used in multiple FIFO commands:
+ */
+
+typedef struct {
+   union {
+      struct {
+         uint16  function;       // SVGA3dFogFunction
+         uint8   type;           // SVGA3dFogType
+         uint8   base;           // SVGA3dFogBase
+      };
+      uint32     uintValue;
+   };
+} SVGA3dFogMode;
+
+/*
+ * Uniquely identify one image (a 1D/2D/3D array) from a surface. This
+ * is a surface ID as well as face/mipmap indices.
+ */
+
+typedef
+struct SVGA3dSurfaceImageId {
+   uint32               sid;
+   uint32               face;
+   uint32               mipmap;
+} SVGA3dSurfaceImageId;
+
+typedef
+struct SVGA3dGuestImage {
+   SVGAGuestPtr         ptr;
+
+   /*
+    * A note on interpretation of pitch: This value of pitch is the
+    * number of bytes between vertically adjacent image
+    * blocks. Normally this is the number of bytes between the first
+    * pixel of two adjacent scanlines. With compressed textures,
+    * however, this may represent the number of bytes between
+    * compression blocks rather than between rows of pixels.
+    *
+    * XXX: Compressed textures currently must be tightly packed in guest memory.
+    *
+    * If the image is 1-dimensional, pitch is ignored.
+    *
+    * If 'pitch' is zero, the SVGA3D device calculates a pitch value
+    * assuming each row of blocks is tightly packed.
+    */
+   uint32 pitch;
+} SVGA3dGuestImage;
+
+
+/*
+ * FIFO command format definitions:
+ */
+
+/*
+ * The data size header following cmdNum for every 3d command
+ */
+typedef
+struct {
+   uint32               id;
+   uint32               size;
+} SVGA3dCmdHeader;
+
+/*
+ * A surface is a hierarchy of host VRAM surfaces: 1D, 2D, or 3D, with
+ * optional mipmaps and cube faces.
+ */
+
+typedef
+struct {
+   uint32               width;
+   uint32               height;
+   uint32               depth;
+} SVGA3dSize;
+
+typedef enum {
+   SVGA3D_SURFACE_CUBEMAP              = (1 << 0),
+   SVGA3D_SURFACE_HINT_STATIC          = (1 << 1),
+   SVGA3D_SURFACE_HINT_DYNAMIC         = (1 << 2),
+   SVGA3D_SURFACE_HINT_INDEXBUFFER     = (1 << 3),
+   SVGA3D_SURFACE_HINT_VERTEXBUFFER    = (1 << 4),
+   SVGA3D_SURFACE_HINT_TEXTURE         = (1 << 5),
+   SVGA3D_SURFACE_HINT_RENDERTARGET    = (1 << 6),
+   SVGA3D_SURFACE_HINT_DEPTHSTENCIL    = (1 << 7),
+   SVGA3D_SURFACE_HINT_WRITEONLY       = (1 << 8),
+} SVGA3dSurfaceFlags;
+
+typedef
+struct {
+   uint32               numMipLevels;
+} SVGA3dSurfaceFace;
+
+typedef
+struct {
+   uint32                      sid;
+   SVGA3dSurfaceFlags          surfaceFlags;
+   SVGA3dSurfaceFormat         format;
+   SVGA3dSurfaceFace           face[SVGA3D_MAX_SURFACE_FACES];
+   /*
+    * Followed by an SVGA3dSize structure for each mip level in each face.
+    *
+    * A note on surface sizes: Sizes are always specified in pixels,
+    * even if the true surface size is not a multiple of the minimum
+    * block size of the surface's format. For example, a 3x3x1 DXT1
+    * compressed texture would actually be stored as a 4x4x1 image in
+    * memory.
+    */
+} SVGA3dCmdDefineSurface;       /* SVGA_3D_CMD_SURFACE_DEFINE */
+
+typedef
+struct {
+   uint32               sid;
+} SVGA3dCmdDestroySurface;      /* SVGA_3D_CMD_SURFACE_DESTROY */
+
+typedef
+struct {
+   uint32               cid;
+} SVGA3dCmdDefineContext;       /* SVGA_3D_CMD_CONTEXT_DEFINE */
+
+typedef
+struct {
+   uint32               cid;
+} SVGA3dCmdDestroyContext;      /* SVGA_3D_CMD_CONTEXT_DESTROY */
+
+typedef
+struct {
+   uint32               cid;
+   SVGA3dClearFlag      clearFlag;
+   uint32               color;
+   float                depth;
+   uint32               stencil;
+   /* Followed by variable number of SVGA3dRect structures */
+} SVGA3dCmdClear;               /* SVGA_3D_CMD_CLEAR */
+
+typedef
+struct SVGA3dCopyRect {
+   uint32               x;
+   uint32               y;
+   uint32               w;
+   uint32               h;
+   uint32               srcx;
+   uint32               srcy;
+} SVGA3dCopyRect;
+
+typedef
+struct SVGA3dCopyBox {
+   uint32               x;
+   uint32               y;
+   uint32               z;
+   uint32               w;
+   uint32               h;
+   uint32               d;
+   uint32               srcx;
+   uint32               srcy;
+   uint32               srcz;
+} SVGA3dCopyBox;
+
+typedef
+struct {
+   uint32               x;
+   uint32               y;
+   uint32               w;
+   uint32               h;
+} SVGA3dRect;
+
+typedef
+struct {
+   uint32               x;
+   uint32               y;
+   uint32               z;
+   uint32               w;
+   uint32               h;
+   uint32               d;
+} SVGA3dBox;
+
+typedef
+struct {
+   uint32               x;
+   uint32               y;
+   uint32               z;
+} SVGA3dPoint;
+
+typedef
+struct {
+   SVGA3dLightType      type;
+   SVGA3dBool           inWorldSpace;
+   float                diffuse[4];
+   float                specular[4];
+   float                ambient[4];
+   float                position[4];
+   float                direction[4];
+   float                range;
+   float                falloff;
+   float                attenuation0;
+   float                attenuation1;
+   float                attenuation2;
+   float                theta;
+   float                phi;
+} SVGA3dLightData;
+
+typedef
+struct {
+   uint32               sid;
+   /* Followed by variable number of SVGA3dCopyRect structures */
+} SVGA3dCmdPresent;             /* SVGA_3D_CMD_PRESENT */
+
+typedef
+struct {
+   SVGA3dRenderStateName   state;
+   union {
+      uint32               uintValue;
+      float                floatValue;
+   };
+} SVGA3dRenderState;
+
+typedef
+struct {
+   uint32               cid;
+   /* Followed by variable number of SVGA3dRenderState structures */
+} SVGA3dCmdSetRenderState;      /* SVGA_3D_CMD_SETRENDERSTATE */
+
+typedef
+struct {
+   uint32                 cid;
+   SVGA3dRenderTargetType type;
+   SVGA3dSurfaceImageId   target;
+} SVGA3dCmdSetRenderTarget;     /* SVGA_3D_CMD_SETRENDERTARGET */
+
+typedef
+struct {
+   SVGA3dSurfaceImageId  src;
+   SVGA3dSurfaceImageId  dest;
+   /* Followed by variable number of SVGA3dCopyBox structures */
+} SVGA3dCmdSurfaceCopy;               /* SVGA_3D_CMD_SURFACE_COPY */
+
+typedef
+struct {
+   SVGA3dSurfaceImageId  src;
+   SVGA3dSurfaceImageId  dest;
+   SVGA3dBox             boxSrc;
+   SVGA3dBox             boxDest;
+   SVGA3dStretchBltMode  mode;
+} SVGA3dCmdSurfaceStretchBlt;         /* SVGA_3D_CMD_SURFACE_STRETCHBLT */
+
+typedef
+struct {
+   /*
+    * If the discard flag is present in a surface DMA operation, the host may
+    * discard the contents of the current mipmap level and face of the target
+    * surface before applying the surface DMA contents.
+    */
+   uint32 discard : 1;
+
+   /*
+    * If the unsynchronized flag is present, the host may perform this upload
+    * without syncing to pending reads on this surface.
+    */
+   uint32 unsynchronized : 1;
+
+   /*
+    * Guests *MUST* set the reserved bits to 0 before submitting the command
+    * suffix as future flags may occupy these bits.
+    */
+   uint32 reserved : 30;
+} SVGA3dSurfaceDMAFlags;
+
+typedef
+struct {
+   SVGA3dGuestImage      guest;
+   SVGA3dSurfaceImageId  host;
+   SVGA3dTransferType    transfer;
+   /*
+    * Followed by variable number of SVGA3dCopyBox structures. For consistency
+    * in all clipping logic and coordinate translation, we define the
+    * "source" in each copyBox as the guest image and the
+    * "destination" as the host image, regardless of transfer
+    * direction.
+    *
+    * For efficiency, the SVGA3D device is free to copy more data than
+    * specified. For example, it may round copy boxes outwards such
+    * that they lie on particular alignment boundaries.
+    */
+} SVGA3dCmdSurfaceDMA;                /* SVGA_3D_CMD_SURFACE_DMA */
+
+/*
+ * SVGA3dCmdSurfaceDMASuffix --
+ *
+ *    This is a command suffix that will appear after a SurfaceDMA command in
+ *    the FIFO.  It contains some extra information that hosts may use to
+ *    optimize performance or protect the guest.  This suffix exists to preserve
+ *    backwards compatibility while also allowing for new functionality to be
+ *    implemented.
+ */
+
+typedef
+struct {
+   uint32 suffixSize;
+
+   /*
+    * The maximum offset is used to determine the maximum offset from the
+    * guestPtr base address that will be accessed or written to during this
+    * surfaceDMA.  If the suffix is supported, the host will respect this
+    * boundary while performing surface DMAs.
+    *
+    * Defaults to MAX_UINT32
+    */
+   uint32 maximumOffset;
+
+   /*
+    * A set of flags that describes optimizations that the host may perform
+    * while performing this surface DMA operation.  The guest should never rely
+    * on behaviour that is different when these flags are set for correctness.
+    *
+    * Defaults to 0
+    */
+   SVGA3dSurfaceDMAFlags flags;
+} SVGA3dCmdSurfaceDMASuffix;
+
+/*
+ * SVGA_3D_CMD_DRAW_PRIMITIVES --
+ *
+ *   This command is the SVGA3D device's generic drawing entry point.
+ *   It can draw multiple ranges of primitives, optionally using an
+ *   index buffer, using an arbitrary collection of vertex buffers.
+ *
+ *   Each SVGA3dVertexDecl defines a distinct vertex array to bind
+ *   during this draw call. The declarations specify which surface
+ *   the vertex data lives in, what that vertex data is used for,
+ *   and how to interpret it.
+ *
+ *   Each SVGA3dPrimitiveRange defines a collection of primitives
+ *   to render using the same vertex arrays. An index buffer is
+ *   optional.
+ */
+
+typedef
+struct {
+   /*
+    * A range hint is an optional specification for the range of indices
+    * in an SVGA3dArray that will be used. If 'last' is zero, it is assumed
+    * that the entire array will be used.
+    *
+    * These are only hints. The SVGA3D device may use them for
+    * performance optimization if possible, but it's also allowed to
+    * ignore these values.
+    */
+   uint32               first;
+   uint32               last;
+} SVGA3dArrayRangeHint;
+
+typedef
+struct {
+   /*
+    * Define the origin and shape of a vertex or index array. Both
+    * 'offset' and 'stride' are in bytes. The provided surface will be
+    * reinterpreted as a flat array of bytes in the same format used
+    * by surface DMA operations. To avoid unnecessary conversions, the
+    * surface should be created with the SVGA3D_BUFFER format.
+    *
+    * Index 0 in the array starts 'offset' bytes into the surface.
+    * Index 1 begins at byte 'offset + stride', etc. Array indices may
+    * not be negative.
+    */
+   uint32               surfaceId;
+   uint32               offset;
+   uint32               stride;
+} SVGA3dArray;
+
+typedef
+struct {
+   /*
+    * Describe a vertex array's data type, and define how it is to be
+    * used by the fixed function pipeline or the vertex shader. It
+    * isn't useful to have two VertexDecls with the same
+    * VertexArrayIdentity in one draw call.
+    */
+   SVGA3dDeclType       type;
+   SVGA3dDeclMethod     method;
+   SVGA3dDeclUsage      usage;
+   uint32               usageIndex;
+} SVGA3dVertexArrayIdentity;
+
+typedef
+struct {
+   SVGA3dVertexArrayIdentity  identity;
+   SVGA3dArray                array;
+   SVGA3dArrayRangeHint       rangeHint;
+} SVGA3dVertexDecl;
+
+typedef
+struct {
+   /*
+    * Define a group of primitives to render, from sequential indices.
+    *
+    * The value of 'primitiveType' and 'primitiveCount' imply the
+    * total number of vertices that will be rendered.
+    */
+   SVGA3dPrimitiveType  primType;
+   uint32               primitiveCount;
+
+   /*
+    * Optional index buffer. If indexArray.surfaceId is
+    * SVGA3D_INVALID_ID, we render without an index buffer. Rendering
+    * without an index buffer is identical to rendering with an index
+    * buffer containing the sequence [0, 1, 2, 3, ...].
+    *
+    * If an index buffer is in use, indexWidth specifies the width in
+    * bytes of each index value. It must be less than or equal to
+    * indexArray.stride.
+    *
+    * (Currently, the SVGA3D device requires index buffers to be tightly
+    * packed. In other words, indexWidth == indexArray.stride)
+    */
+   SVGA3dArray          indexArray;
+   uint32               indexWidth;
+
+   /*
+    * Optional index bias. This number is added to all indices from
+    * indexArray before they are used as vertex array indices. This
+    * can be used in multiple ways:
+    *
+    *  - When not using an indexArray, this bias can be used to
+    *    specify where in the vertex arrays to begin rendering.
+    *
+    *  - A positive number here is equivalent to increasing the
+    *    offset in each vertex array.
+    *
+    *  - A negative number can be used to render using a small
+    *    vertex array and an index buffer that contains large
+    *    values. This may be used by some applications that
+    *    crop a vertex buffer without modifying their index
+    *    buffer.
+    *
+    * Note that rendering with a negative bias value may be slower and
+    * use more memory than rendering with a positive or zero bias.
+    */
+   int32                indexBias;
+} SVGA3dPrimitiveRange;
+
+typedef
+struct {
+   uint32               cid;
+   uint32               numVertexDecls;
+   uint32               numRanges;
+
+   /*
+    * There are two variable size arrays after the
+    * SVGA3dCmdDrawPrimitives structure. In order,
+    * they are:
+    *
+    * 1. SVGA3dVertexDecl, quantity 'numVertexDecls'
+    * 2. SVGA3dPrimitiveRange, quantity 'numRanges'
+    * 3. Optionally, SVGA3dVertexDivisor, quantity 'numVertexDecls' (contains
+    *    the frequency divisor for this the corresponding vertex decl)
+    */
+} SVGA3dCmdDrawPrimitives;      /* SVGA_3D_CMD_DRAWPRIMITIVES */
+
+typedef
+struct {
+   uint32                   stage;
+   SVGA3dTextureStateName   name;
+   union {
+      uint32                value;
+      float                 floatValue;
+   };
+} SVGA3dTextureState;
+
+typedef
+struct {
+   uint32               cid;
+   /* Followed by variable number of SVGA3dTextureState structures */
+} SVGA3dCmdSetTextureState;      /* SVGA_3D_CMD_SETTEXTURESTATE */
+
+typedef
+struct {
+   uint32                   cid;
+   SVGA3dTransformType      type;
+   float                    matrix[16];
+} SVGA3dCmdSetTransform;          /* SVGA_3D_CMD_SETTRANSFORM */
+
+typedef
+struct {
+   float                min;
+   float                max;
+} SVGA3dZRange;
+
+typedef
+struct {
+   uint32               cid;
+   SVGA3dZRange         zRange;
+} SVGA3dCmdSetZRange;             /* SVGA_3D_CMD_SETZRANGE */
+
+typedef
+struct {
+   float                diffuse[4];
+   float                ambient[4];
+   float                specular[4];
+   float                emissive[4];
+   float                shininess;
+} SVGA3dMaterial;
+
+typedef
+struct {
+   uint32               cid;
+   SVGA3dFace           face;
+   SVGA3dMaterial       material;
+} SVGA3dCmdSetMaterial;           /* SVGA_3D_CMD_SETMATERIAL */
+
+typedef
+struct {
+   uint32               cid;
+   uint32               index;
+   SVGA3dLightData      data;
+} SVGA3dCmdSetLightData;           /* SVGA_3D_CMD_SETLIGHTDATA */
+
+typedef
+struct {
+   uint32               cid;
+   uint32               index;
+   uint32               enabled;
+} SVGA3dCmdSetLightEnabled;      /* SVGA_3D_CMD_SETLIGHTENABLED */
+
+typedef
+struct {
+   uint32               cid;
+   SVGA3dRect           rect;
+} SVGA3dCmdSetViewport;           /* SVGA_3D_CMD_SETVIEWPORT */
+
+typedef
+struct {
+   uint32               cid;
+   SVGA3dRect           rect;
+} SVGA3dCmdSetScissorRect;         /* SVGA_3D_CMD_SETSCISSORRECT */
+
+typedef
+struct {
+   uint32               cid;
+   uint32               index;
+   float                plane[4];
+} SVGA3dCmdSetClipPlane;           /* SVGA_3D_CMD_SETCLIPPLANE */
+
+typedef
+struct {
+   uint32               cid;
+   uint32               shid;
+   SVGA3dShaderType     type;
+   /* Followed by variable number of DWORDs for shader bycode */
+} SVGA3dCmdDefineShader;           /* SVGA_3D_CMD_SHADER_DEFINE */
+
+typedef
+struct {
+   uint32               cid;
+   uint32               shid;
+   SVGA3dShaderType     type;
+} SVGA3dCmdDestroyShader;         /* SVGA_3D_CMD_SHADER_DESTROY */
+
+typedef
+struct {
+   uint32                  cid;
+   uint32                  reg;     /* register number */
+   SVGA3dShaderType        type;
+   SVGA3dShaderConstType   ctype;
+   uint32                  values[4];
+} SVGA3dCmdSetShaderConst;        /* SVGA_3D_CMD_SET_SHADER_CONST */
+
+typedef
+struct {
+   uint32               cid;
+   SVGA3dShaderType     type;
+   uint32               shid;
+} SVGA3dCmdSetShader;             /* SVGA_3D_CMD_SET_SHADER */
+
+typedef
+struct {
+   uint32               cid;
+   SVGA3dQueryType      type;
+} SVGA3dCmdBeginQuery;           /* SVGA_3D_CMD_BEGIN_QUERY */
+
+typedef
+struct {
+   uint32               cid;
+   SVGA3dQueryType      type;
+   SVGAGuestPtr         guestResult;  /* Points to an SVGA3dQueryResult structure */
+} SVGA3dCmdEndQuery;                  /* SVGA_3D_CMD_END_QUERY */
+
+typedef
+struct {
+   uint32               cid;          /* Same parameters passed to END_QUERY */
+   SVGA3dQueryType      type;
+   SVGAGuestPtr         guestResult;
+} SVGA3dCmdWaitForQuery;              /* SVGA_3D_CMD_WAIT_FOR_QUERY */
+
+typedef
+struct {
+   uint32               totalSize;    /* Set by guest before query is ended. */
+   SVGA3dQueryState     state;        /* Set by host or guest. See SVGA3dQueryState. */
+   union {                            /* Set by host on exit from PENDING state */
+      uint32            result32;
+   };
+} SVGA3dQueryResult;
+
+/*
+ * SVGA_3D_CMD_BLIT_SURFACE_TO_SCREEN --
+ *
+ *    This is a blit from an SVGA3D surface to a Screen Object. Just
+ *    like GMR-to-screen blits, this blit may be directed at a
+ *    specific screen or to the virtual coordinate space.
+ *
+ *    The blit copies from a rectangular region of an SVGA3D surface
+ *    image to a rectangular region of a screen or screens.
+ *
+ *    This command takes an optional variable-length list of clipping
+ *    rectangles after the body of the command. If no rectangles are
+ *    specified, there is no clipping region. The entire destRect is
+ *    drawn to. If one or more rectangles are included, they describe
+ *    a clipping region. The clip rectangle coordinates are measured
+ *    relative to the top-left corner of destRect.
+ *
+ *    This clipping region serves multiple purposes:
+ *
+ *      - It can be used to perform an irregularly shaped blit more
+ *        efficiently than by issuing many separate blit commands.
+ *
+ *      - It is equivalent to allowing blits with non-integer
+ *        source coordinates. You could blit just one half-pixel
+ *        of a source, for example, by specifying a larger
+ *        destination rectangle than you need, then removing
+ *        part of it using a clip rectangle.
+ *
+ * Availability:
+ *    SVGA_FIFO_CAP_SCREEN_OBJECT
+ *
+ * Limitations:
+ *
+ *    - Currently, no backend supports blits from a mipmap or face
+ *      other than the first one.
+ */
+
+typedef
+struct {
+   SVGA3dSurfaceImageId srcImage;
+   SVGASignedRect       srcRect;
+   uint32               destScreenId; /* Screen ID or SVGA_ID_INVALID for virt. coords */
+   SVGASignedRect       destRect;     /* Supports scaling if src/rest different size */
+   /* Clipping: zero or more SVGASignedRects follow */
+} SVGA3dCmdBlitSurfaceToScreen;         /* SVGA_3D_CMD_BLIT_SURFACE_TO_SCREEN */
+
+
+/*
+ * Capability query index.
+ *
+ * Notes:
+ *
+ *   1. SVGA3D_DEVCAP_MAX_TEXTURES reflects the maximum number of
+ *      fixed-function texture units available. Each of these units
+ *      work in both FFP and Shader modes, and they support texture
+ *      transforms and texture coordinates. The host may have additional
+ *      texture image units that are only usable with shaders.
+ *
+ *   2. The BUFFER_FORMAT capabilities are deprecated, and they always
+ *      return TRUE. Even on physical hardware that does not support
+ *      these formats natively, the SVGA3D device will provide an emulation
+ *      which should be invisible to the guest OS.
+ *
+ *      In general, the SVGA3D device should support any operation on
+ *      any surface format, it just may perform some of these
+ *      operations in software depending on the capabilities of the
+ *      available physical hardware.
+ *
+ *      XXX: In the future, we will add capabilities that describe in
+ *      detail what formats are supported in hardware for what kinds
+ *      of operations.
+ */
+
+typedef enum {
+   SVGA3D_DEVCAP_3D                                = 0,
+   SVGA3D_DEVCAP_MAX_LIGHTS                        = 1,
+   SVGA3D_DEVCAP_MAX_TEXTURES                      = 2,  /* See note (1) */
+   SVGA3D_DEVCAP_MAX_CLIP_PLANES                   = 3,
+   SVGA3D_DEVCAP_VERTEX_SHADER_VERSION             = 4,
+   SVGA3D_DEVCAP_VERTEX_SHADER                     = 5,
+   SVGA3D_DEVCAP_FRAGMENT_SHADER_VERSION           = 6,
+   SVGA3D_DEVCAP_FRAGMENT_SHADER                   = 7,
+   SVGA3D_DEVCAP_MAX_RENDER_TARGETS                = 8,
+   SVGA3D_DEVCAP_S23E8_TEXTURES                    = 9,
+   SVGA3D_DEVCAP_S10E5_TEXTURES                    = 10,
+   SVGA3D_DEVCAP_MAX_FIXED_VERTEXBLEND             = 11,
+   SVGA3D_DEVCAP_D16_BUFFER_FORMAT                 = 12, /* See note (2) */
+   SVGA3D_DEVCAP_D24S8_BUFFER_FORMAT               = 13, /* See note (2) */
+   SVGA3D_DEVCAP_D24X8_BUFFER_FORMAT               = 14, /* See note (2) */
+   SVGA3D_DEVCAP_QUERY_TYPES                       = 15,
+   SVGA3D_DEVCAP_TEXTURE_GRADIENT_SAMPLING         = 16,
+   SVGA3D_DEVCAP_MAX_POINT_SIZE                    = 17,
+   SVGA3D_DEVCAP_MAX_SHADER_TEXTURES               = 18,
+   SVGA3D_DEVCAP_MAX_TEXTURE_WIDTH                 = 19,
+   SVGA3D_DEVCAP_MAX_TEXTURE_HEIGHT                = 20,
+   SVGA3D_DEVCAP_MAX_VOLUME_EXTENT                 = 21,
+   SVGA3D_DEVCAP_MAX_TEXTURE_REPEAT                = 22,
+   SVGA3D_DEVCAP_MAX_TEXTURE_ASPECT_RATIO          = 23,
+   SVGA3D_DEVCAP_MAX_TEXTURE_ANISOTROPY            = 24,
+   SVGA3D_DEVCAP_MAX_PRIMITIVE_COUNT               = 25,
+   SVGA3D_DEVCAP_MAX_VERTEX_INDEX                  = 26,
+   SVGA3D_DEVCAP_MAX_VERTEX_SHADER_INSTRUCTIONS    = 27,
+   SVGA3D_DEVCAP_MAX_FRAGMENT_SHADER_INSTRUCTIONS  = 28,
+   SVGA3D_DEVCAP_MAX_VERTEX_SHADER_TEMPS           = 29,
+   SVGA3D_DEVCAP_MAX_FRAGMENT_SHADER_TEMPS         = 30,
+   SVGA3D_DEVCAP_TEXTURE_OPS                       = 31,
+   SVGA3D_DEVCAP_SURFACEFMT_X8R8G8B8               = 32,
+   SVGA3D_DEVCAP_SURFACEFMT_A8R8G8B8               = 33,
+   SVGA3D_DEVCAP_SURFACEFMT_A2R10G10B10            = 34,
+   SVGA3D_DEVCAP_SURFACEFMT_X1R5G5B5               = 35,
+   SVGA3D_DEVCAP_SURFACEFMT_A1R5G5B5               = 36,
+   SVGA3D_DEVCAP_SURFACEFMT_A4R4G4B4               = 37,
+   SVGA3D_DEVCAP_SURFACEFMT_R5G6B5                 = 38,
+   SVGA3D_DEVCAP_SURFACEFMT_LUMINANCE16            = 39,
+   SVGA3D_DEVCAP_SURFACEFMT_LUMINANCE8_ALPHA8      = 40,
+   SVGA3D_DEVCAP_SURFACEFMT_ALPHA8                 = 41,
+   SVGA3D_DEVCAP_SURFACEFMT_LUMINANCE8             = 42,
+   SVGA3D_DEVCAP_SURFACEFMT_Z_D16                  = 43,
+   SVGA3D_DEVCAP_SURFACEFMT_Z_D24S8                = 44,
+   SVGA3D_DEVCAP_SURFACEFMT_Z_D24X8                = 45,
+   SVGA3D_DEVCAP_SURFACEFMT_DXT1                   = 46,
+   SVGA3D_DEVCAP_SURFACEFMT_DXT2                   = 47,
+   SVGA3D_DEVCAP_SURFACEFMT_DXT3                   = 48,
+   SVGA3D_DEVCAP_SURFACEFMT_DXT4                   = 49,
+   SVGA3D_DEVCAP_SURFACEFMT_DXT5                   = 50,
+   SVGA3D_DEVCAP_SURFACEFMT_BUMPX8L8V8U8           = 51,
+   SVGA3D_DEVCAP_SURFACEFMT_A2W10V10U10            = 52,
+   SVGA3D_DEVCAP_SURFACEFMT_BUMPU8V8               = 53,
+   SVGA3D_DEVCAP_SURFACEFMT_Q8W8V8U8               = 54,
+   SVGA3D_DEVCAP_SURFACEFMT_CxV8U8                 = 55,
+   SVGA3D_DEVCAP_SURFACEFMT_R_S10E5                = 56,
+   SVGA3D_DEVCAP_SURFACEFMT_R_S23E8                = 57,
+   SVGA3D_DEVCAP_SURFACEFMT_RG_S10E5               = 58,
+   SVGA3D_DEVCAP_SURFACEFMT_RG_S23E8               = 59,
+   SVGA3D_DEVCAP_SURFACEFMT_ARGB_S10E5             = 60,
+   SVGA3D_DEVCAP_SURFACEFMT_ARGB_S23E8             = 61,
+   SVGA3D_DEVCAP_MAX_VERTEX_SHADER_TEXTURES        = 63,
+
+   /*
+    * Note that MAX_SIMULTANEOUS_RENDER_TARGETS is a maximum count of color
+    * render targets.  This does no include the depth or stencil targets.
+    */
+   SVGA3D_DEVCAP_MAX_SIMULTANEOUS_RENDER_TARGETS   = 64,
+
+   SVGA3D_DEVCAP_SURFACEFMT_V16U16                 = 65,
+   SVGA3D_DEVCAP_SURFACEFMT_G16R16                 = 66,
+   SVGA3D_DEVCAP_SURFACEFMT_A16B16G16R16           = 67,
+   SVGA3D_DEVCAP_SURFACEFMT_UYVY                   = 68,
+   SVGA3D_DEVCAP_SURFACEFMT_YUY2                   = 69,
+
+   /*
+    * Don't add new caps into the previous section; the values in this
+    * enumeration must not change. You can put new values right before
+    * SVGA3D_DEVCAP_MAX.
+    */
+   SVGA3D_DEVCAP_MAX                                  /* This must be the last index. */
+} SVGA3dDevCapIndex;
+
+typedef union {
+   Bool   b;
+   uint32 u;
+   int32  i;
+   float  f;
+} SVGA3dDevCapResult;
+
+#endif /* _SVGA3D_REG_H_ */
diff --git a/drivers/gpu/drm/vmwgfx/svga_escape.h b/drivers/gpu/drm/vmwgfx/svga_escape.h
new file mode 100644
index 000000000000..7b85e9b8c854
--- /dev/null
+++ b/drivers/gpu/drm/vmwgfx/svga_escape.h
@@ -0,0 +1,89 @@
+/**********************************************************
+ * Copyright 2007-2009 VMware, Inc.  All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use, copy,
+ * modify, merge, publish, distribute, sublicense, and/or sell copies
+ * of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ **********************************************************/
+
+/*
+ * svga_escape.h --
+ *
+ *    Definitions for our own (vendor-specific) SVGA Escape commands.
+ */
+
+#ifndef _SVGA_ESCAPE_H_
+#define _SVGA_ESCAPE_H_
+
+
+/*
+ * Namespace IDs for the escape command
+ */
+
+#define SVGA_ESCAPE_NSID_VMWARE 0x00000000
+#define SVGA_ESCAPE_NSID_DEVEL  0xFFFFFFFF
+
+
+/*
+ * Within SVGA_ESCAPE_NSID_VMWARE, we multiplex commands according to
+ * the first DWORD of escape data (after the nsID and size). As a
+ * guideline we're using the high word and low word as a major and
+ * minor command number, respectively.
+ *
+ * Major command number allocation:
+ *
+ *   0000: Reserved
+ *   0001: SVGA_ESCAPE_VMWARE_LOG (svga_binary_logger.h)
+ *   0002: SVGA_ESCAPE_VMWARE_VIDEO (svga_overlay.h)
+ *   0003: SVGA_ESCAPE_VMWARE_HINT (svga_escape.h)
+ */
+
+#define SVGA_ESCAPE_VMWARE_MAJOR_MASK  0xFFFF0000
+
+
+/*
+ * SVGA Hint commands.
+ *
+ * These escapes let the SVGA driver provide optional information to
+ * he host about the state of the guest or guest applications. The
+ * host can use these hints to make user interface or performance
+ * decisions.
+ *
+ * Notes:
+ *
+ *   - SVGA_ESCAPE_VMWARE_HINT_FULLSCREEN is deprecated for guests
+ *     that use the SVGA Screen Object extension. Instead of sending
+ *     this escape, use the SVGA_SCREEN_FULLSCREEN_HINT flag on your
+ *     Screen Object.
+ */
+
+#define SVGA_ESCAPE_VMWARE_HINT               0x00030000
+#define SVGA_ESCAPE_VMWARE_HINT_FULLSCREEN    0x00030001  // Deprecated
+
+typedef
+struct {
+   uint32 command;
+   uint32 fullscreen;
+   struct {
+      int32 x, y;
+   } monitorPosition;
+} SVGAEscapeHintFullscreen;
+
+#endif /* _SVGA_ESCAPE_H_ */
diff --git a/drivers/gpu/drm/vmwgfx/svga_overlay.h b/drivers/gpu/drm/vmwgfx/svga_overlay.h
new file mode 100644
index 000000000000..f753d73c14b4
--- /dev/null
+++ b/drivers/gpu/drm/vmwgfx/svga_overlay.h
@@ -0,0 +1,201 @@
+/**********************************************************
+ * Copyright 2007-2009 VMware, Inc.  All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use, copy,
+ * modify, merge, publish, distribute, sublicense, and/or sell copies
+ * of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ **********************************************************/
+
+/*
+ * svga_overlay.h --
+ *
+ *    Definitions for video-overlay support.
+ */
+
+#ifndef _SVGA_OVERLAY_H_
+#define _SVGA_OVERLAY_H_
+
+#include "svga_reg.h"
+
+/*
+ * Video formats we support
+ */
+
+#define VMWARE_FOURCC_YV12 0x32315659 // 'Y' 'V' '1' '2'
+#define VMWARE_FOURCC_YUY2 0x32595559 // 'Y' 'U' 'Y' '2'
+#define VMWARE_FOURCC_UYVY 0x59565955 // 'U' 'Y' 'V' 'Y'
+
+typedef enum {
+   SVGA_OVERLAY_FORMAT_INVALID = 0,
+   SVGA_OVERLAY_FORMAT_YV12 = VMWARE_FOURCC_YV12,
+   SVGA_OVERLAY_FORMAT_YUY2 = VMWARE_FOURCC_YUY2,
+   SVGA_OVERLAY_FORMAT_UYVY = VMWARE_FOURCC_UYVY,
+} SVGAOverlayFormat;
+
+#define SVGA_VIDEO_COLORKEY_MASK             0x00ffffff
+
+#define SVGA_ESCAPE_VMWARE_VIDEO             0x00020000
+
+#define SVGA_ESCAPE_VMWARE_VIDEO_SET_REGS    0x00020001
+        /* FIFO escape layout:
+         * Type, Stream Id, (Register Id, Value) pairs */
+
+#define SVGA_ESCAPE_VMWARE_VIDEO_FLUSH       0x00020002
+        /* FIFO escape layout:
+         * Type, Stream Id */
+
+typedef
+struct SVGAEscapeVideoSetRegs {
+   struct {
+      uint32 cmdType;
+      uint32 streamId;
+   } header;
+
+   // May include zero or more items.
+   struct {
+      uint32 registerId;
+      uint32 value;
+   } items[1];
+} SVGAEscapeVideoSetRegs;
+
+typedef
+struct SVGAEscapeVideoFlush {
+   uint32 cmdType;
+   uint32 streamId;
+} SVGAEscapeVideoFlush;
+
+
+/*
+ * Struct definitions for the video overlay commands built on
+ * SVGAFifoCmdEscape.
+ */
+typedef
+struct {
+   uint32 command;
+   uint32 overlay;
+} SVGAFifoEscapeCmdVideoBase;
+
+typedef
+struct {
+   SVGAFifoEscapeCmdVideoBase videoCmd;
+} SVGAFifoEscapeCmdVideoFlush;
+
+typedef
+struct {
+   SVGAFifoEscapeCmdVideoBase videoCmd;
+   struct {
+      uint32 regId;
+      uint32 value;
+   } items[1];
+} SVGAFifoEscapeCmdVideoSetRegs;
+
+typedef
+struct {
+   SVGAFifoEscapeCmdVideoBase videoCmd;
+   struct {
+      uint32 regId;
+      uint32 value;
+   } items[SVGA_VIDEO_NUM_REGS];
+} SVGAFifoEscapeCmdVideoSetAllRegs;
+
+
+/*
+ *----------------------------------------------------------------------
+ *
+ * VMwareVideoGetAttributes --
+ *
+ *      Computes the size, pitches and offsets for YUV frames.
+ *
+ * Results:
+ *      TRUE on success; otherwise FALSE on failure.
+ *
+ * Side effects:
+ *      Pitches and offsets for the given YUV frame are put in 'pitches'
+ *      and 'offsets' respectively. They are both optional though.
+ *
+ *----------------------------------------------------------------------
+ */
+
+static inline bool
+VMwareVideoGetAttributes(const SVGAOverlayFormat format,    // IN
+                         uint32 *width,                     // IN / OUT
+                         uint32 *height,                    // IN / OUT
+                         uint32 *size,                      // OUT
+                         uint32 *pitches,                   // OUT (optional)
+                         uint32 *offsets)                   // OUT (optional)
+{
+    int tmp;
+
+    *width = (*width + 1) & ~1;
+
+    if (offsets) {
+        offsets[0] = 0;
+    }
+
+    switch (format) {
+    case VMWARE_FOURCC_YV12:
+       *height = (*height + 1) & ~1;
+       *size = (*width + 3) & ~3;
+
+       if (pitches) {
+          pitches[0] = *size;
+       }
+
+       *size *= *height;
+
+       if (offsets) {
+          offsets[1] = *size;
+       }
+
+       tmp = ((*width >> 1) + 3) & ~3;
+
+       if (pitches) {
+          pitches[1] = pitches[2] = tmp;
+       }
+
+       tmp *= (*height >> 1);
+       *size += tmp;
+
+       if (offsets) {
+          offsets[2] = *size;
+       }
+
+       *size += tmp;
+       break;
+
+    case VMWARE_FOURCC_YUY2:
+    case VMWARE_FOURCC_UYVY:
+       *size = *width * 2;
+
+       if (pitches) {
+          pitches[0] = *size;
+       }
+
+       *size *= *height;
+       break;
+
+    default:
+       return false;
+    }
+
+    return true;
+}
+
+#endif // _SVGA_OVERLAY_H_
diff --git a/drivers/gpu/drm/vmwgfx/svga_reg.h b/drivers/gpu/drm/vmwgfx/svga_reg.h
new file mode 100644
index 000000000000..1b96c2ec07dd
--- /dev/null
+++ b/drivers/gpu/drm/vmwgfx/svga_reg.h
@@ -0,0 +1,1346 @@
+/**********************************************************
+ * Copyright 1998-2009 VMware, Inc.  All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use, copy,
+ * modify, merge, publish, distribute, sublicense, and/or sell copies
+ * of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ **********************************************************/
+
+/*
+ * svga_reg.h --
+ *
+ *    Virtual hardware definitions for the VMware SVGA II device.
+ */
+
+#ifndef _SVGA_REG_H_
+#define _SVGA_REG_H_
+
+/*
+ * PCI device IDs.
+ */
+#define PCI_VENDOR_ID_VMWARE            0x15AD
+#define PCI_DEVICE_ID_VMWARE_SVGA2      0x0405
+
+/*
+ * Legal values for the SVGA_REG_CURSOR_ON register in old-fashioned
+ * cursor bypass mode. This is still supported, but no new guest
+ * drivers should use it.
+ */
+#define SVGA_CURSOR_ON_HIDE            0x0   /* Must be 0 to maintain backward compatibility */
+#define SVGA_CURSOR_ON_SHOW            0x1   /* Must be 1 to maintain backward compatibility */
+#define SVGA_CURSOR_ON_REMOVE_FROM_FB  0x2   /* Remove the cursor from the framebuffer because we need to see what's under it */
+#define SVGA_CURSOR_ON_RESTORE_TO_FB   0x3   /* Put the cursor back in the framebuffer so the user can see it */
+
+/*
+ * The maximum framebuffer size that can traced for e.g. guests in VESA mode.
+ * The changeMap in the monitor is proportional to this number. Therefore, we'd
+ * like to keep it as small as possible to reduce monitor overhead (using
+ * SVGA_VRAM_MAX_SIZE for this increases the size of the shared area by over
+ * 4k!).
+ *
+ * NB: For compatibility reasons, this value must be greater than 0xff0000.
+ *     See bug 335072.
+ */
+#define SVGA_FB_MAX_TRACEABLE_SIZE      0x1000000
+
+#define SVGA_MAX_PSEUDOCOLOR_DEPTH      8
+#define SVGA_MAX_PSEUDOCOLORS           (1 << SVGA_MAX_PSEUDOCOLOR_DEPTH)
+#define SVGA_NUM_PALETTE_REGS           (3 * SVGA_MAX_PSEUDOCOLORS)
+
+#define SVGA_MAGIC         0x900000UL
+#define SVGA_MAKE_ID(ver)  (SVGA_MAGIC << 8 | (ver))
+
+/* Version 2 let the address of the frame buffer be unsigned on Win32 */
+#define SVGA_VERSION_2     2
+#define SVGA_ID_2          SVGA_MAKE_ID(SVGA_VERSION_2)
+
+/* Version 1 has new registers starting with SVGA_REG_CAPABILITIES so
+   PALETTE_BASE has moved */
+#define SVGA_VERSION_1     1
+#define SVGA_ID_1          SVGA_MAKE_ID(SVGA_VERSION_1)
+
+/* Version 0 is the initial version */
+#define SVGA_VERSION_0     0
+#define SVGA_ID_0          SVGA_MAKE_ID(SVGA_VERSION_0)
+
+/* "Invalid" value for all SVGA IDs. (Version ID, screen object ID, surface ID...) */
+#define SVGA_ID_INVALID    0xFFFFFFFF
+
+/* Port offsets, relative to BAR0 */
+#define SVGA_INDEX_PORT         0x0
+#define SVGA_VALUE_PORT         0x1
+#define SVGA_BIOS_PORT          0x2
+#define SVGA_IRQSTATUS_PORT     0x8
+
+/*
+ * Interrupt source flags for IRQSTATUS_PORT and IRQMASK.
+ *
+ * Interrupts are only supported when the
+ * SVGA_CAP_IRQMASK capability is present.
+ */
+#define SVGA_IRQFLAG_ANY_FENCE            0x1    /* Any fence was passed */
+#define SVGA_IRQFLAG_FIFO_PROGRESS        0x2    /* Made forward progress in the FIFO */
+#define SVGA_IRQFLAG_FENCE_GOAL           0x4    /* SVGA_FIFO_FENCE_GOAL reached */
+
+/*
+ * Registers
+ */
+
+enum {
+   SVGA_REG_ID = 0,
+   SVGA_REG_ENABLE = 1,
+   SVGA_REG_WIDTH = 2,
+   SVGA_REG_HEIGHT = 3,
+   SVGA_REG_MAX_WIDTH = 4,
+   SVGA_REG_MAX_HEIGHT = 5,
+   SVGA_REG_DEPTH = 6,
+   SVGA_REG_BITS_PER_PIXEL = 7,       /* Current bpp in the guest */
+   SVGA_REG_PSEUDOCOLOR = 8,
+   SVGA_REG_RED_MASK = 9,
+   SVGA_REG_GREEN_MASK = 10,
+   SVGA_REG_BLUE_MASK = 11,
+   SVGA_REG_BYTES_PER_LINE = 12,
+   SVGA_REG_FB_START = 13,            /* (Deprecated) */
+   SVGA_REG_FB_OFFSET = 14,
+   SVGA_REG_VRAM_SIZE = 15,
+   SVGA_REG_FB_SIZE = 16,
+
+   /* ID 0 implementation only had the above registers, then the palette */
+
+   SVGA_REG_CAPABILITIES = 17,
+   SVGA_REG_MEM_START = 18,           /* (Deprecated) */
+   SVGA_REG_MEM_SIZE = 19,
+   SVGA_REG_CONFIG_DONE = 20,         /* Set when memory area configured */
+   SVGA_REG_SYNC = 21,                /* See "FIFO Synchronization Registers" */
+   SVGA_REG_BUSY = 22,                /* See "FIFO Synchronization Registers" */
+   SVGA_REG_GUEST_ID = 23,            /* Set guest OS identifier */
+   SVGA_REG_CURSOR_ID = 24,           /* (Deprecated) */
+   SVGA_REG_CURSOR_X = 25,            /* (Deprecated) */
+   SVGA_REG_CURSOR_Y = 26,            /* (Deprecated) */
+   SVGA_REG_CURSOR_ON = 27,           /* (Deprecated) */
+   SVGA_REG_HOST_BITS_PER_PIXEL = 28, /* (Deprecated) */
+   SVGA_REG_SCRATCH_SIZE = 29,        /* Number of scratch registers */
+   SVGA_REG_MEM_REGS = 30,            /* Number of FIFO registers */
+   SVGA_REG_NUM_DISPLAYS = 31,        /* (Deprecated) */
+   SVGA_REG_PITCHLOCK = 32,           /* Fixed pitch for all modes */
+   SVGA_REG_IRQMASK = 33,             /* Interrupt mask */
+
+   /* Legacy multi-monitor support */
+   SVGA_REG_NUM_GUEST_DISPLAYS = 34,/* Number of guest displays in X/Y direction */
+   SVGA_REG_DISPLAY_ID = 35,        /* Display ID for the following display attributes */
+   SVGA_REG_DISPLAY_IS_PRIMARY = 36,/* Whether this is a primary display */
+   SVGA_REG_DISPLAY_POSITION_X = 37,/* The display position x */
+   SVGA_REG_DISPLAY_POSITION_Y = 38,/* The display position y */
+   SVGA_REG_DISPLAY_WIDTH = 39,     /* The display's width */
+   SVGA_REG_DISPLAY_HEIGHT = 40,    /* The display's height */
+
+   /* See "Guest memory regions" below. */
+   SVGA_REG_GMR_ID = 41,
+   SVGA_REG_GMR_DESCRIPTOR = 42,
+   SVGA_REG_GMR_MAX_IDS = 43,
+   SVGA_REG_GMR_MAX_DESCRIPTOR_LENGTH = 44,
+
+   SVGA_REG_TRACES = 45,            /* Enable trace-based updates even when FIFO is on */
+   SVGA_REG_TOP = 46,               /* Must be 1 more than the last register */
+
+   SVGA_PALETTE_BASE = 1024,        /* Base of SVGA color map */
+   /* Next 768 (== 256*3) registers exist for colormap */
+
+   SVGA_SCRATCH_BASE = SVGA_PALETTE_BASE + SVGA_NUM_PALETTE_REGS
+                                    /* Base of scratch registers */
+   /* Next reg[SVGA_REG_SCRATCH_SIZE] registers exist for scratch usage:
+      First 4 are reserved for VESA BIOS Extension; any remaining are for
+      the use of the current SVGA driver. */
+};
+
+
+/*
+ * Guest memory regions (GMRs):
+ *
+ * This is a new memory mapping feature available in SVGA devices
+ * which have the SVGA_CAP_GMR bit set. Previously, there were two
+ * fixed memory regions available with which to share data between the
+ * device and the driver: the FIFO ('MEM') and the framebuffer. GMRs
+ * are our name for an extensible way of providing arbitrary DMA
+ * buffers for use between the driver and the SVGA device. They are a
+ * new alternative to framebuffer memory, usable for both 2D and 3D
+ * graphics operations.
+ *
+ * Since GMR mapping must be done synchronously with guest CPU
+ * execution, we use a new pair of SVGA registers:
+ *
+ *   SVGA_REG_GMR_ID --
+ *
+ *     Read/write.
+ *     This register holds the 32-bit ID (a small positive integer)
+ *     of a GMR to create, delete, or redefine. Writing this register
+ *     has no side-effects.
+ *
+ *   SVGA_REG_GMR_DESCRIPTOR --
+ *
+ *     Write-only.
+ *     Writing this register will create, delete, or redefine the GMR
+ *     specified by the above ID register. If this register is zero,
+ *     the GMR is deleted. Any pointers into this GMR (including those
+ *     currently being processed by FIFO commands) will be
+ *     synchronously invalidated.
+ *
+ *     If this register is nonzero, it must be the physical page
+ *     number (PPN) of a data structure which describes the physical
+ *     layout of the memory region this GMR should describe. The
+ *     descriptor structure will be read synchronously by the SVGA
+ *     device when this register is written. The descriptor need not
+ *     remain allocated for the lifetime of the GMR.
+ *
+ *     The guest driver should write SVGA_REG_GMR_ID first, then
+ *     SVGA_REG_GMR_DESCRIPTOR.
+ *
+ *   SVGA_REG_GMR_MAX_IDS --
+ *
+ *     Read-only.
+ *     The SVGA device may choose to support a maximum number of
+ *     user-defined GMR IDs. This register holds the number of supported
+ *     IDs. (The maximum supported ID plus 1)
+ *
+ *   SVGA_REG_GMR_MAX_DESCRIPTOR_LENGTH --
+ *
+ *     Read-only.
+ *     The SVGA device may choose to put a limit on the total number
+ *     of SVGAGuestMemDescriptor structures it will read when defining
+ *     a single GMR.
+ *
+ * The descriptor structure is an array of SVGAGuestMemDescriptor
+ * structures. Each structure may do one of three things:
+ *
+ *   - Terminate the GMR descriptor list.
+ *     (ppn==0, numPages==0)
+ *
+ *   - Add a PPN or range of PPNs to the GMR's virtual address space.
+ *     (ppn != 0, numPages != 0)
+ *
+ *   - Provide the PPN of the next SVGAGuestMemDescriptor, in order to
+ *     support multi-page GMR descriptor tables without forcing the
+ *     driver to allocate physically contiguous memory.
+ *     (ppn != 0, numPages == 0)
+ *
+ * Note that each physical page of SVGAGuestMemDescriptor structures
+ * can describe at least 2MB of guest memory. If the driver needs to
+ * use more than one page of descriptor structures, it must use one of
+ * its SVGAGuestMemDescriptors to point to an additional page.  The
+ * device will never automatically cross a page boundary.
+ *
+ * Once the driver has described a GMR, it is immediately available
+ * for use via any FIFO command that uses an SVGAGuestPtr structure.
+ * These pointers include a GMR identifier plus an offset into that
+ * GMR.
+ *
+ * The driver must check the SVGA_CAP_GMR bit before using the GMR
+ * registers.
+ */
+
+/*
+ * Special GMR IDs, allowing SVGAGuestPtrs to point to framebuffer
+ * memory as well.  In the future, these IDs could even be used to
+ * allow legacy memory regions to be redefined by the guest as GMRs.
+ *
+ * Using the guest framebuffer (GFB) at BAR1 for general purpose DMA
+ * is being phased out. Please try to use user-defined GMRs whenever
+ * possible.
+ */
+#define SVGA_GMR_NULL         ((uint32) -1)
+#define SVGA_GMR_FRAMEBUFFER  ((uint32) -2)  // Guest Framebuffer (GFB)
+
+typedef
+struct SVGAGuestMemDescriptor {
+   uint32 ppn;
+   uint32 numPages;
+} SVGAGuestMemDescriptor;
+
+typedef
+struct SVGAGuestPtr {
+   uint32 gmrId;
+   uint32 offset;
+} SVGAGuestPtr;
+
+
+/*
+ * SVGAGMRImageFormat --
+ *
+ *    This is a packed representation of the source 2D image format
+ *    for a GMR-to-screen blit. Currently it is defined as an encoding
+ *    of the screen's color depth and bits-per-pixel, however, 16 bits
+ *    are reserved for future use to identify other encodings (such as
+ *    RGBA or higher-precision images).
+ *
+ *    Currently supported formats:
+ *
+ *       bpp depth  Format Name
+ *       --- -----  -----------
+ *        32    24  32-bit BGRX
+ *        24    24  24-bit BGR
+ *        16    16  RGB 5-6-5
+ *        16    15  RGB 5-5-5
+ *
+ */
+
+typedef
+struct SVGAGMRImageFormat {
+   union {
+      struct {
+         uint32 bitsPerPixel : 8;
+         uint32 colorDepth   : 8;
+         uint32 reserved     : 16;  // Must be zero
+      };
+
+      uint32 value;
+   };
+} SVGAGMRImageFormat;
+
+/*
+ * SVGAColorBGRX --
+ *
+ *    A 24-bit color format (BGRX), which does not depend on the
+ *    format of the legacy guest framebuffer (GFB) or the current
+ *    GMRFB state.
+ */
+
+typedef
+struct SVGAColorBGRX {
+   union {
+      struct {
+         uint32 b : 8;
+         uint32 g : 8;
+         uint32 r : 8;
+         uint32 x : 8;  // Unused
+      };
+
+      uint32 value;
+   };
+} SVGAColorBGRX;
+
+
+/*
+ * SVGASignedRect --
+ * SVGASignedPoint --
+ *
+ *    Signed rectangle and point primitives. These are used by the new
+ *    2D primitives for drawing to Screen Objects, which can occupy a
+ *    signed virtual coordinate space.
+ *
+ *    SVGASignedRect specifies a half-open interval: the (left, top)
+ *    pixel is part of the rectangle, but the (right, bottom) pixel is
+ *    not.
+ */
+
+typedef
+struct SVGASignedRect {
+   int32  left;
+   int32  top;
+   int32  right;
+   int32  bottom;
+} SVGASignedRect;
+
+typedef
+struct SVGASignedPoint {
+   int32  x;
+   int32  y;
+} SVGASignedPoint;
+
+
+/*
+ *  Capabilities
+ *
+ *  Note the holes in the bitfield. Missing bits have been deprecated,
+ *  and must not be reused. Those capabilities will never be reported
+ *  by new versions of the SVGA device.
+ */
+
+#define SVGA_CAP_NONE               0x00000000
+#define SVGA_CAP_RECT_COPY          0x00000002
+#define SVGA_CAP_CURSOR             0x00000020
+#define SVGA_CAP_CURSOR_BYPASS      0x00000040   // Legacy (Use Cursor Bypass 3 instead)
+#define SVGA_CAP_CURSOR_BYPASS_2    0x00000080   // Legacy (Use Cursor Bypass 3 instead)
+#define SVGA_CAP_8BIT_EMULATION     0x00000100
+#define SVGA_CAP_ALPHA_CURSOR       0x00000200
+#define SVGA_CAP_3D                 0x00004000
+#define SVGA_CAP_EXTENDED_FIFO      0x00008000
+#define SVGA_CAP_MULTIMON           0x00010000   // Legacy multi-monitor support
+#define SVGA_CAP_PITCHLOCK          0x00020000
+#define SVGA_CAP_IRQMASK            0x00040000
+#define SVGA_CAP_DISPLAY_TOPOLOGY   0x00080000   // Legacy multi-monitor support
+#define SVGA_CAP_GMR                0x00100000
+#define SVGA_CAP_TRACES             0x00200000
+
+
+/*
+ * FIFO register indices.
+ *
+ * The FIFO is a chunk of device memory mapped into guest physmem.  It
+ * is always treated as 32-bit words.
+ *
+ * The guest driver gets to decide how to partition it between
+ * - FIFO registers (there are always at least 4, specifying where the
+ *   following data area is and how much data it contains; there may be
+ *   more registers following these, depending on the FIFO protocol
+ *   version in use)
+ * - FIFO data, written by the guest and slurped out by the VMX.
+ * These indices are 32-bit word offsets into the FIFO.
+ */
+
+enum {
+   /*
+    * Block 1 (basic registers): The originally defined FIFO registers.
+    * These exist and are valid for all versions of the FIFO protocol.
+    */
+
+   SVGA_FIFO_MIN = 0,
+   SVGA_FIFO_MAX,       /* The distance from MIN to MAX must be at least 10K */
+   SVGA_FIFO_NEXT_CMD,
+   SVGA_FIFO_STOP,
+
+   /*
+    * Block 2 (extended registers): Mandatory registers for the extended
+    * FIFO.  These exist if the SVGA caps register includes
+    * SVGA_CAP_EXTENDED_FIFO; some of them are valid only if their
+    * associated capability bit is enabled.
+    *
+    * Note that when originally defined, SVGA_CAP_EXTENDED_FIFO implied
+    * support only for (FIFO registers) CAPABILITIES, FLAGS, and FENCE.
+    * This means that the guest has to test individually (in most cases
+    * using FIFO caps) for the presence of registers after this; the VMX
+    * can define "extended FIFO" to mean whatever it wants, and currently
+    * won't enable it unless there's room for that set and much more.
+    */
+
+   SVGA_FIFO_CAPABILITIES = 4,
+   SVGA_FIFO_FLAGS,
+   // Valid with SVGA_FIFO_CAP_FENCE:
+   SVGA_FIFO_FENCE,
+
+   /*
+    * Block 3a (optional extended registers): Additional registers for the
+    * extended FIFO, whose presence isn't actually implied by
+    * SVGA_CAP_EXTENDED_FIFO; these exist if SVGA_FIFO_MIN is high enough to
+    * leave room for them.
+    *
+    * These in block 3a, the VMX currently considers mandatory for the
+    * extended FIFO.
+    */
+
+   // Valid if exists (i.e. if extended FIFO enabled):
+   SVGA_FIFO_3D_HWVERSION,       /* See SVGA3dHardwareVersion in svga3d_reg.h */
+   // Valid with SVGA_FIFO_CAP_PITCHLOCK:
+   SVGA_FIFO_PITCHLOCK,
+
+   // Valid with SVGA_FIFO_CAP_CURSOR_BYPASS_3:
+   SVGA_FIFO_CURSOR_ON,          /* Cursor bypass 3 show/hide register */
+   SVGA_FIFO_CURSOR_X,           /* Cursor bypass 3 x register */
+   SVGA_FIFO_CURSOR_Y,           /* Cursor bypass 3 y register */
+   SVGA_FIFO_CURSOR_COUNT,       /* Incremented when any of the other 3 change */
+   SVGA_FIFO_CURSOR_LAST_UPDATED,/* Last time the host updated the cursor */
+
+   // Valid with SVGA_FIFO_CAP_RESERVE:
+   SVGA_FIFO_RESERVED,           /* Bytes past NEXT_CMD with real contents */
+
+   /*
+    * Valid with SVGA_FIFO_CAP_SCREEN_OBJECT:
+    *
+    * By default this is SVGA_ID_INVALID, to indicate that the cursor
+    * coordinates are specified relative to the virtual root. If this
+    * is set to a specific screen ID, cursor position is reinterpreted
+    * as a signed offset relative to that screen's origin. This is the
+    * only way to place the cursor on a non-rooted screen.
+    */
+   SVGA_FIFO_CURSOR_SCREEN_ID,
+
+   /*
+    * XXX: The gap here, up until SVGA_FIFO_3D_CAPS, can be used for new
+    * registers, but this must be done carefully and with judicious use of
+    * capability bits, since comparisons based on SVGA_FIFO_MIN aren't
+    * enough to tell you whether the register exists: we've shipped drivers
+    * and products that used SVGA_FIFO_3D_CAPS but didn't know about some of
+    * the earlier ones.  The actual order of introduction was:
+    * - PITCHLOCK
+    * - 3D_CAPS
+    * - CURSOR_* (cursor bypass 3)
+    * - RESERVED
+    * So, code that wants to know whether it can use any of the
+    * aforementioned registers, or anything else added after PITCHLOCK and
+    * before 3D_CAPS, needs to reason about something other than
+    * SVGA_FIFO_MIN.
+    */
+
+   /*
+    * 3D caps block space; valid with 3D hardware version >=
+    * SVGA3D_HWVERSION_WS6_B1.
+    */
+   SVGA_FIFO_3D_CAPS      = 32,
+   SVGA_FIFO_3D_CAPS_LAST = 32 + 255,
+
+   /*
+    * End of VMX's current definition of "extended-FIFO registers".
+    * Registers before here are always enabled/disabled as a block; either
+    * the extended FIFO is enabled and includes all preceding registers, or
+    * it's disabled entirely.
+    *
+    * Block 3b (truly optional extended registers): Additional registers for
+    * the extended FIFO, which the VMX already knows how to enable and
+    * disable with correct granularity.
+    *
+    * Registers after here exist if and only if the guest SVGA driver
+    * sets SVGA_FIFO_MIN high enough to leave room for them.
+    */
+
+   // Valid if register exists:
+   SVGA_FIFO_GUEST_3D_HWVERSION, /* Guest driver's 3D version */
+   SVGA_FIFO_FENCE_GOAL,         /* Matching target for SVGA_IRQFLAG_FENCE_GOAL */
+   SVGA_FIFO_BUSY,               /* See "FIFO Synchronization Registers" */
+
+   /*
+    * Always keep this last.  This defines the maximum number of
+    * registers we know about.  At power-on, this value is placed in
+    * the SVGA_REG_MEM_REGS register, and we expect the guest driver
+    * to allocate this much space in FIFO memory for registers.
+    */
+    SVGA_FIFO_NUM_REGS
+};
+
+
+/*
+ * Definition of registers included in extended FIFO support.
+ *
+ * The guest SVGA driver gets to allocate the FIFO between registers
+ * and data.  It must always allocate at least 4 registers, but old
+ * drivers stopped there.
+ *
+ * The VMX will enable extended FIFO support if and only if the guest
+ * left enough room for all registers defined as part of the mandatory
+ * set for the extended FIFO.
+ *
+ * Note that the guest drivers typically allocate the FIFO only at
+ * initialization time, not at mode switches, so it's likely that the
+ * number of FIFO registers won't change without a reboot.
+ *
+ * All registers less than this value are guaranteed to be present if
+ * svgaUser->fifo.extended is set. Any later registers must be tested
+ * individually for compatibility at each use (in the VMX).
+ *
+ * This value is used only by the VMX, so it can change without
+ * affecting driver compatibility; keep it that way?
+ */
+#define SVGA_FIFO_EXTENDED_MANDATORY_REGS  (SVGA_FIFO_3D_CAPS_LAST + 1)
+
+
+/*
+ * FIFO Synchronization Registers
+ *
+ *  This explains the relationship between the various FIFO
+ *  sync-related registers in IOSpace and in FIFO space.
+ *
+ *  SVGA_REG_SYNC --
+ *
+ *       The SYNC register can be used in two different ways by the guest:
+ *
+ *         1. If the guest wishes to fully sync (drain) the FIFO,
+ *            it will write once to SYNC then poll on the BUSY
+ *            register. The FIFO is sync'ed once BUSY is zero.
+ *
+ *         2. If the guest wants to asynchronously wake up the host,
+ *            it will write once to SYNC without polling on BUSY.
+ *            Ideally it will do this after some new commands have
+ *            been placed in the FIFO, and after reading a zero
+ *            from SVGA_FIFO_BUSY.
+ *
+ *       (1) is the original behaviour that SYNC was designed to
+ *       support.  Originally, a write to SYNC would implicitly
+ *       trigger a read from BUSY. This causes us to synchronously
+ *       process the FIFO.
+ *
+ *       This behaviour has since been changed so that writing SYNC
+ *       will *not* implicitly cause a read from BUSY. Instead, it
+ *       makes a channel call which asynchronously wakes up the MKS
+ *       thread.
+ *
+ *       New guests can use this new behaviour to implement (2)
+ *       efficiently. This lets guests get the host's attention
+ *       without waiting for the MKS to poll, which gives us much
+ *       better CPU utilization on SMP hosts and on UP hosts while
+ *       we're blocked on the host GPU.
+ *
+ *       Old guests shouldn't notice the behaviour change. SYNC was
+ *       never guaranteed to process the entire FIFO, since it was
+ *       bounded to a particular number of CPU cycles. Old guests will
+ *       still loop on the BUSY register until the FIFO is empty.
+ *
+ *       Writing to SYNC currently has the following side-effects:
+ *
+ *         - Sets SVGA_REG_BUSY to TRUE (in the monitor)
+ *         - Asynchronously wakes up the MKS thread for FIFO processing
+ *         - The value written to SYNC is recorded as a "reason", for
+ *           stats purposes.
+ *
+ *       If SVGA_FIFO_BUSY is available, drivers are advised to only
+ *       write to SYNC if SVGA_FIFO_BUSY is FALSE. Drivers should set
+ *       SVGA_FIFO_BUSY to TRUE after writing to SYNC. The MKS will
+ *       eventually set SVGA_FIFO_BUSY on its own, but this approach
+ *       lets the driver avoid sending multiple asynchronous wakeup
+ *       messages to the MKS thread.
+ *
+ *  SVGA_REG_BUSY --
+ *
+ *       This register is set to TRUE when SVGA_REG_SYNC is written,
+ *       and it reads as FALSE when the FIFO has been completely
+ *       drained.
+ *
+ *       Every read from this register causes us to synchronously
+ *       process FIFO commands. There is no guarantee as to how many
+ *       commands each read will process.
+ *
+ *       CPU time spent processing FIFO commands will be billed to
+ *       the guest.
+ *
+ *       New drivers should avoid using this register unless they
+ *       need to guarantee that the FIFO is completely drained. It
+ *       is overkill for performing a sync-to-fence. Older drivers
+ *       will use this register for any type of synchronization.
+ *
+ *  SVGA_FIFO_BUSY --
+ *
+ *       This register is a fast way for the guest driver to check
+ *       whether the FIFO is already being processed. It reads and
+ *       writes at normal RAM speeds, with no monitor intervention.
+ *
+ *       If this register reads as TRUE, the host is guaranteeing that
+ *       any new commands written into the FIFO will be noticed before
+ *       the MKS goes back to sleep.
+ *
+ *       If this register reads as FALSE, no such guarantee can be
+ *       made.
+ *
+ *       The guest should use this register to quickly determine
+ *       whether or not it needs to wake up the host. If the guest
+ *       just wrote a command or group of commands that it would like
+ *       the host to begin processing, it should:
+ *
+ *         1. Read SVGA_FIFO_BUSY. If it reads as TRUE, no further
+ *            action is necessary.
+ *
+ *         2. Write TRUE to SVGA_FIFO_BUSY. This informs future guest
+ *            code that we've already sent a SYNC to the host and we
+ *            don't need to send a duplicate.
+ *
+ *         3. Write a reason to SVGA_REG_SYNC. This will send an
+ *            asynchronous wakeup to the MKS thread.
+ */
+
+
+/*
+ * FIFO Capabilities
+ *
+ *      Fence -- Fence register and command are supported
+ *      Accel Front -- Front buffer only commands are supported
+ *      Pitch Lock -- Pitch lock register is supported
+ *      Video -- SVGA Video overlay units are supported
+ *      Escape -- Escape command is supported
+ *
+ * XXX: Add longer descriptions for each capability, including a list
+ *      of the new features that each capability provides.
+ *
+ * SVGA_FIFO_CAP_SCREEN_OBJECT --
+ *
+ *    Provides dynamic multi-screen rendering, for improved Unity and
+ *    multi-monitor modes. With Screen Object, the guest can
+ *    dynamically create and destroy 'screens', which can represent
+ *    Unity windows or virtual monitors. Screen Object also provides
+ *    strong guarantees that DMA operations happen only when
+ *    guest-initiated. Screen Object deprecates the BAR1 guest
+ *    framebuffer (GFB) and all commands that work only with the GFB.
+ *
+ *    New registers:
+ *       FIFO_CURSOR_SCREEN_ID, VIDEO_DATA_GMRID, VIDEO_DST_SCREEN_ID
+ *
+ *    New 2D commands:
+ *       DEFINE_SCREEN, DESTROY_SCREEN, DEFINE_GMRFB, BLIT_GMRFB_TO_SCREEN,
+ *       BLIT_SCREEN_TO_GMRFB, ANNOTATION_FILL, ANNOTATION_COPY
+ *
+ *    New 3D commands:
+ *       BLIT_SURFACE_TO_SCREEN
+ *
+ *    New guarantees:
+ *
+ *       - The host will not read or write guest memory, including the GFB,
+ *         except when explicitly initiated by a DMA command.
+ *
+ *       - All DMA, including legacy DMA like UPDATE and PRESENT_READBACK,
+ *         is guaranteed to complete before any subsequent FENCEs.
+ *
+ *       - All legacy commands which affect a Screen (UPDATE, PRESENT,
+ *         PRESENT_READBACK) as well as new Screen blit commands will
+ *         all behave consistently as blits, and memory will be read
+ *         or written in FIFO order.
+ *
+ *         For example, if you PRESENT from one SVGA3D surface to multiple
+ *         places on the screen, the data copied will always be from the
+ *         SVGA3D surface at the time the PRESENT was issued in the FIFO.
+ *         This was not necessarily true on devices without Screen Object.
+ *
+ *         This means that on devices that support Screen Object, the
+ *         PRESENT_READBACK command should not be necessary unless you
+ *         actually want to read back the results of 3D rendering into
+ *         system memory. (And for that, the BLIT_SCREEN_TO_GMRFB
+ *         command provides a strict superset of functionality.)
+ *
+ *       - When a screen is resized, either using Screen Object commands or
+ *         legacy multimon registers, its contents are preserved.
+ */
+
+#define SVGA_FIFO_CAP_NONE                  0
+#define SVGA_FIFO_CAP_FENCE             (1<<0)
+#define SVGA_FIFO_CAP_ACCELFRONT        (1<<1)
+#define SVGA_FIFO_CAP_PITCHLOCK         (1<<2)
+#define SVGA_FIFO_CAP_VIDEO             (1<<3)
+#define SVGA_FIFO_CAP_CURSOR_BYPASS_3   (1<<4)
+#define SVGA_FIFO_CAP_ESCAPE            (1<<5)
+#define SVGA_FIFO_CAP_RESERVE           (1<<6)
+#define SVGA_FIFO_CAP_SCREEN_OBJECT     (1<<7)
+
+
+/*
+ * FIFO Flags
+ *
+ *      Accel Front -- Driver should use front buffer only commands
+ */
+
+#define SVGA_FIFO_FLAG_NONE                 0
+#define SVGA_FIFO_FLAG_ACCELFRONT       (1<<0)
+#define SVGA_FIFO_FLAG_RESERVED        (1<<31) // Internal use only
+
+/*
+ * FIFO reservation sentinel value
+ */
+
+#define SVGA_FIFO_RESERVED_UNKNOWN      0xffffffff
+
+
+/*
+ * Video overlay support
+ */
+
+#define SVGA_NUM_OVERLAY_UNITS 32
+
+
+/*
+ * Video capabilities that the guest is currently using
+ */
+
+#define SVGA_VIDEO_FLAG_COLORKEY        0x0001
+
+
+/*
+ * Offsets for the video overlay registers
+ */
+
+enum {
+   SVGA_VIDEO_ENABLED = 0,
+   SVGA_VIDEO_FLAGS,
+   SVGA_VIDEO_DATA_OFFSET,
+   SVGA_VIDEO_FORMAT,
+   SVGA_VIDEO_COLORKEY,
+   SVGA_VIDEO_SIZE,          // Deprecated
+   SVGA_VIDEO_WIDTH,
+   SVGA_VIDEO_HEIGHT,
+   SVGA_VIDEO_SRC_X,
+   SVGA_VIDEO_SRC_Y,
+   SVGA_VIDEO_SRC_WIDTH,
+   SVGA_VIDEO_SRC_HEIGHT,
+   SVGA_VIDEO_DST_X,         // Signed int32
+   SVGA_VIDEO_DST_Y,         // Signed int32
+   SVGA_VIDEO_DST_WIDTH,
+   SVGA_VIDEO_DST_HEIGHT,
+   SVGA_VIDEO_PITCH_1,
+   SVGA_VIDEO_PITCH_2,
+   SVGA_VIDEO_PITCH_3,
+   SVGA_VIDEO_DATA_GMRID,    // Optional, defaults to SVGA_GMR_FRAMEBUFFER
+   SVGA_VIDEO_DST_SCREEN_ID, // Optional, defaults to virtual coords (SVGA_ID_INVALID)
+   SVGA_VIDEO_NUM_REGS
+};
+
+
+/*
+ * SVGA Overlay Units
+ *
+ *      width and height relate to the entire source video frame.
+ *      srcX, srcY, srcWidth and srcHeight represent subset of the source
+ *      video frame to be displayed.
+ */
+
+typedef struct SVGAOverlayUnit {
+   uint32 enabled;
+   uint32 flags;
+   uint32 dataOffset;
+   uint32 format;
+   uint32 colorKey;
+   uint32 size;
+   uint32 width;
+   uint32 height;
+   uint32 srcX;
+   uint32 srcY;
+   uint32 srcWidth;
+   uint32 srcHeight;
+   int32  dstX;
+   int32  dstY;
+   uint32 dstWidth;
+   uint32 dstHeight;
+   uint32 pitches[3];
+   uint32 dataGMRId;
+   uint32 dstScreenId;
+} SVGAOverlayUnit;
+
+
+/*
+ * SVGAScreenObject --
+ *
+ *    This is a new way to represent a guest's multi-monitor screen or
+ *    Unity window. Screen objects are only supported if the
+ *    SVGA_FIFO_CAP_SCREEN_OBJECT capability bit is set.
+ *
+ *    If Screen Objects are supported, they can be used to fully
+ *    replace the functionality provided by the framebuffer registers
+ *    (SVGA_REG_WIDTH, HEIGHT, etc.) and by SVGA_CAP_DISPLAY_TOPOLOGY.
+ *
+ *    The screen object is a struct with guaranteed binary
+ *    compatibility. New flags can be added, and the struct may grow,
+ *    but existing fields must retain their meaning.
+ *
+ */
+
+#define SVGA_SCREEN_HAS_ROOT    (1 << 0)  // Screen is present in the virtual coord space
+#define SVGA_SCREEN_IS_PRIMARY  (1 << 1)  // Guest considers this screen to be 'primary'
+#define SVGA_SCREEN_FULLSCREEN_HINT (1 << 2)   // Guest is running a fullscreen app here
+
+typedef
+struct SVGAScreenObject {
+   uint32 structSize;   // sizeof(SVGAScreenObject)
+   uint32 id;
+   uint32 flags;
+   struct {
+      uint32 width;
+      uint32 height;
+   } size;
+   struct {
+      int32 x;
+      int32 y;
+   } root;              // Only used if SVGA_SCREEN_HAS_ROOT is set.
+} SVGAScreenObject;
+
+
+/*
+ *  Commands in the command FIFO:
+ *
+ *  Command IDs defined below are used for the traditional 2D FIFO
+ *  communication (not all commands are available for all versions of the
+ *  SVGA FIFO protocol).
+ *
+ *  Note the holes in the command ID numbers: These commands have been
+ *  deprecated, and the old IDs must not be reused.
+ *
+ *  Command IDs from 1000 to 1999 are reserved for use by the SVGA3D
+ *  protocol.
+ *
+ *  Each command's parameters are described by the comments and
+ *  structs below.
+ */
+
+typedef enum {
+   SVGA_CMD_INVALID_CMD           = 0,
+   SVGA_CMD_UPDATE                = 1,
+   SVGA_CMD_RECT_COPY             = 3,
+   SVGA_CMD_DEFINE_CURSOR         = 19,
+   SVGA_CMD_DEFINE_ALPHA_CURSOR   = 22,
+   SVGA_CMD_UPDATE_VERBOSE        = 25,
+   SVGA_CMD_FRONT_ROP_FILL        = 29,
+   SVGA_CMD_FENCE                 = 30,
+   SVGA_CMD_ESCAPE                = 33,
+   SVGA_CMD_DEFINE_SCREEN         = 34,
+   SVGA_CMD_DESTROY_SCREEN        = 35,
+   SVGA_CMD_DEFINE_GMRFB          = 36,
+   SVGA_CMD_BLIT_GMRFB_TO_SCREEN  = 37,
+   SVGA_CMD_BLIT_SCREEN_TO_GMRFB  = 38,
+   SVGA_CMD_ANNOTATION_FILL       = 39,
+   SVGA_CMD_ANNOTATION_COPY       = 40,
+   SVGA_CMD_MAX
+} SVGAFifoCmdId;
+
+#define SVGA_CMD_MAX_ARGS           64
+
+
+/*
+ * SVGA_CMD_UPDATE --
+ *
+ *    This is a DMA transfer which copies from the Guest Framebuffer
+ *    (GFB) at BAR1 + SVGA_REG_FB_OFFSET to any screens which
+ *    intersect with the provided virtual rectangle.
+ *
+ *    This command does not support using arbitrary guest memory as a
+ *    data source- it only works with the pre-defined GFB memory.
+ *    This command also does not support signed virtual coordinates.
+ *    If you have defined screens (using SVGA_CMD_DEFINE_SCREEN) with
+ *    negative root x/y coordinates, the negative portion of those
+ *    screens will not be reachable by this command.
+ *
+ *    This command is not necessary when using framebuffer
+ *    traces. Traces are automatically enabled if the SVGA FIFO is
+ *    disabled, and you may explicitly enable/disable traces using
+ *    SVGA_REG_TRACES. With traces enabled, any write to the GFB will
+ *    automatically act as if a subsequent SVGA_CMD_UPDATE was issued.
+ *
+ *    Traces and SVGA_CMD_UPDATE are the only supported ways to render
+ *    pseudocolor screen updates. The newer Screen Object commands
+ *    only support true color formats.
+ *
+ * Availability:
+ *    Always available.
+ */
+
+typedef
+struct {
+   uint32 x;
+   uint32 y;
+   uint32 width;
+   uint32 height;
+} SVGAFifoCmdUpdate;
+
+
+/*
+ * SVGA_CMD_RECT_COPY --
+ *
+ *    Perform a rectangular DMA transfer from one area of the GFB to
+ *    another, and copy the result to any screens which intersect it.
+ *
+ * Availability:
+ *    SVGA_CAP_RECT_COPY
+ */
+
+typedef
+struct {
+   uint32 srcX;
+   uint32 srcY;
+   uint32 destX;
+   uint32 destY;
+   uint32 width;
+   uint32 height;
+} SVGAFifoCmdRectCopy;
+
+
+/*
+ * SVGA_CMD_DEFINE_CURSOR --
+ *
+ *    Provide a new cursor image, as an AND/XOR mask.
+ *
+ *    The recommended way to position the cursor overlay is by using
+ *    the SVGA_FIFO_CURSOR_* registers, supported by the
+ *    SVGA_FIFO_CAP_CURSOR_BYPASS_3 capability.
+ *
+ * Availability:
+ *    SVGA_CAP_CURSOR
+ */
+
+typedef
+struct {
+   uint32 id;             // Reserved, must be zero.
+   uint32 hotspotX;
+   uint32 hotspotY;
+   uint32 width;
+   uint32 height;
+   uint32 andMaskDepth;   // Value must be 1 or equal to BITS_PER_PIXEL
+   uint32 xorMaskDepth;   // Value must be 1 or equal to BITS_PER_PIXEL
+   /*
+    * Followed by scanline data for AND mask, then XOR mask.
+    * Each scanline is padded to a 32-bit boundary.
+   */
+} SVGAFifoCmdDefineCursor;
+
+
+/*
+ * SVGA_CMD_DEFINE_ALPHA_CURSOR --
+ *
+ *    Provide a new cursor image, in 32-bit BGRA format.
+ *
+ *    The recommended way to position the cursor overlay is by using
+ *    the SVGA_FIFO_CURSOR_* registers, supported by the
+ *    SVGA_FIFO_CAP_CURSOR_BYPASS_3 capability.
+ *
+ * Availability:
+ *    SVGA_CAP_ALPHA_CURSOR
+ */
+
+typedef
+struct {
+   uint32 id;             // Reserved, must be zero.
+   uint32 hotspotX;
+   uint32 hotspotY;
+   uint32 width;
+   uint32 height;
+   /* Followed by scanline data */
+} SVGAFifoCmdDefineAlphaCursor;
+
+
+/*
+ * SVGA_CMD_UPDATE_VERBOSE --
+ *
+ *    Just like SVGA_CMD_UPDATE, but also provide a per-rectangle
+ *    'reason' value, an opaque cookie which is used by internal
+ *    debugging tools. Third party drivers should not use this
+ *    command.
+ *
+ * Availability:
+ *    SVGA_CAP_EXTENDED_FIFO
+ */
+
+typedef
+struct {
+   uint32 x;
+   uint32 y;
+   uint32 width;
+   uint32 height;
+   uint32 reason;
+} SVGAFifoCmdUpdateVerbose;
+
+
+/*
+ * SVGA_CMD_FRONT_ROP_FILL --
+ *
+ *    This is a hint which tells the SVGA device that the driver has
+ *    just filled a rectangular region of the GFB with a solid
+ *    color. Instead of reading these pixels from the GFB, the device
+ *    can assume that they all equal 'color'. This is primarily used
+ *    for remote desktop protocols.
+ *
+ * Availability:
+ *    SVGA_FIFO_CAP_ACCELFRONT
+ */
+
+#define  SVGA_ROP_COPY                    0x03
+
+typedef
+struct {
+   uint32 color;     // In the same format as the GFB
+   uint32 x;
+   uint32 y;
+   uint32 width;
+   uint32 height;
+   uint32 rop;       // Must be SVGA_ROP_COPY
+} SVGAFifoCmdFrontRopFill;
+
+
+/*
+ * SVGA_CMD_FENCE --
+ *
+ *    Insert a synchronization fence.  When the SVGA device reaches
+ *    this command, it will copy the 'fence' value into the
+ *    SVGA_FIFO_FENCE register. It will also compare the fence against
+ *    SVGA_FIFO_FENCE_GOAL. If the fence matches the goal and the
+ *    SVGA_IRQFLAG_FENCE_GOAL interrupt is enabled, the device will
+ *    raise this interrupt.
+ *
+ * Availability:
+ *    SVGA_FIFO_FENCE for this command,
+ *    SVGA_CAP_IRQMASK for SVGA_FIFO_FENCE_GOAL.
+ */
+
+typedef
+struct {
+   uint32 fence;
+} SVGAFifoCmdFence;
+
+
+/*
+ * SVGA_CMD_ESCAPE --
+ *
+ *    Send an extended or vendor-specific variable length command.
+ *    This is used for video overlay, third party plugins, and
+ *    internal debugging tools. See svga_escape.h
+ *
+ * Availability:
+ *    SVGA_FIFO_CAP_ESCAPE
+ */
+
+typedef
+struct {
+   uint32 nsid;
+   uint32 size;
+   /* followed by 'size' bytes of data */
+} SVGAFifoCmdEscape;
+
+
+/*
+ * SVGA_CMD_DEFINE_SCREEN --
+ *
+ *    Define or redefine an SVGAScreenObject. See the description of
+ *    SVGAScreenObject above.  The video driver is responsible for
+ *    generating new screen IDs. They should be small positive
+ *    integers. The virtual device will have an implementation
+ *    specific upper limit on the number of screen IDs
+ *    supported. Drivers are responsible for recycling IDs. The first
+ *    valid ID is zero.
+ *
+ *    - Interaction with other registers:
+ *
+ *    For backwards compatibility, when the GFB mode registers (WIDTH,
+ *    HEIGHT, PITCHLOCK, BITS_PER_PIXEL) are modified, the SVGA device
+ *    deletes all screens other than screen #0, and redefines screen
+ *    #0 according to the specified mode. Drivers that use
+ *    SVGA_CMD_DEFINE_SCREEN should destroy or redefine screen #0.
+ *
+ *    If you use screen objects, do not use the legacy multi-mon
+ *    registers (SVGA_REG_NUM_GUEST_DISPLAYS, SVGA_REG_DISPLAY_*).
+ *
+ * Availability:
+ *    SVGA_FIFO_CAP_SCREEN_OBJECT
+ */
+
+typedef
+struct {
+   SVGAScreenObject screen;   // Variable-length according to version
+} SVGAFifoCmdDefineScreen;
+
+
+/*
+ * SVGA_CMD_DESTROY_SCREEN --
+ *
+ *    Destroy an SVGAScreenObject. Its ID is immediately available for
+ *    re-use.
+ *
+ * Availability:
+ *    SVGA_FIFO_CAP_SCREEN_OBJECT
+ */
+
+typedef
+struct {
+   uint32 screenId;
+} SVGAFifoCmdDestroyScreen;
+
+
+/*
+ * SVGA_CMD_DEFINE_GMRFB --
+ *
+ *    This command sets a piece of SVGA device state called the
+ *    Guest Memory Region Framebuffer, or GMRFB. The GMRFB is a
+ *    piece of light-weight state which identifies the location and
+ *    format of an image in guest memory or in BAR1. The GMRFB has
+ *    an arbitrary size, and it doesn't need to match the geometry
+ *    of the GFB or any screen object.
+ *
+ *    The GMRFB can be redefined as often as you like. You could
+ *    always use the same GMRFB, you could redefine it before
+ *    rendering from a different guest screen, or you could even
+ *    redefine it before every blit.
+ *
+ *    There are multiple ways to use this command. The simplest way is
+ *    to use it to move the framebuffer either to elsewhere in the GFB
+ *    (BAR1) memory region, or to a user-defined GMR. This lets a
+ *    driver use a framebuffer allocated entirely out of normal system
+ *    memory, which we encourage.
+ *
+ *    Another way to use this command is to set up a ring buffer of
+ *    updates in GFB memory. If a driver wants to ensure that no
+ *    frames are skipped by the SVGA device, it is important that the
+ *    driver not modify the source data for a blit until the device is
+ *    done processing the command. One efficient way to accomplish
+ *    this is to use a ring of small DMA buffers. Each buffer is used
+ *    for one blit, then we move on to the next buffer in the
+ *    ring. The FENCE mechanism is used to protect each buffer from
+ *    re-use until the device is finished with that buffer's
+ *    corresponding blit.
+ *
+ *    This command does not affect the meaning of SVGA_CMD_UPDATE.
+ *    UPDATEs always occur from the legacy GFB memory area. This
+ *    command has no support for pseudocolor GMRFBs. Currently only
+ *    true-color 15, 16, and 24-bit depths are supported. Future
+ *    devices may expose capabilities for additional framebuffer
+ *    formats.
+ *
+ *    The default GMRFB value is undefined. Drivers must always send
+ *    this command at least once before performing any blit from the
+ *    GMRFB.
+ *
+ * Availability:
+ *    SVGA_FIFO_CAP_SCREEN_OBJECT
+ */
+
+typedef
+struct {
+   SVGAGuestPtr        ptr;
+   uint32              bytesPerLine;
+   SVGAGMRImageFormat  format;
+} SVGAFifoCmdDefineGMRFB;
+
+
+/*
+ * SVGA_CMD_BLIT_GMRFB_TO_SCREEN --
+ *
+ *    This is a guest-to-host blit. It performs a DMA operation to
+ *    copy a rectangular region of pixels from the current GMRFB to
+ *    one or more Screen Objects.
+ *
+ *    The destination coordinate may be specified relative to a
+ *    screen's origin (if a screen ID is specified) or relative to the
+ *    virtual coordinate system's origin (if the screen ID is
+ *    SVGA_ID_INVALID). The actual destination may span zero or more
+ *    screens, in the case of a virtual destination rect or a rect
+ *    which extends off the edge of the specified screen.
+ *
+ *    This command writes to the screen's "base layer": the underlying
+ *    framebuffer which exists below any cursor or video overlays. No
+ *    action is necessary to explicitly hide or update any overlays
+ *    which exist on top of the updated region.
+ *
+ *    The SVGA device is guaranteed to finish reading from the GMRFB
+ *    by the time any subsequent FENCE commands are reached.
+ *
+ *    This command consumes an annotation. See the
+ *    SVGA_CMD_ANNOTATION_* commands for details.
+ *
+ * Availability:
+ *    SVGA_FIFO_CAP_SCREEN_OBJECT
+ */
+
+typedef
+struct {
+   SVGASignedPoint  srcOrigin;
+   SVGASignedRect   destRect;
+   uint32           destScreenId;
+} SVGAFifoCmdBlitGMRFBToScreen;
+
+
+/*
+ * SVGA_CMD_BLIT_SCREEN_TO_GMRFB --
+ *
+ *    This is a host-to-guest blit. It performs a DMA operation to
+ *    copy a rectangular region of pixels from a single Screen Object
+ *    back to the current GMRFB.
+ *
+ *    Usage note: This command should be used rarely. It will
+ *    typically be inefficient, but it is necessary for some types of
+ *    synchronization between 3D (GPU) and 2D (CPU) rendering into
+ *    overlapping areas of a screen.
+ *
+ *    The source coordinate is specified relative to a screen's
+ *    origin. The provided screen ID must be valid. If any parameters
+ *    are invalid, the resulting pixel values are undefined.
+ *
+ *    This command reads the screen's "base layer". Overlays like
+ *    video and cursor are not included, but any data which was sent
+ *    using a blit-to-screen primitive will be available, no matter
+ *    whether the data's original source was the GMRFB or the 3D
+ *    acceleration hardware.
+ *
+ *    Note that our guest-to-host blits and host-to-guest blits aren't
+ *    symmetric in their current implementation. While the parameters
+ *    are identical, host-to-guest blits are a lot less featureful.
+ *    They do not support clipping: If the source parameters don't
+ *    fully fit within a screen, the blit fails. They must originate
+ *    from exactly one screen. Virtual coordinates are not directly
+ *    supported.
+ *
+ *    Host-to-guest blits do support the same set of GMRFB formats
+ *    offered by guest-to-host blits.
+ *
+ *    The SVGA device is guaranteed to finish writing to the GMRFB by
+ *    the time any subsequent FENCE commands are reached.
+ *
+ * Availability:
+ *    SVGA_FIFO_CAP_SCREEN_OBJECT
+ */
+
+typedef
+struct {
+   SVGASignedPoint  destOrigin;
+   SVGASignedRect   srcRect;
+   uint32           srcScreenId;
+} SVGAFifoCmdBlitScreenToGMRFB;
+
+
+/*
+ * SVGA_CMD_ANNOTATION_FILL --
+ *
+ *    This is a blit annotation. This command stores a small piece of
+ *    device state which is consumed by the next blit-to-screen
+ *    command. The state is only cleared by commands which are
+ *    specifically documented as consuming an annotation. Other
+ *    commands (such as ESCAPEs for debugging) may intervene between
+ *    the annotation and its associated blit.
+ *
+ *    This annotation is a promise about the contents of the next
+ *    blit: The video driver is guaranteeing that all pixels in that
+ *    blit will have the same value, specified here as a color in
+ *    SVGAColorBGRX format.
+ *
+ *    The SVGA device can still render the blit correctly even if it
+ *    ignores this annotation, but the annotation may allow it to
+ *    perform the blit more efficiently, for example by ignoring the
+ *    source data and performing a fill in hardware.
+ *
+ *    This annotation is most important for performance when the
+ *    user's display is being remoted over a network connection.
+ *
+ * Availability:
+ *    SVGA_FIFO_CAP_SCREEN_OBJECT
+ */
+
+typedef
+struct {
+   SVGAColorBGRX  color;
+} SVGAFifoCmdAnnotationFill;
+
+
+/*
+ * SVGA_CMD_ANNOTATION_COPY --
+ *
+ *    This is a blit annotation. See SVGA_CMD_ANNOTATION_FILL for more
+ *    information about annotations.
+ *
+ *    This annotation is a promise about the contents of the next
+ *    blit: The video driver is guaranteeing that all pixels in that
+ *    blit will have the same value as those which already exist at an
+ *    identically-sized region on the same or a different screen.
+ *
+ *    Note that the source pixels for the COPY in this annotation are
+ *    sampled before applying the anqnotation's associated blit. They
+ *    are allowed to overlap with the blit's destination pixels.
+ *
+ *    The copy source rectangle is specified the same way as the blit
+ *    destination: it can be a rectangle which spans zero or more
+ *    screens, specified relative to either a screen or to the virtual
+ *    coordinate system's origin. If the source rectangle includes
+ *    pixels which are not from exactly one screen, the results are
+ *    undefined.
+ *
+ * Availability:
+ *    SVGA_FIFO_CAP_SCREEN_OBJECT
+ */
+
+typedef
+struct {
+   SVGASignedPoint  srcOrigin;
+   uint32           srcScreenId;
+} SVGAFifoCmdAnnotationCopy;
+
+#endif
diff --git a/drivers/gpu/drm/vmwgfx/svga_types.h b/drivers/gpu/drm/vmwgfx/svga_types.h
new file mode 100644
index 000000000000..55836dedcfc2
--- /dev/null
+++ b/drivers/gpu/drm/vmwgfx/svga_types.h
@@ -0,0 +1,45 @@
+/**************************************************************************
+ *
+ * Copyright © 2009 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * Silly typedefs for the svga headers. Currently the headers are shared
+ * between all components that talk to svga. And as such the headers are
+ * are in a completely different style and use weird defines.
+ *
+ * This file lets all the ugly be prefixed with svga*.
+ */
+
+#ifndef _SVGA_TYPES_H_
+#define _SVGA_TYPES_H_
+
+typedef uint16_t uint16;
+typedef uint32_t uint32;
+typedef uint8_t uint8;
+typedef int32_t int32;
+typedef bool Bool;
+
+#endif
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_buffer.c b/drivers/gpu/drm/vmwgfx/vmwgfx_buffer.c
new file mode 100644
index 000000000000..825ebe3d89d5
--- /dev/null
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_buffer.c
@@ -0,0 +1,252 @@
+/**************************************************************************
+ *
+ * Copyright © 2009 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "vmwgfx_drv.h"
+#include "ttm/ttm_bo_driver.h"
+#include "ttm/ttm_placement.h"
+
+static uint32_t vram_placement_flags = TTM_PL_FLAG_VRAM |
+	TTM_PL_FLAG_CACHED;
+
+static uint32_t vram_ne_placement_flags = TTM_PL_FLAG_VRAM |
+	TTM_PL_FLAG_CACHED |
+	TTM_PL_FLAG_NO_EVICT;
+
+static uint32_t sys_placement_flags = TTM_PL_FLAG_SYSTEM |
+	TTM_PL_FLAG_CACHED;
+
+struct ttm_placement vmw_vram_placement = {
+	.fpfn = 0,
+	.lpfn = 0,
+	.num_placement = 1,
+	.placement = &vram_placement_flags,
+	.num_busy_placement = 1,
+	.busy_placement = &vram_placement_flags
+};
+
+struct ttm_placement vmw_vram_sys_placement = {
+	.fpfn = 0,
+	.lpfn = 0,
+	.num_placement = 1,
+	.placement = &vram_placement_flags,
+	.num_busy_placement = 1,
+	.busy_placement = &sys_placement_flags
+};
+
+struct ttm_placement vmw_vram_ne_placement = {
+	.fpfn = 0,
+	.lpfn = 0,
+	.num_placement = 1,
+	.placement = &vram_ne_placement_flags,
+	.num_busy_placement = 1,
+	.busy_placement = &vram_ne_placement_flags
+};
+
+struct ttm_placement vmw_sys_placement = {
+	.fpfn = 0,
+	.lpfn = 0,
+	.num_placement = 1,
+	.placement = &sys_placement_flags,
+	.num_busy_placement = 1,
+	.busy_placement = &sys_placement_flags
+};
+
+struct vmw_ttm_backend {
+	struct ttm_backend backend;
+};
+
+static int vmw_ttm_populate(struct ttm_backend *backend,
+			    unsigned long num_pages, struct page **pages,
+			    struct page *dummy_read_page)
+{
+	return 0;
+}
+
+static int vmw_ttm_bind(struct ttm_backend *backend, struct ttm_mem_reg *bo_mem)
+{
+	return 0;
+}
+
+static int vmw_ttm_unbind(struct ttm_backend *backend)
+{
+	return 0;
+}
+
+static void vmw_ttm_clear(struct ttm_backend *backend)
+{
+}
+
+static void vmw_ttm_destroy(struct ttm_backend *backend)
+{
+	struct vmw_ttm_backend *vmw_be =
+	    container_of(backend, struct vmw_ttm_backend, backend);
+
+	kfree(vmw_be);
+}
+
+static struct ttm_backend_func vmw_ttm_func = {
+	.populate = vmw_ttm_populate,
+	.clear = vmw_ttm_clear,
+	.bind = vmw_ttm_bind,
+	.unbind = vmw_ttm_unbind,
+	.destroy = vmw_ttm_destroy,
+};
+
+struct ttm_backend *vmw_ttm_backend_init(struct ttm_bo_device *bdev)
+{
+	struct vmw_ttm_backend *vmw_be;
+
+	vmw_be = kmalloc(sizeof(*vmw_be), GFP_KERNEL);
+	if (!vmw_be)
+		return NULL;
+
+	vmw_be->backend.func = &vmw_ttm_func;
+
+	return &vmw_be->backend;
+}
+
+int vmw_invalidate_caches(struct ttm_bo_device *bdev, uint32_t flags)
+{
+	return 0;
+}
+
+int vmw_init_mem_type(struct ttm_bo_device *bdev, uint32_t type,
+		      struct ttm_mem_type_manager *man)
+{
+	struct vmw_private *dev_priv =
+	    container_of(bdev, struct vmw_private, bdev);
+
+	switch (type) {
+	case TTM_PL_SYSTEM:
+		/* System memory */
+
+		man->flags = TTM_MEMTYPE_FLAG_MAPPABLE;
+		man->available_caching = TTM_PL_MASK_CACHING;
+		man->default_caching = TTM_PL_FLAG_CACHED;
+		break;
+	case TTM_PL_VRAM:
+		/* "On-card" video ram */
+		man->gpu_offset = 0;
+		man->io_offset = dev_priv->vram_start;
+		man->io_size = dev_priv->vram_size;
+		man->flags = TTM_MEMTYPE_FLAG_FIXED |
+		    TTM_MEMTYPE_FLAG_NEEDS_IOREMAP | TTM_MEMTYPE_FLAG_MAPPABLE;
+		man->io_addr = NULL;
+		man->available_caching = TTM_PL_MASK_CACHING;
+		man->default_caching = TTM_PL_FLAG_WC;
+		break;
+	default:
+		DRM_ERROR("Unsupported memory type %u\n", (unsigned)type);
+		return -EINVAL;
+	}
+	return 0;
+}
+
+void vmw_evict_flags(struct ttm_buffer_object *bo,
+		     struct ttm_placement *placement)
+{
+	*placement = vmw_sys_placement;
+}
+
+/**
+ * FIXME: Proper access checks on buffers.
+ */
+
+static int vmw_verify_access(struct ttm_buffer_object *bo, struct file *filp)
+{
+	return 0;
+}
+
+static void vmw_move_notify(struct ttm_buffer_object *bo,
+		     struct ttm_mem_reg *new_mem)
+{
+	if (new_mem->mem_type != TTM_PL_SYSTEM)
+		vmw_dmabuf_gmr_unbind(bo);
+}
+
+static void vmw_swap_notify(struct ttm_buffer_object *bo)
+{
+	vmw_dmabuf_gmr_unbind(bo);
+}
+
+/**
+ * FIXME: We're using the old vmware polling method to sync.
+ * Do this with fences instead.
+ */
+
+static void *vmw_sync_obj_ref(void *sync_obj)
+{
+	return sync_obj;
+}
+
+static void vmw_sync_obj_unref(void **sync_obj)
+{
+	*sync_obj = NULL;
+}
+
+static int vmw_sync_obj_flush(void *sync_obj, void *sync_arg)
+{
+	struct vmw_private *dev_priv = (struct vmw_private *)sync_arg;
+
+	mutex_lock(&dev_priv->hw_mutex);
+	vmw_write(dev_priv, SVGA_REG_SYNC, SVGA_SYNC_GENERIC);
+	mutex_unlock(&dev_priv->hw_mutex);
+	return 0;
+}
+
+static bool vmw_sync_obj_signaled(void *sync_obj, void *sync_arg)
+{
+	struct vmw_private *dev_priv = (struct vmw_private *)sync_arg;
+	uint32_t sequence = (unsigned long) sync_obj;
+
+	return vmw_fence_signaled(dev_priv, sequence);
+}
+
+static int vmw_sync_obj_wait(void *sync_obj, void *sync_arg,
+			     bool lazy, bool interruptible)
+{
+	struct vmw_private *dev_priv = (struct vmw_private *)sync_arg;
+	uint32_t sequence = (unsigned long) sync_obj;
+
+	return vmw_wait_fence(dev_priv, false, sequence, false, 3*HZ);
+}
+
+struct ttm_bo_driver vmw_bo_driver = {
+	.create_ttm_backend_entry = vmw_ttm_backend_init,
+	.invalidate_caches = vmw_invalidate_caches,
+	.init_mem_type = vmw_init_mem_type,
+	.evict_flags = vmw_evict_flags,
+	.move = NULL,
+	.verify_access = vmw_verify_access,
+	.sync_obj_signaled = vmw_sync_obj_signaled,
+	.sync_obj_wait = vmw_sync_obj_wait,
+	.sync_obj_flush = vmw_sync_obj_flush,
+	.sync_obj_unref = vmw_sync_obj_unref,
+	.sync_obj_ref = vmw_sync_obj_ref,
+	.move_notify = vmw_move_notify,
+	.swap_notify = vmw_swap_notify
+};
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
new file mode 100644
index 000000000000..0c9c0811f42d
--- /dev/null
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
@@ -0,0 +1,783 @@
+/**************************************************************************
+ *
+ * Copyright © 2009 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "drmP.h"
+#include "vmwgfx_drv.h"
+#include "ttm/ttm_placement.h"
+#include "ttm/ttm_bo_driver.h"
+#include "ttm/ttm_object.h"
+#include "ttm/ttm_module.h"
+
+#define VMWGFX_DRIVER_NAME "vmwgfx"
+#define VMWGFX_DRIVER_DESC "Linux drm driver for VMware graphics devices"
+#define VMWGFX_CHIP_SVGAII 0
+#define VMW_FB_RESERVATION 0
+
+/**
+ * Fully encoded drm commands. Might move to vmw_drm.h
+ */
+
+#define DRM_IOCTL_VMW_GET_PARAM					\
+	DRM_IOWR(DRM_COMMAND_BASE + DRM_VMW_GET_PARAM,		\
+		 struct drm_vmw_getparam_arg)
+#define DRM_IOCTL_VMW_ALLOC_DMABUF				\
+	DRM_IOWR(DRM_COMMAND_BASE + DRM_VMW_ALLOC_DMABUF,	\
+		union drm_vmw_alloc_dmabuf_arg)
+#define DRM_IOCTL_VMW_UNREF_DMABUF				\
+	DRM_IOW(DRM_COMMAND_BASE + DRM_VMW_UNREF_DMABUF,	\
+		struct drm_vmw_unref_dmabuf_arg)
+#define DRM_IOCTL_VMW_CURSOR_BYPASS				\
+	DRM_IOW(DRM_COMMAND_BASE + DRM_VMW_CURSOR_BYPASS,	\
+		 struct drm_vmw_cursor_bypass_arg)
+
+#define DRM_IOCTL_VMW_CONTROL_STREAM				\
+	DRM_IOW(DRM_COMMAND_BASE + DRM_VMW_CONTROL_STREAM,	\
+		 struct drm_vmw_control_stream_arg)
+#define DRM_IOCTL_VMW_CLAIM_STREAM				\
+	DRM_IOR(DRM_COMMAND_BASE + DRM_VMW_CLAIM_STREAM,	\
+		 struct drm_vmw_stream_arg)
+#define DRM_IOCTL_VMW_UNREF_STREAM				\
+	DRM_IOW(DRM_COMMAND_BASE + DRM_VMW_UNREF_STREAM,	\
+		 struct drm_vmw_stream_arg)
+
+#define DRM_IOCTL_VMW_CREATE_CONTEXT				\
+	DRM_IOR(DRM_COMMAND_BASE + DRM_VMW_CREATE_CONTEXT,	\
+		struct drm_vmw_context_arg)
+#define DRM_IOCTL_VMW_UNREF_CONTEXT				\
+	DRM_IOW(DRM_COMMAND_BASE + DRM_VMW_UNREF_CONTEXT,	\
+		struct drm_vmw_context_arg)
+#define DRM_IOCTL_VMW_CREATE_SURFACE				\
+	DRM_IOWR(DRM_COMMAND_BASE + DRM_VMW_CREATE_SURFACE,	\
+		 union drm_vmw_surface_create_arg)
+#define DRM_IOCTL_VMW_UNREF_SURFACE				\
+	DRM_IOW(DRM_COMMAND_BASE + DRM_VMW_UNREF_SURFACE,	\
+		 struct drm_vmw_surface_arg)
+#define DRM_IOCTL_VMW_REF_SURFACE				\
+	DRM_IOWR(DRM_COMMAND_BASE + DRM_VMW_REF_SURFACE,	\
+		 union drm_vmw_surface_reference_arg)
+#define DRM_IOCTL_VMW_EXECBUF					\
+	DRM_IOW(DRM_COMMAND_BASE + DRM_VMW_EXECBUF,		\
+		struct drm_vmw_execbuf_arg)
+#define DRM_IOCTL_VMW_FIFO_DEBUG				\
+	DRM_IOWR(DRM_COMMAND_BASE + DRM_VMW_FIFO_DEBUG,		\
+		 struct drm_vmw_fifo_debug_arg)
+#define DRM_IOCTL_VMW_FENCE_WAIT				\
+	DRM_IOWR(DRM_COMMAND_BASE + DRM_VMW_FENCE_WAIT,		\
+		 struct drm_vmw_fence_wait_arg)
+
+
+/**
+ * The core DRM version of this macro doesn't account for
+ * DRM_COMMAND_BASE.
+ */
+
+#define VMW_IOCTL_DEF(ioctl, func, flags) \
+	[DRM_IOCTL_NR(ioctl) - DRM_COMMAND_BASE] = {ioctl, flags, func}
+
+/**
+ * Ioctl definitions.
+ */
+
+static struct drm_ioctl_desc vmw_ioctls[] = {
+	VMW_IOCTL_DEF(DRM_IOCTL_VMW_GET_PARAM, vmw_getparam_ioctl,
+		      DRM_AUTH | DRM_UNLOCKED),
+	VMW_IOCTL_DEF(DRM_IOCTL_VMW_ALLOC_DMABUF, vmw_dmabuf_alloc_ioctl,
+		      DRM_AUTH | DRM_UNLOCKED),
+	VMW_IOCTL_DEF(DRM_IOCTL_VMW_UNREF_DMABUF, vmw_dmabuf_unref_ioctl,
+		      DRM_AUTH | DRM_UNLOCKED),
+	VMW_IOCTL_DEF(DRM_IOCTL_VMW_CURSOR_BYPASS,
+		      vmw_kms_cursor_bypass_ioctl,
+		      DRM_MASTER | DRM_CONTROL_ALLOW | DRM_UNLOCKED),
+
+	VMW_IOCTL_DEF(DRM_IOCTL_VMW_CONTROL_STREAM, vmw_overlay_ioctl,
+		      DRM_MASTER | DRM_CONTROL_ALLOW | DRM_UNLOCKED),
+	VMW_IOCTL_DEF(DRM_IOCTL_VMW_CLAIM_STREAM, vmw_stream_claim_ioctl,
+		      DRM_MASTER | DRM_CONTROL_ALLOW | DRM_UNLOCKED),
+	VMW_IOCTL_DEF(DRM_IOCTL_VMW_UNREF_STREAM, vmw_stream_unref_ioctl,
+		      DRM_MASTER | DRM_CONTROL_ALLOW | DRM_UNLOCKED),
+
+	VMW_IOCTL_DEF(DRM_IOCTL_VMW_CREATE_CONTEXT, vmw_context_define_ioctl,
+		      DRM_AUTH | DRM_UNLOCKED),
+	VMW_IOCTL_DEF(DRM_IOCTL_VMW_UNREF_CONTEXT, vmw_context_destroy_ioctl,
+		      DRM_AUTH | DRM_UNLOCKED),
+	VMW_IOCTL_DEF(DRM_IOCTL_VMW_CREATE_SURFACE, vmw_surface_define_ioctl,
+		      DRM_AUTH | DRM_UNLOCKED),
+	VMW_IOCTL_DEF(DRM_IOCTL_VMW_UNREF_SURFACE, vmw_surface_destroy_ioctl,
+		      DRM_AUTH | DRM_UNLOCKED),
+	VMW_IOCTL_DEF(DRM_IOCTL_VMW_REF_SURFACE, vmw_surface_reference_ioctl,
+		      DRM_AUTH | DRM_UNLOCKED),
+	VMW_IOCTL_DEF(DRM_IOCTL_VMW_EXECBUF, vmw_execbuf_ioctl,
+		      DRM_AUTH | DRM_UNLOCKED),
+	VMW_IOCTL_DEF(DRM_IOCTL_VMW_FIFO_DEBUG, vmw_fifo_debug_ioctl,
+		      DRM_AUTH | DRM_ROOT_ONLY | DRM_MASTER | DRM_UNLOCKED),
+	VMW_IOCTL_DEF(DRM_IOCTL_VMW_FENCE_WAIT, vmw_fence_wait_ioctl,
+		      DRM_AUTH | DRM_UNLOCKED)
+};
+
+static struct pci_device_id vmw_pci_id_list[] = {
+	{0x15ad, 0x0405, PCI_ANY_ID, PCI_ANY_ID, 0, 0, VMWGFX_CHIP_SVGAII},
+	{0, 0, 0}
+};
+
+static char *vmw_devname = "vmwgfx";
+
+static int vmw_probe(struct pci_dev *, const struct pci_device_id *);
+static void vmw_master_init(struct vmw_master *);
+static int vmwgfx_pm_notifier(struct notifier_block *nb, unsigned long val,
+			      void *ptr);
+
+static void vmw_print_capabilities(uint32_t capabilities)
+{
+	DRM_INFO("Capabilities:\n");
+	if (capabilities & SVGA_CAP_RECT_COPY)
+		DRM_INFO("  Rect copy.\n");
+	if (capabilities & SVGA_CAP_CURSOR)
+		DRM_INFO("  Cursor.\n");
+	if (capabilities & SVGA_CAP_CURSOR_BYPASS)
+		DRM_INFO("  Cursor bypass.\n");
+	if (capabilities & SVGA_CAP_CURSOR_BYPASS_2)
+		DRM_INFO("  Cursor bypass 2.\n");
+	if (capabilities & SVGA_CAP_8BIT_EMULATION)
+		DRM_INFO("  8bit emulation.\n");
+	if (capabilities & SVGA_CAP_ALPHA_CURSOR)
+		DRM_INFO("  Alpha cursor.\n");
+	if (capabilities & SVGA_CAP_3D)
+		DRM_INFO("  3D.\n");
+	if (capabilities & SVGA_CAP_EXTENDED_FIFO)
+		DRM_INFO("  Extended Fifo.\n");
+	if (capabilities & SVGA_CAP_MULTIMON)
+		DRM_INFO("  Multimon.\n");
+	if (capabilities & SVGA_CAP_PITCHLOCK)
+		DRM_INFO("  Pitchlock.\n");
+	if (capabilities & SVGA_CAP_IRQMASK)
+		DRM_INFO("  Irq mask.\n");
+	if (capabilities & SVGA_CAP_DISPLAY_TOPOLOGY)
+		DRM_INFO("  Display Topology.\n");
+	if (capabilities & SVGA_CAP_GMR)
+		DRM_INFO("  GMR.\n");
+	if (capabilities & SVGA_CAP_TRACES)
+		DRM_INFO("  Traces.\n");
+}
+
+static int vmw_request_device(struct vmw_private *dev_priv)
+{
+	int ret;
+
+	vmw_kms_save_vga(dev_priv);
+
+	ret = vmw_fifo_init(dev_priv, &dev_priv->fifo);
+	if (unlikely(ret != 0)) {
+		DRM_ERROR("Unable to initialize FIFO.\n");
+		return ret;
+	}
+
+	return 0;
+}
+
+static void vmw_release_device(struct vmw_private *dev_priv)
+{
+	vmw_fifo_release(dev_priv, &dev_priv->fifo);
+	vmw_kms_restore_vga(dev_priv);
+}
+
+
+static int vmw_driver_load(struct drm_device *dev, unsigned long chipset)
+{
+	struct vmw_private *dev_priv;
+	int ret;
+	uint32_t svga_id;
+
+	dev_priv = kzalloc(sizeof(*dev_priv), GFP_KERNEL);
+	if (unlikely(dev_priv == NULL)) {
+		DRM_ERROR("Failed allocating a device private struct.\n");
+		return -ENOMEM;
+	}
+	memset(dev_priv, 0, sizeof(*dev_priv));
+
+	dev_priv->dev = dev;
+	dev_priv->vmw_chipset = chipset;
+	dev_priv->last_read_sequence = (uint32_t) -100;
+	mutex_init(&dev_priv->hw_mutex);
+	mutex_init(&dev_priv->cmdbuf_mutex);
+	rwlock_init(&dev_priv->resource_lock);
+	idr_init(&dev_priv->context_idr);
+	idr_init(&dev_priv->surface_idr);
+	idr_init(&dev_priv->stream_idr);
+	ida_init(&dev_priv->gmr_ida);
+	mutex_init(&dev_priv->init_mutex);
+	init_waitqueue_head(&dev_priv->fence_queue);
+	init_waitqueue_head(&dev_priv->fifo_queue);
+	atomic_set(&dev_priv->fence_queue_waiters, 0);
+	atomic_set(&dev_priv->fifo_queue_waiters, 0);
+	INIT_LIST_HEAD(&dev_priv->gmr_lru);
+
+	dev_priv->io_start = pci_resource_start(dev->pdev, 0);
+	dev_priv->vram_start = pci_resource_start(dev->pdev, 1);
+	dev_priv->mmio_start = pci_resource_start(dev->pdev, 2);
+
+	mutex_lock(&dev_priv->hw_mutex);
+
+	vmw_write(dev_priv, SVGA_REG_ID, SVGA_ID_2);
+	svga_id = vmw_read(dev_priv, SVGA_REG_ID);
+	if (svga_id != SVGA_ID_2) {
+		ret = -ENOSYS;
+		DRM_ERROR("Unsuported SVGA ID 0x%x\n", svga_id);
+		mutex_unlock(&dev_priv->hw_mutex);
+		goto out_err0;
+	}
+
+	dev_priv->capabilities = vmw_read(dev_priv, SVGA_REG_CAPABILITIES);
+
+	if (dev_priv->capabilities & SVGA_CAP_GMR) {
+		dev_priv->max_gmr_descriptors =
+			vmw_read(dev_priv,
+				 SVGA_REG_GMR_MAX_DESCRIPTOR_LENGTH);
+		dev_priv->max_gmr_ids =
+			vmw_read(dev_priv, SVGA_REG_GMR_MAX_IDS);
+	}
+
+	dev_priv->vram_size = vmw_read(dev_priv, SVGA_REG_VRAM_SIZE);
+	dev_priv->mmio_size = vmw_read(dev_priv, SVGA_REG_MEM_SIZE);
+	dev_priv->fb_max_width = vmw_read(dev_priv, SVGA_REG_MAX_WIDTH);
+	dev_priv->fb_max_height = vmw_read(dev_priv, SVGA_REG_MAX_HEIGHT);
+
+	mutex_unlock(&dev_priv->hw_mutex);
+
+	vmw_print_capabilities(dev_priv->capabilities);
+
+	if (dev_priv->capabilities & SVGA_CAP_GMR) {
+		DRM_INFO("Max GMR ids is %u\n",
+			 (unsigned)dev_priv->max_gmr_ids);
+		DRM_INFO("Max GMR descriptors is %u\n",
+			 (unsigned)dev_priv->max_gmr_descriptors);
+	}
+	DRM_INFO("VRAM at 0x%08x size is %u kiB\n",
+		 dev_priv->vram_start, dev_priv->vram_size / 1024);
+	DRM_INFO("MMIO at 0x%08x size is %u kiB\n",
+		 dev_priv->mmio_start, dev_priv->mmio_size / 1024);
+
+	ret = vmw_ttm_global_init(dev_priv);
+	if (unlikely(ret != 0))
+		goto out_err0;
+
+
+	vmw_master_init(&dev_priv->fbdev_master);
+	ttm_lock_set_kill(&dev_priv->fbdev_master.lock, false, SIGTERM);
+	dev_priv->active_master = &dev_priv->fbdev_master;
+
+
+	ret = ttm_bo_device_init(&dev_priv->bdev,
+				 dev_priv->bo_global_ref.ref.object,
+				 &vmw_bo_driver, VMWGFX_FILE_PAGE_OFFSET,
+				 false);
+	if (unlikely(ret != 0)) {
+		DRM_ERROR("Failed initializing TTM buffer object driver.\n");
+		goto out_err1;
+	}
+
+	ret = ttm_bo_init_mm(&dev_priv->bdev, TTM_PL_VRAM,
+			     (dev_priv->vram_size >> PAGE_SHIFT));
+	if (unlikely(ret != 0)) {
+		DRM_ERROR("Failed initializing memory manager for VRAM.\n");
+		goto out_err2;
+	}
+
+	dev_priv->mmio_mtrr = drm_mtrr_add(dev_priv->mmio_start,
+					   dev_priv->mmio_size, DRM_MTRR_WC);
+
+	dev_priv->mmio_virt = ioremap_wc(dev_priv->mmio_start,
+					 dev_priv->mmio_size);
+
+	if (unlikely(dev_priv->mmio_virt == NULL)) {
+		ret = -ENOMEM;
+		DRM_ERROR("Failed mapping MMIO.\n");
+		goto out_err3;
+	}
+
+	dev_priv->tdev = ttm_object_device_init
+	    (dev_priv->mem_global_ref.object, 12);
+
+	if (unlikely(dev_priv->tdev == NULL)) {
+		DRM_ERROR("Unable to initialize TTM object management.\n");
+		ret = -ENOMEM;
+		goto out_err4;
+	}
+
+	dev->dev_private = dev_priv;
+
+	if (!dev->devname)
+		dev->devname = vmw_devname;
+
+	if (dev_priv->capabilities & SVGA_CAP_IRQMASK) {
+		ret = drm_irq_install(dev);
+		if (unlikely(ret != 0)) {
+			DRM_ERROR("Failed installing irq: %d\n", ret);
+			goto out_no_irq;
+		}
+	}
+
+	ret = pci_request_regions(dev->pdev, "vmwgfx probe");
+	dev_priv->stealth = (ret != 0);
+	if (dev_priv->stealth) {
+		/**
+		 * Request at least the mmio PCI resource.
+		 */
+
+		DRM_INFO("It appears like vesafb is loaded. "
+			 "Ignore above error if any.\n");
+		ret = pci_request_region(dev->pdev, 2, "vmwgfx stealth probe");
+		if (unlikely(ret != 0)) {
+			DRM_ERROR("Failed reserving the SVGA MMIO resource.\n");
+			goto out_no_device;
+		}
+	}
+	ret = vmw_request_device(dev_priv);
+	if (unlikely(ret != 0))
+		goto out_no_device;
+	vmw_kms_init(dev_priv);
+	vmw_overlay_init(dev_priv);
+	vmw_fb_init(dev_priv);
+
+	dev_priv->pm_nb.notifier_call = vmwgfx_pm_notifier;
+	register_pm_notifier(&dev_priv->pm_nb);
+
+	DRM_INFO("%s", vmw_fifo_have_3d(dev_priv) ? "Have 3D\n" : "No 3D\n");
+
+	return 0;
+
+out_no_device:
+	if (dev_priv->capabilities & SVGA_CAP_IRQMASK)
+		drm_irq_uninstall(dev_priv->dev);
+	if (dev->devname == vmw_devname)
+		dev->devname = NULL;
+out_no_irq:
+	ttm_object_device_release(&dev_priv->tdev);
+out_err4:
+	iounmap(dev_priv->mmio_virt);
+out_err3:
+	drm_mtrr_del(dev_priv->mmio_mtrr, dev_priv->mmio_start,
+		     dev_priv->mmio_size, DRM_MTRR_WC);
+	(void)ttm_bo_clean_mm(&dev_priv->bdev, TTM_PL_VRAM);
+out_err2:
+	(void)ttm_bo_device_release(&dev_priv->bdev);
+out_err1:
+	vmw_ttm_global_release(dev_priv);
+out_err0:
+	ida_destroy(&dev_priv->gmr_ida);
+	idr_destroy(&dev_priv->surface_idr);
+	idr_destroy(&dev_priv->context_idr);
+	idr_destroy(&dev_priv->stream_idr);
+	kfree(dev_priv);
+	return ret;
+}
+
+static int vmw_driver_unload(struct drm_device *dev)
+{
+	struct vmw_private *dev_priv = vmw_priv(dev);
+
+	DRM_INFO(VMWGFX_DRIVER_NAME " unload.\n");
+
+	unregister_pm_notifier(&dev_priv->pm_nb);
+
+	vmw_fb_close(dev_priv);
+	vmw_kms_close(dev_priv);
+	vmw_overlay_close(dev_priv);
+	vmw_release_device(dev_priv);
+	if (dev_priv->stealth)
+		pci_release_region(dev->pdev, 2);
+	else
+		pci_release_regions(dev->pdev);
+
+	if (dev_priv->capabilities & SVGA_CAP_IRQMASK)
+		drm_irq_uninstall(dev_priv->dev);
+	if (dev->devname == vmw_devname)
+		dev->devname = NULL;
+	ttm_object_device_release(&dev_priv->tdev);
+	iounmap(dev_priv->mmio_virt);
+	drm_mtrr_del(dev_priv->mmio_mtrr, dev_priv->mmio_start,
+		     dev_priv->mmio_size, DRM_MTRR_WC);
+	(void)ttm_bo_clean_mm(&dev_priv->bdev, TTM_PL_VRAM);
+	(void)ttm_bo_device_release(&dev_priv->bdev);
+	vmw_ttm_global_release(dev_priv);
+	ida_destroy(&dev_priv->gmr_ida);
+	idr_destroy(&dev_priv->surface_idr);
+	idr_destroy(&dev_priv->context_idr);
+	idr_destroy(&dev_priv->stream_idr);
+
+	kfree(dev_priv);
+
+	return 0;
+}
+
+static void vmw_postclose(struct drm_device *dev,
+			 struct drm_file *file_priv)
+{
+	struct vmw_fpriv *vmw_fp;
+
+	vmw_fp = vmw_fpriv(file_priv);
+	ttm_object_file_release(&vmw_fp->tfile);
+	if (vmw_fp->locked_master)
+		drm_master_put(&vmw_fp->locked_master);
+	kfree(vmw_fp);
+}
+
+static int vmw_driver_open(struct drm_device *dev, struct drm_file *file_priv)
+{
+	struct vmw_private *dev_priv = vmw_priv(dev);
+	struct vmw_fpriv *vmw_fp;
+	int ret = -ENOMEM;
+
+	vmw_fp = kzalloc(sizeof(*vmw_fp), GFP_KERNEL);
+	if (unlikely(vmw_fp == NULL))
+		return ret;
+
+	vmw_fp->tfile = ttm_object_file_init(dev_priv->tdev, 10);
+	if (unlikely(vmw_fp->tfile == NULL))
+		goto out_no_tfile;
+
+	file_priv->driver_priv = vmw_fp;
+
+	if (unlikely(dev_priv->bdev.dev_mapping == NULL))
+		dev_priv->bdev.dev_mapping =
+			file_priv->filp->f_path.dentry->d_inode->i_mapping;
+
+	return 0;
+
+out_no_tfile:
+	kfree(vmw_fp);
+	return ret;
+}
+
+static long vmw_unlocked_ioctl(struct file *filp, unsigned int cmd,
+			       unsigned long arg)
+{
+	struct drm_file *file_priv = filp->private_data;
+	struct drm_device *dev = file_priv->minor->dev;
+	unsigned int nr = DRM_IOCTL_NR(cmd);
+
+	/*
+	 * Do extra checking on driver private ioctls.
+	 */
+
+	if ((nr >= DRM_COMMAND_BASE) && (nr < DRM_COMMAND_END)
+	    && (nr < DRM_COMMAND_BASE + dev->driver->num_ioctls)) {
+		struct drm_ioctl_desc *ioctl =
+		    &vmw_ioctls[nr - DRM_COMMAND_BASE];
+
+		if (unlikely(ioctl->cmd != cmd)) {
+			DRM_ERROR("Invalid command format, ioctl %d\n",
+				  nr - DRM_COMMAND_BASE);
+			return -EINVAL;
+		}
+	}
+
+	return drm_ioctl(filp, cmd, arg);
+}
+
+static int vmw_firstopen(struct drm_device *dev)
+{
+	struct vmw_private *dev_priv = vmw_priv(dev);
+	dev_priv->is_opened = true;
+
+	return 0;
+}
+
+static void vmw_lastclose(struct drm_device *dev)
+{
+	struct vmw_private *dev_priv = vmw_priv(dev);
+	struct drm_crtc *crtc;
+	struct drm_mode_set set;
+	int ret;
+
+	/**
+	 * Do nothing on the lastclose call from drm_unload.
+	 */
+
+	if (!dev_priv->is_opened)
+		return;
+
+	dev_priv->is_opened = false;
+	set.x = 0;
+	set.y = 0;
+	set.fb = NULL;
+	set.mode = NULL;
+	set.connectors = NULL;
+	set.num_connectors = 0;
+
+	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
+		set.crtc = crtc;
+		ret = crtc->funcs->set_config(&set);
+		WARN_ON(ret != 0);
+	}
+
+}
+
+static void vmw_master_init(struct vmw_master *vmaster)
+{
+	ttm_lock_init(&vmaster->lock);
+}
+
+static int vmw_master_create(struct drm_device *dev,
+			     struct drm_master *master)
+{
+	struct vmw_master *vmaster;
+
+	DRM_INFO("Master create.\n");
+	vmaster = kzalloc(sizeof(*vmaster), GFP_KERNEL);
+	if (unlikely(vmaster == NULL))
+		return -ENOMEM;
+
+	ttm_lock_init(&vmaster->lock);
+	ttm_lock_set_kill(&vmaster->lock, true, SIGTERM);
+	master->driver_priv = vmaster;
+
+	return 0;
+}
+
+static void vmw_master_destroy(struct drm_device *dev,
+			       struct drm_master *master)
+{
+	struct vmw_master *vmaster = vmw_master(master);
+
+	DRM_INFO("Master destroy.\n");
+	master->driver_priv = NULL;
+	kfree(vmaster);
+}
+
+
+static int vmw_master_set(struct drm_device *dev,
+			  struct drm_file *file_priv,
+			  bool from_open)
+{
+	struct vmw_private *dev_priv = vmw_priv(dev);
+	struct vmw_fpriv *vmw_fp = vmw_fpriv(file_priv);
+	struct vmw_master *active = dev_priv->active_master;
+	struct vmw_master *vmaster = vmw_master(file_priv->master);
+	int ret = 0;
+
+	DRM_INFO("Master set.\n");
+
+	if (active) {
+		BUG_ON(active != &dev_priv->fbdev_master);
+		ret = ttm_vt_lock(&active->lock, false, vmw_fp->tfile);
+		if (unlikely(ret != 0))
+			goto out_no_active_lock;
+
+		ttm_lock_set_kill(&active->lock, true, SIGTERM);
+		ret = ttm_bo_evict_mm(&dev_priv->bdev, TTM_PL_VRAM);
+		if (unlikely(ret != 0)) {
+			DRM_ERROR("Unable to clean VRAM on "
+				  "master drop.\n");
+		}
+
+		dev_priv->active_master = NULL;
+	}
+
+	ttm_lock_set_kill(&vmaster->lock, false, SIGTERM);
+	if (!from_open) {
+		ttm_vt_unlock(&vmaster->lock);
+		BUG_ON(vmw_fp->locked_master != file_priv->master);
+		drm_master_put(&vmw_fp->locked_master);
+	}
+
+	dev_priv->active_master = vmaster;
+
+	return 0;
+
+out_no_active_lock:
+	vmw_release_device(dev_priv);
+	return ret;
+}
+
+static void vmw_master_drop(struct drm_device *dev,
+			    struct drm_file *file_priv,
+			    bool from_release)
+{
+	struct vmw_private *dev_priv = vmw_priv(dev);
+	struct vmw_fpriv *vmw_fp = vmw_fpriv(file_priv);
+	struct vmw_master *vmaster = vmw_master(file_priv->master);
+	int ret;
+
+	DRM_INFO("Master drop.\n");
+
+	/**
+	 * Make sure the master doesn't disappear while we have
+	 * it locked.
+	 */
+
+	vmw_fp->locked_master = drm_master_get(file_priv->master);
+	ret = ttm_vt_lock(&vmaster->lock, false, vmw_fp->tfile);
+
+	if (unlikely((ret != 0))) {
+		DRM_ERROR("Unable to lock TTM at VT switch.\n");
+		drm_master_put(&vmw_fp->locked_master);
+	}
+
+	ttm_lock_set_kill(&vmaster->lock, true, SIGTERM);
+
+	dev_priv->active_master = &dev_priv->fbdev_master;
+	ttm_lock_set_kill(&dev_priv->fbdev_master.lock, false, SIGTERM);
+	ttm_vt_unlock(&dev_priv->fbdev_master.lock);
+
+	vmw_fb_on(dev_priv);
+}
+
+
+static void vmw_remove(struct pci_dev *pdev)
+{
+	struct drm_device *dev = pci_get_drvdata(pdev);
+
+	drm_put_dev(dev);
+}
+
+static int vmwgfx_pm_notifier(struct notifier_block *nb, unsigned long val,
+			      void *ptr)
+{
+	struct vmw_private *dev_priv =
+		container_of(nb, struct vmw_private, pm_nb);
+	struct vmw_master *vmaster = dev_priv->active_master;
+
+	switch (val) {
+	case PM_HIBERNATION_PREPARE:
+	case PM_SUSPEND_PREPARE:
+		ttm_suspend_lock(&vmaster->lock);
+
+		/**
+		 * This empties VRAM and unbinds all GMR bindings.
+		 * Buffer contents is moved to swappable memory.
+		 */
+		ttm_bo_swapout_all(&dev_priv->bdev);
+		break;
+	case PM_POST_HIBERNATION:
+	case PM_POST_SUSPEND:
+		ttm_suspend_unlock(&vmaster->lock);
+		break;
+	case PM_RESTORE_PREPARE:
+		break;
+	case PM_POST_RESTORE:
+		break;
+	default:
+		break;
+	}
+	return 0;
+}
+
+/**
+ * These might not be needed with the virtual SVGA device.
+ */
+
+int vmw_pci_suspend(struct pci_dev *pdev, pm_message_t state)
+{
+	pci_save_state(pdev);
+	pci_disable_device(pdev);
+	pci_set_power_state(pdev, PCI_D3hot);
+	return 0;
+}
+
+int vmw_pci_resume(struct pci_dev *pdev)
+{
+	pci_set_power_state(pdev, PCI_D0);
+	pci_restore_state(pdev);
+	return pci_enable_device(pdev);
+}
+
+static struct drm_driver driver = {
+	.driver_features = DRIVER_HAVE_IRQ | DRIVER_IRQ_SHARED |
+	DRIVER_MODESET,
+	.load = vmw_driver_load,
+	.unload = vmw_driver_unload,
+	.firstopen = vmw_firstopen,
+	.lastclose = vmw_lastclose,
+	.irq_preinstall = vmw_irq_preinstall,
+	.irq_postinstall = vmw_irq_postinstall,
+	.irq_uninstall = vmw_irq_uninstall,
+	.irq_handler = vmw_irq_handler,
+	.reclaim_buffers_locked = NULL,
+	.get_map_ofs = drm_core_get_map_ofs,
+	.get_reg_ofs = drm_core_get_reg_ofs,
+	.ioctls = vmw_ioctls,
+	.num_ioctls = DRM_ARRAY_SIZE(vmw_ioctls),
+	.dma_quiescent = NULL,	/*vmw_dma_quiescent, */
+	.master_create = vmw_master_create,
+	.master_destroy = vmw_master_destroy,
+	.master_set = vmw_master_set,
+	.master_drop = vmw_master_drop,
+	.open = vmw_driver_open,
+	.postclose = vmw_postclose,
+	.fops = {
+		 .owner = THIS_MODULE,
+		 .open = drm_open,
+		 .release = drm_release,
+		 .unlocked_ioctl = vmw_unlocked_ioctl,
+		 .mmap = vmw_mmap,
+		 .poll = drm_poll,
+		 .fasync = drm_fasync,
+#if defined(CONFIG_COMPAT)
+		 .compat_ioctl = drm_compat_ioctl,
+#endif
+		 },
+	.pci_driver = {
+		       .name = VMWGFX_DRIVER_NAME,
+		       .id_table = vmw_pci_id_list,
+		       .probe = vmw_probe,
+		       .remove = vmw_remove,
+		       .suspend = vmw_pci_suspend,
+		       .resume = vmw_pci_resume
+		       },
+	.name = VMWGFX_DRIVER_NAME,
+	.desc = VMWGFX_DRIVER_DESC,
+	.date = VMWGFX_DRIVER_DATE,
+	.major = VMWGFX_DRIVER_MAJOR,
+	.minor = VMWGFX_DRIVER_MINOR,
+	.patchlevel = VMWGFX_DRIVER_PATCHLEVEL
+};
+
+static int vmw_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
+{
+	return drm_get_dev(pdev, ent, &driver);
+}
+
+static int __init vmwgfx_init(void)
+{
+	int ret;
+	ret = drm_init(&driver);
+	if (ret)
+		DRM_ERROR("Failed initializing DRM.\n");
+	return ret;
+}
+
+static void __exit vmwgfx_exit(void)
+{
+	drm_exit(&driver);
+}
+
+module_init(vmwgfx_init);
+module_exit(vmwgfx_exit);
+
+MODULE_AUTHOR("VMware Inc. and others");
+MODULE_DESCRIPTION("Standalone drm driver for the VMware SVGA device");
+MODULE_LICENSE("GPL and additional rights");
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h
new file mode 100644
index 000000000000..356dc935ec13
--- /dev/null
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h
@@ -0,0 +1,521 @@
+/**************************************************************************
+ *
+ * Copyright © 2009 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef _VMWGFX_DRV_H_
+#define _VMWGFX_DRV_H_
+
+#include "vmwgfx_reg.h"
+#include "drmP.h"
+#include "vmwgfx_drm.h"
+#include "drm_hashtab.h"
+#include "linux/suspend.h"
+#include "ttm/ttm_bo_driver.h"
+#include "ttm/ttm_object.h"
+#include "ttm/ttm_lock.h"
+#include "ttm/ttm_execbuf_util.h"
+#include "ttm/ttm_module.h"
+
+#define VMWGFX_DRIVER_DATE "20100209"
+#define VMWGFX_DRIVER_MAJOR 1
+#define VMWGFX_DRIVER_MINOR 0
+#define VMWGFX_DRIVER_PATCHLEVEL 0
+#define VMWGFX_FILE_PAGE_OFFSET 0x00100000
+#define VMWGFX_FIFO_STATIC_SIZE (1024*1024)
+#define VMWGFX_MAX_RELOCATIONS 2048
+#define VMWGFX_MAX_GMRS 2048
+
+struct vmw_fpriv {
+	struct drm_master *locked_master;
+	struct ttm_object_file *tfile;
+};
+
+struct vmw_dma_buffer {
+	struct ttm_buffer_object base;
+	struct list_head validate_list;
+	struct list_head gmr_lru;
+	uint32_t gmr_id;
+	bool gmr_bound;
+	uint32_t cur_validate_node;
+	bool on_validate_list;
+};
+
+struct vmw_resource {
+	struct kref kref;
+	struct vmw_private *dev_priv;
+	struct idr *idr;
+	int id;
+	enum ttm_object_type res_type;
+	bool avail;
+	void (*hw_destroy) (struct vmw_resource *res);
+	void (*res_free) (struct vmw_resource *res);
+
+	/* TODO is a generic snooper needed? */
+#if 0
+	void (*snoop)(struct vmw_resource *res,
+		      struct ttm_object_file *tfile,
+		      SVGA3dCmdHeader *header);
+	void *snoop_priv;
+#endif
+};
+
+struct vmw_cursor_snooper {
+	struct drm_crtc *crtc;
+	size_t age;
+	uint32_t *image;
+};
+
+struct vmw_surface {
+	struct vmw_resource res;
+	uint32_t flags;
+	uint32_t format;
+	uint32_t mip_levels[DRM_VMW_MAX_SURFACE_FACES];
+	struct drm_vmw_size *sizes;
+	uint32_t num_sizes;
+
+	bool scanout;
+
+	/* TODO so far just a extra pointer */
+	struct vmw_cursor_snooper snooper;
+};
+
+struct vmw_fifo_state {
+	unsigned long reserved_size;
+	__le32 *dynamic_buffer;
+	__le32 *static_buffer;
+	__le32 *last_buffer;
+	uint32_t last_data_size;
+	uint32_t last_buffer_size;
+	bool last_buffer_add;
+	unsigned long static_buffer_size;
+	bool using_bounce_buffer;
+	uint32_t capabilities;
+	struct mutex fifo_mutex;
+	struct rw_semaphore rwsem;
+};
+
+struct vmw_relocation {
+	SVGAGuestPtr *location;
+	uint32_t index;
+};
+
+struct vmw_sw_context{
+	struct ida bo_list;
+	uint32_t last_cid;
+	bool cid_valid;
+	uint32_t last_sid;
+	uint32_t sid_translation;
+	bool sid_valid;
+	struct ttm_object_file *tfile;
+	struct list_head validate_nodes;
+	struct vmw_relocation relocs[VMWGFX_MAX_RELOCATIONS];
+	uint32_t cur_reloc;
+	struct ttm_validate_buffer val_bufs[VMWGFX_MAX_GMRS];
+	uint32_t cur_val_buf;
+};
+
+struct vmw_legacy_display;
+struct vmw_overlay;
+
+struct vmw_master {
+	struct ttm_lock lock;
+};
+
+struct vmw_private {
+	struct ttm_bo_device bdev;
+	struct ttm_bo_global_ref bo_global_ref;
+	struct ttm_global_reference mem_global_ref;
+
+	struct vmw_fifo_state fifo;
+
+	struct drm_device *dev;
+	unsigned long vmw_chipset;
+	unsigned int io_start;
+	uint32_t vram_start;
+	uint32_t vram_size;
+	uint32_t mmio_start;
+	uint32_t mmio_size;
+	uint32_t fb_max_width;
+	uint32_t fb_max_height;
+	__le32 __iomem *mmio_virt;
+	int mmio_mtrr;
+	uint32_t capabilities;
+	uint32_t max_gmr_descriptors;
+	uint32_t max_gmr_ids;
+	struct mutex hw_mutex;
+
+	/*
+	 * VGA registers.
+	 */
+
+	uint32_t vga_width;
+	uint32_t vga_height;
+	uint32_t vga_depth;
+	uint32_t vga_bpp;
+	uint32_t vga_pseudo;
+	uint32_t vga_red_mask;
+	uint32_t vga_blue_mask;
+	uint32_t vga_green_mask;
+
+	/*
+	 * Framebuffer info.
+	 */
+
+	void *fb_info;
+	struct vmw_legacy_display *ldu_priv;
+	struct vmw_overlay *overlay_priv;
+
+	/*
+	 * Context and surface management.
+	 */
+
+	rwlock_t resource_lock;
+	struct idr context_idr;
+	struct idr surface_idr;
+	struct idr stream_idr;
+
+	/*
+	 * Block lastclose from racing with firstopen.
+	 */
+
+	struct mutex init_mutex;
+
+	/*
+	 * A resource manager for kernel-only surfaces and
+	 * contexts.
+	 */
+
+	struct ttm_object_device *tdev;
+
+	/*
+	 * Fencing and IRQs.
+	 */
+
+	atomic_t fence_seq;
+	wait_queue_head_t fence_queue;
+	wait_queue_head_t fifo_queue;
+	atomic_t fence_queue_waiters;
+	atomic_t fifo_queue_waiters;
+	uint32_t last_read_sequence;
+	spinlock_t irq_lock;
+
+	/*
+	 * Device state
+	 */
+
+	uint32_t traces_state;
+	uint32_t enable_state;
+	uint32_t config_done_state;
+
+	/**
+	 * Execbuf
+	 */
+	/**
+	 * Protected by the cmdbuf mutex.
+	 */
+
+	struct vmw_sw_context ctx;
+	uint32_t val_seq;
+	struct mutex cmdbuf_mutex;
+
+	/**
+	 * GMR management. Protected by the lru spinlock.
+	 */
+
+	struct ida gmr_ida;
+	struct list_head gmr_lru;
+
+
+	/**
+	 * Operating mode.
+	 */
+
+	bool stealth;
+	bool is_opened;
+
+	/**
+	 * Master management.
+	 */
+
+	struct vmw_master *active_master;
+	struct vmw_master fbdev_master;
+	struct notifier_block pm_nb;
+};
+
+static inline struct vmw_private *vmw_priv(struct drm_device *dev)
+{
+	return (struct vmw_private *)dev->dev_private;
+}
+
+static inline struct vmw_fpriv *vmw_fpriv(struct drm_file *file_priv)
+{
+	return (struct vmw_fpriv *)file_priv->driver_priv;
+}
+
+static inline struct vmw_master *vmw_master(struct drm_master *master)
+{
+	return (struct vmw_master *) master->driver_priv;
+}
+
+static inline void vmw_write(struct vmw_private *dev_priv,
+			     unsigned int offset, uint32_t value)
+{
+	outl(offset, dev_priv->io_start + VMWGFX_INDEX_PORT);
+	outl(value, dev_priv->io_start + VMWGFX_VALUE_PORT);
+}
+
+static inline uint32_t vmw_read(struct vmw_private *dev_priv,
+				unsigned int offset)
+{
+	uint32_t val;
+
+	outl(offset, dev_priv->io_start + VMWGFX_INDEX_PORT);
+	val = inl(dev_priv->io_start + VMWGFX_VALUE_PORT);
+	return val;
+}
+
+/**
+ * GMR utilities - vmwgfx_gmr.c
+ */
+
+extern int vmw_gmr_bind(struct vmw_private *dev_priv,
+			struct ttm_buffer_object *bo);
+extern void vmw_gmr_unbind(struct vmw_private *dev_priv, int gmr_id);
+
+/**
+ * Resource utilities - vmwgfx_resource.c
+ */
+
+extern struct vmw_resource *vmw_context_alloc(struct vmw_private *dev_priv);
+extern void vmw_resource_unreference(struct vmw_resource **p_res);
+extern struct vmw_resource *vmw_resource_reference(struct vmw_resource *res);
+extern int vmw_context_destroy_ioctl(struct drm_device *dev, void *data,
+				     struct drm_file *file_priv);
+extern int vmw_context_define_ioctl(struct drm_device *dev, void *data,
+				    struct drm_file *file_priv);
+extern int vmw_context_check(struct vmw_private *dev_priv,
+			     struct ttm_object_file *tfile,
+			     int id);
+extern void vmw_surface_res_free(struct vmw_resource *res);
+extern int vmw_surface_init(struct vmw_private *dev_priv,
+			    struct vmw_surface *srf,
+			    void (*res_free) (struct vmw_resource *res));
+extern int vmw_user_surface_lookup_handle(struct vmw_private *dev_priv,
+					  struct ttm_object_file *tfile,
+					  uint32_t handle,
+					  struct vmw_surface **out);
+extern int vmw_surface_destroy_ioctl(struct drm_device *dev, void *data,
+				     struct drm_file *file_priv);
+extern int vmw_surface_define_ioctl(struct drm_device *dev, void *data,
+				    struct drm_file *file_priv);
+extern int vmw_surface_reference_ioctl(struct drm_device *dev, void *data,
+				       struct drm_file *file_priv);
+extern int vmw_surface_check(struct vmw_private *dev_priv,
+			     struct ttm_object_file *tfile,
+			     uint32_t handle, int *id);
+extern void vmw_dmabuf_bo_free(struct ttm_buffer_object *bo);
+extern int vmw_dmabuf_init(struct vmw_private *dev_priv,
+			   struct vmw_dma_buffer *vmw_bo,
+			   size_t size, struct ttm_placement *placement,
+			   bool interuptable,
+			   void (*bo_free) (struct ttm_buffer_object *bo));
+extern int vmw_dmabuf_alloc_ioctl(struct drm_device *dev, void *data,
+				  struct drm_file *file_priv);
+extern int vmw_dmabuf_unref_ioctl(struct drm_device *dev, void *data,
+				  struct drm_file *file_priv);
+extern uint32_t vmw_dmabuf_validate_node(struct ttm_buffer_object *bo,
+					 uint32_t cur_validate_node);
+extern void vmw_dmabuf_validate_clear(struct ttm_buffer_object *bo);
+extern int vmw_user_dmabuf_lookup(struct ttm_object_file *tfile,
+				  uint32_t id, struct vmw_dma_buffer **out);
+extern uint32_t vmw_dmabuf_gmr(struct ttm_buffer_object *bo);
+extern void vmw_dmabuf_set_gmr(struct ttm_buffer_object *bo, uint32_t id);
+extern int vmw_gmr_id_alloc(struct vmw_private *dev_priv, uint32_t *p_id);
+extern int vmw_dmabuf_to_start_of_vram(struct vmw_private *vmw_priv,
+				       struct vmw_dma_buffer *bo);
+extern int vmw_dmabuf_from_vram(struct vmw_private *vmw_priv,
+				struct vmw_dma_buffer *bo);
+extern void vmw_dmabuf_gmr_unbind(struct ttm_buffer_object *bo);
+extern int vmw_stream_claim_ioctl(struct drm_device *dev, void *data,
+				  struct drm_file *file_priv);
+extern int vmw_stream_unref_ioctl(struct drm_device *dev, void *data,
+				  struct drm_file *file_priv);
+extern int vmw_user_stream_lookup(struct vmw_private *dev_priv,
+				  struct ttm_object_file *tfile,
+				  uint32_t *inout_id,
+				  struct vmw_resource **out);
+
+
+/**
+ * Misc Ioctl functionality - vmwgfx_ioctl.c
+ */
+
+extern int vmw_getparam_ioctl(struct drm_device *dev, void *data,
+			      struct drm_file *file_priv);
+extern int vmw_fifo_debug_ioctl(struct drm_device *dev, void *data,
+				struct drm_file *file_priv);
+
+/**
+ * Fifo utilities - vmwgfx_fifo.c
+ */
+
+extern int vmw_fifo_init(struct vmw_private *dev_priv,
+			 struct vmw_fifo_state *fifo);
+extern void vmw_fifo_release(struct vmw_private *dev_priv,
+			     struct vmw_fifo_state *fifo);
+extern void *vmw_fifo_reserve(struct vmw_private *dev_priv, uint32_t bytes);
+extern void vmw_fifo_commit(struct vmw_private *dev_priv, uint32_t bytes);
+extern int vmw_fifo_send_fence(struct vmw_private *dev_priv,
+			       uint32_t *sequence);
+extern void vmw_fifo_ping_host(struct vmw_private *dev_priv, uint32_t reason);
+extern int vmw_fifo_mmap(struct file *filp, struct vm_area_struct *vma);
+extern bool vmw_fifo_have_3d(struct vmw_private *dev_priv);
+
+/**
+ * TTM glue - vmwgfx_ttm_glue.c
+ */
+
+extern int vmw_ttm_global_init(struct vmw_private *dev_priv);
+extern void vmw_ttm_global_release(struct vmw_private *dev_priv);
+extern int vmw_mmap(struct file *filp, struct vm_area_struct *vma);
+
+/**
+ * TTM buffer object driver - vmwgfx_buffer.c
+ */
+
+extern struct ttm_placement vmw_vram_placement;
+extern struct ttm_placement vmw_vram_ne_placement;
+extern struct ttm_placement vmw_vram_sys_placement;
+extern struct ttm_placement vmw_sys_placement;
+extern struct ttm_bo_driver vmw_bo_driver;
+extern int vmw_dma_quiescent(struct drm_device *dev);
+
+/**
+ * Command submission - vmwgfx_execbuf.c
+ */
+
+extern int vmw_execbuf_ioctl(struct drm_device *dev, void *data,
+			     struct drm_file *file_priv);
+
+/**
+ * IRQs and wating - vmwgfx_irq.c
+ */
+
+extern irqreturn_t vmw_irq_handler(DRM_IRQ_ARGS);
+extern int vmw_wait_fence(struct vmw_private *dev_priv, bool lazy,
+			  uint32_t sequence, bool interruptible,
+			  unsigned long timeout);
+extern void vmw_irq_preinstall(struct drm_device *dev);
+extern int vmw_irq_postinstall(struct drm_device *dev);
+extern void vmw_irq_uninstall(struct drm_device *dev);
+extern bool vmw_fence_signaled(struct vmw_private *dev_priv,
+			       uint32_t sequence);
+extern int vmw_fence_wait_ioctl(struct drm_device *dev, void *data,
+				struct drm_file *file_priv);
+extern int vmw_fallback_wait(struct vmw_private *dev_priv,
+			     bool lazy,
+			     bool fifo_idle,
+			     uint32_t sequence,
+			     bool interruptible,
+			     unsigned long timeout);
+
+/**
+ * Kernel framebuffer - vmwgfx_fb.c
+ */
+
+int vmw_fb_init(struct vmw_private *vmw_priv);
+int vmw_fb_close(struct vmw_private *dev_priv);
+int vmw_fb_off(struct vmw_private *vmw_priv);
+int vmw_fb_on(struct vmw_private *vmw_priv);
+
+/**
+ * Kernel modesetting - vmwgfx_kms.c
+ */
+
+int vmw_kms_init(struct vmw_private *dev_priv);
+int vmw_kms_close(struct vmw_private *dev_priv);
+int vmw_kms_save_vga(struct vmw_private *vmw_priv);
+int vmw_kms_restore_vga(struct vmw_private *vmw_priv);
+int vmw_kms_cursor_bypass_ioctl(struct drm_device *dev, void *data,
+				struct drm_file *file_priv);
+void vmw_kms_cursor_post_execbuf(struct vmw_private *dev_priv);
+void vmw_kms_cursor_snoop(struct vmw_surface *srf,
+			  struct ttm_object_file *tfile,
+			  struct ttm_buffer_object *bo,
+			  SVGA3dCmdHeader *header);
+
+/**
+ * Overlay control - vmwgfx_overlay.c
+ */
+
+int vmw_overlay_init(struct vmw_private *dev_priv);
+int vmw_overlay_close(struct vmw_private *dev_priv);
+int vmw_overlay_ioctl(struct drm_device *dev, void *data,
+		      struct drm_file *file_priv);
+int vmw_overlay_stop_all(struct vmw_private *dev_priv);
+int vmw_overlay_resume_all(struct vmw_private *dev_priv);
+int vmw_overlay_pause_all(struct vmw_private *dev_priv);
+int vmw_overlay_claim(struct vmw_private *dev_priv, uint32_t *out);
+int vmw_overlay_unref(struct vmw_private *dev_priv, uint32_t stream_id);
+int vmw_overlay_num_overlays(struct vmw_private *dev_priv);
+int vmw_overlay_num_free_overlays(struct vmw_private *dev_priv);
+
+/**
+ * Inline helper functions
+ */
+
+static inline void vmw_surface_unreference(struct vmw_surface **srf)
+{
+	struct vmw_surface *tmp_srf = *srf;
+	struct vmw_resource *res = &tmp_srf->res;
+	*srf = NULL;
+
+	vmw_resource_unreference(&res);
+}
+
+static inline struct vmw_surface *vmw_surface_reference(struct vmw_surface *srf)
+{
+	(void) vmw_resource_reference(&srf->res);
+	return srf;
+}
+
+static inline void vmw_dmabuf_unreference(struct vmw_dma_buffer **buf)
+{
+	struct vmw_dma_buffer *tmp_buf = *buf;
+	struct ttm_buffer_object *bo = &tmp_buf->base;
+	*buf = NULL;
+
+	ttm_bo_unref(&bo);
+}
+
+static inline struct vmw_dma_buffer *vmw_dmabuf_reference(struct vmw_dma_buffer *buf)
+{
+	if (ttm_bo_reference(&buf->base))
+		return buf;
+	return NULL;
+}
+
+#endif
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c b/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c
new file mode 100644
index 000000000000..0897359b3e4e
--- /dev/null
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c
@@ -0,0 +1,716 @@
+/**************************************************************************
+ *
+ * Copyright © 2009 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "vmwgfx_drv.h"
+#include "vmwgfx_reg.h"
+#include "ttm/ttm_bo_api.h"
+#include "ttm/ttm_placement.h"
+
+static int vmw_cmd_invalid(struct vmw_private *dev_priv,
+			   struct vmw_sw_context *sw_context,
+			   SVGA3dCmdHeader *header)
+{
+	return capable(CAP_SYS_ADMIN) ? : -EINVAL;
+}
+
+static int vmw_cmd_ok(struct vmw_private *dev_priv,
+		      struct vmw_sw_context *sw_context,
+		      SVGA3dCmdHeader *header)
+{
+	return 0;
+}
+
+static int vmw_cmd_cid_check(struct vmw_private *dev_priv,
+			     struct vmw_sw_context *sw_context,
+			     SVGA3dCmdHeader *header)
+{
+	struct vmw_cid_cmd {
+		SVGA3dCmdHeader header;
+		__le32 cid;
+	} *cmd;
+	int ret;
+
+	cmd = container_of(header, struct vmw_cid_cmd, header);
+	if (likely(sw_context->cid_valid && cmd->cid == sw_context->last_cid))
+		return 0;
+
+	ret = vmw_context_check(dev_priv, sw_context->tfile, cmd->cid);
+	if (unlikely(ret != 0)) {
+		DRM_ERROR("Could not find or use context %u\n",
+			  (unsigned) cmd->cid);
+		return ret;
+	}
+
+	sw_context->last_cid = cmd->cid;
+	sw_context->cid_valid = true;
+
+	return 0;
+}
+
+static int vmw_cmd_sid_check(struct vmw_private *dev_priv,
+			     struct vmw_sw_context *sw_context,
+			     uint32_t *sid)
+{
+	if (*sid == SVGA3D_INVALID_ID)
+		return 0;
+
+	if (unlikely((!sw_context->sid_valid  ||
+		      *sid != sw_context->last_sid))) {
+		int real_id;
+		int ret = vmw_surface_check(dev_priv, sw_context->tfile,
+					    *sid, &real_id);
+
+		if (unlikely(ret != 0)) {
+			DRM_ERROR("Could ot find or use surface 0x%08x "
+				  "address 0x%08lx\n",
+				  (unsigned int) *sid,
+				  (unsigned long) sid);
+			return ret;
+		}
+
+		sw_context->last_sid = *sid;
+		sw_context->sid_valid = true;
+		*sid = real_id;
+		sw_context->sid_translation = real_id;
+	} else
+		*sid = sw_context->sid_translation;
+
+	return 0;
+}
+
+
+static int vmw_cmd_set_render_target_check(struct vmw_private *dev_priv,
+					   struct vmw_sw_context *sw_context,
+					   SVGA3dCmdHeader *header)
+{
+	struct vmw_sid_cmd {
+		SVGA3dCmdHeader header;
+		SVGA3dCmdSetRenderTarget body;
+	} *cmd;
+	int ret;
+
+	ret = vmw_cmd_cid_check(dev_priv, sw_context, header);
+	if (unlikely(ret != 0))
+		return ret;
+
+	cmd = container_of(header, struct vmw_sid_cmd, header);
+	ret = vmw_cmd_sid_check(dev_priv, sw_context, &cmd->body.target.sid);
+	return ret;
+}
+
+static int vmw_cmd_surface_copy_check(struct vmw_private *dev_priv,
+				      struct vmw_sw_context *sw_context,
+				      SVGA3dCmdHeader *header)
+{
+	struct vmw_sid_cmd {
+		SVGA3dCmdHeader header;
+		SVGA3dCmdSurfaceCopy body;
+	} *cmd;
+	int ret;
+
+	cmd = container_of(header, struct vmw_sid_cmd, header);
+	ret = vmw_cmd_sid_check(dev_priv, sw_context, &cmd->body.src.sid);
+	if (unlikely(ret != 0))
+		return ret;
+	return vmw_cmd_sid_check(dev_priv, sw_context, &cmd->body.dest.sid);
+}
+
+static int vmw_cmd_stretch_blt_check(struct vmw_private *dev_priv,
+				     struct vmw_sw_context *sw_context,
+				     SVGA3dCmdHeader *header)
+{
+	struct vmw_sid_cmd {
+		SVGA3dCmdHeader header;
+		SVGA3dCmdSurfaceStretchBlt body;
+	} *cmd;
+	int ret;
+
+	cmd = container_of(header, struct vmw_sid_cmd, header);
+	ret = vmw_cmd_sid_check(dev_priv, sw_context, &cmd->body.src.sid);
+	if (unlikely(ret != 0))
+		return ret;
+	return vmw_cmd_sid_check(dev_priv, sw_context, &cmd->body.dest.sid);
+}
+
+static int vmw_cmd_blt_surf_screen_check(struct vmw_private *dev_priv,
+					 struct vmw_sw_context *sw_context,
+					 SVGA3dCmdHeader *header)
+{
+	struct vmw_sid_cmd {
+		SVGA3dCmdHeader header;
+		SVGA3dCmdBlitSurfaceToScreen body;
+	} *cmd;
+
+	cmd = container_of(header, struct vmw_sid_cmd, header);
+	return vmw_cmd_sid_check(dev_priv, sw_context, &cmd->body.srcImage.sid);
+}
+
+static int vmw_cmd_present_check(struct vmw_private *dev_priv,
+				 struct vmw_sw_context *sw_context,
+				 SVGA3dCmdHeader *header)
+{
+	struct vmw_sid_cmd {
+		SVGA3dCmdHeader header;
+		SVGA3dCmdPresent body;
+	} *cmd;
+
+	cmd = container_of(header, struct vmw_sid_cmd, header);
+	return vmw_cmd_sid_check(dev_priv, sw_context, &cmd->body.sid);
+}
+
+static int vmw_translate_guest_ptr(struct vmw_private *dev_priv,
+				   struct vmw_sw_context *sw_context,
+				   SVGAGuestPtr *ptr,
+				   struct vmw_dma_buffer **vmw_bo_p)
+{
+	struct vmw_dma_buffer *vmw_bo = NULL;
+	struct ttm_buffer_object *bo;
+	uint32_t handle = ptr->gmrId;
+	struct vmw_relocation *reloc;
+	uint32_t cur_validate_node;
+	struct ttm_validate_buffer *val_buf;
+	int ret;
+
+	ret = vmw_user_dmabuf_lookup(sw_context->tfile, handle, &vmw_bo);
+	if (unlikely(ret != 0)) {
+		DRM_ERROR("Could not find or use GMR region.\n");
+		return -EINVAL;
+	}
+	bo = &vmw_bo->base;
+
+	if (unlikely(sw_context->cur_reloc >= VMWGFX_MAX_RELOCATIONS)) {
+		DRM_ERROR("Max number relocations per submission"
+			  " exceeded\n");
+		ret = -EINVAL;
+		goto out_no_reloc;
+	}
+
+	reloc = &sw_context->relocs[sw_context->cur_reloc++];
+	reloc->location = ptr;
+
+	cur_validate_node = vmw_dmabuf_validate_node(bo, sw_context->cur_val_buf);
+	if (unlikely(cur_validate_node >= VMWGFX_MAX_GMRS)) {
+		DRM_ERROR("Max number of DMA buffers per submission"
+			  " exceeded.\n");
+		ret = -EINVAL;
+		goto out_no_reloc;
+	}
+
+	reloc->index = cur_validate_node;
+	if (unlikely(cur_validate_node == sw_context->cur_val_buf)) {
+		val_buf = &sw_context->val_bufs[cur_validate_node];
+		val_buf->bo = ttm_bo_reference(bo);
+		val_buf->new_sync_obj_arg = (void *) dev_priv;
+		list_add_tail(&val_buf->head, &sw_context->validate_nodes);
+		++sw_context->cur_val_buf;
+	}
+	*vmw_bo_p = vmw_bo;
+	return 0;
+
+out_no_reloc:
+	vmw_dmabuf_unreference(&vmw_bo);
+	vmw_bo_p = NULL;
+	return ret;
+}
+
+static int vmw_cmd_end_query(struct vmw_private *dev_priv,
+			     struct vmw_sw_context *sw_context,
+			     SVGA3dCmdHeader *header)
+{
+	struct vmw_dma_buffer *vmw_bo;
+	struct vmw_query_cmd {
+		SVGA3dCmdHeader header;
+		SVGA3dCmdEndQuery q;
+	} *cmd;
+	int ret;
+
+	cmd = container_of(header, struct vmw_query_cmd, header);
+	ret = vmw_cmd_cid_check(dev_priv, sw_context, header);
+	if (unlikely(ret != 0))
+		return ret;
+
+	ret = vmw_translate_guest_ptr(dev_priv, sw_context,
+				      &cmd->q.guestResult,
+				      &vmw_bo);
+	if (unlikely(ret != 0))
+		return ret;
+
+	vmw_dmabuf_unreference(&vmw_bo);
+	return 0;
+}
+
+static int vmw_cmd_wait_query(struct vmw_private *dev_priv,
+			      struct vmw_sw_context *sw_context,
+			      SVGA3dCmdHeader *header)
+{
+	struct vmw_dma_buffer *vmw_bo;
+	struct vmw_query_cmd {
+		SVGA3dCmdHeader header;
+		SVGA3dCmdWaitForQuery q;
+	} *cmd;
+	int ret;
+
+	cmd = container_of(header, struct vmw_query_cmd, header);
+	ret = vmw_cmd_cid_check(dev_priv, sw_context, header);
+	if (unlikely(ret != 0))
+		return ret;
+
+	ret = vmw_translate_guest_ptr(dev_priv, sw_context,
+				      &cmd->q.guestResult,
+				      &vmw_bo);
+	if (unlikely(ret != 0))
+		return ret;
+
+	vmw_dmabuf_unreference(&vmw_bo);
+	return 0;
+}
+
+
+static int vmw_cmd_dma(struct vmw_private *dev_priv,
+		       struct vmw_sw_context *sw_context,
+		       SVGA3dCmdHeader *header)
+{
+	struct vmw_dma_buffer *vmw_bo = NULL;
+	struct ttm_buffer_object *bo;
+	struct vmw_surface *srf = NULL;
+	struct vmw_dma_cmd {
+		SVGA3dCmdHeader header;
+		SVGA3dCmdSurfaceDMA dma;
+	} *cmd;
+	int ret;
+
+	cmd = container_of(header, struct vmw_dma_cmd, header);
+	ret = vmw_translate_guest_ptr(dev_priv, sw_context,
+				      &cmd->dma.guest.ptr,
+				      &vmw_bo);
+	if (unlikely(ret != 0))
+		return ret;
+
+	bo = &vmw_bo->base;
+	ret = vmw_user_surface_lookup_handle(dev_priv, sw_context->tfile,
+					     cmd->dma.host.sid, &srf);
+	if (ret) {
+		DRM_ERROR("could not find surface\n");
+		goto out_no_reloc;
+	}
+
+	/**
+	 * Patch command stream with device SID.
+	 */
+
+	cmd->dma.host.sid = srf->res.id;
+	vmw_kms_cursor_snoop(srf, sw_context->tfile, bo, header);
+	/**
+	 * FIXME: May deadlock here when called from the
+	 * command parsing code.
+	 */
+	vmw_surface_unreference(&srf);
+
+out_no_reloc:
+	vmw_dmabuf_unreference(&vmw_bo);
+	return ret;
+}
+
+static int vmw_cmd_draw(struct vmw_private *dev_priv,
+			struct vmw_sw_context *sw_context,
+			SVGA3dCmdHeader *header)
+{
+	struct vmw_draw_cmd {
+		SVGA3dCmdHeader header;
+		SVGA3dCmdDrawPrimitives body;
+	} *cmd;
+	SVGA3dVertexDecl *decl = (SVGA3dVertexDecl *)(
+		(unsigned long)header + sizeof(*cmd));
+	SVGA3dPrimitiveRange *range;
+	uint32_t i;
+	uint32_t maxnum;
+	int ret;
+
+	ret = vmw_cmd_cid_check(dev_priv, sw_context, header);
+	if (unlikely(ret != 0))
+		return ret;
+
+	cmd = container_of(header, struct vmw_draw_cmd, header);
+	maxnum = (header->size - sizeof(cmd->body)) / sizeof(*decl);
+
+	if (unlikely(cmd->body.numVertexDecls > maxnum)) {
+		DRM_ERROR("Illegal number of vertex declarations.\n");
+		return -EINVAL;
+	}
+
+	for (i = 0; i < cmd->body.numVertexDecls; ++i, ++decl) {
+		ret = vmw_cmd_sid_check(dev_priv, sw_context,
+					&decl->array.surfaceId);
+		if (unlikely(ret != 0))
+			return ret;
+	}
+
+	maxnum = (header->size - sizeof(cmd->body) -
+		  cmd->body.numVertexDecls * sizeof(*decl)) / sizeof(*range);
+	if (unlikely(cmd->body.numRanges > maxnum)) {
+		DRM_ERROR("Illegal number of index ranges.\n");
+		return -EINVAL;
+	}
+
+	range = (SVGA3dPrimitiveRange *) decl;
+	for (i = 0; i < cmd->body.numRanges; ++i, ++range) {
+		ret = vmw_cmd_sid_check(dev_priv, sw_context,
+					&range->indexArray.surfaceId);
+		if (unlikely(ret != 0))
+			return ret;
+	}
+	return 0;
+}
+
+
+static int vmw_cmd_tex_state(struct vmw_private *dev_priv,
+			     struct vmw_sw_context *sw_context,
+			     SVGA3dCmdHeader *header)
+{
+	struct vmw_tex_state_cmd {
+		SVGA3dCmdHeader header;
+		SVGA3dCmdSetTextureState state;
+	};
+
+	SVGA3dTextureState *last_state = (SVGA3dTextureState *)
+	  ((unsigned long) header + header->size + sizeof(header));
+	SVGA3dTextureState *cur_state = (SVGA3dTextureState *)
+		((unsigned long) header + sizeof(struct vmw_tex_state_cmd));
+	int ret;
+
+	ret = vmw_cmd_cid_check(dev_priv, sw_context, header);
+	if (unlikely(ret != 0))
+		return ret;
+
+	for (; cur_state < last_state; ++cur_state) {
+		if (likely(cur_state->name != SVGA3D_TS_BIND_TEXTURE))
+			continue;
+
+		ret = vmw_cmd_sid_check(dev_priv, sw_context,
+					&cur_state->value);
+		if (unlikely(ret != 0))
+			return ret;
+	}
+
+	return 0;
+}
+
+
+typedef int (*vmw_cmd_func) (struct vmw_private *,
+			     struct vmw_sw_context *,
+			     SVGA3dCmdHeader *);
+
+#define VMW_CMD_DEF(cmd, func) \
+	[cmd - SVGA_3D_CMD_BASE] = func
+
+static vmw_cmd_func vmw_cmd_funcs[SVGA_3D_CMD_MAX] = {
+	VMW_CMD_DEF(SVGA_3D_CMD_SURFACE_DEFINE, &vmw_cmd_invalid),
+	VMW_CMD_DEF(SVGA_3D_CMD_SURFACE_DESTROY, &vmw_cmd_invalid),
+	VMW_CMD_DEF(SVGA_3D_CMD_SURFACE_COPY, &vmw_cmd_surface_copy_check),
+	VMW_CMD_DEF(SVGA_3D_CMD_SURFACE_STRETCHBLT, &vmw_cmd_stretch_blt_check),
+	VMW_CMD_DEF(SVGA_3D_CMD_SURFACE_DMA, &vmw_cmd_dma),
+	VMW_CMD_DEF(SVGA_3D_CMD_CONTEXT_DEFINE, &vmw_cmd_invalid),
+	VMW_CMD_DEF(SVGA_3D_CMD_CONTEXT_DESTROY, &vmw_cmd_invalid),
+	VMW_CMD_DEF(SVGA_3D_CMD_SETTRANSFORM, &vmw_cmd_cid_check),
+	VMW_CMD_DEF(SVGA_3D_CMD_SETZRANGE, &vmw_cmd_cid_check),
+	VMW_CMD_DEF(SVGA_3D_CMD_SETRENDERSTATE, &vmw_cmd_cid_check),
+	VMW_CMD_DEF(SVGA_3D_CMD_SETRENDERTARGET,
+		    &vmw_cmd_set_render_target_check),
+	VMW_CMD_DEF(SVGA_3D_CMD_SETTEXTURESTATE, &vmw_cmd_tex_state),
+	VMW_CMD_DEF(SVGA_3D_CMD_SETMATERIAL, &vmw_cmd_cid_check),
+	VMW_CMD_DEF(SVGA_3D_CMD_SETLIGHTDATA, &vmw_cmd_cid_check),
+	VMW_CMD_DEF(SVGA_3D_CMD_SETLIGHTENABLED, &vmw_cmd_cid_check),
+	VMW_CMD_DEF(SVGA_3D_CMD_SETVIEWPORT, &vmw_cmd_cid_check),
+	VMW_CMD_DEF(SVGA_3D_CMD_SETCLIPPLANE, &vmw_cmd_cid_check),
+	VMW_CMD_DEF(SVGA_3D_CMD_CLEAR, &vmw_cmd_cid_check),
+	VMW_CMD_DEF(SVGA_3D_CMD_PRESENT, &vmw_cmd_present_check),
+	VMW_CMD_DEF(SVGA_3D_CMD_SHADER_DEFINE, &vmw_cmd_cid_check),
+	VMW_CMD_DEF(SVGA_3D_CMD_SHADER_DESTROY, &vmw_cmd_cid_check),
+	VMW_CMD_DEF(SVGA_3D_CMD_SET_SHADER, &vmw_cmd_cid_check),
+	VMW_CMD_DEF(SVGA_3D_CMD_SET_SHADER_CONST, &vmw_cmd_cid_check),
+	VMW_CMD_DEF(SVGA_3D_CMD_DRAW_PRIMITIVES, &vmw_cmd_draw),
+	VMW_CMD_DEF(SVGA_3D_CMD_SETSCISSORRECT, &vmw_cmd_cid_check),
+	VMW_CMD_DEF(SVGA_3D_CMD_BEGIN_QUERY, &vmw_cmd_cid_check),
+	VMW_CMD_DEF(SVGA_3D_CMD_END_QUERY, &vmw_cmd_end_query),
+	VMW_CMD_DEF(SVGA_3D_CMD_WAIT_FOR_QUERY, &vmw_cmd_wait_query),
+	VMW_CMD_DEF(SVGA_3D_CMD_PRESENT_READBACK, &vmw_cmd_ok),
+	VMW_CMD_DEF(SVGA_3D_CMD_BLIT_SURFACE_TO_SCREEN,
+		    &vmw_cmd_blt_surf_screen_check)
+};
+
+static int vmw_cmd_check(struct vmw_private *dev_priv,
+			 struct vmw_sw_context *sw_context,
+			 void *buf, uint32_t *size)
+{
+	uint32_t cmd_id;
+	uint32_t size_remaining = *size;
+	SVGA3dCmdHeader *header = (SVGA3dCmdHeader *) buf;
+	int ret;
+
+	cmd_id = ((uint32_t *)buf)[0];
+	if (cmd_id == SVGA_CMD_UPDATE) {
+		*size = 5 << 2;
+		return 0;
+	}
+
+	cmd_id = le32_to_cpu(header->id);
+	*size = le32_to_cpu(header->size) + sizeof(SVGA3dCmdHeader);
+
+	cmd_id -= SVGA_3D_CMD_BASE;
+	if (unlikely(*size > size_remaining))
+		goto out_err;
+
+	if (unlikely(cmd_id >= SVGA_3D_CMD_MAX - SVGA_3D_CMD_BASE))
+		goto out_err;
+
+	ret = vmw_cmd_funcs[cmd_id](dev_priv, sw_context, header);
+	if (unlikely(ret != 0))
+		goto out_err;
+
+	return 0;
+out_err:
+	DRM_ERROR("Illegal / Invalid SVGA3D command: %d\n",
+		  cmd_id + SVGA_3D_CMD_BASE);
+	return -EINVAL;
+}
+
+static int vmw_cmd_check_all(struct vmw_private *dev_priv,
+			     struct vmw_sw_context *sw_context,
+			     void *buf, uint32_t size)
+{
+	int32_t cur_size = size;
+	int ret;
+
+	while (cur_size > 0) {
+		size = cur_size;
+		ret = vmw_cmd_check(dev_priv, sw_context, buf, &size);
+		if (unlikely(ret != 0))
+			return ret;
+		buf = (void *)((unsigned long) buf + size);
+		cur_size -= size;
+	}
+
+	if (unlikely(cur_size != 0)) {
+		DRM_ERROR("Command verifier out of sync.\n");
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static void vmw_free_relocations(struct vmw_sw_context *sw_context)
+{
+	sw_context->cur_reloc = 0;
+}
+
+static void vmw_apply_relocations(struct vmw_sw_context *sw_context)
+{
+	uint32_t i;
+	struct vmw_relocation *reloc;
+	struct ttm_validate_buffer *validate;
+	struct ttm_buffer_object *bo;
+
+	for (i = 0; i < sw_context->cur_reloc; ++i) {
+		reloc = &sw_context->relocs[i];
+		validate = &sw_context->val_bufs[reloc->index];
+		bo = validate->bo;
+		reloc->location->offset += bo->offset;
+		reloc->location->gmrId = vmw_dmabuf_gmr(bo);
+	}
+	vmw_free_relocations(sw_context);
+}
+
+static void vmw_clear_validations(struct vmw_sw_context *sw_context)
+{
+	struct ttm_validate_buffer *entry, *next;
+
+	list_for_each_entry_safe(entry, next, &sw_context->validate_nodes,
+				 head) {
+		list_del(&entry->head);
+		vmw_dmabuf_validate_clear(entry->bo);
+		ttm_bo_unref(&entry->bo);
+		sw_context->cur_val_buf--;
+	}
+	BUG_ON(sw_context->cur_val_buf != 0);
+}
+
+static int vmw_validate_single_buffer(struct vmw_private *dev_priv,
+				      struct ttm_buffer_object *bo)
+{
+	int ret;
+
+	if (vmw_dmabuf_gmr(bo) != SVGA_GMR_NULL)
+		return 0;
+
+	/**
+	 * Put BO in VRAM, only if there is space.
+	 */
+
+	ret = ttm_bo_validate(bo, &vmw_vram_sys_placement, true, false);
+	if (unlikely(ret == -ERESTARTSYS))
+		return ret;
+
+	/**
+	 * Otherwise, set it up as GMR.
+	 */
+
+	if (vmw_dmabuf_gmr(bo) != SVGA_GMR_NULL)
+		return 0;
+
+	ret = vmw_gmr_bind(dev_priv, bo);
+	if (likely(ret == 0 || ret == -ERESTARTSYS))
+		return ret;
+
+	/**
+	 * If that failed, try VRAM again, this time evicting
+	 * previous contents.
+	 */
+
+	ret = ttm_bo_validate(bo, &vmw_vram_placement, true, false);
+	return ret;
+}
+
+
+static int vmw_validate_buffers(struct vmw_private *dev_priv,
+				struct vmw_sw_context *sw_context)
+{
+	struct ttm_validate_buffer *entry;
+	int ret;
+
+	list_for_each_entry(entry, &sw_context->validate_nodes, head) {
+		ret = vmw_validate_single_buffer(dev_priv, entry->bo);
+		if (unlikely(ret != 0))
+			return ret;
+	}
+	return 0;
+}
+
+int vmw_execbuf_ioctl(struct drm_device *dev, void *data,
+		      struct drm_file *file_priv)
+{
+	struct vmw_private *dev_priv = vmw_priv(dev);
+	struct drm_vmw_execbuf_arg *arg = (struct drm_vmw_execbuf_arg *)data;
+	struct drm_vmw_fence_rep fence_rep;
+	struct drm_vmw_fence_rep __user *user_fence_rep;
+	int ret;
+	void *user_cmd;
+	void *cmd;
+	uint32_t sequence;
+	struct vmw_sw_context *sw_context = &dev_priv->ctx;
+	struct vmw_master *vmaster = vmw_master(file_priv->master);
+
+	ret = ttm_read_lock(&vmaster->lock, true);
+	if (unlikely(ret != 0))
+		return ret;
+
+	ret = mutex_lock_interruptible(&dev_priv->cmdbuf_mutex);
+	if (unlikely(ret != 0)) {
+		ret = -ERESTARTSYS;
+		goto out_no_cmd_mutex;
+	}
+
+	cmd = vmw_fifo_reserve(dev_priv, arg->command_size);
+	if (unlikely(cmd == NULL)) {
+		DRM_ERROR("Failed reserving fifo space for commands.\n");
+		ret = -ENOMEM;
+		goto out_unlock;
+	}
+
+	user_cmd = (void __user *)(unsigned long)arg->commands;
+	ret = copy_from_user(cmd, user_cmd, arg->command_size);
+
+	if (unlikely(ret != 0)) {
+		DRM_ERROR("Failed copying commands.\n");
+		goto out_commit;
+	}
+
+	sw_context->tfile = vmw_fpriv(file_priv)->tfile;
+	sw_context->cid_valid = false;
+	sw_context->sid_valid = false;
+	sw_context->cur_reloc = 0;
+	sw_context->cur_val_buf = 0;
+
+	INIT_LIST_HEAD(&sw_context->validate_nodes);
+
+	ret = vmw_cmd_check_all(dev_priv, sw_context, cmd, arg->command_size);
+	if (unlikely(ret != 0))
+		goto out_err;
+	ret = ttm_eu_reserve_buffers(&sw_context->validate_nodes,
+				     dev_priv->val_seq++);
+	if (unlikely(ret != 0))
+		goto out_err;
+
+	ret = vmw_validate_buffers(dev_priv, sw_context);
+	if (unlikely(ret != 0))
+		goto out_err;
+
+	vmw_apply_relocations(sw_context);
+	vmw_fifo_commit(dev_priv, arg->command_size);
+
+	ret = vmw_fifo_send_fence(dev_priv, &sequence);
+
+	ttm_eu_fence_buffer_objects(&sw_context->validate_nodes,
+				    (void *)(unsigned long) sequence);
+	vmw_clear_validations(sw_context);
+	mutex_unlock(&dev_priv->cmdbuf_mutex);
+
+	/*
+	 * This error is harmless, because if fence submission fails,
+	 * vmw_fifo_send_fence will sync.
+	 */
+
+	if (ret != 0)
+		DRM_ERROR("Fence submission error. Syncing.\n");
+
+	fence_rep.error = ret;
+	fence_rep.fence_seq = (uint64_t) sequence;
+
+	user_fence_rep = (struct drm_vmw_fence_rep __user *)
+	    (unsigned long)arg->fence_rep;
+
+	/*
+	 * copy_to_user errors will be detected by user space not
+	 * seeing fence_rep::error filled in.
+	 */
+
+	ret = copy_to_user(user_fence_rep, &fence_rep, sizeof(fence_rep));
+
+	vmw_kms_cursor_post_execbuf(dev_priv);
+	ttm_read_unlock(&vmaster->lock);
+	return 0;
+out_err:
+	vmw_free_relocations(sw_context);
+	ttm_eu_backoff_reservation(&sw_context->validate_nodes);
+	vmw_clear_validations(sw_context);
+out_commit:
+	vmw_fifo_commit(dev_priv, 0);
+out_unlock:
+	mutex_unlock(&dev_priv->cmdbuf_mutex);
+out_no_cmd_mutex:
+	ttm_read_unlock(&vmaster->lock);
+	return ret;
+}
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_fb.c b/drivers/gpu/drm/vmwgfx/vmwgfx_fb.c
new file mode 100644
index 000000000000..a93367041cdc
--- /dev/null
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_fb.c
@@ -0,0 +1,737 @@
+/**************************************************************************
+ *
+ * Copyright © 2007 David Airlie
+ * Copyright © 2009 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "drmP.h"
+#include "vmwgfx_drv.h"
+
+#include "ttm/ttm_placement.h"
+
+#define VMW_DIRTY_DELAY (HZ / 30)
+
+struct vmw_fb_par {
+	struct vmw_private *vmw_priv;
+
+	void *vmalloc;
+
+	struct vmw_dma_buffer *vmw_bo;
+	struct ttm_bo_kmap_obj map;
+
+	u32 pseudo_palette[17];
+
+	unsigned depth;
+	unsigned bpp;
+
+	unsigned max_width;
+	unsigned max_height;
+
+	void *bo_ptr;
+	unsigned bo_size;
+	bool bo_iowrite;
+
+	struct {
+		spinlock_t lock;
+		bool active;
+		unsigned x1;
+		unsigned y1;
+		unsigned x2;
+		unsigned y2;
+	} dirty;
+};
+
+static int vmw_fb_setcolreg(unsigned regno, unsigned red, unsigned green,
+			    unsigned blue, unsigned transp,
+			    struct fb_info *info)
+{
+	struct vmw_fb_par *par = info->par;
+	u32 *pal = par->pseudo_palette;
+
+	if (regno > 15) {
+		DRM_ERROR("Bad regno %u.\n", regno);
+		return 1;
+	}
+
+	switch (par->depth) {
+	case 24:
+	case 32:
+		pal[regno] = ((red & 0xff00) << 8) |
+			      (green & 0xff00) |
+			     ((blue  & 0xff00) >> 8);
+		break;
+	default:
+		DRM_ERROR("Bad depth %u, bpp %u.\n", par->depth, par->bpp);
+		return 1;
+	}
+
+	return 0;
+}
+
+static int vmw_fb_check_var(struct fb_var_screeninfo *var,
+			    struct fb_info *info)
+{
+	int depth = var->bits_per_pixel;
+	struct vmw_fb_par *par = info->par;
+	struct vmw_private *vmw_priv = par->vmw_priv;
+
+	switch (var->bits_per_pixel) {
+	case 32:
+		depth = (var->transp.length > 0) ? 32 : 24;
+		break;
+	default:
+		DRM_ERROR("Bad bpp %u.\n", var->bits_per_pixel);
+		return -EINVAL;
+	}
+
+	switch (depth) {
+	case 24:
+		var->red.offset = 16;
+		var->green.offset = 8;
+		var->blue.offset = 0;
+		var->red.length = 8;
+		var->green.length = 8;
+		var->blue.length = 8;
+		var->transp.length = 0;
+		var->transp.offset = 0;
+		break;
+	case 32:
+		var->red.offset = 16;
+		var->green.offset = 8;
+		var->blue.offset = 0;
+		var->red.length = 8;
+		var->green.length = 8;
+		var->blue.length = 8;
+		var->transp.length = 8;
+		var->transp.offset = 24;
+		break;
+	default:
+		DRM_ERROR("Bad depth %u.\n", depth);
+		return -EINVAL;
+	}
+
+	/* without multimon its hard to resize */
+	if (!(vmw_priv->capabilities & SVGA_CAP_MULTIMON) &&
+	    (var->xres != par->max_width ||
+	     var->yres != par->max_height)) {
+		DRM_ERROR("Tried to resize, but we don't have multimon\n");
+		return -EINVAL;
+	}
+
+	if (var->xres > par->max_width ||
+	    var->yres > par->max_height) {
+		DRM_ERROR("Requested geom can not fit in framebuffer\n");
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static int vmw_fb_set_par(struct fb_info *info)
+{
+	struct vmw_fb_par *par = info->par;
+	struct vmw_private *vmw_priv = par->vmw_priv;
+
+	if (vmw_priv->capabilities & SVGA_CAP_MULTIMON) {
+		vmw_write(vmw_priv, SVGA_REG_NUM_GUEST_DISPLAYS, 1);
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_ID, 0);
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_IS_PRIMARY, true);
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_POSITION_X, 0);
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_POSITION_Y, 0);
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_WIDTH, 0);
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_HEIGHT, 0);
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_ID, SVGA_ID_INVALID);
+
+		vmw_write(vmw_priv, SVGA_REG_ENABLE, 1);
+		vmw_write(vmw_priv, SVGA_REG_WIDTH, par->max_width);
+		vmw_write(vmw_priv, SVGA_REG_HEIGHT, par->max_height);
+		vmw_write(vmw_priv, SVGA_REG_BITS_PER_PIXEL, par->bpp);
+		vmw_write(vmw_priv, SVGA_REG_DEPTH, par->depth);
+		vmw_write(vmw_priv, SVGA_REG_RED_MASK, 0x00ff0000);
+		vmw_write(vmw_priv, SVGA_REG_GREEN_MASK, 0x0000ff00);
+		vmw_write(vmw_priv, SVGA_REG_BLUE_MASK, 0x000000ff);
+
+		/* TODO check if pitch and offset changes */
+
+		vmw_write(vmw_priv, SVGA_REG_NUM_GUEST_DISPLAYS, 1);
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_ID, 0);
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_IS_PRIMARY, true);
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_POSITION_X, info->var.xoffset);
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_POSITION_Y, info->var.yoffset);
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_WIDTH, info->var.xres);
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_HEIGHT, info->var.yres);
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_ID, SVGA_ID_INVALID);
+	} else {
+		vmw_write(vmw_priv, SVGA_REG_WIDTH, info->var.xres);
+		vmw_write(vmw_priv, SVGA_REG_HEIGHT, info->var.yres);
+
+		/* TODO check if pitch and offset changes */
+	}
+
+	return 0;
+}
+
+static int vmw_fb_pan_display(struct fb_var_screeninfo *var,
+			      struct fb_info *info)
+{
+	return 0;
+}
+
+static int vmw_fb_blank(int blank, struct fb_info *info)
+{
+	return 0;
+}
+
+/*
+ * Dirty code
+ */
+
+static void vmw_fb_dirty_flush(struct vmw_fb_par *par)
+{
+	struct vmw_private *vmw_priv = par->vmw_priv;
+	struct fb_info *info = vmw_priv->fb_info;
+	int stride = (info->fix.line_length / 4);
+	int *src = (int *)info->screen_base;
+	__le32 __iomem *vram_mem = par->bo_ptr;
+	unsigned long flags;
+	unsigned x, y, w, h;
+	int i, k;
+	struct {
+		uint32_t header;
+		SVGAFifoCmdUpdate body;
+	} *cmd;
+
+	spin_lock_irqsave(&par->dirty.lock, flags);
+	if (!par->dirty.active) {
+		spin_unlock_irqrestore(&par->dirty.lock, flags);
+		return;
+	}
+	x = par->dirty.x1;
+	y = par->dirty.y1;
+	w = min(par->dirty.x2, info->var.xres) - x;
+	h = min(par->dirty.y2, info->var.yres) - y;
+	par->dirty.x1 = par->dirty.x2 = 0;
+	par->dirty.y1 = par->dirty.y2 = 0;
+	spin_unlock_irqrestore(&par->dirty.lock, flags);
+
+	for (i = y * stride; i < info->fix.smem_len / 4; i += stride) {
+		for (k = i+x; k < i+x+w && k < info->fix.smem_len / 4; k++)
+			iowrite32(src[k], vram_mem + k);
+	}
+
+#if 0
+	DRM_INFO("%s, (%u, %u) (%ux%u)\n", __func__, x, y, w, h);
+#endif
+
+	cmd = vmw_fifo_reserve(vmw_priv, sizeof(*cmd));
+	if (unlikely(cmd == NULL)) {
+		DRM_ERROR("Fifo reserve failed.\n");
+		return;
+	}
+
+	cmd->header = cpu_to_le32(SVGA_CMD_UPDATE);
+	cmd->body.x = cpu_to_le32(x);
+	cmd->body.y = cpu_to_le32(y);
+	cmd->body.width = cpu_to_le32(w);
+	cmd->body.height = cpu_to_le32(h);
+	vmw_fifo_commit(vmw_priv, sizeof(*cmd));
+}
+
+static void vmw_fb_dirty_mark(struct vmw_fb_par *par,
+			      unsigned x1, unsigned y1,
+			      unsigned width, unsigned height)
+{
+	struct fb_info *info = par->vmw_priv->fb_info;
+	unsigned long flags;
+	unsigned x2 = x1 + width;
+	unsigned y2 = y1 + height;
+
+	spin_lock_irqsave(&par->dirty.lock, flags);
+	if (par->dirty.x1 == par->dirty.x2) {
+		par->dirty.x1 = x1;
+		par->dirty.y1 = y1;
+		par->dirty.x2 = x2;
+		par->dirty.y2 = y2;
+		/* if we are active start the dirty work
+		 * we share the work with the defio system */
+		if (par->dirty.active)
+			schedule_delayed_work(&info->deferred_work, VMW_DIRTY_DELAY);
+	} else {
+		if (x1 < par->dirty.x1)
+			par->dirty.x1 = x1;
+		if (y1 < par->dirty.y1)
+			par->dirty.y1 = y1;
+		if (x2 > par->dirty.x2)
+			par->dirty.x2 = x2;
+		if (y2 > par->dirty.y2)
+			par->dirty.y2 = y2;
+	}
+	spin_unlock_irqrestore(&par->dirty.lock, flags);
+}
+
+static void vmw_deferred_io(struct fb_info *info,
+			    struct list_head *pagelist)
+{
+	struct vmw_fb_par *par = info->par;
+	unsigned long start, end, min, max;
+	unsigned long flags;
+	struct page *page;
+	int y1, y2;
+
+	min = ULONG_MAX;
+	max = 0;
+	list_for_each_entry(page, pagelist, lru) {
+		start = page->index << PAGE_SHIFT;
+		end = start + PAGE_SIZE - 1;
+		min = min(min, start);
+		max = max(max, end);
+	}
+
+	if (min < max) {
+		y1 = min / info->fix.line_length;
+		y2 = (max / info->fix.line_length) + 1;
+
+		spin_lock_irqsave(&par->dirty.lock, flags);
+		par->dirty.x1 = 0;
+		par->dirty.y1 = y1;
+		par->dirty.x2 = info->var.xres;
+		par->dirty.y2 = y2;
+		spin_unlock_irqrestore(&par->dirty.lock, flags);
+	}
+
+	vmw_fb_dirty_flush(par);
+};
+
+struct fb_deferred_io vmw_defio = {
+	.delay		= VMW_DIRTY_DELAY,
+	.deferred_io	= vmw_deferred_io,
+};
+
+/*
+ * Draw code
+ */
+
+static void vmw_fb_fillrect(struct fb_info *info, const struct fb_fillrect *rect)
+{
+	cfb_fillrect(info, rect);
+	vmw_fb_dirty_mark(info->par, rect->dx, rect->dy,
+			  rect->width, rect->height);
+}
+
+static void vmw_fb_copyarea(struct fb_info *info, const struct fb_copyarea *region)
+{
+	cfb_copyarea(info, region);
+	vmw_fb_dirty_mark(info->par, region->dx, region->dy,
+			  region->width, region->height);
+}
+
+static void vmw_fb_imageblit(struct fb_info *info, const struct fb_image *image)
+{
+	cfb_imageblit(info, image);
+	vmw_fb_dirty_mark(info->par, image->dx, image->dy,
+			  image->width, image->height);
+}
+
+/*
+ * Bring up code
+ */
+
+static struct fb_ops vmw_fb_ops = {
+	.owner = THIS_MODULE,
+	.fb_check_var = vmw_fb_check_var,
+	.fb_set_par = vmw_fb_set_par,
+	.fb_setcolreg = vmw_fb_setcolreg,
+	.fb_fillrect = vmw_fb_fillrect,
+	.fb_copyarea = vmw_fb_copyarea,
+	.fb_imageblit = vmw_fb_imageblit,
+	.fb_pan_display = vmw_fb_pan_display,
+	.fb_blank = vmw_fb_blank,
+};
+
+static int vmw_fb_create_bo(struct vmw_private *vmw_priv,
+			    size_t size, struct vmw_dma_buffer **out)
+{
+	struct vmw_dma_buffer *vmw_bo;
+	struct ttm_placement ne_placement = vmw_vram_ne_placement;
+	int ret;
+
+	ne_placement.lpfn = (size + PAGE_SIZE - 1) >> PAGE_SHIFT;
+
+	/* interuptable? */
+	ret = ttm_write_lock(&vmw_priv->fbdev_master.lock, false);
+	if (unlikely(ret != 0))
+		return ret;
+
+	vmw_bo = kmalloc(sizeof(*vmw_bo), GFP_KERNEL);
+	if (!vmw_bo)
+		goto err_unlock;
+
+	ret = vmw_dmabuf_init(vmw_priv, vmw_bo, size,
+			      &ne_placement,
+			      false,
+			      &vmw_dmabuf_bo_free);
+	if (unlikely(ret != 0))
+		goto err_unlock; /* init frees the buffer on failure */
+
+	*out = vmw_bo;
+
+	ttm_write_unlock(&vmw_priv->fbdev_master.lock);
+
+	return 0;
+
+err_unlock:
+	ttm_write_unlock(&vmw_priv->fbdev_master.lock);
+	return ret;
+}
+
+int vmw_fb_init(struct vmw_private *vmw_priv)
+{
+	struct device *device = &vmw_priv->dev->pdev->dev;
+	struct vmw_fb_par *par;
+	struct fb_info *info;
+	unsigned initial_width, initial_height;
+	unsigned fb_width, fb_height;
+	unsigned fb_bbp, fb_depth, fb_offset, fb_pitch, fb_size;
+	int ret;
+
+	initial_width = 800;
+	initial_height = 600;
+
+	fb_bbp = 32;
+	fb_depth = 24;
+
+	if (vmw_priv->capabilities & SVGA_CAP_MULTIMON) {
+		fb_width = min(vmw_priv->fb_max_width, (unsigned)2048);
+		fb_height = min(vmw_priv->fb_max_height, (unsigned)2048);
+	} else {
+		fb_width = min(vmw_priv->fb_max_width, initial_width);
+		fb_height = min(vmw_priv->fb_max_height, initial_height);
+	}
+
+	initial_width = min(fb_width, initial_width);
+	initial_height = min(fb_height, initial_height);
+
+	vmw_write(vmw_priv, SVGA_REG_WIDTH, fb_width);
+	vmw_write(vmw_priv, SVGA_REG_HEIGHT, fb_height);
+	vmw_write(vmw_priv, SVGA_REG_BITS_PER_PIXEL, fb_bbp);
+	vmw_write(vmw_priv, SVGA_REG_DEPTH, fb_depth);
+	vmw_write(vmw_priv, SVGA_REG_RED_MASK, 0x00ff0000);
+	vmw_write(vmw_priv, SVGA_REG_GREEN_MASK, 0x0000ff00);
+	vmw_write(vmw_priv, SVGA_REG_BLUE_MASK, 0x000000ff);
+
+	fb_size = vmw_read(vmw_priv, SVGA_REG_FB_SIZE);
+	fb_offset = vmw_read(vmw_priv, SVGA_REG_FB_OFFSET);
+	fb_pitch = vmw_read(vmw_priv, SVGA_REG_BYTES_PER_LINE);
+
+	DRM_DEBUG("width  %u\n", vmw_read(vmw_priv, SVGA_REG_MAX_WIDTH));
+	DRM_DEBUG("height %u\n", vmw_read(vmw_priv, SVGA_REG_MAX_HEIGHT));
+	DRM_DEBUG("width  %u\n", vmw_read(vmw_priv, SVGA_REG_WIDTH));
+	DRM_DEBUG("height %u\n", vmw_read(vmw_priv, SVGA_REG_HEIGHT));
+	DRM_DEBUG("bpp    %u\n", vmw_read(vmw_priv, SVGA_REG_BITS_PER_PIXEL));
+	DRM_DEBUG("depth  %u\n", vmw_read(vmw_priv, SVGA_REG_DEPTH));
+	DRM_DEBUG("bpl    %u\n", vmw_read(vmw_priv, SVGA_REG_BYTES_PER_LINE));
+	DRM_DEBUG("r mask %08x\n", vmw_read(vmw_priv, SVGA_REG_RED_MASK));
+	DRM_DEBUG("g mask %08x\n", vmw_read(vmw_priv, SVGA_REG_GREEN_MASK));
+	DRM_DEBUG("b mask %08x\n", vmw_read(vmw_priv, SVGA_REG_BLUE_MASK));
+	DRM_DEBUG("fb_offset 0x%08x\n", fb_offset);
+	DRM_DEBUG("fb_pitch  %u\n", fb_pitch);
+	DRM_DEBUG("fb_size   %u kiB\n", fb_size / 1024);
+
+	info = framebuffer_alloc(sizeof(*par), device);
+	if (!info)
+		return -ENOMEM;
+
+	/*
+	 * Par
+	 */
+	vmw_priv->fb_info = info;
+	par = info->par;
+	par->vmw_priv = vmw_priv;
+	par->depth = fb_depth;
+	par->bpp = fb_bbp;
+	par->vmalloc = NULL;
+	par->max_width = fb_width;
+	par->max_height = fb_height;
+
+	/*
+	 * Create buffers and alloc memory
+	 */
+	par->vmalloc = vmalloc(fb_size);
+	if (unlikely(par->vmalloc == NULL)) {
+		ret = -ENOMEM;
+		goto err_free;
+	}
+
+	ret = vmw_fb_create_bo(vmw_priv, fb_size, &par->vmw_bo);
+	if (unlikely(ret != 0))
+		goto err_free;
+
+	ret = ttm_bo_kmap(&par->vmw_bo->base,
+			  0,
+			  par->vmw_bo->base.num_pages,
+			  &par->map);
+	if (unlikely(ret != 0))
+		goto err_unref;
+	par->bo_ptr = ttm_kmap_obj_virtual(&par->map, &par->bo_iowrite);
+	par->bo_size = fb_size;
+
+	/*
+	 * Fixed and var
+	 */
+	strcpy(info->fix.id, "svgadrmfb");
+	info->fix.type = FB_TYPE_PACKED_PIXELS;
+	info->fix.visual = FB_VISUAL_TRUECOLOR;
+	info->fix.type_aux = 0;
+	info->fix.xpanstep = 1; /* doing it in hw */
+	info->fix.ypanstep = 1; /* doing it in hw */
+	info->fix.ywrapstep = 0;
+	info->fix.accel = FB_ACCEL_NONE;
+	info->fix.line_length = fb_pitch;
+
+	info->fix.smem_start = 0;
+	info->fix.smem_len = fb_size;
+
+	info->fix.mmio_start = 0;
+	info->fix.mmio_len = 0;
+
+	info->pseudo_palette = par->pseudo_palette;
+	info->screen_base = par->vmalloc;
+	info->screen_size = fb_size;
+
+	info->flags = FBINFO_DEFAULT;
+	info->fbops = &vmw_fb_ops;
+
+	/* 24 depth per default */
+	info->var.red.offset = 16;
+	info->var.green.offset = 8;
+	info->var.blue.offset = 0;
+	info->var.red.length = 8;
+	info->var.green.length = 8;
+	info->var.blue.length = 8;
+	info->var.transp.offset = 0;
+	info->var.transp.length = 0;
+
+	info->var.xres_virtual = fb_width;
+	info->var.yres_virtual = fb_height;
+	info->var.bits_per_pixel = par->bpp;
+	info->var.xoffset = 0;
+	info->var.yoffset = 0;
+	info->var.activate = FB_ACTIVATE_NOW;
+	info->var.height = -1;
+	info->var.width = -1;
+
+	info->var.xres = initial_width;
+	info->var.yres = initial_height;
+
+#if 0
+	info->pixmap.size = 64*1024;
+	info->pixmap.buf_align = 8;
+	info->pixmap.access_align = 32;
+	info->pixmap.flags = FB_PIXMAP_SYSTEM;
+	info->pixmap.scan_align = 1;
+#else
+	info->pixmap.size = 0;
+	info->pixmap.buf_align = 8;
+	info->pixmap.access_align = 32;
+	info->pixmap.flags = FB_PIXMAP_SYSTEM;
+	info->pixmap.scan_align = 1;
+#endif
+
+	info->aperture_base = vmw_priv->vram_start;
+	info->aperture_size = vmw_priv->vram_size;
+
+	/*
+	 * Dirty & Deferred IO
+	 */
+	par->dirty.x1 = par->dirty.x2 = 0;
+	par->dirty.y1 = par->dirty.y1 = 0;
+	par->dirty.active = true;
+	spin_lock_init(&par->dirty.lock);
+	info->fbdefio = &vmw_defio;
+	fb_deferred_io_init(info);
+
+	ret = register_framebuffer(info);
+	if (unlikely(ret != 0))
+		goto err_defio;
+
+	return 0;
+
+err_defio:
+	fb_deferred_io_cleanup(info);
+	ttm_bo_kunmap(&par->map);
+err_unref:
+	ttm_bo_unref((struct ttm_buffer_object **)&par->vmw_bo);
+err_free:
+	vfree(par->vmalloc);
+	framebuffer_release(info);
+	vmw_priv->fb_info = NULL;
+
+	return ret;
+}
+
+int vmw_fb_close(struct vmw_private *vmw_priv)
+{
+	struct fb_info *info;
+	struct vmw_fb_par *par;
+	struct ttm_buffer_object *bo;
+
+	if (!vmw_priv->fb_info)
+		return 0;
+
+	info = vmw_priv->fb_info;
+	par = info->par;
+	bo = &par->vmw_bo->base;
+	par->vmw_bo = NULL;
+
+	/* ??? order */
+	fb_deferred_io_cleanup(info);
+	unregister_framebuffer(info);
+
+	ttm_bo_kunmap(&par->map);
+	ttm_bo_unref(&bo);
+
+	vfree(par->vmalloc);
+	framebuffer_release(info);
+
+	return 0;
+}
+
+int vmw_dmabuf_from_vram(struct vmw_private *vmw_priv,
+			 struct vmw_dma_buffer *vmw_bo)
+{
+	struct ttm_buffer_object *bo = &vmw_bo->base;
+	int ret = 0;
+
+	ret = ttm_bo_reserve(bo, false, false, false, 0);
+	if (unlikely(ret != 0))
+		return ret;
+
+	ret = ttm_bo_validate(bo, &vmw_sys_placement, false, false);
+	ttm_bo_unreserve(bo);
+
+	return ret;
+}
+
+int vmw_dmabuf_to_start_of_vram(struct vmw_private *vmw_priv,
+				struct vmw_dma_buffer *vmw_bo)
+{
+	struct ttm_buffer_object *bo = &vmw_bo->base;
+	struct ttm_placement ne_placement = vmw_vram_ne_placement;
+	int ret = 0;
+
+	ne_placement.lpfn = bo->num_pages;
+
+	/* interuptable? */
+	ret = ttm_write_lock(&vmw_priv->active_master->lock, false);
+	if (unlikely(ret != 0))
+		return ret;
+
+	ret = ttm_bo_reserve(bo, false, false, false, 0);
+	if (unlikely(ret != 0))
+		goto err_unlock;
+
+	ret = ttm_bo_validate(bo, &ne_placement, false, false);
+	ttm_bo_unreserve(bo);
+err_unlock:
+	ttm_write_unlock(&vmw_priv->active_master->lock);
+
+	return ret;
+}
+
+int vmw_fb_off(struct vmw_private *vmw_priv)
+{
+	struct fb_info *info;
+	struct vmw_fb_par *par;
+	unsigned long flags;
+
+	if (!vmw_priv->fb_info)
+		return -EINVAL;
+
+	info = vmw_priv->fb_info;
+	par = info->par;
+
+	spin_lock_irqsave(&par->dirty.lock, flags);
+	par->dirty.active = false;
+	spin_unlock_irqrestore(&par->dirty.lock, flags);
+
+	flush_scheduled_work();
+
+	par->bo_ptr = NULL;
+	ttm_bo_kunmap(&par->map);
+
+	vmw_dmabuf_from_vram(vmw_priv, par->vmw_bo);
+
+	return 0;
+}
+
+int vmw_fb_on(struct vmw_private *vmw_priv)
+{
+	struct fb_info *info;
+	struct vmw_fb_par *par;
+	unsigned long flags;
+	bool dummy;
+	int ret;
+
+	if (!vmw_priv->fb_info)
+		return -EINVAL;
+
+	info = vmw_priv->fb_info;
+	par = info->par;
+
+	/* we are already active */
+	if (par->bo_ptr != NULL)
+		return 0;
+
+	/* Make sure that all overlays are stoped when we take over */
+	vmw_overlay_stop_all(vmw_priv);
+
+	ret = vmw_dmabuf_to_start_of_vram(vmw_priv, par->vmw_bo);
+	if (unlikely(ret != 0)) {
+		DRM_ERROR("could not move buffer to start of VRAM\n");
+		goto err_no_buffer;
+	}
+
+	ret = ttm_bo_kmap(&par->vmw_bo->base,
+			  0,
+			  par->vmw_bo->base.num_pages,
+			  &par->map);
+	BUG_ON(ret != 0);
+	par->bo_ptr = ttm_kmap_obj_virtual(&par->map, &dummy);
+
+	spin_lock_irqsave(&par->dirty.lock, flags);
+	par->dirty.active = true;
+	spin_unlock_irqrestore(&par->dirty.lock, flags);
+
+err_no_buffer:
+	vmw_fb_set_par(info);
+
+	vmw_fb_dirty_mark(par, 0, 0, info->var.xres, info->var.yres);
+
+	/* If there already was stuff dirty we wont
+	 * schedule a new work, so lets do it now */
+	schedule_delayed_work(&info->deferred_work, 0);
+
+	return 0;
+}
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_fifo.c b/drivers/gpu/drm/vmwgfx/vmwgfx_fifo.c
new file mode 100644
index 000000000000..39d43a01d846
--- /dev/null
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_fifo.c
@@ -0,0 +1,538 @@
+/**************************************************************************
+ *
+ * Copyright © 2009 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "vmwgfx_drv.h"
+#include "drmP.h"
+#include "ttm/ttm_placement.h"
+
+bool vmw_fifo_have_3d(struct vmw_private *dev_priv)
+{
+	__le32 __iomem *fifo_mem = dev_priv->mmio_virt;
+	uint32_t fifo_min, hwversion;
+
+	fifo_min = ioread32(fifo_mem  + SVGA_FIFO_MIN);
+	if (fifo_min <= SVGA_FIFO_3D_HWVERSION * sizeof(unsigned int))
+		return false;
+
+	hwversion = ioread32(fifo_mem + SVGA_FIFO_3D_HWVERSION);
+	if (hwversion == 0)
+		return false;
+
+	if (hwversion < SVGA3D_HWVERSION_WS65_B1)
+		return false;
+
+	return true;
+}
+
+int vmw_fifo_init(struct vmw_private *dev_priv, struct vmw_fifo_state *fifo)
+{
+	__le32 __iomem *fifo_mem = dev_priv->mmio_virt;
+	uint32_t max;
+	uint32_t min;
+	uint32_t dummy;
+	int ret;
+
+	fifo->static_buffer_size = VMWGFX_FIFO_STATIC_SIZE;
+	fifo->static_buffer = vmalloc(fifo->static_buffer_size);
+	if (unlikely(fifo->static_buffer == NULL))
+		return -ENOMEM;
+
+	fifo->last_buffer_size = VMWGFX_FIFO_STATIC_SIZE;
+	fifo->last_data_size = 0;
+	fifo->last_buffer_add = false;
+	fifo->last_buffer = vmalloc(fifo->last_buffer_size);
+	if (unlikely(fifo->last_buffer == NULL)) {
+		ret = -ENOMEM;
+		goto out_err;
+	}
+
+	fifo->dynamic_buffer = NULL;
+	fifo->reserved_size = 0;
+	fifo->using_bounce_buffer = false;
+
+	mutex_init(&fifo->fifo_mutex);
+	init_rwsem(&fifo->rwsem);
+
+	/*
+	 * Allow mapping the first page read-only to user-space.
+	 */
+
+	DRM_INFO("width %d\n", vmw_read(dev_priv, SVGA_REG_WIDTH));
+	DRM_INFO("height %d\n", vmw_read(dev_priv, SVGA_REG_HEIGHT));
+	DRM_INFO("bpp %d\n", vmw_read(dev_priv, SVGA_REG_BITS_PER_PIXEL));
+
+	mutex_lock(&dev_priv->hw_mutex);
+	dev_priv->enable_state = vmw_read(dev_priv, SVGA_REG_ENABLE);
+	dev_priv->config_done_state = vmw_read(dev_priv, SVGA_REG_CONFIG_DONE);
+	vmw_write(dev_priv, SVGA_REG_ENABLE, 1);
+
+	min = 4;
+	if (dev_priv->capabilities & SVGA_CAP_EXTENDED_FIFO)
+		min = vmw_read(dev_priv, SVGA_REG_MEM_REGS);
+	min <<= 2;
+
+	if (min < PAGE_SIZE)
+		min = PAGE_SIZE;
+
+	iowrite32(min, fifo_mem + SVGA_FIFO_MIN);
+	iowrite32(dev_priv->mmio_size, fifo_mem + SVGA_FIFO_MAX);
+	wmb();
+	iowrite32(min,  fifo_mem + SVGA_FIFO_NEXT_CMD);
+	iowrite32(min,  fifo_mem + SVGA_FIFO_STOP);
+	iowrite32(0, fifo_mem + SVGA_FIFO_BUSY);
+	mb();
+
+	vmw_write(dev_priv, SVGA_REG_CONFIG_DONE, 1);
+	mutex_unlock(&dev_priv->hw_mutex);
+
+	max = ioread32(fifo_mem + SVGA_FIFO_MAX);
+	min = ioread32(fifo_mem  + SVGA_FIFO_MIN);
+	fifo->capabilities = ioread32(fifo_mem + SVGA_FIFO_CAPABILITIES);
+
+	DRM_INFO("Fifo max 0x%08x min 0x%08x cap 0x%08x\n",
+		 (unsigned int) max,
+		 (unsigned int) min,
+		 (unsigned int) fifo->capabilities);
+
+	atomic_set(&dev_priv->fence_seq, dev_priv->last_read_sequence);
+	iowrite32(dev_priv->last_read_sequence, fifo_mem + SVGA_FIFO_FENCE);
+
+	return vmw_fifo_send_fence(dev_priv, &dummy);
+out_err:
+	vfree(fifo->static_buffer);
+	fifo->static_buffer = NULL;
+	return ret;
+}
+
+void vmw_fifo_ping_host(struct vmw_private *dev_priv, uint32_t reason)
+{
+	__le32 __iomem *fifo_mem = dev_priv->mmio_virt;
+
+	mutex_lock(&dev_priv->hw_mutex);
+
+	if (unlikely(ioread32(fifo_mem + SVGA_FIFO_BUSY) == 0)) {
+		iowrite32(1, fifo_mem + SVGA_FIFO_BUSY);
+		vmw_write(dev_priv, SVGA_REG_SYNC, reason);
+	}
+
+	mutex_unlock(&dev_priv->hw_mutex);
+}
+
+void vmw_fifo_release(struct vmw_private *dev_priv, struct vmw_fifo_state *fifo)
+{
+	__le32 __iomem *fifo_mem = dev_priv->mmio_virt;
+
+	mutex_lock(&dev_priv->hw_mutex);
+
+	while (vmw_read(dev_priv, SVGA_REG_BUSY) != 0)
+		vmw_write(dev_priv, SVGA_REG_SYNC, SVGA_SYNC_GENERIC);
+
+	dev_priv->last_read_sequence = ioread32(fifo_mem + SVGA_FIFO_FENCE);
+
+	vmw_write(dev_priv, SVGA_REG_CONFIG_DONE,
+		  dev_priv->config_done_state);
+	vmw_write(dev_priv, SVGA_REG_ENABLE,
+		  dev_priv->enable_state);
+
+	mutex_unlock(&dev_priv->hw_mutex);
+
+	if (likely(fifo->last_buffer != NULL)) {
+		vfree(fifo->last_buffer);
+		fifo->last_buffer = NULL;
+	}
+
+	if (likely(fifo->static_buffer != NULL)) {
+		vfree(fifo->static_buffer);
+		fifo->static_buffer = NULL;
+	}
+
+	if (likely(fifo->dynamic_buffer != NULL)) {
+		vfree(fifo->dynamic_buffer);
+		fifo->dynamic_buffer = NULL;
+	}
+}
+
+static bool vmw_fifo_is_full(struct vmw_private *dev_priv, uint32_t bytes)
+{
+	__le32 __iomem *fifo_mem = dev_priv->mmio_virt;
+	uint32_t max = ioread32(fifo_mem + SVGA_FIFO_MAX);
+	uint32_t next_cmd = ioread32(fifo_mem + SVGA_FIFO_NEXT_CMD);
+	uint32_t min = ioread32(fifo_mem + SVGA_FIFO_MIN);
+	uint32_t stop = ioread32(fifo_mem + SVGA_FIFO_STOP);
+
+	return ((max - next_cmd) + (stop - min) <= bytes);
+}
+
+static int vmw_fifo_wait_noirq(struct vmw_private *dev_priv,
+			       uint32_t bytes, bool interruptible,
+			       unsigned long timeout)
+{
+	int ret = 0;
+	unsigned long end_jiffies = jiffies + timeout;
+	DEFINE_WAIT(__wait);
+
+	DRM_INFO("Fifo wait noirq.\n");
+
+	for (;;) {
+		prepare_to_wait(&dev_priv->fifo_queue, &__wait,
+				(interruptible) ?
+				TASK_INTERRUPTIBLE : TASK_UNINTERRUPTIBLE);
+		if (!vmw_fifo_is_full(dev_priv, bytes))
+			break;
+		if (time_after_eq(jiffies, end_jiffies)) {
+			ret = -EBUSY;
+			DRM_ERROR("SVGA device lockup.\n");
+			break;
+		}
+		schedule_timeout(1);
+		if (interruptible && signal_pending(current)) {
+			ret = -ERESTARTSYS;
+			break;
+		}
+	}
+	finish_wait(&dev_priv->fifo_queue, &__wait);
+	wake_up_all(&dev_priv->fifo_queue);
+	DRM_INFO("Fifo noirq exit.\n");
+	return ret;
+}
+
+static int vmw_fifo_wait(struct vmw_private *dev_priv,
+			 uint32_t bytes, bool interruptible,
+			 unsigned long timeout)
+{
+	long ret = 1L;
+	unsigned long irq_flags;
+
+	if (likely(!vmw_fifo_is_full(dev_priv, bytes)))
+		return 0;
+
+	vmw_fifo_ping_host(dev_priv, SVGA_SYNC_FIFOFULL);
+	if (!(dev_priv->capabilities & SVGA_CAP_IRQMASK))
+		return vmw_fifo_wait_noirq(dev_priv, bytes,
+					   interruptible, timeout);
+
+	mutex_lock(&dev_priv->hw_mutex);
+	if (atomic_add_return(1, &dev_priv->fifo_queue_waiters) > 0) {
+		spin_lock_irqsave(&dev_priv->irq_lock, irq_flags);
+		outl(SVGA_IRQFLAG_FIFO_PROGRESS,
+		     dev_priv->io_start + VMWGFX_IRQSTATUS_PORT);
+		vmw_write(dev_priv, SVGA_REG_IRQMASK,
+			  vmw_read(dev_priv, SVGA_REG_IRQMASK) |
+			  SVGA_IRQFLAG_FIFO_PROGRESS);
+		spin_unlock_irqrestore(&dev_priv->irq_lock, irq_flags);
+	}
+	mutex_unlock(&dev_priv->hw_mutex);
+
+	if (interruptible)
+		ret = wait_event_interruptible_timeout
+		    (dev_priv->fifo_queue,
+		     !vmw_fifo_is_full(dev_priv, bytes), timeout);
+	else
+		ret = wait_event_timeout
+		    (dev_priv->fifo_queue,
+		     !vmw_fifo_is_full(dev_priv, bytes), timeout);
+
+	if (unlikely(ret == 0))
+		ret = -EBUSY;
+	else if (likely(ret > 0))
+		ret = 0;
+
+	mutex_lock(&dev_priv->hw_mutex);
+	if (atomic_dec_and_test(&dev_priv->fifo_queue_waiters)) {
+		spin_lock_irqsave(&dev_priv->irq_lock, irq_flags);
+		vmw_write(dev_priv, SVGA_REG_IRQMASK,
+			  vmw_read(dev_priv, SVGA_REG_IRQMASK) &
+			  ~SVGA_IRQFLAG_FIFO_PROGRESS);
+		spin_unlock_irqrestore(&dev_priv->irq_lock, irq_flags);
+	}
+	mutex_unlock(&dev_priv->hw_mutex);
+
+	return ret;
+}
+
+void *vmw_fifo_reserve(struct vmw_private *dev_priv, uint32_t bytes)
+{
+	struct vmw_fifo_state *fifo_state = &dev_priv->fifo;
+	__le32 __iomem *fifo_mem = dev_priv->mmio_virt;
+	uint32_t max;
+	uint32_t min;
+	uint32_t next_cmd;
+	uint32_t reserveable = fifo_state->capabilities & SVGA_FIFO_CAP_RESERVE;
+	int ret;
+
+	mutex_lock(&fifo_state->fifo_mutex);
+	max = ioread32(fifo_mem + SVGA_FIFO_MAX);
+	min = ioread32(fifo_mem + SVGA_FIFO_MIN);
+	next_cmd = ioread32(fifo_mem + SVGA_FIFO_NEXT_CMD);
+
+	if (unlikely(bytes >= (max - min)))
+		goto out_err;
+
+	BUG_ON(fifo_state->reserved_size != 0);
+	BUG_ON(fifo_state->dynamic_buffer != NULL);
+
+	fifo_state->reserved_size = bytes;
+
+	while (1) {
+		uint32_t stop = ioread32(fifo_mem + SVGA_FIFO_STOP);
+		bool need_bounce = false;
+		bool reserve_in_place = false;
+
+		if (next_cmd >= stop) {
+			if (likely((next_cmd + bytes < max ||
+				    (next_cmd + bytes == max && stop > min))))
+				reserve_in_place = true;
+
+			else if (vmw_fifo_is_full(dev_priv, bytes)) {
+				ret = vmw_fifo_wait(dev_priv, bytes,
+						    false, 3 * HZ);
+				if (unlikely(ret != 0))
+					goto out_err;
+			} else
+				need_bounce = true;
+
+		} else {
+
+			if (likely((next_cmd + bytes < stop)))
+				reserve_in_place = true;
+			else {
+				ret = vmw_fifo_wait(dev_priv, bytes,
+						    false, 3 * HZ);
+				if (unlikely(ret != 0))
+					goto out_err;
+			}
+		}
+
+		if (reserve_in_place) {
+			if (reserveable || bytes <= sizeof(uint32_t)) {
+				fifo_state->using_bounce_buffer = false;
+
+				if (reserveable)
+					iowrite32(bytes, fifo_mem +
+						  SVGA_FIFO_RESERVED);
+				return fifo_mem + (next_cmd >> 2);
+			} else {
+				need_bounce = true;
+			}
+		}
+
+		if (need_bounce) {
+			fifo_state->using_bounce_buffer = true;
+			if (bytes < fifo_state->static_buffer_size)
+				return fifo_state->static_buffer;
+			else {
+				fifo_state->dynamic_buffer = vmalloc(bytes);
+				return fifo_state->dynamic_buffer;
+			}
+		}
+	}
+out_err:
+	fifo_state->reserved_size = 0;
+	mutex_unlock(&fifo_state->fifo_mutex);
+	return NULL;
+}
+
+static void vmw_fifo_res_copy(struct vmw_fifo_state *fifo_state,
+			      __le32 __iomem *fifo_mem,
+			      uint32_t next_cmd,
+			      uint32_t max, uint32_t min, uint32_t bytes)
+{
+	uint32_t chunk_size = max - next_cmd;
+	uint32_t rest;
+	uint32_t *buffer = (fifo_state->dynamic_buffer != NULL) ?
+	    fifo_state->dynamic_buffer : fifo_state->static_buffer;
+
+	if (bytes < chunk_size)
+		chunk_size = bytes;
+
+	iowrite32(bytes, fifo_mem + SVGA_FIFO_RESERVED);
+	mb();
+	memcpy_toio(fifo_mem + (next_cmd >> 2), buffer, chunk_size);
+	rest = bytes - chunk_size;
+	if (rest)
+		memcpy_toio(fifo_mem + (min >> 2), buffer + (chunk_size >> 2),
+			    rest);
+}
+
+static void vmw_fifo_slow_copy(struct vmw_fifo_state *fifo_state,
+			       __le32 __iomem *fifo_mem,
+			       uint32_t next_cmd,
+			       uint32_t max, uint32_t min, uint32_t bytes)
+{
+	uint32_t *buffer = (fifo_state->dynamic_buffer != NULL) ?
+	    fifo_state->dynamic_buffer : fifo_state->static_buffer;
+
+	while (bytes > 0) {
+		iowrite32(*buffer++, fifo_mem + (next_cmd >> 2));
+		next_cmd += sizeof(uint32_t);
+		if (unlikely(next_cmd == max))
+			next_cmd = min;
+		mb();
+		iowrite32(next_cmd, fifo_mem + SVGA_FIFO_NEXT_CMD);
+		mb();
+		bytes -= sizeof(uint32_t);
+	}
+}
+
+void vmw_fifo_commit(struct vmw_private *dev_priv, uint32_t bytes)
+{
+	struct vmw_fifo_state *fifo_state = &dev_priv->fifo;
+	__le32 __iomem *fifo_mem = dev_priv->mmio_virt;
+	uint32_t next_cmd = ioread32(fifo_mem + SVGA_FIFO_NEXT_CMD);
+	uint32_t max = ioread32(fifo_mem + SVGA_FIFO_MAX);
+	uint32_t min = ioread32(fifo_mem + SVGA_FIFO_MIN);
+	bool reserveable = fifo_state->capabilities & SVGA_FIFO_CAP_RESERVE;
+
+	BUG_ON((bytes & 3) != 0);
+	BUG_ON(bytes > fifo_state->reserved_size);
+
+	fifo_state->reserved_size = 0;
+
+	if (fifo_state->using_bounce_buffer) {
+		if (reserveable)
+			vmw_fifo_res_copy(fifo_state, fifo_mem,
+					  next_cmd, max, min, bytes);
+		else
+			vmw_fifo_slow_copy(fifo_state, fifo_mem,
+					   next_cmd, max, min, bytes);
+
+		if (fifo_state->dynamic_buffer) {
+			vfree(fifo_state->dynamic_buffer);
+			fifo_state->dynamic_buffer = NULL;
+		}
+
+	}
+
+	down_write(&fifo_state->rwsem);
+	if (fifo_state->using_bounce_buffer || reserveable) {
+		next_cmd += bytes;
+		if (next_cmd >= max)
+			next_cmd -= max - min;
+		mb();
+		iowrite32(next_cmd, fifo_mem + SVGA_FIFO_NEXT_CMD);
+	}
+
+	if (reserveable)
+		iowrite32(0, fifo_mem + SVGA_FIFO_RESERVED);
+	mb();
+	up_write(&fifo_state->rwsem);
+	vmw_fifo_ping_host(dev_priv, SVGA_SYNC_GENERIC);
+	mutex_unlock(&fifo_state->fifo_mutex);
+}
+
+int vmw_fifo_send_fence(struct vmw_private *dev_priv, uint32_t *sequence)
+{
+	struct vmw_fifo_state *fifo_state = &dev_priv->fifo;
+	struct svga_fifo_cmd_fence *cmd_fence;
+	void *fm;
+	int ret = 0;
+	uint32_t bytes = sizeof(__le32) + sizeof(*cmd_fence);
+
+	fm = vmw_fifo_reserve(dev_priv, bytes);
+	if (unlikely(fm == NULL)) {
+		*sequence = atomic_read(&dev_priv->fence_seq);
+		ret = -ENOMEM;
+		(void)vmw_fallback_wait(dev_priv, false, true, *sequence,
+					false, 3*HZ);
+		goto out_err;
+	}
+
+	do {
+		*sequence = atomic_add_return(1, &dev_priv->fence_seq);
+	} while (*sequence == 0);
+
+	if (!(fifo_state->capabilities & SVGA_FIFO_CAP_FENCE)) {
+
+		/*
+		 * Don't request hardware to send a fence. The
+		 * waiting code in vmwgfx_irq.c will emulate this.
+		 */
+
+		vmw_fifo_commit(dev_priv, 0);
+		return 0;
+	}
+
+	*(__le32 *) fm = cpu_to_le32(SVGA_CMD_FENCE);
+	cmd_fence = (struct svga_fifo_cmd_fence *)
+	    ((unsigned long)fm + sizeof(__le32));
+
+	iowrite32(*sequence, &cmd_fence->fence);
+	fifo_state->last_buffer_add = true;
+	vmw_fifo_commit(dev_priv, bytes);
+	fifo_state->last_buffer_add = false;
+
+out_err:
+	return ret;
+}
+
+/**
+ * Map the first page of the FIFO read-only to user-space.
+ */
+
+static int vmw_fifo_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
+{
+	int ret;
+	unsigned long address = (unsigned long)vmf->virtual_address;
+
+	if (address != vma->vm_start)
+		return VM_FAULT_SIGBUS;
+
+	ret = vm_insert_pfn(vma, address, vma->vm_pgoff);
+	if (likely(ret == -EBUSY || ret == 0))
+		return VM_FAULT_NOPAGE;
+	else if (ret == -ENOMEM)
+		return VM_FAULT_OOM;
+
+	return VM_FAULT_SIGBUS;
+}
+
+static struct vm_operations_struct vmw_fifo_vm_ops = {
+	.fault = vmw_fifo_vm_fault,
+	.open = NULL,
+	.close = NULL
+};
+
+int vmw_fifo_mmap(struct file *filp, struct vm_area_struct *vma)
+{
+	struct drm_file *file_priv;
+	struct vmw_private *dev_priv;
+
+	file_priv = (struct drm_file *)filp->private_data;
+	dev_priv = vmw_priv(file_priv->minor->dev);
+
+	if (vma->vm_pgoff != (dev_priv->mmio_start >> PAGE_SHIFT) ||
+	    (vma->vm_end - vma->vm_start) != PAGE_SIZE)
+		return -EINVAL;
+
+	vma->vm_flags &= ~(VM_WRITE | VM_MAYWRITE);
+	vma->vm_flags |= VM_IO | VM_PFNMAP | VM_DONTEXPAND | VM_SHARED;
+	vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
+	vma->vm_page_prot = ttm_io_prot(TTM_PL_FLAG_UNCACHED,
+					vma->vm_page_prot);
+	vma->vm_ops = &vmw_fifo_vm_ops;
+	return 0;
+}
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_gmr.c b/drivers/gpu/drm/vmwgfx/vmwgfx_gmr.c
new file mode 100644
index 000000000000..5f8908a5d7fd
--- /dev/null
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_gmr.c
@@ -0,0 +1,213 @@
+/**************************************************************************
+ *
+ * Copyright © 2009 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "vmwgfx_drv.h"
+#include "drmP.h"
+#include "ttm/ttm_bo_driver.h"
+
+/**
+ * FIXME: Adjust to the ttm lowmem / highmem storage to minimize
+ * the number of used descriptors.
+ */
+
+static int vmw_gmr_build_descriptors(struct list_head *desc_pages,
+				     struct page *pages[],
+				     unsigned long num_pages)
+{
+	struct page *page, *next;
+	struct svga_guest_mem_descriptor *page_virtual = NULL;
+	struct svga_guest_mem_descriptor *desc_virtual = NULL;
+	unsigned int desc_per_page;
+	unsigned long prev_pfn;
+	unsigned long pfn;
+	int ret;
+
+	desc_per_page = PAGE_SIZE /
+	    sizeof(struct svga_guest_mem_descriptor) - 1;
+
+	while (likely(num_pages != 0)) {
+		page = alloc_page(__GFP_HIGHMEM);
+		if (unlikely(page == NULL)) {
+			ret = -ENOMEM;
+			goto out_err;
+		}
+
+		list_add_tail(&page->lru, desc_pages);
+
+		/*
+		 * Point previous page terminating descriptor to this
+		 * page before unmapping it.
+		 */
+
+		if (likely(page_virtual != NULL)) {
+			desc_virtual->ppn = page_to_pfn(page);
+			kunmap_atomic(page_virtual, KM_USER0);
+		}
+
+		page_virtual = kmap_atomic(page, KM_USER0);
+		desc_virtual = page_virtual - 1;
+		prev_pfn = ~(0UL);
+
+		while (likely(num_pages != 0)) {
+			pfn = page_to_pfn(*pages);
+
+			if (pfn != prev_pfn + 1) {
+
+				if (desc_virtual - page_virtual ==
+				    desc_per_page - 1)
+					break;
+
+				(++desc_virtual)->ppn = cpu_to_le32(pfn);
+				desc_virtual->num_pages = cpu_to_le32(1);
+			} else {
+				uint32_t tmp =
+				    le32_to_cpu(desc_virtual->num_pages);
+				desc_virtual->num_pages = cpu_to_le32(tmp + 1);
+			}
+			prev_pfn = pfn;
+			--num_pages;
+			++pages;
+		}
+
+		(++desc_virtual)->ppn = cpu_to_le32(0);
+		desc_virtual->num_pages = cpu_to_le32(0);
+	}
+
+	if (likely(page_virtual != NULL))
+		kunmap_atomic(page_virtual, KM_USER0);
+
+	return 0;
+out_err:
+	list_for_each_entry_safe(page, next, desc_pages, lru) {
+		list_del_init(&page->lru);
+		__free_page(page);
+	}
+	return ret;
+}
+
+static inline void vmw_gmr_free_descriptors(struct list_head *desc_pages)
+{
+	struct page *page, *next;
+
+	list_for_each_entry_safe(page, next, desc_pages, lru) {
+		list_del_init(&page->lru);
+		__free_page(page);
+	}
+}
+
+static void vmw_gmr_fire_descriptors(struct vmw_private *dev_priv,
+				     int gmr_id, struct list_head *desc_pages)
+{
+	struct page *page;
+
+	if (unlikely(list_empty(desc_pages)))
+		return;
+
+	page = list_entry(desc_pages->next, struct page, lru);
+
+	mutex_lock(&dev_priv->hw_mutex);
+
+	vmw_write(dev_priv, SVGA_REG_GMR_ID, gmr_id);
+	wmb();
+	vmw_write(dev_priv, SVGA_REG_GMR_DESCRIPTOR, page_to_pfn(page));
+	mb();
+
+	mutex_unlock(&dev_priv->hw_mutex);
+
+}
+
+/**
+ * FIXME: Adjust to the ttm lowmem / highmem storage to minimize
+ * the number of used descriptors.
+ */
+
+static unsigned long vmw_gmr_count_descriptors(struct page *pages[],
+					       unsigned long num_pages)
+{
+	unsigned long prev_pfn = ~(0UL);
+	unsigned long pfn;
+	unsigned long descriptors = 0;
+
+	while (num_pages--) {
+		pfn = page_to_pfn(*pages++);
+		if (prev_pfn + 1 != pfn)
+			++descriptors;
+		prev_pfn = pfn;
+	}
+
+	return descriptors;
+}
+
+int vmw_gmr_bind(struct vmw_private *dev_priv,
+		 struct ttm_buffer_object *bo)
+{
+	struct ttm_tt *ttm = bo->ttm;
+	unsigned long descriptors;
+	int ret;
+	uint32_t id;
+	struct list_head desc_pages;
+
+	if (!(dev_priv->capabilities & SVGA_CAP_GMR))
+		return -EINVAL;
+
+	ret = ttm_tt_populate(ttm);
+	if (unlikely(ret != 0))
+		return ret;
+
+	descriptors = vmw_gmr_count_descriptors(ttm->pages, ttm->num_pages);
+	if (unlikely(descriptors > dev_priv->max_gmr_descriptors))
+		return -EINVAL;
+
+	INIT_LIST_HEAD(&desc_pages);
+	ret = vmw_gmr_build_descriptors(&desc_pages, ttm->pages,
+					ttm->num_pages);
+	if (unlikely(ret != 0))
+		return ret;
+
+	ret = vmw_gmr_id_alloc(dev_priv, &id);
+	if (unlikely(ret != 0))
+		goto out_no_id;
+
+	vmw_gmr_fire_descriptors(dev_priv, id, &desc_pages);
+	vmw_gmr_free_descriptors(&desc_pages);
+	vmw_dmabuf_set_gmr(bo, id);
+	return 0;
+
+out_no_id:
+	vmw_gmr_free_descriptors(&desc_pages);
+	return ret;
+}
+
+void vmw_gmr_unbind(struct vmw_private *dev_priv, int gmr_id)
+{
+	mutex_lock(&dev_priv->hw_mutex);
+	vmw_write(dev_priv, SVGA_REG_GMR_ID, gmr_id);
+	wmb();
+	vmw_write(dev_priv, SVGA_REG_GMR_DESCRIPTOR, 0);
+	mb();
+	mutex_unlock(&dev_priv->hw_mutex);
+}
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ioctl.c b/drivers/gpu/drm/vmwgfx/vmwgfx_ioctl.c
new file mode 100644
index 000000000000..1c7a316454d8
--- /dev/null
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ioctl.c
@@ -0,0 +1,87 @@
+/**************************************************************************
+ *
+ * Copyright © 2009 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "vmwgfx_drv.h"
+#include "vmwgfx_drm.h"
+
+int vmw_getparam_ioctl(struct drm_device *dev, void *data,
+		       struct drm_file *file_priv)
+{
+	struct vmw_private *dev_priv = vmw_priv(dev);
+	struct drm_vmw_getparam_arg *param =
+	    (struct drm_vmw_getparam_arg *)data;
+
+	switch (param->param) {
+	case DRM_VMW_PARAM_NUM_STREAMS:
+		param->value = vmw_overlay_num_overlays(dev_priv);
+		break;
+	case DRM_VMW_PARAM_NUM_FREE_STREAMS:
+		param->value = vmw_overlay_num_free_overlays(dev_priv);
+		break;
+	case DRM_VMW_PARAM_3D:
+		param->value = vmw_fifo_have_3d(dev_priv) ? 1 : 0;
+		break;
+	case DRM_VMW_PARAM_FIFO_OFFSET:
+		param->value = dev_priv->mmio_start;
+		break;
+	case DRM_VMW_PARAM_HW_CAPS:
+		param->value = dev_priv->capabilities;
+		break;
+	case DRM_VMW_PARAM_FIFO_CAPS:
+		param->value = dev_priv->fifo.capabilities;
+		break;
+	default:
+		DRM_ERROR("Illegal vmwgfx get param request: %d\n",
+			  param->param);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+int vmw_fifo_debug_ioctl(struct drm_device *dev, void *data,
+			 struct drm_file *file_priv)
+{
+	struct vmw_private *dev_priv = vmw_priv(dev);
+	struct vmw_fifo_state *fifo_state = &dev_priv->fifo;
+	struct drm_vmw_fifo_debug_arg *arg =
+	    (struct drm_vmw_fifo_debug_arg *)data;
+	__le32 __user *buffer = (__le32 __user *)
+	    (unsigned long)arg->debug_buffer;
+
+	if (unlikely(fifo_state->last_buffer == NULL))
+		return -EINVAL;
+
+	if (arg->debug_buffer_size < fifo_state->last_data_size) {
+		arg->used_size = arg->debug_buffer_size;
+		arg->did_not_fit = 1;
+	} else {
+		arg->used_size = fifo_state->last_data_size;
+		arg->did_not_fit = 0;
+	}
+	return copy_to_user(buffer, fifo_state->last_buffer, arg->used_size);
+}
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_irq.c b/drivers/gpu/drm/vmwgfx/vmwgfx_irq.c
new file mode 100644
index 000000000000..4d7cb5393860
--- /dev/null
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_irq.c
@@ -0,0 +1,286 @@
+/**************************************************************************
+ *
+ * Copyright © 2009 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "drmP.h"
+#include "vmwgfx_drv.h"
+
+#define VMW_FENCE_WRAP (1 << 24)
+
+irqreturn_t vmw_irq_handler(DRM_IRQ_ARGS)
+{
+	struct drm_device *dev = (struct drm_device *)arg;
+	struct vmw_private *dev_priv = vmw_priv(dev);
+	uint32_t status;
+
+	spin_lock(&dev_priv->irq_lock);
+	status = inl(dev_priv->io_start + VMWGFX_IRQSTATUS_PORT);
+	spin_unlock(&dev_priv->irq_lock);
+
+	if (status & SVGA_IRQFLAG_ANY_FENCE)
+		wake_up_all(&dev_priv->fence_queue);
+	if (status & SVGA_IRQFLAG_FIFO_PROGRESS)
+		wake_up_all(&dev_priv->fifo_queue);
+
+	if (likely(status)) {
+		outl(status, dev_priv->io_start + VMWGFX_IRQSTATUS_PORT);
+		return IRQ_HANDLED;
+	}
+
+	return IRQ_NONE;
+}
+
+static bool vmw_fifo_idle(struct vmw_private *dev_priv, uint32_t sequence)
+{
+	uint32_t busy;
+
+	mutex_lock(&dev_priv->hw_mutex);
+	busy = vmw_read(dev_priv, SVGA_REG_BUSY);
+	mutex_unlock(&dev_priv->hw_mutex);
+
+	return (busy == 0);
+}
+
+
+bool vmw_fence_signaled(struct vmw_private *dev_priv,
+			uint32_t sequence)
+{
+	__le32 __iomem *fifo_mem = dev_priv->mmio_virt;
+	struct vmw_fifo_state *fifo_state;
+	bool ret;
+
+	if (likely(dev_priv->last_read_sequence - sequence < VMW_FENCE_WRAP))
+		return true;
+
+	dev_priv->last_read_sequence = ioread32(fifo_mem + SVGA_FIFO_FENCE);
+	if (likely(dev_priv->last_read_sequence - sequence < VMW_FENCE_WRAP))
+		return true;
+
+	fifo_state = &dev_priv->fifo;
+	if (!(fifo_state->capabilities & SVGA_FIFO_CAP_FENCE) &&
+	    vmw_fifo_idle(dev_priv, sequence))
+		return true;
+
+	/**
+	 * Then check if the sequence is higher than what we've actually
+	 * emitted. Then the fence is stale and signaled.
+	 */
+
+	ret = ((atomic_read(&dev_priv->fence_seq) - sequence)
+	       > VMW_FENCE_WRAP);
+
+	return ret;
+}
+
+int vmw_fallback_wait(struct vmw_private *dev_priv,
+		      bool lazy,
+		      bool fifo_idle,
+		      uint32_t sequence,
+		      bool interruptible,
+		      unsigned long timeout)
+{
+	struct vmw_fifo_state *fifo_state = &dev_priv->fifo;
+
+	uint32_t count = 0;
+	uint32_t signal_seq;
+	int ret;
+	unsigned long end_jiffies = jiffies + timeout;
+	bool (*wait_condition)(struct vmw_private *, uint32_t);
+	DEFINE_WAIT(__wait);
+
+	wait_condition = (fifo_idle) ? &vmw_fifo_idle :
+		&vmw_fence_signaled;
+
+	/**
+	 * Block command submission while waiting for idle.
+	 */
+
+	if (fifo_idle)
+		down_read(&fifo_state->rwsem);
+	signal_seq = atomic_read(&dev_priv->fence_seq);
+	ret = 0;
+
+	for (;;) {
+		prepare_to_wait(&dev_priv->fence_queue, &__wait,
+				(interruptible) ?
+				TASK_INTERRUPTIBLE : TASK_UNINTERRUPTIBLE);
+		if (wait_condition(dev_priv, sequence))
+			break;
+		if (time_after_eq(jiffies, end_jiffies)) {
+			DRM_ERROR("SVGA device lockup.\n");
+			break;
+		}
+		if (lazy)
+			schedule_timeout(1);
+		else if ((++count & 0x0F) == 0) {
+			/**
+			 * FIXME: Use schedule_hr_timeout here for
+			 * newer kernels and lower CPU utilization.
+			 */
+
+			__set_current_state(TASK_RUNNING);
+			schedule();
+			__set_current_state((interruptible) ?
+					    TASK_INTERRUPTIBLE :
+					    TASK_UNINTERRUPTIBLE);
+		}
+		if (interruptible && signal_pending(current)) {
+			ret = -ERESTARTSYS;
+			break;
+		}
+	}
+	finish_wait(&dev_priv->fence_queue, &__wait);
+	if (ret == 0 && fifo_idle) {
+		__le32 __iomem *fifo_mem = dev_priv->mmio_virt;
+		iowrite32(signal_seq, fifo_mem + SVGA_FIFO_FENCE);
+	}
+	wake_up_all(&dev_priv->fence_queue);
+	if (fifo_idle)
+		up_read(&fifo_state->rwsem);
+
+	return ret;
+}
+
+int vmw_wait_fence(struct vmw_private *dev_priv,
+		   bool lazy, uint32_t sequence,
+		   bool interruptible, unsigned long timeout)
+{
+	long ret;
+	unsigned long irq_flags;
+	struct vmw_fifo_state *fifo = &dev_priv->fifo;
+
+	if (likely(dev_priv->last_read_sequence - sequence < VMW_FENCE_WRAP))
+		return 0;
+
+	if (likely(vmw_fence_signaled(dev_priv, sequence)))
+		return 0;
+
+	vmw_fifo_ping_host(dev_priv, SVGA_SYNC_GENERIC);
+
+	if (!(fifo->capabilities & SVGA_FIFO_CAP_FENCE))
+		return vmw_fallback_wait(dev_priv, lazy, true, sequence,
+					 interruptible, timeout);
+
+	if (!(dev_priv->capabilities & SVGA_CAP_IRQMASK))
+		return vmw_fallback_wait(dev_priv, lazy, false, sequence,
+					 interruptible, timeout);
+
+	mutex_lock(&dev_priv->hw_mutex);
+	if (atomic_add_return(1, &dev_priv->fence_queue_waiters) > 0) {
+		spin_lock_irqsave(&dev_priv->irq_lock, irq_flags);
+		outl(SVGA_IRQFLAG_ANY_FENCE,
+		     dev_priv->io_start + VMWGFX_IRQSTATUS_PORT);
+		vmw_write(dev_priv, SVGA_REG_IRQMASK,
+			  vmw_read(dev_priv, SVGA_REG_IRQMASK) |
+			  SVGA_IRQFLAG_ANY_FENCE);
+		spin_unlock_irqrestore(&dev_priv->irq_lock, irq_flags);
+	}
+	mutex_unlock(&dev_priv->hw_mutex);
+
+	if (interruptible)
+		ret = wait_event_interruptible_timeout
+		    (dev_priv->fence_queue,
+		     vmw_fence_signaled(dev_priv, sequence),
+		     timeout);
+	else
+		ret = wait_event_timeout
+		    (dev_priv->fence_queue,
+		     vmw_fence_signaled(dev_priv, sequence),
+		     timeout);
+
+	if (unlikely(ret == 0))
+		ret = -EBUSY;
+	else if (likely(ret > 0))
+		ret = 0;
+
+	mutex_lock(&dev_priv->hw_mutex);
+	if (atomic_dec_and_test(&dev_priv->fence_queue_waiters)) {
+		spin_lock_irqsave(&dev_priv->irq_lock, irq_flags);
+		vmw_write(dev_priv, SVGA_REG_IRQMASK,
+			  vmw_read(dev_priv, SVGA_REG_IRQMASK) &
+			  ~SVGA_IRQFLAG_ANY_FENCE);
+		spin_unlock_irqrestore(&dev_priv->irq_lock, irq_flags);
+	}
+	mutex_unlock(&dev_priv->hw_mutex);
+
+	return ret;
+}
+
+void vmw_irq_preinstall(struct drm_device *dev)
+{
+	struct vmw_private *dev_priv = vmw_priv(dev);
+	uint32_t status;
+
+	if (!(dev_priv->capabilities & SVGA_CAP_IRQMASK))
+		return;
+
+	spin_lock_init(&dev_priv->irq_lock);
+	status = inl(dev_priv->io_start + VMWGFX_IRQSTATUS_PORT);
+	outl(status, dev_priv->io_start + VMWGFX_IRQSTATUS_PORT);
+}
+
+int vmw_irq_postinstall(struct drm_device *dev)
+{
+	return 0;
+}
+
+void vmw_irq_uninstall(struct drm_device *dev)
+{
+	struct vmw_private *dev_priv = vmw_priv(dev);
+	uint32_t status;
+
+	if (!(dev_priv->capabilities & SVGA_CAP_IRQMASK))
+		return;
+
+	mutex_lock(&dev_priv->hw_mutex);
+	vmw_write(dev_priv, SVGA_REG_IRQMASK, 0);
+	mutex_unlock(&dev_priv->hw_mutex);
+
+	status = inl(dev_priv->io_start + VMWGFX_IRQSTATUS_PORT);
+	outl(status, dev_priv->io_start + VMWGFX_IRQSTATUS_PORT);
+}
+
+#define VMW_FENCE_WAIT_TIMEOUT 3*HZ;
+
+int vmw_fence_wait_ioctl(struct drm_device *dev, void *data,
+			 struct drm_file *file_priv)
+{
+	struct drm_vmw_fence_wait_arg *arg =
+	    (struct drm_vmw_fence_wait_arg *)data;
+	unsigned long timeout;
+
+	if (!arg->cookie_valid) {
+		arg->cookie_valid = 1;
+		arg->kernel_cookie = jiffies + VMW_FENCE_WAIT_TIMEOUT;
+	}
+
+	timeout = jiffies;
+	if (time_after_eq(timeout, (unsigned long)arg->kernel_cookie))
+		return -EBUSY;
+
+	timeout = (unsigned long)arg->kernel_cookie - timeout;
+	return vmw_wait_fence(vmw_priv(dev), true, arg->sequence, true, timeout);
+}
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c
new file mode 100644
index 000000000000..31f9afed0a63
--- /dev/null
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c
@@ -0,0 +1,880 @@
+/**************************************************************************
+ *
+ * Copyright © 2009 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "vmwgfx_kms.h"
+
+/* Might need a hrtimer here? */
+#define VMWGFX_PRESENT_RATE ((HZ / 60 > 0) ? HZ / 60 : 1)
+
+
+void vmw_display_unit_cleanup(struct vmw_display_unit *du)
+{
+	if (du->cursor_surface)
+		vmw_surface_unreference(&du->cursor_surface);
+	if (du->cursor_dmabuf)
+		vmw_dmabuf_unreference(&du->cursor_dmabuf);
+	drm_crtc_cleanup(&du->crtc);
+	drm_encoder_cleanup(&du->encoder);
+	drm_connector_cleanup(&du->connector);
+}
+
+/*
+ * Display Unit Cursor functions
+ */
+
+int vmw_cursor_update_image(struct vmw_private *dev_priv,
+			    u32 *image, u32 width, u32 height,
+			    u32 hotspotX, u32 hotspotY)
+{
+	struct {
+		u32 cmd;
+		SVGAFifoCmdDefineAlphaCursor cursor;
+	} *cmd;
+	u32 image_size = width * height * 4;
+	u32 cmd_size = sizeof(*cmd) + image_size;
+
+	if (!image)
+		return -EINVAL;
+
+	cmd = vmw_fifo_reserve(dev_priv, cmd_size);
+	if (unlikely(cmd == NULL)) {
+		DRM_ERROR("Fifo reserve failed.\n");
+		return -ENOMEM;
+	}
+
+	memset(cmd, 0, sizeof(*cmd));
+
+	memcpy(&cmd[1], image, image_size);
+
+	cmd->cmd = cpu_to_le32(SVGA_CMD_DEFINE_ALPHA_CURSOR);
+	cmd->cursor.id = cpu_to_le32(0);
+	cmd->cursor.width = cpu_to_le32(width);
+	cmd->cursor.height = cpu_to_le32(height);
+	cmd->cursor.hotspotX = cpu_to_le32(hotspotX);
+	cmd->cursor.hotspotY = cpu_to_le32(hotspotY);
+
+	vmw_fifo_commit(dev_priv, cmd_size);
+
+	return 0;
+}
+
+void vmw_cursor_update_position(struct vmw_private *dev_priv,
+				bool show, int x, int y)
+{
+	__le32 __iomem *fifo_mem = dev_priv->mmio_virt;
+	uint32_t count;
+
+	iowrite32(show ? 1 : 0, fifo_mem + SVGA_FIFO_CURSOR_ON);
+	iowrite32(x, fifo_mem + SVGA_FIFO_CURSOR_X);
+	iowrite32(y, fifo_mem + SVGA_FIFO_CURSOR_Y);
+	count = ioread32(fifo_mem + SVGA_FIFO_CURSOR_COUNT);
+	iowrite32(++count, fifo_mem + SVGA_FIFO_CURSOR_COUNT);
+}
+
+int vmw_du_crtc_cursor_set(struct drm_crtc *crtc, struct drm_file *file_priv,
+			   uint32_t handle, uint32_t width, uint32_t height)
+{
+	struct vmw_private *dev_priv = vmw_priv(crtc->dev);
+	struct ttm_object_file *tfile = vmw_fpriv(file_priv)->tfile;
+	struct vmw_display_unit *du = vmw_crtc_to_du(crtc);
+	struct vmw_surface *surface = NULL;
+	struct vmw_dma_buffer *dmabuf = NULL;
+	int ret;
+
+	if (handle) {
+		ret = vmw_user_surface_lookup_handle(dev_priv, tfile,
+						     handle, &surface);
+		if (!ret) {
+			if (!surface->snooper.image) {
+				DRM_ERROR("surface not suitable for cursor\n");
+				return -EINVAL;
+			}
+		} else {
+			ret = vmw_user_dmabuf_lookup(tfile,
+						     handle, &dmabuf);
+			if (ret) {
+				DRM_ERROR("failed to find surface or dmabuf: %i\n", ret);
+				return -EINVAL;
+			}
+		}
+	}
+
+	/* takedown old cursor */
+	if (du->cursor_surface) {
+		du->cursor_surface->snooper.crtc = NULL;
+		vmw_surface_unreference(&du->cursor_surface);
+	}
+	if (du->cursor_dmabuf)
+		vmw_dmabuf_unreference(&du->cursor_dmabuf);
+
+	/* setup new image */
+	if (surface) {
+		/* vmw_user_surface_lookup takes one reference */
+		du->cursor_surface = surface;
+
+		du->cursor_surface->snooper.crtc = crtc;
+		du->cursor_age = du->cursor_surface->snooper.age;
+		vmw_cursor_update_image(dev_priv, surface->snooper.image,
+					64, 64, du->hotspot_x, du->hotspot_y);
+	} else if (dmabuf) {
+		struct ttm_bo_kmap_obj map;
+		unsigned long kmap_offset;
+		unsigned long kmap_num;
+		void *virtual;
+		bool dummy;
+
+		/* vmw_user_surface_lookup takes one reference */
+		du->cursor_dmabuf = dmabuf;
+
+		kmap_offset = 0;
+		kmap_num = (64*64*4) >> PAGE_SHIFT;
+
+		ret = ttm_bo_reserve(&dmabuf->base, true, false, false, 0);
+		if (unlikely(ret != 0)) {
+			DRM_ERROR("reserve failed\n");
+			return -EINVAL;
+		}
+
+		ret = ttm_bo_kmap(&dmabuf->base, kmap_offset, kmap_num, &map);
+		if (unlikely(ret != 0))
+			goto err_unreserve;
+
+		virtual = ttm_kmap_obj_virtual(&map, &dummy);
+		vmw_cursor_update_image(dev_priv, virtual, 64, 64,
+					du->hotspot_x, du->hotspot_y);
+
+		ttm_bo_kunmap(&map);
+err_unreserve:
+		ttm_bo_unreserve(&dmabuf->base);
+
+	} else {
+		vmw_cursor_update_position(dev_priv, false, 0, 0);
+		return 0;
+	}
+
+	vmw_cursor_update_position(dev_priv, true, du->cursor_x, du->cursor_y);
+
+	return 0;
+}
+
+int vmw_du_crtc_cursor_move(struct drm_crtc *crtc, int x, int y)
+{
+	struct vmw_private *dev_priv = vmw_priv(crtc->dev);
+	struct vmw_display_unit *du = vmw_crtc_to_du(crtc);
+	bool shown = du->cursor_surface || du->cursor_dmabuf ? true : false;
+
+	du->cursor_x = x + crtc->x;
+	du->cursor_y = y + crtc->y;
+
+	vmw_cursor_update_position(dev_priv, shown,
+				   du->cursor_x, du->cursor_y);
+
+	return 0;
+}
+
+void vmw_kms_cursor_snoop(struct vmw_surface *srf,
+			  struct ttm_object_file *tfile,
+			  struct ttm_buffer_object *bo,
+			  SVGA3dCmdHeader *header)
+{
+	struct ttm_bo_kmap_obj map;
+	unsigned long kmap_offset;
+	unsigned long kmap_num;
+	SVGA3dCopyBox *box;
+	unsigned box_count;
+	void *virtual;
+	bool dummy;
+	struct vmw_dma_cmd {
+		SVGA3dCmdHeader header;
+		SVGA3dCmdSurfaceDMA dma;
+	} *cmd;
+	int ret;
+
+	cmd = container_of(header, struct vmw_dma_cmd, header);
+
+	/* No snooper installed */
+	if (!srf->snooper.image)
+		return;
+
+	if (cmd->dma.host.face != 0 || cmd->dma.host.mipmap != 0) {
+		DRM_ERROR("face and mipmap for cursors should never != 0\n");
+		return;
+	}
+
+	if (cmd->header.size < 64) {
+		DRM_ERROR("at least one full copy box must be given\n");
+		return;
+	}
+
+	box = (SVGA3dCopyBox *)&cmd[1];
+	box_count = (cmd->header.size - sizeof(SVGA3dCmdSurfaceDMA)) /
+			sizeof(SVGA3dCopyBox);
+
+	if (cmd->dma.guest.pitch != (64 * 4) ||
+	    cmd->dma.guest.ptr.offset % PAGE_SIZE ||
+	    box->x != 0    || box->y != 0    || box->z != 0    ||
+	    box->srcx != 0 || box->srcy != 0 || box->srcz != 0 ||
+	    box->w != 64   || box->h != 64   || box->d != 1    ||
+	    box_count != 1) {
+		/* TODO handle none page aligned offsets */
+		/* TODO handle partial uploads and pitch != 256 */
+		/* TODO handle more then one copy (size != 64) */
+		DRM_ERROR("lazy programer, cant handle wierd stuff\n");
+		return;
+	}
+
+	kmap_offset = cmd->dma.guest.ptr.offset >> PAGE_SHIFT;
+	kmap_num = (64*64*4) >> PAGE_SHIFT;
+
+	ret = ttm_bo_reserve(bo, true, false, false, 0);
+	if (unlikely(ret != 0)) {
+		DRM_ERROR("reserve failed\n");
+		return;
+	}
+
+	ret = ttm_bo_kmap(bo, kmap_offset, kmap_num, &map);
+	if (unlikely(ret != 0))
+		goto err_unreserve;
+
+	virtual = ttm_kmap_obj_virtual(&map, &dummy);
+
+	memcpy(srf->snooper.image, virtual, 64*64*4);
+	srf->snooper.age++;
+
+	/* we can't call this function from this function since execbuf has
+	 * reserved fifo space.
+	 *
+	 * if (srf->snooper.crtc)
+	 *	vmw_ldu_crtc_cursor_update_image(dev_priv,
+	 *					 srf->snooper.image, 64, 64,
+	 *					 du->hotspot_x, du->hotspot_y);
+	 */
+
+	ttm_bo_kunmap(&map);
+err_unreserve:
+	ttm_bo_unreserve(bo);
+}
+
+void vmw_kms_cursor_post_execbuf(struct vmw_private *dev_priv)
+{
+	struct drm_device *dev = dev_priv->dev;
+	struct vmw_display_unit *du;
+	struct drm_crtc *crtc;
+
+	mutex_lock(&dev->mode_config.mutex);
+
+	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
+		du = vmw_crtc_to_du(crtc);
+		if (!du->cursor_surface ||
+		    du->cursor_age == du->cursor_surface->snooper.age)
+			continue;
+
+		du->cursor_age = du->cursor_surface->snooper.age;
+		vmw_cursor_update_image(dev_priv,
+					du->cursor_surface->snooper.image,
+					64, 64, du->hotspot_x, du->hotspot_y);
+	}
+
+	mutex_unlock(&dev->mode_config.mutex);
+}
+
+/*
+ * Generic framebuffer code
+ */
+
+int vmw_framebuffer_create_handle(struct drm_framebuffer *fb,
+				  struct drm_file *file_priv,
+				  unsigned int *handle)
+{
+	if (handle)
+		handle = 0;
+
+	return 0;
+}
+
+/*
+ * Surface framebuffer code
+ */
+
+#define vmw_framebuffer_to_vfbs(x) \
+	container_of(x, struct vmw_framebuffer_surface, base.base)
+
+struct vmw_framebuffer_surface {
+	struct vmw_framebuffer base;
+	struct vmw_surface *surface;
+	struct delayed_work d_work;
+	struct mutex work_lock;
+	bool present_fs;
+};
+
+void vmw_framebuffer_surface_destroy(struct drm_framebuffer *framebuffer)
+{
+	struct vmw_framebuffer_surface *vfb =
+		vmw_framebuffer_to_vfbs(framebuffer);
+
+	cancel_delayed_work_sync(&vfb->d_work);
+	drm_framebuffer_cleanup(framebuffer);
+	vmw_surface_unreference(&vfb->surface);
+
+	kfree(framebuffer);
+}
+
+static void vmw_framebuffer_present_fs_callback(struct work_struct *work)
+{
+	struct delayed_work *d_work =
+		container_of(work, struct delayed_work, work);
+	struct vmw_framebuffer_surface *vfbs =
+		container_of(d_work, struct vmw_framebuffer_surface, d_work);
+	struct vmw_surface *surf = vfbs->surface;
+	struct drm_framebuffer *framebuffer = &vfbs->base.base;
+	struct vmw_private *dev_priv = vmw_priv(framebuffer->dev);
+
+	struct {
+		SVGA3dCmdHeader header;
+		SVGA3dCmdPresent body;
+		SVGA3dCopyRect cr;
+	} *cmd;
+
+	mutex_lock(&vfbs->work_lock);
+	if (!vfbs->present_fs)
+		goto out_unlock;
+
+	cmd = vmw_fifo_reserve(dev_priv, sizeof(*cmd));
+	if (unlikely(cmd == NULL))
+		goto out_resched;
+
+	cmd->header.id = cpu_to_le32(SVGA_3D_CMD_PRESENT);
+	cmd->header.size = cpu_to_le32(sizeof(cmd->body) + sizeof(cmd->cr));
+	cmd->body.sid = cpu_to_le32(surf->res.id);
+	cmd->cr.x = cpu_to_le32(0);
+	cmd->cr.y = cpu_to_le32(0);
+	cmd->cr.srcx = cmd->cr.x;
+	cmd->cr.srcy = cmd->cr.y;
+	cmd->cr.w = cpu_to_le32(framebuffer->width);
+	cmd->cr.h = cpu_to_le32(framebuffer->height);
+	vfbs->present_fs = false;
+	vmw_fifo_commit(dev_priv, sizeof(*cmd));
+out_resched:
+	/**
+	 * Will not re-add if already pending.
+	 */
+	schedule_delayed_work(&vfbs->d_work, VMWGFX_PRESENT_RATE);
+out_unlock:
+	mutex_unlock(&vfbs->work_lock);
+}
+
+
+int vmw_framebuffer_surface_dirty(struct drm_framebuffer *framebuffer,
+				  unsigned flags, unsigned color,
+				  struct drm_clip_rect *clips,
+				  unsigned num_clips)
+{
+	struct vmw_private *dev_priv = vmw_priv(framebuffer->dev);
+	struct vmw_framebuffer_surface *vfbs =
+		vmw_framebuffer_to_vfbs(framebuffer);
+	struct vmw_surface *surf = vfbs->surface;
+	struct drm_clip_rect norect;
+	SVGA3dCopyRect *cr;
+	int i, inc = 1;
+
+	struct {
+		SVGA3dCmdHeader header;
+		SVGA3dCmdPresent body;
+		SVGA3dCopyRect cr;
+	} *cmd;
+
+	if (!num_clips ||
+	    !(dev_priv->fifo.capabilities &
+	      SVGA_FIFO_CAP_SCREEN_OBJECT)) {
+		int ret;
+
+		mutex_lock(&vfbs->work_lock);
+		vfbs->present_fs = true;
+		ret = schedule_delayed_work(&vfbs->d_work, VMWGFX_PRESENT_RATE);
+		mutex_unlock(&vfbs->work_lock);
+		if (ret) {
+			/**
+			 * No work pending, Force immediate present.
+			 */
+			vmw_framebuffer_present_fs_callback(&vfbs->d_work.work);
+		}
+		return 0;
+	}
+
+	if (!num_clips) {
+		num_clips = 1;
+		clips = &norect;
+		norect.x1 = norect.y1 = 0;
+		norect.x2 = framebuffer->width;
+		norect.y2 = framebuffer->height;
+	} else if (flags & DRM_MODE_FB_DIRTY_ANNOTATE_COPY) {
+		num_clips /= 2;
+		inc = 2; /* skip source rects */
+	}
+
+	cmd = vmw_fifo_reserve(dev_priv, sizeof(*cmd) + (num_clips - 1) * sizeof(cmd->cr));
+	if (unlikely(cmd == NULL)) {
+		DRM_ERROR("Fifo reserve failed.\n");
+		return -ENOMEM;
+	}
+
+	memset(cmd, 0, sizeof(*cmd));
+
+	cmd->header.id = cpu_to_le32(SVGA_3D_CMD_PRESENT);
+	cmd->header.size = cpu_to_le32(sizeof(cmd->body) + num_clips * sizeof(cmd->cr));
+	cmd->body.sid = cpu_to_le32(surf->res.id);
+
+	for (i = 0, cr = &cmd->cr; i < num_clips; i++, cr++, clips += inc) {
+		cr->x = cpu_to_le16(clips->x1);
+		cr->y = cpu_to_le16(clips->y1);
+		cr->srcx = cr->x;
+		cr->srcy = cr->y;
+		cr->w = cpu_to_le16(clips->x2 - clips->x1);
+		cr->h = cpu_to_le16(clips->y2 - clips->y1);
+	}
+
+	vmw_fifo_commit(dev_priv, sizeof(*cmd) + (num_clips - 1) * sizeof(cmd->cr));
+
+	return 0;
+}
+
+static struct drm_framebuffer_funcs vmw_framebuffer_surface_funcs = {
+	.destroy = vmw_framebuffer_surface_destroy,
+	.dirty = vmw_framebuffer_surface_dirty,
+	.create_handle = vmw_framebuffer_create_handle,
+};
+
+int vmw_kms_new_framebuffer_surface(struct vmw_private *dev_priv,
+				    struct vmw_surface *surface,
+				    struct vmw_framebuffer **out,
+				    unsigned width, unsigned height)
+
+{
+	struct drm_device *dev = dev_priv->dev;
+	struct vmw_framebuffer_surface *vfbs;
+	int ret;
+
+	vfbs = kzalloc(sizeof(*vfbs), GFP_KERNEL);
+	if (!vfbs) {
+		ret = -ENOMEM;
+		goto out_err1;
+	}
+
+	ret = drm_framebuffer_init(dev, &vfbs->base.base,
+				   &vmw_framebuffer_surface_funcs);
+	if (ret)
+		goto out_err2;
+
+	if (!vmw_surface_reference(surface)) {
+		DRM_ERROR("failed to reference surface %p\n", surface);
+		goto out_err3;
+	}
+
+	/* XXX get the first 3 from the surface info */
+	vfbs->base.base.bits_per_pixel = 32;
+	vfbs->base.base.pitch = width * 32 / 4;
+	vfbs->base.base.depth = 24;
+	vfbs->base.base.width = width;
+	vfbs->base.base.height = height;
+	vfbs->base.pin = NULL;
+	vfbs->base.unpin = NULL;
+	vfbs->surface = surface;
+	mutex_init(&vfbs->work_lock);
+	INIT_DELAYED_WORK(&vfbs->d_work, &vmw_framebuffer_present_fs_callback);
+	*out = &vfbs->base;
+
+	return 0;
+
+out_err3:
+	drm_framebuffer_cleanup(&vfbs->base.base);
+out_err2:
+	kfree(vfbs);
+out_err1:
+	return ret;
+}
+
+/*
+ * Dmabuf framebuffer code
+ */
+
+#define vmw_framebuffer_to_vfbd(x) \
+	container_of(x, struct vmw_framebuffer_dmabuf, base.base)
+
+struct vmw_framebuffer_dmabuf {
+	struct vmw_framebuffer base;
+	struct vmw_dma_buffer *buffer;
+};
+
+void vmw_framebuffer_dmabuf_destroy(struct drm_framebuffer *framebuffer)
+{
+	struct vmw_framebuffer_dmabuf *vfbd =
+		vmw_framebuffer_to_vfbd(framebuffer);
+
+	drm_framebuffer_cleanup(framebuffer);
+	vmw_dmabuf_unreference(&vfbd->buffer);
+
+	kfree(vfbd);
+}
+
+int vmw_framebuffer_dmabuf_dirty(struct drm_framebuffer *framebuffer,
+				 unsigned flags, unsigned color,
+				 struct drm_clip_rect *clips,
+				 unsigned num_clips)
+{
+	struct vmw_private *dev_priv = vmw_priv(framebuffer->dev);
+	struct drm_clip_rect norect;
+	struct {
+		uint32_t header;
+		SVGAFifoCmdUpdate body;
+	} *cmd;
+	int i, increment = 1;
+
+	if (!num_clips) {
+		num_clips = 1;
+		clips = &norect;
+		norect.x1 = norect.y1 = 0;
+		norect.x2 = framebuffer->width;
+		norect.y2 = framebuffer->height;
+	} else if (flags & DRM_MODE_FB_DIRTY_ANNOTATE_COPY) {
+		num_clips /= 2;
+		increment = 2;
+	}
+
+	cmd = vmw_fifo_reserve(dev_priv, sizeof(*cmd) * num_clips);
+	if (unlikely(cmd == NULL)) {
+		DRM_ERROR("Fifo reserve failed.\n");
+		return -ENOMEM;
+	}
+
+	for (i = 0; i < num_clips; i++, clips += increment) {
+		cmd[i].header = cpu_to_le32(SVGA_CMD_UPDATE);
+		cmd[i].body.x = cpu_to_le32(clips->x1);
+		cmd[i].body.y = cpu_to_le32(clips->y1);
+		cmd[i].body.width = cpu_to_le32(clips->x2 - clips->x1);
+		cmd[i].body.height = cpu_to_le32(clips->y2 - clips->y1);
+	}
+
+	vmw_fifo_commit(dev_priv, sizeof(*cmd) * num_clips);
+
+	return 0;
+}
+
+static struct drm_framebuffer_funcs vmw_framebuffer_dmabuf_funcs = {
+	.destroy = vmw_framebuffer_dmabuf_destroy,
+	.dirty = vmw_framebuffer_dmabuf_dirty,
+	.create_handle = vmw_framebuffer_create_handle,
+};
+
+static int vmw_framebuffer_dmabuf_pin(struct vmw_framebuffer *vfb)
+{
+	struct vmw_private *dev_priv = vmw_priv(vfb->base.dev);
+	struct vmw_framebuffer_dmabuf *vfbd =
+		vmw_framebuffer_to_vfbd(&vfb->base);
+	int ret;
+
+	vmw_overlay_pause_all(dev_priv);
+
+	ret = vmw_dmabuf_to_start_of_vram(dev_priv, vfbd->buffer);
+
+	if (dev_priv->capabilities & SVGA_CAP_MULTIMON) {
+		vmw_write(dev_priv, SVGA_REG_NUM_GUEST_DISPLAYS, 1);
+		vmw_write(dev_priv, SVGA_REG_DISPLAY_ID, 0);
+		vmw_write(dev_priv, SVGA_REG_DISPLAY_IS_PRIMARY, true);
+		vmw_write(dev_priv, SVGA_REG_DISPLAY_POSITION_X, 0);
+		vmw_write(dev_priv, SVGA_REG_DISPLAY_POSITION_Y, 0);
+		vmw_write(dev_priv, SVGA_REG_DISPLAY_WIDTH, 0);
+		vmw_write(dev_priv, SVGA_REG_DISPLAY_HEIGHT, 0);
+		vmw_write(dev_priv, SVGA_REG_DISPLAY_ID, SVGA_ID_INVALID);
+
+		vmw_write(dev_priv, SVGA_REG_ENABLE, 1);
+		vmw_write(dev_priv, SVGA_REG_WIDTH, vfb->base.width);
+		vmw_write(dev_priv, SVGA_REG_HEIGHT, vfb->base.height);
+		vmw_write(dev_priv, SVGA_REG_BITS_PER_PIXEL, vfb->base.bits_per_pixel);
+		vmw_write(dev_priv, SVGA_REG_DEPTH, vfb->base.depth);
+		vmw_write(dev_priv, SVGA_REG_RED_MASK, 0x00ff0000);
+		vmw_write(dev_priv, SVGA_REG_GREEN_MASK, 0x0000ff00);
+		vmw_write(dev_priv, SVGA_REG_BLUE_MASK, 0x000000ff);
+	} else
+		WARN_ON(true);
+
+	vmw_overlay_resume_all(dev_priv);
+
+	return 0;
+}
+
+static int vmw_framebuffer_dmabuf_unpin(struct vmw_framebuffer *vfb)
+{
+	struct vmw_private *dev_priv = vmw_priv(vfb->base.dev);
+	struct vmw_framebuffer_dmabuf *vfbd =
+		vmw_framebuffer_to_vfbd(&vfb->base);
+
+	if (!vfbd->buffer) {
+		WARN_ON(!vfbd->buffer);
+		return 0;
+	}
+
+	return vmw_dmabuf_from_vram(dev_priv, vfbd->buffer);
+}
+
+int vmw_kms_new_framebuffer_dmabuf(struct vmw_private *dev_priv,
+				   struct vmw_dma_buffer *dmabuf,
+				   struct vmw_framebuffer **out,
+				   unsigned width, unsigned height)
+
+{
+	struct drm_device *dev = dev_priv->dev;
+	struct vmw_framebuffer_dmabuf *vfbd;
+	int ret;
+
+	vfbd = kzalloc(sizeof(*vfbd), GFP_KERNEL);
+	if (!vfbd) {
+		ret = -ENOMEM;
+		goto out_err1;
+	}
+
+	ret = drm_framebuffer_init(dev, &vfbd->base.base,
+				   &vmw_framebuffer_dmabuf_funcs);
+	if (ret)
+		goto out_err2;
+
+	if (!vmw_dmabuf_reference(dmabuf)) {
+		DRM_ERROR("failed to reference dmabuf %p\n", dmabuf);
+		goto out_err3;
+	}
+
+	/* XXX get the first 3 from the surface info */
+	vfbd->base.base.bits_per_pixel = 32;
+	vfbd->base.base.pitch = width * 32 / 4;
+	vfbd->base.base.depth = 24;
+	vfbd->base.base.width = width;
+	vfbd->base.base.height = height;
+	vfbd->base.pin = vmw_framebuffer_dmabuf_pin;
+	vfbd->base.unpin = vmw_framebuffer_dmabuf_unpin;
+	vfbd->buffer = dmabuf;
+	*out = &vfbd->base;
+
+	return 0;
+
+out_err3:
+	drm_framebuffer_cleanup(&vfbd->base.base);
+out_err2:
+	kfree(vfbd);
+out_err1:
+	return ret;
+}
+
+/*
+ * Generic Kernel modesetting functions
+ */
+
+static struct drm_framebuffer *vmw_kms_fb_create(struct drm_device *dev,
+						 struct drm_file *file_priv,
+						 struct drm_mode_fb_cmd *mode_cmd)
+{
+	struct vmw_private *dev_priv = vmw_priv(dev);
+	struct ttm_object_file *tfile = vmw_fpriv(file_priv)->tfile;
+	struct vmw_framebuffer *vfb = NULL;
+	struct vmw_surface *surface = NULL;
+	struct vmw_dma_buffer *bo = NULL;
+	int ret;
+
+	ret = vmw_user_surface_lookup_handle(dev_priv, tfile,
+					     mode_cmd->handle, &surface);
+	if (ret)
+		goto try_dmabuf;
+
+	if (!surface->scanout)
+		goto err_not_scanout;
+
+	ret = vmw_kms_new_framebuffer_surface(dev_priv, surface, &vfb,
+					      mode_cmd->width, mode_cmd->height);
+
+	/* vmw_user_surface_lookup takes one ref so does new_fb */
+	vmw_surface_unreference(&surface);
+
+	if (ret) {
+		DRM_ERROR("failed to create vmw_framebuffer: %i\n", ret);
+		return NULL;
+	}
+	return &vfb->base;
+
+try_dmabuf:
+	DRM_INFO("%s: trying buffer\n", __func__);
+
+	ret = vmw_user_dmabuf_lookup(tfile, mode_cmd->handle, &bo);
+	if (ret) {
+		DRM_ERROR("failed to find buffer: %i\n", ret);
+		return NULL;
+	}
+
+	ret = vmw_kms_new_framebuffer_dmabuf(dev_priv, bo, &vfb,
+					     mode_cmd->width, mode_cmd->height);
+
+	/* vmw_user_dmabuf_lookup takes one ref so does new_fb */
+	vmw_dmabuf_unreference(&bo);
+
+	if (ret) {
+		DRM_ERROR("failed to create vmw_framebuffer: %i\n", ret);
+		return NULL;
+	}
+
+	return &vfb->base;
+
+err_not_scanout:
+	DRM_ERROR("surface not marked as scanout\n");
+	/* vmw_user_surface_lookup takes one ref */
+	vmw_surface_unreference(&surface);
+
+	return NULL;
+}
+
+static int vmw_kms_fb_changed(struct drm_device *dev)
+{
+	return 0;
+}
+
+static struct drm_mode_config_funcs vmw_kms_funcs = {
+	.fb_create = vmw_kms_fb_create,
+	.fb_changed = vmw_kms_fb_changed,
+};
+
+int vmw_kms_init(struct vmw_private *dev_priv)
+{
+	struct drm_device *dev = dev_priv->dev;
+	int ret;
+
+	drm_mode_config_init(dev);
+	dev->mode_config.funcs = &vmw_kms_funcs;
+	dev->mode_config.min_width = 1;
+	dev->mode_config.min_height = 1;
+	dev->mode_config.max_width = dev_priv->fb_max_width;
+	dev->mode_config.max_height = dev_priv->fb_max_height;
+
+	ret = vmw_kms_init_legacy_display_system(dev_priv);
+
+	return 0;
+}
+
+int vmw_kms_close(struct vmw_private *dev_priv)
+{
+	/*
+	 * Docs says we should take the lock before calling this function
+	 * but since it destroys encoders and our destructor calls
+	 * drm_encoder_cleanup which takes the lock we deadlock.
+	 */
+	drm_mode_config_cleanup(dev_priv->dev);
+	vmw_kms_close_legacy_display_system(dev_priv);
+	return 0;
+}
+
+int vmw_kms_cursor_bypass_ioctl(struct drm_device *dev, void *data,
+				struct drm_file *file_priv)
+{
+	struct drm_vmw_cursor_bypass_arg *arg = data;
+	struct vmw_display_unit *du;
+	struct drm_mode_object *obj;
+	struct drm_crtc *crtc;
+	int ret = 0;
+
+
+	mutex_lock(&dev->mode_config.mutex);
+	if (arg->flags & DRM_VMW_CURSOR_BYPASS_ALL) {
+
+		list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
+			du = vmw_crtc_to_du(crtc);
+			du->hotspot_x = arg->xhot;
+			du->hotspot_y = arg->yhot;
+		}
+
+		mutex_unlock(&dev->mode_config.mutex);
+		return 0;
+	}
+
+	obj = drm_mode_object_find(dev, arg->crtc_id, DRM_MODE_OBJECT_CRTC);
+	if (!obj) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	crtc = obj_to_crtc(obj);
+	du = vmw_crtc_to_du(crtc);
+
+	du->hotspot_x = arg->xhot;
+	du->hotspot_y = arg->yhot;
+
+out:
+	mutex_unlock(&dev->mode_config.mutex);
+
+	return ret;
+}
+
+int vmw_kms_save_vga(struct vmw_private *vmw_priv)
+{
+	/*
+	 * setup a single multimon monitor with the size
+	 * of 0x0, this stops the UI from resizing when we
+	 * change the framebuffer size
+	 */
+	if (vmw_priv->capabilities & SVGA_CAP_MULTIMON) {
+		vmw_write(vmw_priv, SVGA_REG_NUM_GUEST_DISPLAYS, 1);
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_ID, 0);
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_IS_PRIMARY, true);
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_POSITION_X, 0);
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_POSITION_Y, 0);
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_WIDTH, 0);
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_HEIGHT, 0);
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_ID, SVGA_ID_INVALID);
+	}
+
+	vmw_priv->vga_width = vmw_read(vmw_priv, SVGA_REG_WIDTH);
+	vmw_priv->vga_height = vmw_read(vmw_priv, SVGA_REG_HEIGHT);
+	vmw_priv->vga_bpp = vmw_read(vmw_priv, SVGA_REG_BITS_PER_PIXEL);
+	vmw_priv->vga_depth = vmw_read(vmw_priv, SVGA_REG_DEPTH);
+	vmw_priv->vga_pseudo = vmw_read(vmw_priv, SVGA_REG_PSEUDOCOLOR);
+	vmw_priv->vga_red_mask = vmw_read(vmw_priv, SVGA_REG_RED_MASK);
+	vmw_priv->vga_green_mask = vmw_read(vmw_priv, SVGA_REG_GREEN_MASK);
+	vmw_priv->vga_blue_mask = vmw_read(vmw_priv, SVGA_REG_BLUE_MASK);
+
+	return 0;
+}
+
+int vmw_kms_restore_vga(struct vmw_private *vmw_priv)
+{
+	vmw_write(vmw_priv, SVGA_REG_WIDTH, vmw_priv->vga_width);
+	vmw_write(vmw_priv, SVGA_REG_HEIGHT, vmw_priv->vga_height);
+	vmw_write(vmw_priv, SVGA_REG_BITS_PER_PIXEL, vmw_priv->vga_bpp);
+	vmw_write(vmw_priv, SVGA_REG_DEPTH, vmw_priv->vga_depth);
+	vmw_write(vmw_priv, SVGA_REG_PSEUDOCOLOR, vmw_priv->vga_pseudo);
+	vmw_write(vmw_priv, SVGA_REG_RED_MASK, vmw_priv->vga_red_mask);
+	vmw_write(vmw_priv, SVGA_REG_GREEN_MASK, vmw_priv->vga_green_mask);
+	vmw_write(vmw_priv, SVGA_REG_BLUE_MASK, vmw_priv->vga_blue_mask);
+
+	/* TODO check for multimon */
+	vmw_write(vmw_priv, SVGA_REG_NUM_GUEST_DISPLAYS, 0);
+
+	return 0;
+}
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.h b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.h
new file mode 100644
index 000000000000..8b95249f0531
--- /dev/null
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.h
@@ -0,0 +1,102 @@
+/**************************************************************************
+ *
+ * Copyright © 2009 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef VMWGFX_KMS_H_
+#define VMWGFX_KMS_H_
+
+#include "drmP.h"
+#include "vmwgfx_drv.h"
+
+
+#define vmw_framebuffer_to_vfb(x) \
+	container_of(x, struct vmw_framebuffer, base)
+
+/**
+ * Base class for framebuffers
+ *
+ * @pin is called the when ever a crtc uses this framebuffer
+ * @unpin is called
+ */
+struct vmw_framebuffer {
+	struct drm_framebuffer base;
+	int (*pin)(struct vmw_framebuffer *fb);
+	int (*unpin)(struct vmw_framebuffer *fb);
+};
+
+
+#define vmw_crtc_to_du(x) \
+	container_of(x, struct vmw_display_unit, crtc)
+
+/*
+ * Basic cursor manipulation
+ */
+int vmw_cursor_update_image(struct vmw_private *dev_priv,
+			    u32 *image, u32 width, u32 height,
+			    u32 hotspotX, u32 hotspotY);
+void vmw_cursor_update_position(struct vmw_private *dev_priv,
+				bool show, int x, int y);
+
+/**
+ * Base class display unit.
+ *
+ * Since the SVGA hw doesn't have a concept of a crtc, encoder or connector
+ * so the display unit is all of them at the same time. This is true for both
+ * legacy multimon and screen objects.
+ */
+struct vmw_display_unit {
+	struct drm_crtc crtc;
+	struct drm_encoder encoder;
+	struct drm_connector connector;
+
+	struct vmw_surface *cursor_surface;
+	struct vmw_dma_buffer *cursor_dmabuf;
+	size_t cursor_age;
+
+	int cursor_x;
+	int cursor_y;
+
+	int hotspot_x;
+	int hotspot_y;
+
+	unsigned unit;
+};
+
+/*
+ * Shared display unit functions - vmwgfx_kms.c
+ */
+void vmw_display_unit_cleanup(struct vmw_display_unit *du);
+int vmw_du_crtc_cursor_set(struct drm_crtc *crtc, struct drm_file *file_priv,
+			   uint32_t handle, uint32_t width, uint32_t height);
+int vmw_du_crtc_cursor_move(struct drm_crtc *crtc, int x, int y);
+
+/*
+ * Legacy display unit functions - vmwgfx_ldu.h
+ */
+int vmw_kms_init_legacy_display_system(struct vmw_private *dev_priv);
+int vmw_kms_close_legacy_display_system(struct vmw_private *dev_priv);
+
+#endif
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ldu.c b/drivers/gpu/drm/vmwgfx/vmwgfx_ldu.c
new file mode 100644
index 000000000000..90891593bf6c
--- /dev/null
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ldu.c
@@ -0,0 +1,516 @@
+/**************************************************************************
+ *
+ * Copyright © 2009 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "vmwgfx_kms.h"
+
+#define vmw_crtc_to_ldu(x) \
+	container_of(x, struct vmw_legacy_display_unit, base.crtc)
+#define vmw_encoder_to_ldu(x) \
+	container_of(x, struct vmw_legacy_display_unit, base.encoder)
+#define vmw_connector_to_ldu(x) \
+	container_of(x, struct vmw_legacy_display_unit, base.connector)
+
+struct vmw_legacy_display {
+	struct list_head active;
+
+	unsigned num_active;
+
+	struct vmw_framebuffer *fb;
+};
+
+/**
+ * Display unit using the legacy register interface.
+ */
+struct vmw_legacy_display_unit {
+	struct vmw_display_unit base;
+
+	struct list_head active;
+
+	unsigned unit;
+};
+
+static void vmw_ldu_destroy(struct vmw_legacy_display_unit *ldu)
+{
+	list_del_init(&ldu->active);
+	vmw_display_unit_cleanup(&ldu->base);
+	kfree(ldu);
+}
+
+
+/*
+ * Legacy Display Unit CRTC functions
+ */
+
+static void vmw_ldu_crtc_save(struct drm_crtc *crtc)
+{
+}
+
+static void vmw_ldu_crtc_restore(struct drm_crtc *crtc)
+{
+}
+
+static void vmw_ldu_crtc_gamma_set(struct drm_crtc *crtc,
+				   u16 *r, u16 *g, u16 *b,
+				   uint32_t size)
+{
+}
+
+static void vmw_ldu_crtc_destroy(struct drm_crtc *crtc)
+{
+	vmw_ldu_destroy(vmw_crtc_to_ldu(crtc));
+}
+
+static int vmw_ldu_commit_list(struct vmw_private *dev_priv)
+{
+	struct vmw_legacy_display *lds = dev_priv->ldu_priv;
+	struct vmw_legacy_display_unit *entry;
+	struct drm_crtc *crtc;
+	int i = 0;
+
+	/* to stop the screen from changing size on resize */
+	vmw_write(dev_priv, SVGA_REG_NUM_GUEST_DISPLAYS, 0);
+	for (i = 0; i < lds->num_active; i++) {
+		vmw_write(dev_priv, SVGA_REG_DISPLAY_ID, i);
+		vmw_write(dev_priv, SVGA_REG_DISPLAY_IS_PRIMARY, !i);
+		vmw_write(dev_priv, SVGA_REG_DISPLAY_POSITION_X, 0);
+		vmw_write(dev_priv, SVGA_REG_DISPLAY_POSITION_Y, 0);
+		vmw_write(dev_priv, SVGA_REG_DISPLAY_WIDTH, 0);
+		vmw_write(dev_priv, SVGA_REG_DISPLAY_HEIGHT, 0);
+		vmw_write(dev_priv, SVGA_REG_DISPLAY_ID, SVGA_ID_INVALID);
+	}
+
+	/* Now set the mode */
+	vmw_write(dev_priv, SVGA_REG_NUM_GUEST_DISPLAYS, lds->num_active);
+	i = 0;
+	list_for_each_entry(entry, &lds->active, active) {
+		crtc = &entry->base.crtc;
+
+		vmw_write(dev_priv, SVGA_REG_DISPLAY_ID, i);
+		vmw_write(dev_priv, SVGA_REG_DISPLAY_IS_PRIMARY, !i);
+		vmw_write(dev_priv, SVGA_REG_DISPLAY_POSITION_X, crtc->x);
+		vmw_write(dev_priv, SVGA_REG_DISPLAY_POSITION_Y, crtc->y);
+		vmw_write(dev_priv, SVGA_REG_DISPLAY_WIDTH, crtc->mode.hdisplay);
+		vmw_write(dev_priv, SVGA_REG_DISPLAY_HEIGHT, crtc->mode.vdisplay);
+		vmw_write(dev_priv, SVGA_REG_DISPLAY_ID, SVGA_ID_INVALID);
+
+		i++;
+	}
+
+	return 0;
+}
+
+static int vmw_ldu_del_active(struct vmw_private *vmw_priv,
+			      struct vmw_legacy_display_unit *ldu)
+{
+	struct vmw_legacy_display *ld = vmw_priv->ldu_priv;
+	if (list_empty(&ldu->active))
+		return 0;
+
+	list_del_init(&ldu->active);
+	if (--(ld->num_active) == 0) {
+		BUG_ON(!ld->fb);
+		if (ld->fb->unpin)
+			ld->fb->unpin(ld->fb);
+		ld->fb = NULL;
+	}
+
+	return 0;
+}
+
+static int vmw_ldu_add_active(struct vmw_private *vmw_priv,
+			      struct vmw_legacy_display_unit *ldu,
+			      struct vmw_framebuffer *vfb)
+{
+	struct vmw_legacy_display *ld = vmw_priv->ldu_priv;
+	struct vmw_legacy_display_unit *entry;
+	struct list_head *at;
+
+	if (!list_empty(&ldu->active))
+		return 0;
+
+	at = &ld->active;
+	list_for_each_entry(entry, &ld->active, active) {
+		if (entry->unit > ldu->unit)
+			break;
+
+		at = &entry->active;
+	}
+
+	list_add(&ldu->active, at);
+	if (ld->num_active++ == 0) {
+		BUG_ON(ld->fb);
+		if (vfb->pin)
+			vfb->pin(vfb);
+		ld->fb = vfb;
+	}
+
+	return 0;
+}
+
+static int vmw_ldu_crtc_set_config(struct drm_mode_set *set)
+{
+	struct vmw_private *dev_priv;
+	struct vmw_legacy_display_unit *ldu;
+	struct drm_connector *connector;
+	struct drm_display_mode *mode;
+	struct drm_encoder *encoder;
+	struct vmw_framebuffer *vfb;
+	struct drm_framebuffer *fb;
+	struct drm_crtc *crtc;
+
+	if (!set)
+		return -EINVAL;
+
+	if (!set->crtc)
+		return -EINVAL;
+
+	/* get the ldu */
+	crtc = set->crtc;
+	ldu = vmw_crtc_to_ldu(crtc);
+	vfb = set->fb ? vmw_framebuffer_to_vfb(set->fb) : NULL;
+	dev_priv = vmw_priv(crtc->dev);
+
+	if (set->num_connectors > 1) {
+		DRM_ERROR("to many connectors\n");
+		return -EINVAL;
+	}
+
+	if (set->num_connectors == 1 &&
+	    set->connectors[0] != &ldu->base.connector) {
+		DRM_ERROR("connector doesn't match %p %p\n",
+			set->connectors[0], &ldu->base.connector);
+		return -EINVAL;
+	}
+
+	/* ldu only supports one fb active at the time */
+	if (dev_priv->ldu_priv->fb && vfb &&
+	    dev_priv->ldu_priv->fb != vfb) {
+		DRM_ERROR("Multiple framebuffers not supported\n");
+		return -EINVAL;
+	}
+
+	/* since they always map one to one these are safe */
+	connector = &ldu->base.connector;
+	encoder = &ldu->base.encoder;
+
+	/* should we turn the crtc off? */
+	if (set->num_connectors == 0 || !set->mode || !set->fb) {
+
+		connector->encoder = NULL;
+		encoder->crtc = NULL;
+		crtc->fb = NULL;
+
+		vmw_ldu_del_active(dev_priv, ldu);
+
+		vmw_ldu_commit_list(dev_priv);
+
+		return 0;
+	}
+
+
+	/* we now know we want to set a mode */
+	mode = set->mode;
+	fb = set->fb;
+
+	if (set->x + mode->hdisplay > fb->width ||
+	    set->y + mode->vdisplay > fb->height) {
+		DRM_ERROR("set outside of framebuffer\n");
+		return -EINVAL;
+	}
+
+	vmw_fb_off(dev_priv);
+
+	crtc->fb = fb;
+	encoder->crtc = crtc;
+	connector->encoder = encoder;
+	crtc->x = set->x;
+	crtc->y = set->y;
+	crtc->mode = *mode;
+
+	vmw_ldu_add_active(dev_priv, ldu, vfb);
+
+	vmw_ldu_commit_list(dev_priv);
+
+	return 0;
+}
+
+static struct drm_crtc_funcs vmw_legacy_crtc_funcs = {
+	.save = vmw_ldu_crtc_save,
+	.restore = vmw_ldu_crtc_restore,
+	.cursor_set = vmw_du_crtc_cursor_set,
+	.cursor_move = vmw_du_crtc_cursor_move,
+	.gamma_set = vmw_ldu_crtc_gamma_set,
+	.destroy = vmw_ldu_crtc_destroy,
+	.set_config = vmw_ldu_crtc_set_config,
+};
+
+/*
+ * Legacy Display Unit encoder functions
+ */
+
+static void vmw_ldu_encoder_destroy(struct drm_encoder *encoder)
+{
+	vmw_ldu_destroy(vmw_encoder_to_ldu(encoder));
+}
+
+static struct drm_encoder_funcs vmw_legacy_encoder_funcs = {
+	.destroy = vmw_ldu_encoder_destroy,
+};
+
+/*
+ * Legacy Display Unit connector functions
+ */
+
+static void vmw_ldu_connector_dpms(struct drm_connector *connector, int mode)
+{
+}
+
+static void vmw_ldu_connector_save(struct drm_connector *connector)
+{
+}
+
+static void vmw_ldu_connector_restore(struct drm_connector *connector)
+{
+}
+
+static enum drm_connector_status
+	vmw_ldu_connector_detect(struct drm_connector *connector)
+{
+	/* XXX vmwctrl should control connection status */
+	if (vmw_connector_to_ldu(connector)->base.unit == 0)
+		return connector_status_connected;
+	return connector_status_disconnected;
+}
+
+static struct drm_display_mode vmw_ldu_connector_builtin[] = {
+	/* 640x480@60Hz */
+	{ DRM_MODE("640x480", DRM_MODE_TYPE_DRIVER, 25175, 640, 656,
+		   752, 800, 0, 480, 489, 492, 525, 0,
+		   DRM_MODE_FLAG_NHSYNC | DRM_MODE_FLAG_NVSYNC) },
+	/* 800x600@60Hz */
+	{ DRM_MODE("800x600",
+		   DRM_MODE_TYPE_DRIVER | DRM_MODE_TYPE_PREFERRED,
+		   40000, 800, 840, 968, 1056, 0, 600, 601, 605, 628,
+		   0, DRM_MODE_FLAG_PHSYNC | DRM_MODE_FLAG_PVSYNC) },
+	/* 1024x768@60Hz */
+	{ DRM_MODE("1024x768", DRM_MODE_TYPE_DRIVER, 65000, 1024, 1048,
+		   1184, 1344, 0, 768, 771, 777, 806, 0,
+		   DRM_MODE_FLAG_NHSYNC | DRM_MODE_FLAG_NVSYNC) },
+	/* 1152x864@75Hz */
+	{ DRM_MODE("1152x864", DRM_MODE_TYPE_DRIVER, 108000, 1152, 1216,
+		   1344, 1600, 0, 864, 865, 868, 900, 0,
+		   DRM_MODE_FLAG_PHSYNC | DRM_MODE_FLAG_PVSYNC) },
+	/* 1280x768@60Hz */
+	{ DRM_MODE("1280x768", DRM_MODE_TYPE_DRIVER, 79500, 1280, 1344,
+		   1472, 1664, 0, 768, 771, 778, 798, 0,
+		   DRM_MODE_FLAG_NHSYNC | DRM_MODE_FLAG_PVSYNC) },
+	/* 1280x800@60Hz */
+	{ DRM_MODE("1280x800", DRM_MODE_TYPE_DRIVER, 83500, 1280, 1352,
+		   1480, 1680, 0, 800, 803, 809, 831, 0,
+		   DRM_MODE_FLAG_PHSYNC | DRM_MODE_FLAG_NVSYNC) },
+	/* 1280x960@60Hz */
+	{ DRM_MODE("1280x960", DRM_MODE_TYPE_DRIVER, 108000, 1280, 1376,
+		   1488, 1800, 0, 960, 961, 964, 1000, 0,
+		   DRM_MODE_FLAG_PHSYNC | DRM_MODE_FLAG_PVSYNC) },
+	/* 1280x1024@60Hz */
+	{ DRM_MODE("1280x1024", DRM_MODE_TYPE_DRIVER, 108000, 1280, 1328,
+		   1440, 1688, 0, 1024, 1025, 1028, 1066, 0,
+		   DRM_MODE_FLAG_PHSYNC | DRM_MODE_FLAG_PVSYNC) },
+	/* 1360x768@60Hz */
+	{ DRM_MODE("1360x768", DRM_MODE_TYPE_DRIVER, 85500, 1360, 1424,
+		   1536, 1792, 0, 768, 771, 777, 795, 0,
+		   DRM_MODE_FLAG_PHSYNC | DRM_MODE_FLAG_PVSYNC) },
+	/* 1440x1050@60Hz */
+	{ DRM_MODE("1400x1050", DRM_MODE_TYPE_DRIVER, 121750, 1400, 1488,
+		   1632, 1864, 0, 1050, 1053, 1057, 1089, 0,
+		   DRM_MODE_FLAG_NHSYNC | DRM_MODE_FLAG_PVSYNC) },
+	/* 1440x900@60Hz */
+	{ DRM_MODE("1440x900", DRM_MODE_TYPE_DRIVER, 106500, 1440, 1520,
+		   1672, 1904, 0, 900, 903, 909, 934, 0,
+		   DRM_MODE_FLAG_NHSYNC | DRM_MODE_FLAG_PVSYNC) },
+	/* 1600x1200@60Hz */
+	{ DRM_MODE("1600x1200", DRM_MODE_TYPE_DRIVER, 162000, 1600, 1664,
+		   1856, 2160, 0, 1200, 1201, 1204, 1250, 0,
+		   DRM_MODE_FLAG_PHSYNC | DRM_MODE_FLAG_PVSYNC) },
+	/* 1680x1050@60Hz */
+	{ DRM_MODE("1680x1050", DRM_MODE_TYPE_DRIVER, 146250, 1680, 1784,
+		   1960, 2240, 0, 1050, 1053, 1059, 1089, 0,
+		   DRM_MODE_FLAG_NHSYNC | DRM_MODE_FLAG_PVSYNC) },
+	/* 1792x1344@60Hz */
+	{ DRM_MODE("1792x1344", DRM_MODE_TYPE_DRIVER, 204750, 1792, 1920,
+		   2120, 2448, 0, 1344, 1345, 1348, 1394, 0,
+		   DRM_MODE_FLAG_NHSYNC | DRM_MODE_FLAG_PVSYNC) },
+	/* 1853x1392@60Hz */
+	{ DRM_MODE("1856x1392", DRM_MODE_TYPE_DRIVER, 218250, 1856, 1952,
+		   2176, 2528, 0, 1392, 1393, 1396, 1439, 0,
+		   DRM_MODE_FLAG_NHSYNC | DRM_MODE_FLAG_PVSYNC) },
+	/* 1920x1200@60Hz */
+	{ DRM_MODE("1920x1200", DRM_MODE_TYPE_DRIVER, 193250, 1920, 2056,
+		   2256, 2592, 0, 1200, 1203, 1209, 1245, 0,
+		   DRM_MODE_FLAG_NHSYNC | DRM_MODE_FLAG_PVSYNC) },
+	/* 1920x1440@60Hz */
+	{ DRM_MODE("1920x1440", DRM_MODE_TYPE_DRIVER, 234000, 1920, 2048,
+		   2256, 2600, 0, 1440, 1441, 1444, 1500, 0,
+		   DRM_MODE_FLAG_NHSYNC | DRM_MODE_FLAG_PVSYNC) },
+	/* 2560x1600@60Hz */
+	{ DRM_MODE("2560x1600", DRM_MODE_TYPE_DRIVER, 348500, 2560, 2752,
+		   3032, 3504, 0, 1600, 1603, 1609, 1658, 0,
+		   DRM_MODE_FLAG_NHSYNC | DRM_MODE_FLAG_PVSYNC) },
+	/* Terminate */
+	{ DRM_MODE("", 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0) },
+};
+
+static int vmw_ldu_connector_fill_modes(struct drm_connector *connector,
+					uint32_t max_width, uint32_t max_height)
+{
+	struct drm_device *dev = connector->dev;
+	struct drm_display_mode *mode = NULL;
+	int i;
+
+	for (i = 0; vmw_ldu_connector_builtin[i].type != 0; i++) {
+		if (vmw_ldu_connector_builtin[i].hdisplay > max_width ||
+		    vmw_ldu_connector_builtin[i].vdisplay > max_height)
+			continue;
+
+		mode = drm_mode_duplicate(dev, &vmw_ldu_connector_builtin[i]);
+		if (!mode)
+			return 0;
+		mode->vrefresh = drm_mode_vrefresh(mode);
+
+		drm_mode_probed_add(connector, mode);
+	}
+
+	drm_mode_connector_list_update(connector);
+
+	return 1;
+}
+
+static int vmw_ldu_connector_set_property(struct drm_connector *connector,
+					  struct drm_property *property,
+					  uint64_t val)
+{
+	return 0;
+}
+
+static void vmw_ldu_connector_destroy(struct drm_connector *connector)
+{
+	vmw_ldu_destroy(vmw_connector_to_ldu(connector));
+}
+
+static struct drm_connector_funcs vmw_legacy_connector_funcs = {
+	.dpms = vmw_ldu_connector_dpms,
+	.save = vmw_ldu_connector_save,
+	.restore = vmw_ldu_connector_restore,
+	.detect = vmw_ldu_connector_detect,
+	.fill_modes = vmw_ldu_connector_fill_modes,
+	.set_property = vmw_ldu_connector_set_property,
+	.destroy = vmw_ldu_connector_destroy,
+};
+
+static int vmw_ldu_init(struct vmw_private *dev_priv, unsigned unit)
+{
+	struct vmw_legacy_display_unit *ldu;
+	struct drm_device *dev = dev_priv->dev;
+	struct drm_connector *connector;
+	struct drm_encoder *encoder;
+	struct drm_crtc *crtc;
+
+	ldu = kzalloc(sizeof(*ldu), GFP_KERNEL);
+	if (!ldu)
+		return -ENOMEM;
+
+	ldu->unit = unit;
+	crtc = &ldu->base.crtc;
+	encoder = &ldu->base.encoder;
+	connector = &ldu->base.connector;
+
+	drm_connector_init(dev, connector, &vmw_legacy_connector_funcs,
+			   DRM_MODE_CONNECTOR_LVDS);
+	/* Initial status */
+	if (unit == 0)
+		connector->status = connector_status_connected;
+	else
+		connector->status = connector_status_disconnected;
+
+	drm_encoder_init(dev, encoder, &vmw_legacy_encoder_funcs,
+			 DRM_MODE_ENCODER_LVDS);
+	drm_mode_connector_attach_encoder(connector, encoder);
+	encoder->possible_crtcs = (1 << unit);
+	encoder->possible_clones = 0;
+
+	INIT_LIST_HEAD(&ldu->active);
+
+	drm_crtc_init(dev, crtc, &vmw_legacy_crtc_funcs);
+
+	drm_connector_attach_property(connector,
+				      dev->mode_config.dirty_info_property,
+				      1);
+
+	return 0;
+}
+
+int vmw_kms_init_legacy_display_system(struct vmw_private *dev_priv)
+{
+	if (dev_priv->ldu_priv) {
+		DRM_INFO("ldu system already on\n");
+		return -EINVAL;
+	}
+
+	dev_priv->ldu_priv = kmalloc(GFP_KERNEL, sizeof(*dev_priv->ldu_priv));
+
+	if (!dev_priv->ldu_priv)
+		return -ENOMEM;
+
+	INIT_LIST_HEAD(&dev_priv->ldu_priv->active);
+	dev_priv->ldu_priv->num_active = 0;
+	dev_priv->ldu_priv->fb = NULL;
+
+	drm_mode_create_dirty_info_property(dev_priv->dev);
+
+	vmw_ldu_init(dev_priv, 0);
+	vmw_ldu_init(dev_priv, 1);
+	vmw_ldu_init(dev_priv, 2);
+	vmw_ldu_init(dev_priv, 3);
+	vmw_ldu_init(dev_priv, 4);
+	vmw_ldu_init(dev_priv, 5);
+	vmw_ldu_init(dev_priv, 6);
+	vmw_ldu_init(dev_priv, 7);
+
+	return 0;
+}
+
+int vmw_kms_close_legacy_display_system(struct vmw_private *dev_priv)
+{
+	if (!dev_priv->ldu_priv)
+		return -ENOSYS;
+
+	BUG_ON(!list_empty(&dev_priv->ldu_priv->active));
+
+	kfree(dev_priv->ldu_priv);
+
+	return 0;
+}
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_overlay.c b/drivers/gpu/drm/vmwgfx/vmwgfx_overlay.c
new file mode 100644
index 000000000000..5b6eabeb7f51
--- /dev/null
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_overlay.c
@@ -0,0 +1,625 @@
+/**************************************************************************
+ *
+ * Copyright © 2009 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#include "drmP.h"
+#include "vmwgfx_drv.h"
+
+#include "ttm/ttm_placement.h"
+
+#include "svga_overlay.h"
+#include "svga_escape.h"
+
+#define VMW_MAX_NUM_STREAMS 1
+
+struct vmw_stream {
+	struct vmw_dma_buffer *buf;
+	bool claimed;
+	bool paused;
+	struct drm_vmw_control_stream_arg saved;
+};
+
+/**
+ * Overlay control
+ */
+struct vmw_overlay {
+	/*
+	 * Each stream is a single overlay. In Xv these are called ports.
+	 */
+	struct mutex mutex;
+	struct vmw_stream stream[VMW_MAX_NUM_STREAMS];
+};
+
+static inline struct vmw_overlay *vmw_overlay(struct drm_device *dev)
+{
+	struct vmw_private *dev_priv = vmw_priv(dev);
+	return dev_priv ? dev_priv->overlay_priv : NULL;
+}
+
+struct vmw_escape_header {
+	uint32_t cmd;
+	SVGAFifoCmdEscape body;
+};
+
+struct vmw_escape_video_flush {
+	struct vmw_escape_header escape;
+	SVGAEscapeVideoFlush flush;
+};
+
+static inline void fill_escape(struct vmw_escape_header *header,
+			       uint32_t size)
+{
+	header->cmd = SVGA_CMD_ESCAPE;
+	header->body.nsid = SVGA_ESCAPE_NSID_VMWARE;
+	header->body.size = size;
+}
+
+static inline void fill_flush(struct vmw_escape_video_flush *cmd,
+			      uint32_t stream_id)
+{
+	fill_escape(&cmd->escape, sizeof(cmd->flush));
+	cmd->flush.cmdType = SVGA_ESCAPE_VMWARE_VIDEO_FLUSH;
+	cmd->flush.streamId = stream_id;
+}
+
+/**
+ * Pin or unpin a buffer in vram.
+ *
+ * @dev_priv:  Driver private.
+ * @buf:  DMA buffer to pin or unpin.
+ * @pin:  Pin buffer in vram if true.
+ * @interruptible:  Use interruptible wait.
+ *
+ * Takes the current masters ttm lock in read.
+ *
+ * Returns
+ * -ERESTARTSYS if interrupted by a signal.
+ */
+static int vmw_dmabuf_pin_in_vram(struct vmw_private *dev_priv,
+				  struct vmw_dma_buffer *buf,
+				  bool pin, bool interruptible)
+{
+	struct ttm_buffer_object *bo = &buf->base;
+	struct ttm_placement *overlay_placement = &vmw_vram_placement;
+	int ret;
+
+	ret = ttm_read_lock(&dev_priv->active_master->lock, interruptible);
+	if (unlikely(ret != 0))
+		return ret;
+
+	ret = ttm_bo_reserve(bo, interruptible, false, false, 0);
+	if (unlikely(ret != 0))
+		goto err;
+
+	if (pin)
+		overlay_placement = &vmw_vram_ne_placement;
+
+	ret = ttm_bo_validate(bo, overlay_placement, interruptible, false);
+
+	ttm_bo_unreserve(bo);
+
+err:
+	ttm_read_unlock(&dev_priv->active_master->lock);
+
+	return ret;
+}
+
+/**
+ * Send put command to hw.
+ *
+ * Returns
+ * -ERESTARTSYS if interrupted by a signal.
+ */
+static int vmw_overlay_send_put(struct vmw_private *dev_priv,
+				struct vmw_dma_buffer *buf,
+				struct drm_vmw_control_stream_arg *arg,
+				bool interruptible)
+{
+	struct {
+		struct vmw_escape_header escape;
+		struct {
+			struct {
+				uint32_t cmdType;
+				uint32_t streamId;
+			} header;
+			struct {
+				uint32_t registerId;
+				uint32_t value;
+			} items[SVGA_VIDEO_PITCH_3 + 1];
+		} body;
+		struct vmw_escape_video_flush flush;
+	} *cmds;
+	uint32_t offset;
+	int i, ret;
+
+	for (;;) {
+		cmds = vmw_fifo_reserve(dev_priv, sizeof(*cmds));
+		if (cmds)
+			break;
+
+		ret = vmw_fallback_wait(dev_priv, false, true, 0,
+					interruptible, 3*HZ);
+		if (interruptible && ret == -ERESTARTSYS)
+			return ret;
+		else
+			BUG_ON(ret != 0);
+	}
+
+	fill_escape(&cmds->escape, sizeof(cmds->body));
+	cmds->body.header.cmdType = SVGA_ESCAPE_VMWARE_VIDEO_SET_REGS;
+	cmds->body.header.streamId = arg->stream_id;
+
+	for (i = 0; i <= SVGA_VIDEO_PITCH_3; i++)
+		cmds->body.items[i].registerId = i;
+
+	offset = buf->base.offset + arg->offset;
+
+	cmds->body.items[SVGA_VIDEO_ENABLED].value     = true;
+	cmds->body.items[SVGA_VIDEO_FLAGS].value       = arg->flags;
+	cmds->body.items[SVGA_VIDEO_DATA_OFFSET].value = offset;
+	cmds->body.items[SVGA_VIDEO_FORMAT].value      = arg->format;
+	cmds->body.items[SVGA_VIDEO_COLORKEY].value    = arg->color_key;
+	cmds->body.items[SVGA_VIDEO_SIZE].value        = arg->size;
+	cmds->body.items[SVGA_VIDEO_WIDTH].value       = arg->width;
+	cmds->body.items[SVGA_VIDEO_HEIGHT].value      = arg->height;
+	cmds->body.items[SVGA_VIDEO_SRC_X].value       = arg->src.x;
+	cmds->body.items[SVGA_VIDEO_SRC_Y].value       = arg->src.y;
+	cmds->body.items[SVGA_VIDEO_SRC_WIDTH].value   = arg->src.w;
+	cmds->body.items[SVGA_VIDEO_SRC_HEIGHT].value  = arg->src.h;
+	cmds->body.items[SVGA_VIDEO_DST_X].value       = arg->dst.x;
+	cmds->body.items[SVGA_VIDEO_DST_Y].value       = arg->dst.y;
+	cmds->body.items[SVGA_VIDEO_DST_WIDTH].value   = arg->dst.w;
+	cmds->body.items[SVGA_VIDEO_DST_HEIGHT].value  = arg->dst.h;
+	cmds->body.items[SVGA_VIDEO_PITCH_1].value     = arg->pitch[0];
+	cmds->body.items[SVGA_VIDEO_PITCH_2].value     = arg->pitch[1];
+	cmds->body.items[SVGA_VIDEO_PITCH_3].value     = arg->pitch[2];
+
+	fill_flush(&cmds->flush, arg->stream_id);
+
+	vmw_fifo_commit(dev_priv, sizeof(*cmds));
+
+	return 0;
+}
+
+/**
+ * Send stop command to hw.
+ *
+ * Returns
+ * -ERESTARTSYS if interrupted by a signal.
+ */
+static int vmw_overlay_send_stop(struct vmw_private *dev_priv,
+				 uint32_t stream_id,
+				 bool interruptible)
+{
+	struct {
+		struct vmw_escape_header escape;
+		SVGAEscapeVideoSetRegs body;
+		struct vmw_escape_video_flush flush;
+	} *cmds;
+	int ret;
+
+	for (;;) {
+		cmds = vmw_fifo_reserve(dev_priv, sizeof(*cmds));
+		if (cmds)
+			break;
+
+		ret = vmw_fallback_wait(dev_priv, false, true, 0,
+					interruptible, 3*HZ);
+		if (interruptible && ret == -ERESTARTSYS)
+			return ret;
+		else
+			BUG_ON(ret != 0);
+	}
+
+	fill_escape(&cmds->escape, sizeof(cmds->body));
+	cmds->body.header.cmdType = SVGA_ESCAPE_VMWARE_VIDEO_SET_REGS;
+	cmds->body.header.streamId = stream_id;
+	cmds->body.items[0].registerId = SVGA_VIDEO_ENABLED;
+	cmds->body.items[0].value = false;
+	fill_flush(&cmds->flush, stream_id);
+
+	vmw_fifo_commit(dev_priv, sizeof(*cmds));
+
+	return 0;
+}
+
+/**
+ * Stop or pause a stream.
+ *
+ * If the stream is paused the no evict flag is removed from the buffer
+ * but left in vram. This allows for instance mode_set to evict it
+ * should it need to.
+ *
+ * The caller must hold the overlay lock.
+ *
+ * @stream_id which stream to stop/pause.
+ * @pause true to pause, false to stop completely.
+ */
+static int vmw_overlay_stop(struct vmw_private *dev_priv,
+			    uint32_t stream_id, bool pause,
+			    bool interruptible)
+{
+	struct vmw_overlay *overlay = dev_priv->overlay_priv;
+	struct vmw_stream *stream = &overlay->stream[stream_id];
+	int ret;
+
+	/* no buffer attached the stream is completely stopped */
+	if (!stream->buf)
+		return 0;
+
+	/* If the stream is paused this is already done */
+	if (!stream->paused) {
+		ret = vmw_overlay_send_stop(dev_priv, stream_id,
+					    interruptible);
+		if (ret)
+			return ret;
+
+		/* We just remove the NO_EVICT flag so no -ENOMEM */
+		ret = vmw_dmabuf_pin_in_vram(dev_priv, stream->buf, false,
+					     interruptible);
+		if (interruptible && ret == -ERESTARTSYS)
+			return ret;
+		else
+			BUG_ON(ret != 0);
+	}
+
+	if (!pause) {
+		vmw_dmabuf_unreference(&stream->buf);
+		stream->paused = false;
+	} else {
+		stream->paused = true;
+	}
+
+	return 0;
+}
+
+/**
+ * Update a stream and send any put or stop fifo commands needed.
+ *
+ * The caller must hold the overlay lock.
+ *
+ * Returns
+ * -ENOMEM if buffer doesn't fit in vram.
+ * -ERESTARTSYS if interrupted.
+ */
+static int vmw_overlay_update_stream(struct vmw_private *dev_priv,
+				     struct vmw_dma_buffer *buf,
+				     struct drm_vmw_control_stream_arg *arg,
+				     bool interruptible)
+{
+	struct vmw_overlay *overlay = dev_priv->overlay_priv;
+	struct vmw_stream *stream = &overlay->stream[arg->stream_id];
+	int ret = 0;
+
+	if (!buf)
+		return -EINVAL;
+
+	DRM_DEBUG("   %s: old %p, new %p, %spaused\n", __func__,
+		  stream->buf, buf, stream->paused ? "" : "not ");
+
+	if (stream->buf != buf) {
+		ret = vmw_overlay_stop(dev_priv, arg->stream_id,
+				       false, interruptible);
+		if (ret)
+			return ret;
+	} else if (!stream->paused) {
+		/* If the buffers match and not paused then just send
+		 * the put command, no need to do anything else.
+		 */
+		ret = vmw_overlay_send_put(dev_priv, buf, arg, interruptible);
+		if (ret == 0)
+			stream->saved = *arg;
+		else
+			BUG_ON(!interruptible);
+
+		return ret;
+	}
+
+	/* We don't start the old stream if we are interrupted.
+	 * Might return -ENOMEM if it can't fit the buffer in vram.
+	 */
+	ret = vmw_dmabuf_pin_in_vram(dev_priv, buf, true, interruptible);
+	if (ret)
+		return ret;
+
+	ret = vmw_overlay_send_put(dev_priv, buf, arg, interruptible);
+	if (ret) {
+		/* This one needs to happen no matter what. We only remove
+		 * the NO_EVICT flag so this is safe from -ENOMEM.
+		 */
+		BUG_ON(vmw_dmabuf_pin_in_vram(dev_priv, buf, false, false) != 0);
+		return ret;
+	}
+
+	if (stream->buf != buf)
+		stream->buf = vmw_dmabuf_reference(buf);
+	stream->saved = *arg;
+
+	return 0;
+}
+
+/**
+ * Stop all streams.
+ *
+ * Used by the fb code when starting.
+ *
+ * Takes the overlay lock.
+ */
+int vmw_overlay_stop_all(struct vmw_private *dev_priv)
+{
+	struct vmw_overlay *overlay = dev_priv->overlay_priv;
+	int i, ret;
+
+	if (!overlay)
+		return 0;
+
+	mutex_lock(&overlay->mutex);
+
+	for (i = 0; i < VMW_MAX_NUM_STREAMS; i++) {
+		struct vmw_stream *stream = &overlay->stream[i];
+		if (!stream->buf)
+			continue;
+
+		ret = vmw_overlay_stop(dev_priv, i, false, false);
+		WARN_ON(ret != 0);
+	}
+
+	mutex_unlock(&overlay->mutex);
+
+	return 0;
+}
+
+/**
+ * Try to resume all paused streams.
+ *
+ * Used by the kms code after moving a new scanout buffer to vram.
+ *
+ * Takes the overlay lock.
+ */
+int vmw_overlay_resume_all(struct vmw_private *dev_priv)
+{
+	struct vmw_overlay *overlay = dev_priv->overlay_priv;
+	int i, ret;
+
+	if (!overlay)
+		return 0;
+
+	mutex_lock(&overlay->mutex);
+
+	for (i = 0; i < VMW_MAX_NUM_STREAMS; i++) {
+		struct vmw_stream *stream = &overlay->stream[i];
+		if (!stream->paused)
+			continue;
+
+		ret = vmw_overlay_update_stream(dev_priv, stream->buf,
+						&stream->saved, false);
+		if (ret != 0)
+			DRM_INFO("%s: *warning* failed to resume stream %i\n",
+				 __func__, i);
+	}
+
+	mutex_unlock(&overlay->mutex);
+
+	return 0;
+}
+
+/**
+ * Pauses all active streams.
+ *
+ * Used by the kms code when moving a new scanout buffer to vram.
+ *
+ * Takes the overlay lock.
+ */
+int vmw_overlay_pause_all(struct vmw_private *dev_priv)
+{
+	struct vmw_overlay *overlay = dev_priv->overlay_priv;
+	int i, ret;
+
+	if (!overlay)
+		return 0;
+
+	mutex_lock(&overlay->mutex);
+
+	for (i = 0; i < VMW_MAX_NUM_STREAMS; i++) {
+		if (overlay->stream[i].paused)
+			DRM_INFO("%s: *warning* stream %i already paused\n",
+				 __func__, i);
+		ret = vmw_overlay_stop(dev_priv, i, true, false);
+		WARN_ON(ret != 0);
+	}
+
+	mutex_unlock(&overlay->mutex);
+
+	return 0;
+}
+
+int vmw_overlay_ioctl(struct drm_device *dev, void *data,
+		      struct drm_file *file_priv)
+{
+	struct ttm_object_file *tfile = vmw_fpriv(file_priv)->tfile;
+	struct vmw_private *dev_priv = vmw_priv(dev);
+	struct vmw_overlay *overlay = dev_priv->overlay_priv;
+	struct drm_vmw_control_stream_arg *arg =
+	    (struct drm_vmw_control_stream_arg *)data;
+	struct vmw_dma_buffer *buf;
+	struct vmw_resource *res;
+	int ret;
+
+	if (!overlay)
+		return -ENOSYS;
+
+	ret = vmw_user_stream_lookup(dev_priv, tfile, &arg->stream_id, &res);
+	if (ret)
+		return ret;
+
+	mutex_lock(&overlay->mutex);
+
+	if (!arg->enabled) {
+		ret = vmw_overlay_stop(dev_priv, arg->stream_id, false, true);
+		goto out_unlock;
+	}
+
+	ret = vmw_user_dmabuf_lookup(tfile, arg->handle, &buf);
+	if (ret)
+		goto out_unlock;
+
+	ret = vmw_overlay_update_stream(dev_priv, buf, arg, true);
+
+	vmw_dmabuf_unreference(&buf);
+
+out_unlock:
+	mutex_unlock(&overlay->mutex);
+	vmw_resource_unreference(&res);
+
+	return ret;
+}
+
+int vmw_overlay_num_overlays(struct vmw_private *dev_priv)
+{
+	if (!dev_priv->overlay_priv)
+		return 0;
+
+	return VMW_MAX_NUM_STREAMS;
+}
+
+int vmw_overlay_num_free_overlays(struct vmw_private *dev_priv)
+{
+	struct vmw_overlay *overlay = dev_priv->overlay_priv;
+	int i, k;
+
+	if (!overlay)
+		return 0;
+
+	mutex_lock(&overlay->mutex);
+
+	for (i = 0, k = 0; i < VMW_MAX_NUM_STREAMS; i++)
+		if (!overlay->stream[i].claimed)
+			k++;
+
+	mutex_unlock(&overlay->mutex);
+
+	return k;
+}
+
+int vmw_overlay_claim(struct vmw_private *dev_priv, uint32_t *out)
+{
+	struct vmw_overlay *overlay = dev_priv->overlay_priv;
+	int i;
+
+	if (!overlay)
+		return -ENOSYS;
+
+	mutex_lock(&overlay->mutex);
+
+	for (i = 0; i < VMW_MAX_NUM_STREAMS; i++) {
+
+		if (overlay->stream[i].claimed)
+			continue;
+
+		overlay->stream[i].claimed = true;
+		*out = i;
+		mutex_unlock(&overlay->mutex);
+		return 0;
+	}
+
+	mutex_unlock(&overlay->mutex);
+	return -ESRCH;
+}
+
+int vmw_overlay_unref(struct vmw_private *dev_priv, uint32_t stream_id)
+{
+	struct vmw_overlay *overlay = dev_priv->overlay_priv;
+
+	BUG_ON(stream_id >= VMW_MAX_NUM_STREAMS);
+
+	if (!overlay)
+		return -ENOSYS;
+
+	mutex_lock(&overlay->mutex);
+
+	WARN_ON(!overlay->stream[stream_id].claimed);
+	vmw_overlay_stop(dev_priv, stream_id, false, false);
+	overlay->stream[stream_id].claimed = false;
+
+	mutex_unlock(&overlay->mutex);
+	return 0;
+}
+
+int vmw_overlay_init(struct vmw_private *dev_priv)
+{
+	struct vmw_overlay *overlay;
+	int i;
+
+	if (dev_priv->overlay_priv)
+		return -EINVAL;
+
+	if (!(dev_priv->fifo.capabilities & SVGA_FIFO_CAP_VIDEO) &&
+	     (dev_priv->fifo.capabilities & SVGA_FIFO_CAP_ESCAPE)) {
+		DRM_INFO("hardware doesn't support overlays\n");
+		return -ENOSYS;
+	}
+
+	overlay = kmalloc(GFP_KERNEL, sizeof(*overlay));
+	if (!overlay)
+		return -ENOMEM;
+
+	memset(overlay, 0, sizeof(*overlay));
+	mutex_init(&overlay->mutex);
+	for (i = 0; i < VMW_MAX_NUM_STREAMS; i++) {
+		overlay->stream[i].buf = NULL;
+		overlay->stream[i].paused = false;
+		overlay->stream[i].claimed = false;
+	}
+
+	dev_priv->overlay_priv = overlay;
+
+	return 0;
+}
+
+int vmw_overlay_close(struct vmw_private *dev_priv)
+{
+	struct vmw_overlay *overlay = dev_priv->overlay_priv;
+	bool forgotten_buffer = false;
+	int i;
+
+	if (!overlay)
+		return -ENOSYS;
+
+	for (i = 0; i < VMW_MAX_NUM_STREAMS; i++) {
+		if (overlay->stream[i].buf) {
+			forgotten_buffer = true;
+			vmw_overlay_stop(dev_priv, i, false, false);
+		}
+	}
+
+	WARN_ON(forgotten_buffer);
+
+	dev_priv->overlay_priv = NULL;
+	kfree(overlay);
+
+	return 0;
+}
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_reg.h b/drivers/gpu/drm/vmwgfx/vmwgfx_reg.h
new file mode 100644
index 000000000000..9d0dd3a342eb
--- /dev/null
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_reg.h
@@ -0,0 +1,57 @@
+/**************************************************************************
+ *
+ * Copyright © 2009 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * This file contains virtual hardware defines for kernel space.
+ */
+
+#ifndef _VMWGFX_REG_H_
+#define _VMWGFX_REG_H_
+
+#include <linux/types.h>
+
+#define VMWGFX_INDEX_PORT     0x0
+#define VMWGFX_VALUE_PORT     0x1
+#define VMWGFX_IRQSTATUS_PORT 0x8
+
+struct svga_guest_mem_descriptor {
+	__le32 ppn;
+	__le32 num_pages;
+};
+
+struct svga_fifo_cmd_fence {
+	__le32 fence;
+};
+
+#define SVGA_SYNC_GENERIC         1
+#define SVGA_SYNC_FIFOFULL        2
+
+#include "svga_types.h"
+
+#include "svga3d_reg.h"
+
+#endif
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c b/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c
new file mode 100644
index 000000000000..f8fbbc67a406
--- /dev/null
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c
@@ -0,0 +1,1187 @@
+/**************************************************************************
+ *
+ * Copyright © 2009 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "vmwgfx_drv.h"
+#include "vmwgfx_drm.h"
+#include "ttm/ttm_object.h"
+#include "ttm/ttm_placement.h"
+#include "drmP.h"
+
+#define VMW_RES_CONTEXT ttm_driver_type0
+#define VMW_RES_SURFACE ttm_driver_type1
+#define VMW_RES_STREAM ttm_driver_type2
+
+struct vmw_user_context {
+	struct ttm_base_object base;
+	struct vmw_resource res;
+};
+
+struct vmw_user_surface {
+	struct ttm_base_object base;
+	struct vmw_surface srf;
+};
+
+struct vmw_user_dma_buffer {
+	struct ttm_base_object base;
+	struct vmw_dma_buffer dma;
+};
+
+struct vmw_bo_user_rep {
+	uint32_t handle;
+	uint64_t map_handle;
+};
+
+struct vmw_stream {
+	struct vmw_resource res;
+	uint32_t stream_id;
+};
+
+struct vmw_user_stream {
+	struct ttm_base_object base;
+	struct vmw_stream stream;
+};
+
+static inline struct vmw_dma_buffer *
+vmw_dma_buffer(struct ttm_buffer_object *bo)
+{
+	return container_of(bo, struct vmw_dma_buffer, base);
+}
+
+static inline struct vmw_user_dma_buffer *
+vmw_user_dma_buffer(struct ttm_buffer_object *bo)
+{
+	struct vmw_dma_buffer *vmw_bo = vmw_dma_buffer(bo);
+	return container_of(vmw_bo, struct vmw_user_dma_buffer, dma);
+}
+
+struct vmw_resource *vmw_resource_reference(struct vmw_resource *res)
+{
+	kref_get(&res->kref);
+	return res;
+}
+
+static void vmw_resource_release(struct kref *kref)
+{
+	struct vmw_resource *res =
+	    container_of(kref, struct vmw_resource, kref);
+	struct vmw_private *dev_priv = res->dev_priv;
+
+	idr_remove(res->idr, res->id);
+	write_unlock(&dev_priv->resource_lock);
+
+	if (likely(res->hw_destroy != NULL))
+		res->hw_destroy(res);
+
+	if (res->res_free != NULL)
+		res->res_free(res);
+	else
+		kfree(res);
+
+	write_lock(&dev_priv->resource_lock);
+}
+
+void vmw_resource_unreference(struct vmw_resource **p_res)
+{
+	struct vmw_resource *res = *p_res;
+	struct vmw_private *dev_priv = res->dev_priv;
+
+	*p_res = NULL;
+	write_lock(&dev_priv->resource_lock);
+	kref_put(&res->kref, vmw_resource_release);
+	write_unlock(&dev_priv->resource_lock);
+}
+
+static int vmw_resource_init(struct vmw_private *dev_priv,
+			     struct vmw_resource *res,
+			     struct idr *idr,
+			     enum ttm_object_type obj_type,
+			     void (*res_free) (struct vmw_resource *res))
+{
+	int ret;
+
+	kref_init(&res->kref);
+	res->hw_destroy = NULL;
+	res->res_free = res_free;
+	res->res_type = obj_type;
+	res->idr = idr;
+	res->avail = false;
+	res->dev_priv = dev_priv;
+
+	do {
+		if (unlikely(idr_pre_get(idr, GFP_KERNEL) == 0))
+			return -ENOMEM;
+
+		write_lock(&dev_priv->resource_lock);
+		ret = idr_get_new_above(idr, res, 1, &res->id);
+		write_unlock(&dev_priv->resource_lock);
+
+	} while (ret == -EAGAIN);
+
+	return ret;
+}
+
+/**
+ * vmw_resource_activate
+ *
+ * @res:        Pointer to the newly created resource
+ * @hw_destroy: Destroy function. NULL if none.
+ *
+ * Activate a resource after the hardware has been made aware of it.
+ * Set tye destroy function to @destroy. Typically this frees the
+ * resource and destroys the hardware resources associated with it.
+ * Activate basically means that the function vmw_resource_lookup will
+ * find it.
+ */
+
+static void vmw_resource_activate(struct vmw_resource *res,
+				  void (*hw_destroy) (struct vmw_resource *))
+{
+	struct vmw_private *dev_priv = res->dev_priv;
+
+	write_lock(&dev_priv->resource_lock);
+	res->avail = true;
+	res->hw_destroy = hw_destroy;
+	write_unlock(&dev_priv->resource_lock);
+}
+
+struct vmw_resource *vmw_resource_lookup(struct vmw_private *dev_priv,
+					 struct idr *idr, int id)
+{
+	struct vmw_resource *res;
+
+	read_lock(&dev_priv->resource_lock);
+	res = idr_find(idr, id);
+	if (res && res->avail)
+		kref_get(&res->kref);
+	else
+		res = NULL;
+	read_unlock(&dev_priv->resource_lock);
+
+	if (unlikely(res == NULL))
+		return NULL;
+
+	return res;
+}
+
+/**
+ * Context management:
+ */
+
+static void vmw_hw_context_destroy(struct vmw_resource *res)
+{
+
+	struct vmw_private *dev_priv = res->dev_priv;
+	struct {
+		SVGA3dCmdHeader header;
+		SVGA3dCmdDestroyContext body;
+	} *cmd = vmw_fifo_reserve(dev_priv, sizeof(*cmd));
+
+	if (unlikely(cmd == NULL)) {
+		DRM_ERROR("Failed reserving FIFO space for surface "
+			  "destruction.\n");
+		return;
+	}
+
+	cmd->header.id = cpu_to_le32(SVGA_3D_CMD_CONTEXT_DESTROY);
+	cmd->header.size = cpu_to_le32(sizeof(cmd->body));
+	cmd->body.cid = cpu_to_le32(res->id);
+
+	vmw_fifo_commit(dev_priv, sizeof(*cmd));
+}
+
+static int vmw_context_init(struct vmw_private *dev_priv,
+			    struct vmw_resource *res,
+			    void (*res_free) (struct vmw_resource *res))
+{
+	int ret;
+
+	struct {
+		SVGA3dCmdHeader header;
+		SVGA3dCmdDefineContext body;
+	} *cmd;
+
+	ret = vmw_resource_init(dev_priv, res, &dev_priv->context_idr,
+				VMW_RES_CONTEXT, res_free);
+
+	if (unlikely(ret != 0)) {
+		if (res_free == NULL)
+			kfree(res);
+		else
+			res_free(res);
+		return ret;
+	}
+
+	cmd = vmw_fifo_reserve(dev_priv, sizeof(*cmd));
+	if (unlikely(cmd == NULL)) {
+		DRM_ERROR("Fifo reserve failed.\n");
+		vmw_resource_unreference(&res);
+		return -ENOMEM;
+	}
+
+	cmd->header.id = cpu_to_le32(SVGA_3D_CMD_CONTEXT_DEFINE);
+	cmd->header.size = cpu_to_le32(sizeof(cmd->body));
+	cmd->body.cid = cpu_to_le32(res->id);
+
+	vmw_fifo_commit(dev_priv, sizeof(*cmd));
+	vmw_resource_activate(res, vmw_hw_context_destroy);
+	return 0;
+}
+
+struct vmw_resource *vmw_context_alloc(struct vmw_private *dev_priv)
+{
+	struct vmw_resource *res = kmalloc(sizeof(*res), GFP_KERNEL);
+	int ret;
+
+	if (unlikely(res == NULL))
+		return NULL;
+
+	ret = vmw_context_init(dev_priv, res, NULL);
+	return (ret == 0) ? res : NULL;
+}
+
+/**
+ * User-space context management:
+ */
+
+static void vmw_user_context_free(struct vmw_resource *res)
+{
+	struct vmw_user_context *ctx =
+	    container_of(res, struct vmw_user_context, res);
+
+	kfree(ctx);
+}
+
+/**
+ * This function is called when user space has no more references on the
+ * base object. It releases the base-object's reference on the resource object.
+ */
+
+static void vmw_user_context_base_release(struct ttm_base_object **p_base)
+{
+	struct ttm_base_object *base = *p_base;
+	struct vmw_user_context *ctx =
+	    container_of(base, struct vmw_user_context, base);
+	struct vmw_resource *res = &ctx->res;
+
+	*p_base = NULL;
+	vmw_resource_unreference(&res);
+}
+
+int vmw_context_destroy_ioctl(struct drm_device *dev, void *data,
+			      struct drm_file *file_priv)
+{
+	struct vmw_private *dev_priv = vmw_priv(dev);
+	struct vmw_resource *res;
+	struct vmw_user_context *ctx;
+	struct drm_vmw_context_arg *arg = (struct drm_vmw_context_arg *)data;
+	struct ttm_object_file *tfile = vmw_fpriv(file_priv)->tfile;
+	int ret = 0;
+
+	res = vmw_resource_lookup(dev_priv, &dev_priv->context_idr, arg->cid);
+	if (unlikely(res == NULL))
+		return -EINVAL;
+
+	if (res->res_free != &vmw_user_context_free) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	ctx = container_of(res, struct vmw_user_context, res);
+	if (ctx->base.tfile != tfile && !ctx->base.shareable) {
+		ret = -EPERM;
+		goto out;
+	}
+
+	ttm_ref_object_base_unref(tfile, ctx->base.hash.key, TTM_REF_USAGE);
+out:
+	vmw_resource_unreference(&res);
+	return ret;
+}
+
+int vmw_context_define_ioctl(struct drm_device *dev, void *data,
+			     struct drm_file *file_priv)
+{
+	struct vmw_private *dev_priv = vmw_priv(dev);
+	struct vmw_user_context *ctx = kmalloc(sizeof(*ctx), GFP_KERNEL);
+	struct vmw_resource *res;
+	struct vmw_resource *tmp;
+	struct drm_vmw_context_arg *arg = (struct drm_vmw_context_arg *)data;
+	struct ttm_object_file *tfile = vmw_fpriv(file_priv)->tfile;
+	int ret;
+
+	if (unlikely(ctx == NULL))
+		return -ENOMEM;
+
+	res = &ctx->res;
+	ctx->base.shareable = false;
+	ctx->base.tfile = NULL;
+
+	ret = vmw_context_init(dev_priv, res, vmw_user_context_free);
+	if (unlikely(ret != 0))
+		return ret;
+
+	tmp = vmw_resource_reference(&ctx->res);
+	ret = ttm_base_object_init(tfile, &ctx->base, false, VMW_RES_CONTEXT,
+				   &vmw_user_context_base_release, NULL);
+
+	if (unlikely(ret != 0)) {
+		vmw_resource_unreference(&tmp);
+		goto out_err;
+	}
+
+	arg->cid = res->id;
+out_err:
+	vmw_resource_unreference(&res);
+	return ret;
+
+}
+
+int vmw_context_check(struct vmw_private *dev_priv,
+		      struct ttm_object_file *tfile,
+		      int id)
+{
+	struct vmw_resource *res;
+	int ret = 0;
+
+	read_lock(&dev_priv->resource_lock);
+	res = idr_find(&dev_priv->context_idr, id);
+	if (res && res->avail) {
+		struct vmw_user_context *ctx =
+			container_of(res, struct vmw_user_context, res);
+		if (ctx->base.tfile != tfile && !ctx->base.shareable)
+			ret = -EPERM;
+	} else
+		ret = -EINVAL;
+	read_unlock(&dev_priv->resource_lock);
+
+	return ret;
+}
+
+
+/**
+ * Surface management.
+ */
+
+static void vmw_hw_surface_destroy(struct vmw_resource *res)
+{
+
+	struct vmw_private *dev_priv = res->dev_priv;
+	struct {
+		SVGA3dCmdHeader header;
+		SVGA3dCmdDestroySurface body;
+	} *cmd = vmw_fifo_reserve(dev_priv, sizeof(*cmd));
+
+	if (unlikely(cmd == NULL)) {
+		DRM_ERROR("Failed reserving FIFO space for surface "
+			  "destruction.\n");
+		return;
+	}
+
+	cmd->header.id = cpu_to_le32(SVGA_3D_CMD_SURFACE_DESTROY);
+	cmd->header.size = cpu_to_le32(sizeof(cmd->body));
+	cmd->body.sid = cpu_to_le32(res->id);
+
+	vmw_fifo_commit(dev_priv, sizeof(*cmd));
+}
+
+void vmw_surface_res_free(struct vmw_resource *res)
+{
+	struct vmw_surface *srf = container_of(res, struct vmw_surface, res);
+
+	kfree(srf->sizes);
+	kfree(srf->snooper.image);
+	kfree(srf);
+}
+
+int vmw_surface_init(struct vmw_private *dev_priv,
+		     struct vmw_surface *srf,
+		     void (*res_free) (struct vmw_resource *res))
+{
+	int ret;
+	struct {
+		SVGA3dCmdHeader header;
+		SVGA3dCmdDefineSurface body;
+	} *cmd;
+	SVGA3dSize *cmd_size;
+	struct vmw_resource *res = &srf->res;
+	struct drm_vmw_size *src_size;
+	size_t submit_size;
+	uint32_t cmd_len;
+	int i;
+
+	BUG_ON(res_free == NULL);
+	ret = vmw_resource_init(dev_priv, res, &dev_priv->surface_idr,
+				VMW_RES_SURFACE, res_free);
+
+	if (unlikely(ret != 0)) {
+		res_free(res);
+		return ret;
+	}
+
+	submit_size = sizeof(*cmd) + srf->num_sizes * sizeof(SVGA3dSize);
+	cmd_len = sizeof(cmd->body) + srf->num_sizes * sizeof(SVGA3dSize);
+
+	cmd = vmw_fifo_reserve(dev_priv, submit_size);
+	if (unlikely(cmd == NULL)) {
+		DRM_ERROR("Fifo reserve failed for create surface.\n");
+		vmw_resource_unreference(&res);
+		return -ENOMEM;
+	}
+
+	cmd->header.id = cpu_to_le32(SVGA_3D_CMD_SURFACE_DEFINE);
+	cmd->header.size = cpu_to_le32(cmd_len);
+	cmd->body.sid = cpu_to_le32(res->id);
+	cmd->body.surfaceFlags = cpu_to_le32(srf->flags);
+	cmd->body.format = cpu_to_le32(srf->format);
+	for (i = 0; i < DRM_VMW_MAX_SURFACE_FACES; ++i) {
+		cmd->body.face[i].numMipLevels =
+		    cpu_to_le32(srf->mip_levels[i]);
+	}
+
+	cmd += 1;
+	cmd_size = (SVGA3dSize *) cmd;
+	src_size = srf->sizes;
+
+	for (i = 0; i < srf->num_sizes; ++i, cmd_size++, src_size++) {
+		cmd_size->width = cpu_to_le32(src_size->width);
+		cmd_size->height = cpu_to_le32(src_size->height);
+		cmd_size->depth = cpu_to_le32(src_size->depth);
+	}
+
+	vmw_fifo_commit(dev_priv, submit_size);
+	vmw_resource_activate(res, vmw_hw_surface_destroy);
+	return 0;
+}
+
+static void vmw_user_surface_free(struct vmw_resource *res)
+{
+	struct vmw_surface *srf = container_of(res, struct vmw_surface, res);
+	struct vmw_user_surface *user_srf =
+	    container_of(srf, struct vmw_user_surface, srf);
+
+	kfree(srf->sizes);
+	kfree(srf->snooper.image);
+	kfree(user_srf);
+}
+
+int vmw_user_surface_lookup_handle(struct vmw_private *dev_priv,
+				   struct ttm_object_file *tfile,
+				   uint32_t handle, struct vmw_surface **out)
+{
+	struct vmw_resource *res;
+	struct vmw_surface *srf;
+	struct vmw_user_surface *user_srf;
+	struct ttm_base_object *base;
+	int ret = -EINVAL;
+
+	base = ttm_base_object_lookup(tfile, handle);
+	if (unlikely(base == NULL))
+		return -EINVAL;
+
+	if (unlikely(base->object_type != VMW_RES_SURFACE))
+		goto out_bad_resource;
+
+	user_srf = container_of(base, struct vmw_user_surface, base);
+	srf = &user_srf->srf;
+	res = &srf->res;
+
+	read_lock(&dev_priv->resource_lock);
+
+	if (!res->avail || res->res_free != &vmw_user_surface_free) {
+		read_unlock(&dev_priv->resource_lock);
+		goto out_bad_resource;
+	}
+
+	kref_get(&res->kref);
+	read_unlock(&dev_priv->resource_lock);
+
+	*out = srf;
+	ret = 0;
+
+out_bad_resource:
+	ttm_base_object_unref(&base);
+
+	return ret;
+}
+
+static void vmw_user_surface_base_release(struct ttm_base_object **p_base)
+{
+	struct ttm_base_object *base = *p_base;
+	struct vmw_user_surface *user_srf =
+	    container_of(base, struct vmw_user_surface, base);
+	struct vmw_resource *res = &user_srf->srf.res;
+
+	*p_base = NULL;
+	vmw_resource_unreference(&res);
+}
+
+int vmw_surface_destroy_ioctl(struct drm_device *dev, void *data,
+			      struct drm_file *file_priv)
+{
+	struct drm_vmw_surface_arg *arg = (struct drm_vmw_surface_arg *)data;
+	struct ttm_object_file *tfile = vmw_fpriv(file_priv)->tfile;
+
+	return ttm_ref_object_base_unref(tfile, arg->sid, TTM_REF_USAGE);
+}
+
+int vmw_surface_define_ioctl(struct drm_device *dev, void *data,
+			     struct drm_file *file_priv)
+{
+	struct vmw_private *dev_priv = vmw_priv(dev);
+	struct vmw_user_surface *user_srf =
+	    kmalloc(sizeof(*user_srf), GFP_KERNEL);
+	struct vmw_surface *srf;
+	struct vmw_resource *res;
+	struct vmw_resource *tmp;
+	union drm_vmw_surface_create_arg *arg =
+	    (union drm_vmw_surface_create_arg *)data;
+	struct drm_vmw_surface_create_req *req = &arg->req;
+	struct drm_vmw_surface_arg *rep = &arg->rep;
+	struct ttm_object_file *tfile = vmw_fpriv(file_priv)->tfile;
+	struct drm_vmw_size __user *user_sizes;
+	int ret;
+	int i;
+
+	if (unlikely(user_srf == NULL))
+		return -ENOMEM;
+
+	srf = &user_srf->srf;
+	res = &srf->res;
+
+	srf->flags = req->flags;
+	srf->format = req->format;
+	srf->scanout = req->scanout;
+	memcpy(srf->mip_levels, req->mip_levels, sizeof(srf->mip_levels));
+	srf->num_sizes = 0;
+	for (i = 0; i < DRM_VMW_MAX_SURFACE_FACES; ++i)
+		srf->num_sizes += srf->mip_levels[i];
+
+	if (srf->num_sizes > DRM_VMW_MAX_SURFACE_FACES *
+	    DRM_VMW_MAX_MIP_LEVELS) {
+		ret = -EINVAL;
+		goto out_err0;
+	}
+
+	srf->sizes = kmalloc(srf->num_sizes * sizeof(*srf->sizes), GFP_KERNEL);
+	if (unlikely(srf->sizes == NULL)) {
+		ret = -ENOMEM;
+		goto out_err0;
+	}
+
+	user_sizes = (struct drm_vmw_size __user *)(unsigned long)
+	    req->size_addr;
+
+	ret = copy_from_user(srf->sizes, user_sizes,
+			     srf->num_sizes * sizeof(*srf->sizes));
+	if (unlikely(ret != 0))
+		goto out_err1;
+
+	if (srf->scanout &&
+	    srf->num_sizes == 1 &&
+	    srf->sizes[0].width == 64 &&
+	    srf->sizes[0].height == 64 &&
+	    srf->format == SVGA3D_A8R8G8B8) {
+
+		srf->snooper.image = kmalloc(64 * 64 * 4, GFP_KERNEL);
+		/* clear the image */
+		if (srf->snooper.image) {
+			memset(srf->snooper.image, 0x00, 64 * 64 * 4);
+		} else {
+			DRM_ERROR("Failed to allocate cursor_image\n");
+			ret = -ENOMEM;
+			goto out_err1;
+		}
+	} else {
+		srf->snooper.image = NULL;
+	}
+	srf->snooper.crtc = NULL;
+
+	user_srf->base.shareable = false;
+	user_srf->base.tfile = NULL;
+
+	/**
+	 * From this point, the generic resource management functions
+	 * destroy the object on failure.
+	 */
+
+	ret = vmw_surface_init(dev_priv, srf, vmw_user_surface_free);
+	if (unlikely(ret != 0))
+		return ret;
+
+	tmp = vmw_resource_reference(&srf->res);
+	ret = ttm_base_object_init(tfile, &user_srf->base,
+				   req->shareable, VMW_RES_SURFACE,
+				   &vmw_user_surface_base_release, NULL);
+
+	if (unlikely(ret != 0)) {
+		vmw_resource_unreference(&tmp);
+		vmw_resource_unreference(&res);
+		return ret;
+	}
+
+	rep->sid = user_srf->base.hash.key;
+	if (rep->sid == SVGA3D_INVALID_ID)
+		DRM_ERROR("Created bad Surface ID.\n");
+
+	vmw_resource_unreference(&res);
+	return 0;
+out_err1:
+	kfree(srf->sizes);
+out_err0:
+	kfree(user_srf);
+	return ret;
+}
+
+int vmw_surface_reference_ioctl(struct drm_device *dev, void *data,
+				struct drm_file *file_priv)
+{
+	union drm_vmw_surface_reference_arg *arg =
+	    (union drm_vmw_surface_reference_arg *)data;
+	struct drm_vmw_surface_arg *req = &arg->req;
+	struct drm_vmw_surface_create_req *rep = &arg->rep;
+	struct ttm_object_file *tfile = vmw_fpriv(file_priv)->tfile;
+	struct vmw_surface *srf;
+	struct vmw_user_surface *user_srf;
+	struct drm_vmw_size __user *user_sizes;
+	struct ttm_base_object *base;
+	int ret = -EINVAL;
+
+	base = ttm_base_object_lookup(tfile, req->sid);
+	if (unlikely(base == NULL)) {
+		DRM_ERROR("Could not find surface to reference.\n");
+		return -EINVAL;
+	}
+
+	if (unlikely(base->object_type != VMW_RES_SURFACE))
+		goto out_bad_resource;
+
+	user_srf = container_of(base, struct vmw_user_surface, base);
+	srf = &user_srf->srf;
+
+	ret = ttm_ref_object_add(tfile, &user_srf->base, TTM_REF_USAGE, NULL);
+	if (unlikely(ret != 0)) {
+		DRM_ERROR("Could not add a reference to a surface.\n");
+		goto out_no_reference;
+	}
+
+	rep->flags = srf->flags;
+	rep->format = srf->format;
+	memcpy(rep->mip_levels, srf->mip_levels, sizeof(srf->mip_levels));
+	user_sizes = (struct drm_vmw_size __user *)(unsigned long)
+	    rep->size_addr;
+
+	if (user_sizes)
+		ret = copy_to_user(user_sizes, srf->sizes,
+				   srf->num_sizes * sizeof(*srf->sizes));
+	if (unlikely(ret != 0))
+		DRM_ERROR("copy_to_user failed %p %u\n",
+			  user_sizes, srf->num_sizes);
+out_bad_resource:
+out_no_reference:
+	ttm_base_object_unref(&base);
+
+	return ret;
+}
+
+int vmw_surface_check(struct vmw_private *dev_priv,
+		      struct ttm_object_file *tfile,
+		      uint32_t handle, int *id)
+{
+	struct ttm_base_object *base;
+	struct vmw_user_surface *user_srf;
+
+	int ret = -EPERM;
+
+	base = ttm_base_object_lookup(tfile, handle);
+	if (unlikely(base == NULL))
+		return -EINVAL;
+
+	if (unlikely(base->object_type != VMW_RES_SURFACE))
+		goto out_bad_surface;
+
+	user_srf = container_of(base, struct vmw_user_surface, base);
+	*id = user_srf->srf.res.id;
+	ret = 0;
+
+out_bad_surface:
+	/**
+	 * FIXME: May deadlock here when called from the
+	 * command parsing code.
+	 */
+
+	ttm_base_object_unref(&base);
+	return ret;
+}
+
+/**
+ * Buffer management.
+ */
+
+static size_t vmw_dmabuf_acc_size(struct ttm_bo_global *glob,
+				  unsigned long num_pages)
+{
+	static size_t bo_user_size = ~0;
+
+	size_t page_array_size =
+	    (num_pages * sizeof(void *) + PAGE_SIZE - 1) & PAGE_MASK;
+
+	if (unlikely(bo_user_size == ~0)) {
+		bo_user_size = glob->ttm_bo_extra_size +
+		    ttm_round_pot(sizeof(struct vmw_dma_buffer));
+	}
+
+	return bo_user_size + page_array_size;
+}
+
+void vmw_dmabuf_gmr_unbind(struct ttm_buffer_object *bo)
+{
+	struct vmw_dma_buffer *vmw_bo = vmw_dma_buffer(bo);
+	struct ttm_bo_global *glob = bo->glob;
+	struct vmw_private *dev_priv =
+		container_of(bo->bdev, struct vmw_private, bdev);
+
+	if (vmw_bo->gmr_bound) {
+		vmw_gmr_unbind(dev_priv, vmw_bo->gmr_id);
+		spin_lock(&glob->lru_lock);
+		ida_remove(&dev_priv->gmr_ida, vmw_bo->gmr_id);
+		spin_unlock(&glob->lru_lock);
+		vmw_bo->gmr_bound = false;
+	}
+}
+
+void vmw_dmabuf_bo_free(struct ttm_buffer_object *bo)
+{
+	struct vmw_dma_buffer *vmw_bo = vmw_dma_buffer(bo);
+	struct ttm_bo_global *glob = bo->glob;
+
+	vmw_dmabuf_gmr_unbind(bo);
+	ttm_mem_global_free(glob->mem_glob, bo->acc_size);
+	kfree(vmw_bo);
+}
+
+int vmw_dmabuf_init(struct vmw_private *dev_priv,
+		    struct vmw_dma_buffer *vmw_bo,
+		    size_t size, struct ttm_placement *placement,
+		    bool interruptible,
+		    void (*bo_free) (struct ttm_buffer_object *bo))
+{
+	struct ttm_bo_device *bdev = &dev_priv->bdev;
+	struct ttm_mem_global *mem_glob = bdev->glob->mem_glob;
+	size_t acc_size;
+	int ret;
+
+	BUG_ON(!bo_free);
+
+	acc_size =
+	    vmw_dmabuf_acc_size(bdev->glob,
+				(size + PAGE_SIZE - 1) >> PAGE_SHIFT);
+
+	ret = ttm_mem_global_alloc(mem_glob, acc_size, false, false);
+	if (unlikely(ret != 0)) {
+		/* we must free the bo here as
+		 * ttm_buffer_object_init does so as well */
+		bo_free(&vmw_bo->base);
+		return ret;
+	}
+
+	memset(vmw_bo, 0, sizeof(*vmw_bo));
+
+	INIT_LIST_HEAD(&vmw_bo->gmr_lru);
+	INIT_LIST_HEAD(&vmw_bo->validate_list);
+	vmw_bo->gmr_id = 0;
+	vmw_bo->gmr_bound = false;
+
+	ret = ttm_bo_init(bdev, &vmw_bo->base, size,
+			  ttm_bo_type_device, placement,
+			  0, 0, interruptible,
+			  NULL, acc_size, bo_free);
+	return ret;
+}
+
+static void vmw_user_dmabuf_destroy(struct ttm_buffer_object *bo)
+{
+	struct vmw_user_dma_buffer *vmw_user_bo = vmw_user_dma_buffer(bo);
+	struct ttm_bo_global *glob = bo->glob;
+
+	vmw_dmabuf_gmr_unbind(bo);
+	ttm_mem_global_free(glob->mem_glob, bo->acc_size);
+	kfree(vmw_user_bo);
+}
+
+static void vmw_user_dmabuf_release(struct ttm_base_object **p_base)
+{
+	struct vmw_user_dma_buffer *vmw_user_bo;
+	struct ttm_base_object *base = *p_base;
+	struct ttm_buffer_object *bo;
+
+	*p_base = NULL;
+
+	if (unlikely(base == NULL))
+		return;
+
+	vmw_user_bo = container_of(base, struct vmw_user_dma_buffer, base);
+	bo = &vmw_user_bo->dma.base;
+	ttm_bo_unref(&bo);
+}
+
+int vmw_dmabuf_alloc_ioctl(struct drm_device *dev, void *data,
+			   struct drm_file *file_priv)
+{
+	struct vmw_private *dev_priv = vmw_priv(dev);
+	union drm_vmw_alloc_dmabuf_arg *arg =
+	    (union drm_vmw_alloc_dmabuf_arg *)data;
+	struct drm_vmw_alloc_dmabuf_req *req = &arg->req;
+	struct drm_vmw_dmabuf_rep *rep = &arg->rep;
+	struct vmw_user_dma_buffer *vmw_user_bo;
+	struct ttm_buffer_object *tmp;
+	struct vmw_master *vmaster = vmw_master(file_priv->master);
+	int ret;
+
+	vmw_user_bo = kzalloc(sizeof(*vmw_user_bo), GFP_KERNEL);
+	if (unlikely(vmw_user_bo == NULL))
+		return -ENOMEM;
+
+	ret = ttm_read_lock(&vmaster->lock, true);
+	if (unlikely(ret != 0)) {
+		kfree(vmw_user_bo);
+		return ret;
+	}
+
+	ret = vmw_dmabuf_init(dev_priv, &vmw_user_bo->dma, req->size,
+			      &vmw_vram_sys_placement, true,
+			      &vmw_user_dmabuf_destroy);
+	if (unlikely(ret != 0))
+		return ret;
+
+	tmp = ttm_bo_reference(&vmw_user_bo->dma.base);
+	ret = ttm_base_object_init(vmw_fpriv(file_priv)->tfile,
+				   &vmw_user_bo->base,
+				   false,
+				   ttm_buffer_type,
+				   &vmw_user_dmabuf_release, NULL);
+	if (unlikely(ret != 0)) {
+		ttm_bo_unref(&tmp);
+	} else {
+		rep->handle = vmw_user_bo->base.hash.key;
+		rep->map_handle = vmw_user_bo->dma.base.addr_space_offset;
+		rep->cur_gmr_id = vmw_user_bo->base.hash.key;
+		rep->cur_gmr_offset = 0;
+	}
+	ttm_bo_unref(&tmp);
+
+	ttm_read_unlock(&vmaster->lock);
+
+	return 0;
+}
+
+int vmw_dmabuf_unref_ioctl(struct drm_device *dev, void *data,
+			   struct drm_file *file_priv)
+{
+	struct drm_vmw_unref_dmabuf_arg *arg =
+	    (struct drm_vmw_unref_dmabuf_arg *)data;
+
+	return ttm_ref_object_base_unref(vmw_fpriv(file_priv)->tfile,
+					 arg->handle,
+					 TTM_REF_USAGE);
+}
+
+uint32_t vmw_dmabuf_validate_node(struct ttm_buffer_object *bo,
+				  uint32_t cur_validate_node)
+{
+	struct vmw_dma_buffer *vmw_bo = vmw_dma_buffer(bo);
+
+	if (likely(vmw_bo->on_validate_list))
+		return vmw_bo->cur_validate_node;
+
+	vmw_bo->cur_validate_node = cur_validate_node;
+	vmw_bo->on_validate_list = true;
+
+	return cur_validate_node;
+}
+
+void vmw_dmabuf_validate_clear(struct ttm_buffer_object *bo)
+{
+	struct vmw_dma_buffer *vmw_bo = vmw_dma_buffer(bo);
+
+	vmw_bo->on_validate_list = false;
+}
+
+uint32_t vmw_dmabuf_gmr(struct ttm_buffer_object *bo)
+{
+	struct vmw_dma_buffer *vmw_bo;
+
+	if (bo->mem.mem_type == TTM_PL_VRAM)
+		return SVGA_GMR_FRAMEBUFFER;
+
+	vmw_bo = vmw_dma_buffer(bo);
+
+	return (vmw_bo->gmr_bound) ? vmw_bo->gmr_id : SVGA_GMR_NULL;
+}
+
+void vmw_dmabuf_set_gmr(struct ttm_buffer_object *bo, uint32_t id)
+{
+	struct vmw_dma_buffer *vmw_bo = vmw_dma_buffer(bo);
+	vmw_bo->gmr_bound = true;
+	vmw_bo->gmr_id = id;
+}
+
+int vmw_user_dmabuf_lookup(struct ttm_object_file *tfile,
+			   uint32_t handle, struct vmw_dma_buffer **out)
+{
+	struct vmw_user_dma_buffer *vmw_user_bo;
+	struct ttm_base_object *base;
+
+	base = ttm_base_object_lookup(tfile, handle);
+	if (unlikely(base == NULL)) {
+		printk(KERN_ERR "Invalid buffer object handle 0x%08lx.\n",
+		       (unsigned long)handle);
+		return -ESRCH;
+	}
+
+	if (unlikely(base->object_type != ttm_buffer_type)) {
+		ttm_base_object_unref(&base);
+		printk(KERN_ERR "Invalid buffer object handle 0x%08lx.\n",
+		       (unsigned long)handle);
+		return -EINVAL;
+	}
+
+	vmw_user_bo = container_of(base, struct vmw_user_dma_buffer, base);
+	(void)ttm_bo_reference(&vmw_user_bo->dma.base);
+	ttm_base_object_unref(&base);
+	*out = &vmw_user_bo->dma;
+
+	return 0;
+}
+
+/**
+ * TODO: Implement a gmr id eviction mechanism. Currently we just fail
+ * when we're out of ids, causing GMR space to be allocated
+ * out of VRAM.
+ */
+
+int vmw_gmr_id_alloc(struct vmw_private *dev_priv, uint32_t *p_id)
+{
+	struct ttm_bo_global *glob = dev_priv->bdev.glob;
+	int id;
+	int ret;
+
+	do {
+		if (unlikely(ida_pre_get(&dev_priv->gmr_ida, GFP_KERNEL) == 0))
+			return -ENOMEM;
+
+		spin_lock(&glob->lru_lock);
+		ret = ida_get_new(&dev_priv->gmr_ida, &id);
+		spin_unlock(&glob->lru_lock);
+	} while (ret == -EAGAIN);
+
+	if (unlikely(ret != 0))
+		return ret;
+
+	if (unlikely(id >= dev_priv->max_gmr_ids)) {
+		spin_lock(&glob->lru_lock);
+		ida_remove(&dev_priv->gmr_ida, id);
+		spin_unlock(&glob->lru_lock);
+		return -EBUSY;
+	}
+
+	*p_id = (uint32_t) id;
+	return 0;
+}
+
+/*
+ * Stream managment
+ */
+
+static void vmw_stream_destroy(struct vmw_resource *res)
+{
+	struct vmw_private *dev_priv = res->dev_priv;
+	struct vmw_stream *stream;
+	int ret;
+
+	DRM_INFO("%s: unref\n", __func__);
+	stream = container_of(res, struct vmw_stream, res);
+
+	ret = vmw_overlay_unref(dev_priv, stream->stream_id);
+	WARN_ON(ret != 0);
+}
+
+static int vmw_stream_init(struct vmw_private *dev_priv,
+			   struct vmw_stream *stream,
+			   void (*res_free) (struct vmw_resource *res))
+{
+	struct vmw_resource *res = &stream->res;
+	int ret;
+
+	ret = vmw_resource_init(dev_priv, res, &dev_priv->stream_idr,
+				VMW_RES_STREAM, res_free);
+
+	if (unlikely(ret != 0)) {
+		if (res_free == NULL)
+			kfree(stream);
+		else
+			res_free(&stream->res);
+		return ret;
+	}
+
+	ret = vmw_overlay_claim(dev_priv, &stream->stream_id);
+	if (ret) {
+		vmw_resource_unreference(&res);
+		return ret;
+	}
+
+	DRM_INFO("%s: claimed\n", __func__);
+
+	vmw_resource_activate(&stream->res, vmw_stream_destroy);
+	return 0;
+}
+
+/**
+ * User-space context management:
+ */
+
+static void vmw_user_stream_free(struct vmw_resource *res)
+{
+	struct vmw_user_stream *stream =
+	    container_of(res, struct vmw_user_stream, stream.res);
+
+	kfree(stream);
+}
+
+/**
+ * This function is called when user space has no more references on the
+ * base object. It releases the base-object's reference on the resource object.
+ */
+
+static void vmw_user_stream_base_release(struct ttm_base_object **p_base)
+{
+	struct ttm_base_object *base = *p_base;
+	struct vmw_user_stream *stream =
+	    container_of(base, struct vmw_user_stream, base);
+	struct vmw_resource *res = &stream->stream.res;
+
+	*p_base = NULL;
+	vmw_resource_unreference(&res);
+}
+
+int vmw_stream_unref_ioctl(struct drm_device *dev, void *data,
+			   struct drm_file *file_priv)
+{
+	struct vmw_private *dev_priv = vmw_priv(dev);
+	struct vmw_resource *res;
+	struct vmw_user_stream *stream;
+	struct drm_vmw_stream_arg *arg = (struct drm_vmw_stream_arg *)data;
+	struct ttm_object_file *tfile = vmw_fpriv(file_priv)->tfile;
+	int ret = 0;
+
+	res = vmw_resource_lookup(dev_priv, &dev_priv->stream_idr, arg->stream_id);
+	if (unlikely(res == NULL))
+		return -EINVAL;
+
+	if (res->res_free != &vmw_user_stream_free) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	stream = container_of(res, struct vmw_user_stream, stream.res);
+	if (stream->base.tfile != tfile) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	ttm_ref_object_base_unref(tfile, stream->base.hash.key, TTM_REF_USAGE);
+out:
+	vmw_resource_unreference(&res);
+	return ret;
+}
+
+int vmw_stream_claim_ioctl(struct drm_device *dev, void *data,
+			   struct drm_file *file_priv)
+{
+	struct vmw_private *dev_priv = vmw_priv(dev);
+	struct vmw_user_stream *stream = kmalloc(sizeof(*stream), GFP_KERNEL);
+	struct vmw_resource *res;
+	struct vmw_resource *tmp;
+	struct drm_vmw_stream_arg *arg = (struct drm_vmw_stream_arg *)data;
+	struct ttm_object_file *tfile = vmw_fpriv(file_priv)->tfile;
+	int ret;
+
+	if (unlikely(stream == NULL))
+		return -ENOMEM;
+
+	res = &stream->stream.res;
+	stream->base.shareable = false;
+	stream->base.tfile = NULL;
+
+	ret = vmw_stream_init(dev_priv, &stream->stream, vmw_user_stream_free);
+	if (unlikely(ret != 0))
+		return ret;
+
+	tmp = vmw_resource_reference(res);
+	ret = ttm_base_object_init(tfile, &stream->base, false, VMW_RES_STREAM,
+				   &vmw_user_stream_base_release, NULL);
+
+	if (unlikely(ret != 0)) {
+		vmw_resource_unreference(&tmp);
+		goto out_err;
+	}
+
+	arg->stream_id = res->id;
+out_err:
+	vmw_resource_unreference(&res);
+	return ret;
+}
+
+int vmw_user_stream_lookup(struct vmw_private *dev_priv,
+			   struct ttm_object_file *tfile,
+			   uint32_t *inout_id, struct vmw_resource **out)
+{
+	struct vmw_user_stream *stream;
+	struct vmw_resource *res;
+	int ret;
+
+	res = vmw_resource_lookup(dev_priv, &dev_priv->stream_idr, *inout_id);
+	if (unlikely(res == NULL))
+		return -EINVAL;
+
+	if (res->res_free != &vmw_user_stream_free) {
+		ret = -EINVAL;
+		goto err_ref;
+	}
+
+	stream = container_of(res, struct vmw_user_stream, stream.res);
+	if (stream->base.tfile != tfile) {
+		ret = -EPERM;
+		goto err_ref;
+	}
+
+	*inout_id = stream->stream.stream_id;
+	*out = res;
+	return 0;
+err_ref:
+	vmw_resource_unreference(&res);
+	return ret;
+}
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_glue.c b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_glue.c
new file mode 100644
index 000000000000..e3df4adfb4d8
--- /dev/null
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_glue.c
@@ -0,0 +1,99 @@
+/**************************************************************************
+ *
+ * Copyright © 2009 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "drmP.h"
+#include "vmwgfx_drv.h"
+
+int vmw_mmap(struct file *filp, struct vm_area_struct *vma)
+{
+	struct drm_file *file_priv;
+	struct vmw_private *dev_priv;
+
+	if (unlikely(vma->vm_pgoff < VMWGFX_FILE_PAGE_OFFSET)) {
+		if (vmw_fifo_mmap(filp, vma) == 0)
+			return 0;
+		return drm_mmap(filp, vma);
+	}
+
+	file_priv = (struct drm_file *)filp->private_data;
+	dev_priv = vmw_priv(file_priv->minor->dev);
+	return ttm_bo_mmap(filp, vma, &dev_priv->bdev);
+}
+
+static int vmw_ttm_mem_global_init(struct ttm_global_reference *ref)
+{
+	DRM_INFO("global init.\n");
+	return ttm_mem_global_init(ref->object);
+}
+
+static void vmw_ttm_mem_global_release(struct ttm_global_reference *ref)
+{
+	ttm_mem_global_release(ref->object);
+}
+
+int vmw_ttm_global_init(struct vmw_private *dev_priv)
+{
+	struct ttm_global_reference *global_ref;
+	int ret;
+
+	global_ref = &dev_priv->mem_global_ref;
+	global_ref->global_type = TTM_GLOBAL_TTM_MEM;
+	global_ref->size = sizeof(struct ttm_mem_global);
+	global_ref->init = &vmw_ttm_mem_global_init;
+	global_ref->release = &vmw_ttm_mem_global_release;
+
+	ret = ttm_global_item_ref(global_ref);
+	if (unlikely(ret != 0)) {
+		DRM_ERROR("Failed setting up TTM memory accounting.\n");
+		return ret;
+	}
+
+	dev_priv->bo_global_ref.mem_glob =
+		dev_priv->mem_global_ref.object;
+	global_ref = &dev_priv->bo_global_ref.ref;
+	global_ref->global_type = TTM_GLOBAL_TTM_BO;
+	global_ref->size = sizeof(struct ttm_bo_global);
+	global_ref->init = &ttm_bo_global_init;
+	global_ref->release = &ttm_bo_global_release;
+		ret = ttm_global_item_ref(global_ref);
+
+	if (unlikely(ret != 0)) {
+		DRM_ERROR("Failed setting up TTM buffer objects.\n");
+		goto out_no_bo;
+	}
+
+	return 0;
+out_no_bo:
+	ttm_global_item_unref(&dev_priv->mem_global_ref);
+	return ret;
+}
+
+void vmw_ttm_global_release(struct vmw_private *dev_priv)
+{
+	ttm_global_item_unref(&dev_priv->bo_global_ref.ref);
+	ttm_global_item_unref(&dev_priv->mem_global_ref);
+}
diff --git a/drivers/gpu/vga/Kconfig b/drivers/gpu/vga/Kconfig
index 790e675b13eb..61ab4daf0bbb 100644
--- a/drivers/gpu/vga/Kconfig
+++ b/drivers/gpu/vga/Kconfig
@@ -8,3 +8,22 @@ config VGA_ARB
 	  are accessed at same time they need some kind of coordination. Please
 	  see Documentation/vgaarbiter.txt for more details. Select this to
 	  enable VGA arbiter.
+
+config VGA_ARB_MAX_GPUS
+	int "Maximum number of GPUs"
+	default 16
+	depends on VGA_ARB
+	help
+	  Reserves space in the kernel to maintain resource locking for
+	  multiple GPUS.  The overhead for each GPU is very small.
+
+config VGA_SWITCHEROO
+	bool "Laptop Hybrid Grapics - GPU switching support"
+	depends on X86
+	depends on ACPI
+	help
+	  Many laptops released in 2008/9/10 have two gpus with a multiplxer
+	  to switch between them. This adds support for dynamic switching when
+          X isn't running and delayed switching until the next logoff. This
+	  features is called hybrid graphics, ATI PowerXpress, and Nvidia
+	  HybridPower.
diff --git a/drivers/gpu/vga/Makefile b/drivers/gpu/vga/Makefile
index 7cc8c1ed645b..14ca30b75d0a 100644
--- a/drivers/gpu/vga/Makefile
+++ b/drivers/gpu/vga/Makefile
@@ -1 +1,2 @@
 obj-$(CONFIG_VGA_ARB)  += vgaarb.o
+obj-$(CONFIG_VGA_SWITCHEROO) += vga_switcheroo.o
diff --git a/drivers/gpu/vga/vga_switcheroo.c b/drivers/gpu/vga/vga_switcheroo.c
new file mode 100644
index 000000000000..d6d1149d525d
--- /dev/null
+++ b/drivers/gpu/vga/vga_switcheroo.c
@@ -0,0 +1,450 @@
+/*
+ * Copyright (c) 2010 Red Hat Inc.
+ * Author : Dave Airlie <airlied@redhat.com>
+ *
+ *
+ * Licensed under GPLv2
+ *
+ * vga_switcheroo.c - Support for laptop with dual GPU using one set of outputs
+
+ Switcher interface - methods require for ATPX and DCM
+ - switchto - this throws the output MUX switch
+ - discrete_set_power - sets the power state for the discrete card
+
+ GPU driver interface
+ - set_gpu_state - this should do the equiv of s/r for the card
+		  - this should *not* set the discrete power state
+ - switch_check  - check if the device is in a position to switch now
+ */
+
+#include <linux/module.h>
+#include <linux/dmi.h>
+#include <linux/seq_file.h>
+#include <linux/uaccess.h>
+#include <linux/fs.h>
+#include <linux/debugfs.h>
+#include <linux/fb.h>
+
+#include <linux/pci.h>
+#include <linux/vga_switcheroo.h>
+
+struct vga_switcheroo_client {
+	struct pci_dev *pdev;
+	struct fb_info *fb_info;
+	int pwr_state;
+	void (*set_gpu_state)(struct pci_dev *pdev, enum vga_switcheroo_state);
+	bool (*can_switch)(struct pci_dev *pdev);
+	int id;
+	bool active;
+};
+
+static DEFINE_MUTEX(vgasr_mutex);
+
+struct vgasr_priv {
+
+	bool active;
+	bool delayed_switch_active;
+	enum vga_switcheroo_client_id delayed_client_id;
+
+	struct dentry *debugfs_root;
+	struct dentry *switch_file;
+
+	int registered_clients;
+	struct vga_switcheroo_client clients[VGA_SWITCHEROO_MAX_CLIENTS];
+
+	struct vga_switcheroo_handler *handler;
+};
+
+static int vga_switcheroo_debugfs_init(struct vgasr_priv *priv);
+static void vga_switcheroo_debugfs_fini(struct vgasr_priv *priv);
+
+/* only one switcheroo per system */
+static struct vgasr_priv vgasr_priv;
+
+int vga_switcheroo_register_handler(struct vga_switcheroo_handler *handler)
+{
+	mutex_lock(&vgasr_mutex);
+	if (vgasr_priv.handler) {
+		mutex_unlock(&vgasr_mutex);
+		return -EINVAL;
+	}
+
+	vgasr_priv.handler = handler;
+	mutex_unlock(&vgasr_mutex);
+	return 0;
+}
+EXPORT_SYMBOL(vga_switcheroo_register_handler);
+
+void vga_switcheroo_unregister_handler(void)
+{
+	mutex_lock(&vgasr_mutex);
+	vgasr_priv.handler = NULL;
+	mutex_unlock(&vgasr_mutex);
+}
+EXPORT_SYMBOL(vga_switcheroo_unregister_handler);
+
+static void vga_switcheroo_enable(void)
+{
+	int i;
+	int ret;
+	/* call the handler to init */
+	vgasr_priv.handler->init();
+
+	for (i = 0; i < VGA_SWITCHEROO_MAX_CLIENTS; i++) {
+		ret = vgasr_priv.handler->get_client_id(vgasr_priv.clients[i].pdev);
+		if (ret < 0)
+			return;
+
+		vgasr_priv.clients[i].id = ret;
+	}
+	vga_switcheroo_debugfs_init(&vgasr_priv);
+	vgasr_priv.active = true;
+}
+
+int vga_switcheroo_register_client(struct pci_dev *pdev,
+				   void (*set_gpu_state)(struct pci_dev *pdev, enum vga_switcheroo_state),
+				   bool (*can_switch)(struct pci_dev *pdev))
+{
+	int index;
+
+	mutex_lock(&vgasr_mutex);
+	/* don't do IGD vs DIS here */
+	if (vgasr_priv.registered_clients & 1)
+		index = 1;
+	else
+		index = 0;
+
+	vgasr_priv.clients[index].pwr_state = VGA_SWITCHEROO_ON;
+	vgasr_priv.clients[index].pdev = pdev;
+	vgasr_priv.clients[index].set_gpu_state = set_gpu_state;
+	vgasr_priv.clients[index].can_switch = can_switch;
+	vgasr_priv.clients[index].id = -1;
+	if (pdev->resource[PCI_ROM_RESOURCE].flags & IORESOURCE_ROM_SHADOW)
+		vgasr_priv.clients[index].active = true;
+
+	vgasr_priv.registered_clients |= (1 << index);
+
+	/* if we get two clients + handler */
+	if (vgasr_priv.registered_clients == 0x3 && vgasr_priv.handler) {
+		printk(KERN_INFO "vga_switcheroo: enabled\n");
+		vga_switcheroo_enable();
+	}
+	mutex_unlock(&vgasr_mutex);
+	return 0;
+}
+EXPORT_SYMBOL(vga_switcheroo_register_client);
+
+void vga_switcheroo_unregister_client(struct pci_dev *pdev)
+{
+	int i;
+
+	mutex_lock(&vgasr_mutex);
+	for (i = 0; i < VGA_SWITCHEROO_MAX_CLIENTS; i++) {
+		if (vgasr_priv.clients[i].pdev == pdev) {
+			vgasr_priv.registered_clients &= ~(1 << i);
+			break;
+		}
+	}
+
+	printk(KERN_INFO "vga_switcheroo: disabled\n");
+	vga_switcheroo_debugfs_fini(&vgasr_priv);
+	vgasr_priv.active = false;
+	mutex_unlock(&vgasr_mutex);
+}
+EXPORT_SYMBOL(vga_switcheroo_unregister_client);
+
+void vga_switcheroo_client_fb_set(struct pci_dev *pdev,
+				 struct fb_info *info)
+{
+	int i;
+
+	mutex_lock(&vgasr_mutex);
+	for (i = 0; i < VGA_SWITCHEROO_MAX_CLIENTS; i++) {
+		if (vgasr_priv.clients[i].pdev == pdev) {
+			vgasr_priv.clients[i].fb_info = info;
+			break;
+		}
+	}
+	mutex_unlock(&vgasr_mutex);
+}
+EXPORT_SYMBOL(vga_switcheroo_client_fb_set);
+
+static int vga_switcheroo_show(struct seq_file *m, void *v)
+{
+	int i;
+	mutex_lock(&vgasr_mutex);
+	for (i = 0; i < VGA_SWITCHEROO_MAX_CLIENTS; i++) {
+		seq_printf(m, "%d:%c:%s:%s\n", i,
+			   vgasr_priv.clients[i].active ? '+' : ' ',
+			   vgasr_priv.clients[i].pwr_state ? "Pwr" : "Off",
+			   pci_name(vgasr_priv.clients[i].pdev));
+	}
+	mutex_unlock(&vgasr_mutex);
+	return 0;
+}
+
+static int vga_switcheroo_debugfs_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, vga_switcheroo_show, NULL);
+}
+
+static int vga_switchon(struct vga_switcheroo_client *client)
+{
+	int ret;
+
+	ret = vgasr_priv.handler->power_state(client->id, VGA_SWITCHEROO_ON);
+	/* call the driver callback to turn on device */
+	client->set_gpu_state(client->pdev, VGA_SWITCHEROO_ON);
+	client->pwr_state = VGA_SWITCHEROO_ON;
+	return 0;
+}
+
+static int vga_switchoff(struct vga_switcheroo_client *client)
+{
+	/* call the driver callback to turn off device */
+	client->set_gpu_state(client->pdev, VGA_SWITCHEROO_OFF);
+	vgasr_priv.handler->power_state(client->id, VGA_SWITCHEROO_OFF);
+	client->pwr_state = VGA_SWITCHEROO_OFF;
+	return 0;
+}
+
+static int vga_switchto(struct vga_switcheroo_client *new_client)
+{
+	int ret;
+	int i;
+	struct vga_switcheroo_client *active = NULL;
+
+	if (new_client->active == true)
+		return 0;
+
+	for (i = 0; i < VGA_SWITCHEROO_MAX_CLIENTS; i++) {
+		if (vgasr_priv.clients[i].active == true) {
+			active = &vgasr_priv.clients[i];
+			break;
+		}
+	}
+	if (!active)
+		return 0;
+
+	/* power up the first device */
+	ret = pci_enable_device(new_client->pdev);
+	if (ret)
+		return ret;
+
+	if (new_client->pwr_state == VGA_SWITCHEROO_OFF)
+		vga_switchon(new_client);
+
+	/* swap shadow resource to denote boot VGA device has changed so X starts on new device */
+	active->active = false;
+
+	active->pdev->resource[PCI_ROM_RESOURCE].flags &= ~IORESOURCE_ROM_SHADOW;
+	new_client->pdev->resource[PCI_ROM_RESOURCE].flags |= IORESOURCE_ROM_SHADOW;
+
+	if (new_client->fb_info) {
+		struct fb_event event;
+		event.info = new_client->fb_info;
+		fb_notifier_call_chain(FB_EVENT_REMAP_ALL_CONSOLE, &event);
+	}
+
+	ret = vgasr_priv.handler->switchto(new_client->id);
+	if (ret)
+		return ret;
+
+	if (active->pwr_state == VGA_SWITCHEROO_ON)
+		vga_switchoff(active);
+
+	new_client->active = true;
+	return 0;
+}
+
+static ssize_t
+vga_switcheroo_debugfs_write(struct file *filp, const char __user *ubuf,
+			     size_t cnt, loff_t *ppos)
+{
+	char usercmd[64];
+	const char *pdev_name;
+	int i, ret;
+	bool delay = false, can_switch;
+	int client_id = -1;
+	struct vga_switcheroo_client *client = NULL;
+
+	if (cnt > 63)
+		cnt = 63;
+
+	if (copy_from_user(usercmd, ubuf, cnt))
+		return -EFAULT;
+
+	mutex_lock(&vgasr_mutex);
+
+	if (!vgasr_priv.active)
+		return -EINVAL;
+
+	/* pwr off the device not in use */
+	if (strncmp(usercmd, "OFF", 3) == 0) {
+		for (i = 0; i < VGA_SWITCHEROO_MAX_CLIENTS; i++) {
+			if (vgasr_priv.clients[i].active)
+				continue;
+			if (vgasr_priv.clients[i].pwr_state == VGA_SWITCHEROO_ON)
+				vga_switchoff(&vgasr_priv.clients[i]);
+		}
+		goto out;
+	}
+	/* pwr on the device not in use */
+	if (strncmp(usercmd, "ON", 2) == 0) {
+		for (i = 0; i < VGA_SWITCHEROO_MAX_CLIENTS; i++) {
+			if (vgasr_priv.clients[i].active)
+				continue;
+			if (vgasr_priv.clients[i].pwr_state == VGA_SWITCHEROO_OFF)
+				vga_switchon(&vgasr_priv.clients[i]);
+		}
+		goto out;
+	}
+
+	/* request a delayed switch - test can we switch now */
+	if (strncmp(usercmd, "DIGD", 4) == 0) {
+		client_id = VGA_SWITCHEROO_IGD;
+		delay = true;
+	}
+
+	if (strncmp(usercmd, "DDIS", 4) == 0) {
+		client_id = VGA_SWITCHEROO_DIS;
+		delay = true;
+	}
+
+	if (strncmp(usercmd, "IGD", 3) == 0)
+		client_id = VGA_SWITCHEROO_IGD;
+
+	if (strncmp(usercmd, "DIS", 3) == 0)
+		client_id = VGA_SWITCHEROO_DIS;
+
+	if (client_id == -1)
+		goto out;
+
+	for (i = 0; i < VGA_SWITCHEROO_MAX_CLIENTS; i++) {
+		if (vgasr_priv.clients[i].id == client_id) {
+			client = &vgasr_priv.clients[i];
+			break;
+		}
+	}
+
+	vgasr_priv.delayed_switch_active = false;
+	/* okay we want a switch - test if devices are willing to switch */
+	can_switch = true;
+	for (i = 0; i < VGA_SWITCHEROO_MAX_CLIENTS; i++) {
+		can_switch = vgasr_priv.clients[i].can_switch(vgasr_priv.clients[i].pdev);
+		if (can_switch == false) {
+			printk(KERN_ERR "vga_switcheroo: client %d refused switch\n", i);
+			break;
+		}
+	}
+
+	if (can_switch == false && delay == false)
+		goto out;
+
+	if (can_switch == true) {
+		pdev_name = pci_name(client->pdev);
+		ret = vga_switchto(client);
+		if (ret)
+			printk(KERN_ERR "vga_switcheroo: switching failed %d\n", ret);
+	} else {
+		printk(KERN_INFO "vga_switcheroo: setting delayed switch to client %d\n", client->id);
+		vgasr_priv.delayed_switch_active = true;
+		vgasr_priv.delayed_client_id = client_id;
+
+		/* we should at least power up the card to
+		   make the switch faster */
+		if (client->pwr_state == VGA_SWITCHEROO_OFF)
+			vga_switchon(client);
+	}
+
+out:
+	mutex_unlock(&vgasr_mutex);
+	return cnt;
+}
+
+static const struct file_operations vga_switcheroo_debugfs_fops = {
+	.owner = THIS_MODULE,
+	.open = vga_switcheroo_debugfs_open,
+	.write = vga_switcheroo_debugfs_write,
+	.read = seq_read,
+	.llseek = seq_lseek,
+	.release = single_release,
+};
+
+static void vga_switcheroo_debugfs_fini(struct vgasr_priv *priv)
+{
+	if (priv->switch_file) {
+		debugfs_remove(priv->switch_file);
+		priv->switch_file = NULL;
+	}
+	if (priv->debugfs_root) {
+		debugfs_remove(priv->debugfs_root);
+		priv->debugfs_root = NULL;
+	}
+}
+
+static int vga_switcheroo_debugfs_init(struct vgasr_priv *priv)
+{
+	/* already initialised */
+	if (priv->debugfs_root)
+		return 0;
+	priv->debugfs_root = debugfs_create_dir("vgaswitcheroo", NULL);
+
+	if (!priv->debugfs_root) {
+		printk(KERN_ERR "vga_switcheroo: Cannot create /sys/kernel/debug/vgaswitcheroo\n");
+		goto fail;
+	}
+
+	priv->switch_file = debugfs_create_file("switch", 0644,
+						priv->debugfs_root, NULL, &vga_switcheroo_debugfs_fops);
+	if (!priv->switch_file) {
+		printk(KERN_ERR "vga_switcheroo: cannot create /sys/kernel/debug/vgaswitcheroo/switch\n");
+		goto fail;
+	}
+	return 0;
+fail:
+	vga_switcheroo_debugfs_fini(priv);
+	return -1;
+}
+
+int vga_switcheroo_process_delayed_switch(void)
+{
+	struct vga_switcheroo_client *client = NULL;
+	const char *pdev_name;
+	bool can_switch = true;
+	int i;
+	int ret;
+	int err = -EINVAL;
+
+	mutex_lock(&vgasr_mutex);
+	if (!vgasr_priv.delayed_switch_active)
+		goto err;
+
+	printk(KERN_INFO "vga_switcheroo: processing delayed switch to %d\n", vgasr_priv.delayed_client_id);
+
+	for (i = 0; i < VGA_SWITCHEROO_MAX_CLIENTS; i++) {
+		if (vgasr_priv.clients[i].id == vgasr_priv.delayed_client_id)
+			client = &vgasr_priv.clients[i];
+		can_switch = vgasr_priv.clients[i].can_switch(vgasr_priv.clients[i].pdev);
+		if (can_switch == false) {
+			printk(KERN_ERR "vga_switcheroo: client %d refused switch\n", i);
+			break;
+		}
+	}
+
+	if (can_switch == false || client == NULL)
+		goto err;
+
+	pdev_name = pci_name(client->pdev);
+	ret = vga_switchto(client);
+	if (ret)
+		printk(KERN_ERR "vga_switcheroo: delayed switching failed %d\n", ret);
+
+	vgasr_priv.delayed_switch_active = false;
+	err = 0;
+err:
+	mutex_unlock(&vgasr_mutex);
+	return err;
+}
+EXPORT_SYMBOL(vga_switcheroo_process_delayed_switch);
+
diff --git a/drivers/gpu/vga/vgaarb.c b/drivers/gpu/vga/vgaarb.c
index 1ac0c93603c9..8827814d0735 100644
--- a/drivers/gpu/vga/vgaarb.c
+++ b/drivers/gpu/vga/vgaarb.c
@@ -688,7 +688,7 @@ EXPORT_SYMBOL(vga_client_register);
  * the arbiter.
  */
 
-#define MAX_USER_CARDS         16
+#define MAX_USER_CARDS         CONFIG_VGA_ARB_MAX_GPUS
 #define PCI_INVALID_CARD       ((struct pci_dev *)-1UL)
 
 /*
@@ -954,6 +954,7 @@ static ssize_t vga_arb_write(struct file *file, const char __user * buf,
 		}
 
 	} else if (strncmp(curr_pos, "target ", 7) == 0) {
+		struct pci_bus *pbus;
 		unsigned int domain, bus, devfn;
 		struct vga_device *vgadev;
 
@@ -961,7 +962,7 @@ static ssize_t vga_arb_write(struct file *file, const char __user * buf,
 		remaining -= 7;
 		pr_devel("client 0x%p called 'target'\n", priv);
 		/* if target is default */
-		if (!strncmp(buf, "default", 7))
+		if (!strncmp(curr_pos, "default", 7))
 			pdev = pci_dev_get(vga_default_device());
 		else {
 			if (!vga_pci_str_to_vars(curr_pos, remaining,
@@ -969,18 +970,31 @@ static ssize_t vga_arb_write(struct file *file, const char __user * buf,
 				ret_val = -EPROTO;
 				goto done;
 			}
-
-			pdev = pci_get_bus_and_slot(bus, devfn);
+			pr_devel("vgaarb: %s ==> %x:%x:%x.%x\n", curr_pos,
+				domain, bus, PCI_SLOT(devfn), PCI_FUNC(devfn));
+
+			pbus = pci_find_bus(domain, bus);
+			pr_devel("vgaarb: pbus %p\n", pbus);
+			if (pbus == NULL) {
+				pr_err("vgaarb: invalid PCI domain and/or bus address %x:%x\n",
+					domain, bus);
+				ret_val = -ENODEV;
+				goto done;
+			}
+			pdev = pci_get_slot(pbus, devfn);
+			pr_devel("vgaarb: pdev %p\n", pdev);
 			if (!pdev) {
-				pr_info("vgaarb: invalid PCI address!\n");
+				pr_err("vgaarb: invalid PCI address %x:%x\n",
+					bus, devfn);
 				ret_val = -ENODEV;
 				goto done;
 			}
 		}
 
 		vgadev = vgadev_find(pdev);
+		pr_devel("vgaarb: vgadev %p\n", vgadev);
 		if (vgadev == NULL) {
-			pr_info("vgaarb: this pci device is not a vga device\n");
+			pr_err("vgaarb: this pci device is not a vga device\n");
 			pci_dev_put(pdev);
 			ret_val = -ENODEV;
 			goto done;
@@ -998,7 +1012,8 @@ static ssize_t vga_arb_write(struct file *file, const char __user * buf,
 			}
 		}
 		if (i == MAX_USER_CARDS) {
-			pr_err("vgaarb: maximum user cards number reached!\n");
+			pr_err("vgaarb: maximum user cards (%d) number reached!\n",
+				MAX_USER_CARDS);
 			pci_dev_put(pdev);
 			/* XXX: which value to return? */
 			ret_val =  -ENOMEM;